Digital Room Correction Thread

[Jones_Rush]

I found a directX, parametric-eq plug-in, which allows, in real time, to program an unlimited (but depends on your cpu, I believe) number of filters (unlimited bands). The only problem is that I can only work with it "offline" currently, because I've only been able to activate it with a sound editing program like Cool Edit Pro. Working off line is not that bad, because I can save the corrected .wav file, and then play it with an ASIO player, and enjoy both worlds, but I really want to have an option to run it with an an audio player in real time. I think that the problem is a bit more complicated, because this is not a pure direct X plug in, but a VST plug in. I use Cakewalk VST v4.3 in order to make it compatible with DX, but not all programs can work with it.

http://www.elementalaudio.com/Images/eqmwaveplus.jpg

[rayf888]

Sounds cool. Where can I get it?

[kazushi]

I think MC9.1 supports (or is trying to support) DirectX Host Plugins. I've never used it and don't know anything about it. So, please check MC9 forum for more information.

[Esben]

Winamp has a DirectX Plugin Adapter Module (dsp_adaptx35.dll). You should try downloading that.

[Jones_Rush]

Damn it!!!. I've just found that this plug-in is not what I've wanted!. I thought it uses Fir filters, but to my dismay, it uses IIR filters.

Here is a small explanation about the difference between the two:
taken from here (http://www.seneschal.net/papers/restoration.htm):

IIR vs FIR:
This is a scary concept but here goes: All filters suck, some more than others. Whew, I’ve gotten that off my chest. Now, what’s the bottom line here? Well, filters or EQ can be built in two flavors, either Infinite Impulse Response (IIR) or Finite Impulse Response (FIR). Most filter that we interact with on a daily basis are of the IIR type. They trade off a frequency–dependent amplitude boost or cut in return for some amount of frequency–dependent phase shift, also known as group delay. Unfortunately, that’s a problem…although we expect them to operate in the frequency domain, they also muck up the time domain. To understand this group delay thing, think of your basic two way loudspeaker (see figure 8 below). The tweeter and woofer are both mounted on the front baffle. The tweeter’s tiny voice coil is very close to the baffle while the woofers large cone offsets it’s voice coil back quite a bit from the baffle. Now, the voice coil and cone comprise the motor that makes a speaker move air so broadband sound launched from the woofer is offset in time from the tweeter’s output by the physical displacement between the two devices. That’s why many loudspeakers have slanted back-leaning front baffles, which time aligns the drivers.

http://www.seneschal.net/images/restoration/acousticgroupdelay.jpg

Figure 8 - An simplified version of group delay via frequency-dependent acoustic delay

This is classic group delay, where the high frequencies from the tweeter arrive at your ear before the woofer’s low frequency content. Think of what this “time smear” does to a broadband sound like a tasty kick drum, a signal with both the HF snap of the beater and the LF boom of the shell’s resonant cavity. Anyway, this same time smear or group delay occurs in all IIR filters to a greater or lesser degree. The only difference is, unlike a speaker, in electronics the high frequencies lag while the low frequencies lead.

As to FIR filter, they exhibit a constant group delay regardless of frequency, so no wonky phase shift problems but, they have a different problem: pre–echo. FIR filters have an annoying tendency to present a small amplitude version of their input at the output, before the input has been applied! I know, what have I been smoking? Nothing, old boy, I simply haven’t mentioned that FIR filters cannot be built in the analog world, only in the bits and bytes of a digital implementation. So, a rip in the time–space continuum that I’ve just mentioned is taken for granted in digital signal processing circles…It’s just one more thing a designer must contend with.

The upshot is that FIR filters, since they lack group delay, don’t “sound” like EQ as we know it. Because of their sonic neutrality, they’re usually used only for specialized correction like in mastering, forensics…or restoration! Though they don’t exhibit group delay, they do impose significant latency due to significant computational overhead.

Now, I asked the company who created "Eqium" (the parametric eq IIR plug in) the next question, in order to see if they have a solution for the IIR prblems:

Hello,
I've learned that your Eqium plug-in uses IIR filters (and not Fir filters, like Firium).
IIR filters are notorious for their group delays (phase problems).
I can not accept group delays in my work, so I would like to know if you
know of a program which can solve the group delays that your Eqium software
creates.

Thanks.

This is their answer:

Hello,

Thank you for writing to us. You are correct, Eqium, like most other EQs, utilizes IIR filters which do not provide a linear phase like FIR filters. I do not know of an application that will "correct" this. Though you regard the effect of Eqium's IIR filters as a problem, virtually all EQs (software and hardware) are modeled similarly and thus exhibit the same behavior. It has only been more recently that an alternative (FIR filters) to this traditional type of EQ/filtering has been made widely available.

I hope you will continue to use Firium and that you are able to find uses for Eqium for which its more traditional EQ characteristics are a good match. I hope I was able to answer your question to your satisfaction. If I was not, or if you have any other questions or comments, please let us know. We appreciate your continued interest in our products.

Kind Regards.

--
Customer Service
Elemental Audio Systems
http://www.elementalaudio.com

I'm searching for a Fir filter based parametric eq plug-in, so hard, but can't find it. I've found Fir filter *equalizers*, but that's no good for room mode correction (we need adjustable bands. Constant bands are no good). I simply can't believe no one has creaed a Fir filter parametric eq software.

I'll keep looking...

[jlo]

Well, here's a program to wrap a vst plug to use in directx

http://www.spinaudio.com/cgi-bin/redirect.pl?demos/wr10demo.exe

It will wrap a max of 2 vst plugs cause its demo. Once wrapped, it'll be available in directshow--think zoom player. This way is free. With adaptx, you'll have 2 pay. (adaptx is also available for wmp9 and winamp3).

I had a long reply prepared, but i was automatically logged out. I c now u don't want this plug. There are many directx and vst mastering plugins u can use in directshow. Some of them are free, and others cost as much as $300 just for a 10band eq.

Check this site out for directx plugs:

http://www.thedirectxfiles.com/plugins.htm

Let me how things go.

[Esben]

What about Shibatch Super Equalizer (http://shibatch.sourceforge.net/)?

Shibatch Super Equalizer is a graphic and parametric equalizer plugin for winamp. This plugin uses 16383th order FIR filter with FFT algorithm. It's equalization is very precise. Equalization setting can be done for each channel separately.

I used it alot when I had my previous speakers (Paradigm Monitor 7), and no subwoofer. With the equalizer I gave the low frequencies a huge boost, thus giving me way better and deeper bass.
I friend of me also used it with a pair of CV AL-1000 speakers to make their 15" woofer actually move. :)

[Jones_Rush]

Esben, this looks really good. Thanks!.
Do you know how does this filter deals with the pre-echo problem of Fir filters ? (if at all). Have you found this equalizer to be neutral ?. What are you using now ?.
Can this plug-in work with the ASIO or Kernel streaming plug ins ? (ie. no kmixer's SRC in the way).

[Jones_Rush]

Esben, I've checked it and the "Shibatch Super Equalizer" plug in has no cure for the pre-echo fir filter artifacts, which leads to ringing just before transients or sharp attacks (I have confirmations for this from several sources).

As it seems, the one and only software which has all the features that I need (and much, much more), is Denis Sbragion's Digital Room Correction (DRC) v2.2.0.
I'm currently studying it, and I hope that by tomorrow I'll have first impressions about it. If only the user interface wasn't from the 80's, I would have started using it months ago.

Anyone who wants to give DRC a try, here (http://www2.gol.com/users/pcazeles/index.htm) is a "step by step" guide which will make things easier.

[jlo]

As far as parametric eq--it allows u to adjust the level, frequency, and q(bandwidth)? Is that right?

Have u tried this site. It's only 5 bands, though. Don't know about the echo thing.

http://www.waves.com/htmls/prods/masters.htm

I also know of a 30 band (non-parametric) fir eq.

[Jones_Rush]

As far as parametric eq--it allows u to adjust the level, frequency, and q(bandwidth)? Is that right?
Yes.

Have u tried this site. It's only 5 bands, though. Don't know about the echo thing.

The problem is that the bands are limited to regions, which render them useless for my needs.


I also know of a 30 band (non-parametric) fir eq.
Useless.

Anyway, the problem is not with finding a parametric eq with fir filters (Esben gave me a link to the exact thing that I asked for, in the beginning). The problem is with finding a software which will also take into acount the pre-echo artifacts of the fir filters. The only thing which does everything I need is DRC V.2.2.0, but it turned out to be WAY more complicated than I've expected. The step by step guide is only good for old versions. I think that I'll try to live with SuperEQ's pre-echo artifacts for now, until something better hits the market.

[RayL Jr.]

Originally posted by Yoniza
IIR vs FIR:
This is a scary concept but here goes: All filters suck, some more than others. Whew, I’ve gotten that off my chest.Maybe that's what's keeping ElvisIncognito from posting about using smiley faces - or he is AWOL... :)

[Jones_Rush]

Maybe that's what's keeping ElvisIncognito from posting about using smiley faces - or he is AWOL...
Why ?, does Elvis uses filters ?.

Btw, I think that this sentence ("all filters sux") is a bit misleading. True, both IIR and Fir filters have their limitations, but in some cases, NOT using them, and just living with what your room's acoustics have to offer, is simply unacceptable, by audiophile standards, at least (especially in the bass region - below 100hz). Anyway, Fir filters (and not IIR filters) are definitely the way to go.

[jlo]

Sounds like u need to relax a bit. Have a swig of black label bro.

If ur gonna use superequ and want a better player than winamp, there are a couple directshow filters that allow u to use winamp2.x plugs in ds. One is by the same guy who does ffdshow (heard stability is a bit iffy). The other is called dc-dsp. Alternatively, u can of course use winamp 2.x plugs in media jukebox/media center.

[Jones_Rush]

jlo,
Thanks for your help, but Winamp is a fine player for me, what I asked is a better eq software than supereq.

Anyway, I'm starting to do some real progress with DRC. More on that later.

[jlo]

Yeah, I here ya yoniza. But, what u originally asked for was a player to accept directx plugs. You were willing to use supereq so i was just trying to give ya more player options.

Good luck with drc.
If I ever run across anything with a gui, will let u know.

[Jones_Rush]

Sorry, jlo, you are right, I do change my needs frequently. But don't blame me, since I'm still in the process of learning this subject.

Btw, (new need) does anyone know where can I find a decent MLS measurement software which is both accurate and inexpensive (preferably free) ?. All the softwares which I know, like ETF 5, cost at least $150 USD, which is a lot of money, considering I don't need to use any extra feature except basic MLS.

[RayL Jr.]

Well, this is a sound card with excellent DSP hardware and software For $399 at - http://www.musiciansfriend.com/srs7/sid=030728165438216067202237497678/g=rec/search/detail/base_id/57542

"Its high reliability and sensational price make Luna II possibly the most desirable audio card for computers running Cubase VST, Logic, Nuendo, GigaSampler and all other popular audio software. Luna II is also the only I/O card in a wide price range that provides you with all the important benefits of a solid DSP architecture! Luna’s unique latency-free routing capabilities and the on-board "live" digital mixer (with advanced surround capabilities and effects) make Luna a valuable enhancement for every audio computer. Luna II is even compatible to the huge Pulsar/SCOPE platform DSP plug-in library."

I made a price/features list for Creamware cards:

Creamware Luna 2496 I/O box $425.00
Creamware Luna II - latency-free routing and surround capabilities, on-board digital mixer $399.99
Creamware Luna II EX - + ADAT Expansion Board & Vocodizer plug-in. $595.00
Creamware PowerPulsar II - 15 SHARC DSP chips, digital mixer, effects $1,995.00
Creamware PowerSampler II - 4 channels I/O, MIDI I/O, SFP software, Vocodizer Plug-in $595.00
Creamware PowerSampler II EX - + expansion plate w/2x ADAT I/O and 1x Zlink connector $745.00
Creamware Pulsar II - 6 SHARC DSP chips, digital mixer, effects $995.00
Creamware Scope /SP - 15 sharc DSP chips, extreme DSP power, SCOPE software package $2,995.00
Creamware Scope SRB - 15 Sharc DSP Chips accelerator for Pulsar and scope systems $1,695.00
Edirol DA-2496 - 8x8 audio recording interface $519.00

"LOOK, MOM, NO LATENCY!
Just a few months ago, MIDI latency in most inexpensive sound cards (or interfaces) was so dreadful that most synth-heads scoffed at the idea of playing or controlling a soft synth in real time using MIDI. To combat this problem, CreamWare has created Ultra Low Latency Interface (ULLI) and included it on the Luna II card. As I tinkered with MIDI instruments such as Native Instruments' Battery 1.0 in Logic Audio, Fruityloops' DirectX instrument DreamStation, and Reason's Subtractor synth and NN-19 sampler, I was delighted to find that MIDI latency had all but vanished. Nearly undetectable latency at this price is a serious accomplishment, and CreamWare deserves a pat on its virtual back." - http://remixmag.com/ar/remix_luna_landing_creamware/

"You can still use your favorite DirectX plug-ins inside your audio application (upstream of the Pulsar), and then let Pulsar’s digital mixer handle mixing and any other EQ, compression or other effects you wish the Pulsar to provide." - http://www.google.com/search?hl=en&ie=UTF-8&oe=UTF-8&q=Creamware+DirectX

Tong Chia here at AVS highly recommended them, no KMixer too. :)

[Jones_Rush]

Nice card RayL Jr, real nice.

Anyway, just wanted to let you know that I made the final step today with DRC. I'm now fully capable of operating the entire Room Correction process, starting from the MLS calibration, and ending at the final convolved real time playback, for stereo sources.

I even made some playback tests today, which were all horrible, but later at night I figured out all the mistakes that I did, and tomorrow I'm going to test it for real, for the first time (to bad I only have a Rat Shack to test it with...)

As of now, the good news are that the entire process is cheap, dirt cheap. Basically, all you need are CoolEdit Pro (or 2000) + a decent mic (preferably with a separate pre amp, not like my rat shack) + some other free sharewares , that's it.

The other good news is that the entire process is VERY easy to do and pretty fast, if you have someone who can tell you exactly what to do (I wish I had one), you really don't need to have any special knowledge about anything.

We'll see what happens tomorrow.

[Jones_Rush]

Well, I had to dig pretty dip to get to this thread again, so I'm not sure if anyone is still reading it.

Anyhow, I've promised an update, so here it is:
I've been doing tests all day, and the interim conclusion that I got to is that my rat shack simply wasn't made for room calibration. If the type of sound that I got was in any way representative of the quality of DRC, then I'm sure that its creator, Denis Sbragion, wouldn't have bothered with it beyond v0.001.
I've sent him my findings, and I'm waiting for his comments, but I'm afraid that if I'll want to really test DRC, as an audiophile tool (and not just a gimmick) I'll have to spend money on a better mike. We'll see about that tomorrow.

Part of me is kind of sorry that I started this DRC journey, since I didn't plan on spending more money now. The only problem is that I've spent so much time figuring out how everything works, that I simply can't allow myself to throw everything away and stop.

[Esben]

Please tell how your room correction works? You're not trying to correct over ~120 Hz, are you?

[Jones_Rush]

Esben, everyone, I have great news!!!.
Apparently, DRC outputs two files at the end of its calculations, one should be used as the "equalizing filter" (and the other is the impulse response of the system after correction, and it is useful to see the results achieved after correction), the thing is, I was convolving using the wrong file!. (It appears that my Rat Shack isn't THAT bad afterall).

Denis Sbragion was kind enough to go with me over all the possible problems that might have occurred. He told me that my Rat Shack mike really isn't up to the task of doing an accurate room measure for room correction, but it should sound much better than I've described. Also, according to some test signals that I've sent him, he told me that this mike is going to give me an overly bright image.

He was right, about everything. As soon as I started using the correct filter, the sound improved, dramatically. Also, it exhibits exactly what Denis said, an overly bright presentation, which really makes things sound compressed at the high frequency region. Anyway, I decided not to care for now about the high frequency region (since I can't do anything about it, yet), and test how my system (and especially, my room) sounds from the sweet spot.

My first reaction was, ASTONISHMENT, I kept saying to myself, "this can't be happening!", my room DISAPPEARED!. I first described it to Denis that it's like hearing my speakers in an unechoic room, it was an incredible experience. Later, Denis put it using better words, it's like the sound you get with headphones, only without the imaging being inside your brain (this is so true!!!). I have to warn you though, this stuff is addicting!, I've decided to bring all the songs which sounds less than decent in my room (especially those with heavy bass, like Marcus Miller), and was again, astounded. I actually prefered listening to the corrected signal, with those songs, even WITH the compressed high frquency response that I am getting. I still can't get over this experience. Some people often describe that after some changes they did in their system, it sounds like a veil had been removed from the speakers, well, in this case, it's like the room has been removed from the speakers!. I keep looking at the speakers, then at the room, then again at the speakers, then again at the walls, and just can't get over the experience that there is no apparent interaction between the two, the illusion is THAT good! (at least in my room, and your milage WILL vary, for better or worse, since no two rooms are the same).

I did notice a couple of things, like this process tax your amp to a certain degree (really depends on your room, I guess), but I have a 150WX2 amp, so this wasn't a big issue. Also, you really need to be in the listening spot, if you'll go too far, things start to sound funny.

Last but not least, Denis told me that sub calibration is not a problem, all I need is to let it work during both MLS recordings (for each speaker), and that DRC will do the rest. I can't wait to try it out tomorrow, using a 40hz cut off for the sub, and letting the mains go "large".

I'm going to buy a better mike next week, and if the higher frequencies problems are going to go away, I think we'll just have our first, truly "killer app" for the HTPC's audio domain. I can't thank Denis enough for all his hard work.

If you're going to try DRC for yourself, be patient, there are some bugs in the process that you might encounter (not with the DRC code itself, which doesn't have a graphical interface, yet, but more with the initial and final process).

I'm waiting until my mike is going to arrive next week, and once it will, if it will solve the problems I've mentioned (meaning, it get's my "audiophile approval", for what it's worth), I'm going to write a complete, step by step, updated guide (there is already one completely outdated guide), which is going to make things very easy for anyone who'll want to take advantage of this amazing tool. That's the least I could do for Denis.

P.S
Esben, the correction can only be done on the entire 20hz-20khz region. The correction will always be softer on the higher frequencies, and stronger on the lower frequencies, plus, you can set it manually to perform a much softer correction on the higher frequencies, but you won't be able to make it run only for the 20hz-200hz, for example. At least not with the current version. Believe me, if you mainly listen from your listening spot, and a good mic really solve the higher frequncies tonality, you'll have no problem going with a strong correction, all over the spectrum.

P.S II
Reading my post again, I want to make sure one thing is clear:
The discussion in this forum raises a lot of possible ways to improve our listening quality. We try better cables, different sound cards, different output types (Kernel/Asio vs Kmixer's SRC), different software players etc. But, up to this point, even though I DO notice and appreciate some of the benefits gained from all the tweaks I've mentioed, the only time in my life I've experienced a real HUGE improvement in sound quality, was when I switched my several hunderds $ speakers, to a several thousands $ pair, this was the only time I experienced a trully dramatic change in sound quality, for the better. Until today.

Just wanted to clear that one. Again, it could be that my room has overly problematic acoustics (I didn't go and measure many other rooms, so I can't really compare), so you do understand that my results are highly dependant on that factor. But I'm sure some of you have worse rooms than mine, so I'm sure at least those guys are going to agree with every word I've written here today.

P.S III
It's a good thing that DRC is a freeware, otherwise, some people would have surely made everything in their power to get me banned immediately, after this post, for marketing reasons.

[RayL Jr.]

Originally posted by Yoniza
Well, I had to dig pretty dip to get to this thread again, so I'm not sure if anyone is still reading it.Well I have this thread saved in Excel and book marked. Stuff has been dug up a lot deeper than yours, so I wouldn't worry.

That is great and encouraging news! :D I do have a Fibonacchi ratio room speaker placement in a rectangular 13'x23' room - room cancellation is done very well through the "room nodes" and reflections naturally. Here's 1 link - http://www.audiolinks.nl/speakersetup.htm

I got a flame once from a Tact fan and their DRC (at audioreview), but it didn’t reach my ceramic encased underground bomb shelter (no flamesuit necessary) and they (or him) deleted his post anyway. But I am a big fan of Tact anyway. :)

PS - interesting comment about headphones, I'm a big headphone/2.1-channel fan... I do like 6-8 channels, maybe eventually....

[Karnis]

bump

[drewmc]

Yoniza,
Great post, I have been using Shibach's Super EQ and been looking for something superior also. I am using it for mobile audio correction were the "room" acoustics are far worse than the worst home room Some questions about Denis's DRC software. Does it work on more than two channels? I would like to use it with a 5.1 setup. I can live without this because good stereo is far better than bad surround. The other issue is horsepower. I am using a VIA M1000 computer for space reasons. Right now it can handle the Shibach's plugin, and SRS Circle sound II 5.1 simultaneously.
Do you think it could handle 2 channel DRC?
Looking forward to your guide
Drew

[Jones_Rush]

Does it work on more than two channels?
Not currently, but be assured that it's going to change in the near future.

I am using a VIA M1000 computer for space reasons. Right now it can handle the Shibach's plugin, and SRS Circle sound II 5.1 simultaneously.
Do you think it could handle 2 channel DRC?
Here, make a test, download the RealReverb plugin for Winamp from this (http://www.ressl.com.ar/realreverb/index_old.html) webpage, then run it using this (http://www2.gol.com/users/pcazeles/iirStereo.wav) (direct download, press right mouse key, then "save as") convolution file. If things plays out smoothly (meaning, just without dropped samples, don't expect it to sound good, it will sound awful, since it is calibrated for someone else's room acoustics, plus, it uses an older version of DRC, which didn't have some important features, but this shouldn't affect the authenticity of the test) you are fine. Be sure to use directsound as your output in Winamp, since it won't work with KS/Asio (though those who'll want to take advantage of bit accurate output, will be able to make the convolve process off line with the .wav file of their choice, and in this case, it doesn't matter if you're on pentium 100, I guess).

[drewmc]

RealReverb uses only uses about 30% CPU on my anemic M1000 processor.
Time to get a calibrated mic. Does RealReverb have a version that is not adware?
PS Your room must be really screwed up. That file sounds weird on just a pair of old soundworks

[Jones_Rush]

Time to get a calibrated mic.
The current bang for the buck, unless you want to go DIY, is the Behringer ECM8000 (about $40. It's lower than what I paid for my Rat Shack, but you'll also need to buy a separate phantom power supply, which is another $40). I think it will be best if you can get it from a place which you'll be able to bring it back for a refund, if things won't sound good to you from any reason. It just that I have not yet heard DRC like it should be heard with a good mike, and even though the potential is there, it still doesn't have my "audiophile seal of approval", again, for what its worth. This will have to wait for next week. After a positive test with a good mike, I'll be much more comfortable recommending DRC to other people, especially to those who actually need to spend money in order to get the mike+power supply hardware to enjoy it (and won't be able to get a refund).

Does RealReverb have a version that is not adware?
Not sure, anyway, why pay when you can get it for free ?.

PS Your room must be really screwed up. That file sounds weird on just a pair of old soundworks
It's not my room, but I'm sure you'll get screwy results with my room's convolution filter also. That's just the way it is when you're correcting for the full 20hz-20khz region, at relatively small rooms.

[jcruse]

Yoniza,

I'm eagerly awaiting your step by step guide....

[drewmc]

Good advice on the mic. You really should get a mic preamp instead of the phantom supply. The mic inputs on most sound cards (including the revo) are not very good. Two good choices are the Beringer SHARK DSP110
(one channel) and the M-Audio Audio Buddy (two channel) These will run you around 69.95 and 79.95 respectively.
Why not use the Adware version? It takes up the entire screen at 800x480.

[dwk123]

I've plugged DRC several times in various digital eq discussions, and to my knowledge this is the first time anyone else has actually tried it.

As far as I know, DRC is the only game in town for anyone looking to 'roll their own' system (ie don't have the $$$ for Tact gear), so to a degree quibbling about some shortcomings is pointless - it's WAY better than nothing.

I'm using DRC in conjunction with BruteFIR (and Jack) under Linux to do a full-blown digital xover + room correction system. Output is via a Delta 1010. It 'works' very well for the most part, although my system has been such a moving target that I can't yet conclude exactly how well.

The algorithm Denis uses is actually pretty clever - basically using a windowing process that filters the time-domain impulse with a bandwidth that falls exponentially with time. (in other words, full-bandwidth correction is applied to the initial arrival, and then high frequencies are left progressively less corrected as time evolves) You can play with the decay rates to tweak it to your particular needs. It then inverts the minimum and excess-phase components separately, and you can choose to correct only the minimum phase elements, or both. Finally, you can apply a global target power response to correct for the brightness that Yoniza experienced - I definately agree that if you correct to 'flat' it'll sound too bright.

My results have been mixed, but I'm 99% certain that it's due to the fact that I have full-range dipoles (Carver ribbons) in way too small a room. With a strong set of early reflections, the correction becomes a bit too agressive resulting in a small amount of tunnel-echo effect if you aren't *exactly in the sweet spot. Aside from that, the improvement in articulation is astounding, and imaging/soundstaging is also greatly improved. The bass correction is outstanding - I have a NASTY mode at ~50Hz that it handles pretty darn well.

My system is in pieces right now as I get my crt pj mounted and get my new speakers going (mounted the Carver ribbons on a new baffle that includes 4x7" midbass drivers to bridge the gap down to the NHT 1259's). In the next couple weeks I'll attack the new xover and room correction filters, which *hopefully* will be the final versions.

[dwk123]

Originally posted by drewmc
Good advice on the mic. You really should get a mic preamp instead of the phantom supply. The mic inputs on most sound cards (including the revo) are not very good. Two good choices are the Beringer SHARK DSP110
(one channel) and the M-Audio Audio Buddy (two channel) These will run you around 69.95 and 79.95 respectively.

I'm using a Behringer ECM8000 with an Audio Buddy, and am perfectly content with the results. For about $100 total, it strikes me as a great deal.

[KikeG]

The ECM8000 is a good & cheap choice for these kind of measurements, but it's true that you need a phantom power source to make it work, and a separate mic preamp to have accurate measurements. Professional mic preamps such as the Audio Buddy include both, but also most mixing desks do.

There is a convolution plugin for foobar2000, that is superior to RealRevelb for Winamp, because of the floating point internal operation + dithering of the final signal, which RealReveb lacks.

It's not clear that IIR filters are worse. An impulse based convolution correction system such as the one you are using, resembles more to an IIR filter than to a FIR filter, mostly because FIR filters don't exist in "analog" world, and your room acoustics are part of that world.

[dwk123]

Originally posted by drewmc
Yoniza,
Some questions about Denis's DRC software. Does it work on more than two channels? I would like to use it with a 5.1 setup.

Technically, DRC doesn't really care about channels. It simply takes a measured single-channel impulse response and attempts to correct it to 'flat' (with the appropriate modifiers). There's nothing inherently preventing you from doing multi channel setups by running each channel through independently.
There are two catches, though:
- you don't want the .1 to be flat
- you need a multi-channel convolution engine.

Neither is a show-stopper. BruteFIR (linux only, sorry) handles an arbitrary number of channels, and I'd imagine there is something out there for windows. Trying to correct filtered signals is a bit more of a challenge, but between DRC's 'global envelope' stage, and possibly some creative pre-processing of the signal you should be able to make it work.

[3-way]

Has anyone tried WaveWarp? www.soundslogical.com It looks like it will do all of this and then some, with a graphical interface and digital crossovers. I've downloaded the demo file, but haven't had time to play with it...

[Jones_Rush]

dwk123,
Regarding the high frequency imperfections, in my case, I don't think it's mainly related to the DRC code, but more to the mike itself (Denis saw this problem when I sent him my recorded mls test, before he even saw my DRC results).

Also, since you have ribbons, you probably can be a good judge of high frequency tonality problems. I would like to know if, in your opinion, DRC changes the tonality of the higher frequency region in some way, and if so, to which degree. When you go from an unconvolved signal, to a convolved signal, do you feel like you're compromising some accuracy at the high freq area ?.

Regarding the Revolution's mic input. Well, I'm not sure about the mic's input, but I know that the line-in's input is pretty accurate. When we make RMAA measures, we use it in loop with the outputs, and the results (http://www.nabs.net/cwatson53/M-Audio%20Revolution.htm) are good. Aren't they good enough ?. What makes you think that the mic input is subpar ?.

[RayL Jr.]

Originally posted by drewmc
RealReverb uses only uses about 30% CPU on my anemic M1000 processor.
Time to get a calibrated mic. Does RealReverb have a version that is not adware?
PS Your room must be really screwed up. That file sounds weird on just a pair of old soundworks "Adaware and SpyBot aware" software is not good for a stable HTPC. In fact, not good for any PC. Be careful too - pay versions can give you the same runaround as their "Adaware and SpyBot aware" counterparts. These programs can hide deep in the registry and can take up extra-unneeded CPU power.

Convolution plugin for foobar2000 (as suggested) or other software that is REAL freeware should be much better for your HTPC and CPU. :)

[Jones_Rush]

Where can I find this convolution plugin for foobar2000 ?.

[drewmc]

The micrphone input is inferior to the line input by M-Audio's own addmission. Please see page 26 of the following PDF.
Reviewers Guide (http://www.midiman.net/products/consumer/revolution/revolution-reviewguide.pdf)
QUOTE/while the mic input provides high-quality audio input,it is not as high as the line in or line outputs. Performing a loopback test with a cable plugged in to the mic in jack will degrade test results/QUOTEl

[greenkiwi]

Are there any ActiveX/DirectX filters that can be used for RCS? I'm planning on using MediaCenter for playback of APE files. Preprocessing files doesn't seem like a good solution because I would like to keep all files in their original state. In addition, each time I change to a new equation for RCS, I'd have to redo all the files...

The other things that I'm very interested in is a ActiveX/DirectX filter that could act as a crossover. I would like to biamp my speakers and have the crossover take place before the amps. It would have to provide output mapping from 2 channels to 4 or 6 channels.

This crossover control plus RCS would give people wonderful control over the sound quality of their system. THen you would just need to have a good program to tune the filter and cross over.

kiwi

[JimmyMack]

DO NOT LET THIS THREAD DIE!!!! I absolutely must know how to accomplish this. Please keep us informed of your progress.

[salsbst]

Originally posted by greenkiwi
Are there any ActiveX/DirectX filters that can be used for RCS? I'm planning on using MediaCenter for playback of APE files. Preprocessing files doesn't seem like a good solution because I would like to keep all files in their original state. In addition, each time I change to a new equation for RCS, I'd have to redo all the files...

The other things that I'm very interested in is a ActiveX/DirectX filter that could act as a crossover. I would like to biamp my speakers and have the crossover take place before the amps. It would have to provide output mapping from 2 channels to 4 or 6 channels.

This crossover control plus RCS would give people wonderful control over the sound quality of their system. THen you would just need to have a good program to tune the filter and cross over.

kiwi

I hope that someday J River may make playback through generic DirectX filters possible, but they seem to be really caught up trying to compete with Apple and Microsoft right now for user interface supremacy.

[3-way]

I mentioned WaveWarp a couple of posts ago. I had some time to play with it today. WaveWarp for MATLab will do nearly anything. WaveWarp comes with a number of built-in adjustable digital filters (Blackman, Hanning, etc.) and brickwall EQ. Better yet, you can use MATLab to design filters and WaveWarp to implement them. That means Vanderkooy filters.

WaveWarp can be used as a DirectX plugin, and, would likely work as a DirectX plugin within MC9. But, both MC9 and CoolEdit Pro crashed when I launched the WaveWarp plugin. After some more research (reading through the bulletin board at www.soundslogical.com) it looks like development of the program has slowed or stopped. It requires MATLab 6.0 or 6.1 (the current version of MATLab is 6.5 or later) and the page explicitly states that it will not work with 6.5. Also, the page specifies DirectX 6.0....

If someone can get WaveWarp running, it looks like it's powerful enough to implement crossovers and parametric EQ with brickwall filters (although I think the parametric EQ would need to be designed in MATLab in order to get adjustable frequencies).

If someone could figure out a way to put together DRC and the crossover filters from MATLab, we'd all be set. It looks like dwk123 has done this or something similar through Linux (Brutefir, Jack and DRC). I plan to look into that, but I'm not sure I want to give up all of the plusses of Windows (Girder, MC9, DVDLobby, Holo3dgraph, etc.).

Ideally, the makers of ETF (Doug Plumb) and DRC (Denis Sbraigon) would get together with Don Maurer (http://home.pacbell.net/donwm/), put their pieces together and develop a program that would do it all.

[Phat Phreddy]

I have spoken with Don a few times and pointed him to AVS digital room correction threads in the past... While friendly and informative he does not want to open up the work he has done on speaker correction as he feels he may incorporate it into a commercial project in the future...

[3-way]

I spoke with Don about his software about 12-15 months ago and got the same sense from him. He's put a lot of work into his project and deserves to get something back. Maybe the three of them could put together a commercial product? I'd be willing to pay $500 or more for a top-quality, easy-to-use and well-supported app that combines features of all three. Think of the equipment it would replace! I'm using a Driverack 260 right now (~US$720). It doesn't have digital input, so, I'm foced to go through an extra A/D & D/A conversion. It's limited to six channels, will be outdated in 6 months (arguably, it's outdated already), and, it doesn't have half the features the software program would have, either.

[Jones_Rush]

DO NOT LET THIS THREAD DIE!!!! I absolutely must know how to accomplish this. Please keep us informed of your progress.
Jimmy, don't worry, this thread won't die, I give you my word. The reason I haven't posted in this thread for the last two days is because I had a setback, my main amp broke. This is frustrating because now I need to get a new amp (or fix this one) which will probably cost me much more money than the money I planned for the new mike and phantom power supply.

Anyway, I hoped that I'll be able to buy the mike this week, but now it will have to wait until a later date in August. I promise you this though, I WILL check DRC with a good mike, and if the results will satisfy me, I will write the guide, BEFORE the end of August (unless someone will do it before me). That's the best I can promise for now.

[buns]

I have been looking for something like this over the last while....... while i know very very little about all this (hell ive never even owned a soundcard beyond that in my laptop!), i will be keeping close track and look forward to trying it when there is a clear route to doing it :D

Ad

[greenkiwi]

Originally posted by salsbst
I hope that someday J River may make playback through generic DirectX filters possible, but they seem to be really caught up trying to compete with Apple and Microsoft right now for user interface supremacy.

I thought that some versions of MC 9/9.1 had DirextX Host filters as an option?

I don't know about their current state in the beta 9.1.x versions. I haven't downloaded their latest version, but they do have it listed in their web site:
http://www.musicex.com/cgi-bin/downloads/mcplugins.pl?type=5&start=0&end=10&page=1

It looks like it was updated on 7-18-2003.

For me, RCS and Digital Crossovers that could be implemented through DirectX would be awesome, as they would be usable in many different programs.

kiwi

[salsbst]

Reports are that the DirectX Host doesn't really work reliably yet.

[KikeG]

Originally posted by Yoniza
Where can I find this convolution plugin for foobar2000 ?.

http://www.hydrogenaudio.org/index.php?showtopic=10611

[jlo]

The realreverb winamp2 plug-in can also be used in media center. Also, there is the yet to be released realreverb studio 4.0.0 (http://www.ressl.com.ar/realreverb), which is a directx/vst32 plug-in

I have also read/experienced that the directx host plug-in for mc is problematic. It is still in beta though. U can install it on media center 9 or 9.1. It uses the same adapt-x (http://jobsearch.chez.tiscali.fr/en/fhome.htm) engine that esben mentioned. If u could get a directx plug-in working in winamp 2/3 with their respective adaptx plug-ins, mc should be able to do it also-once the plug-in is stable enough.

[Jones_Rush]

Guys, I got a recommendation from Denis about this (http://www.behringer.com/02_products/prodindex_ub.cfm?id=UB802&lang=eng) mic pre amp.
Do any of you have experience with this product ?, the price is very good, and the features look amazing too. Also, the tone control existence can be overcome by a tweak Denis offered me. It cost less than the "Audio Buddy", can this be a better deal ?.

Audio Buddy Specifications:
- Frequency Response: 5hz to 50khz, +0, -3dB
- Mic Gain: 60dB Max.
- Mic Input Impedance: 1K Ohm
- Instrument Gain: 40dB Max.
- Instrument Input Impedance: 100K Ohms
- Power Supply: 9V AC, 300mA

UB802 specs:
- Frequency Response: 10hz to 150khz, +0, -1dB
- Mic Gain: 60dB Max.
- Mic Input Impedance: 2.6K Ohm balanced,
Rest of the specs can be found at the link above.

[3-way]

Yoniza,

A question for you:

1. At a command prompt, I typed "drc normal.drc" and the program went to work, spitting out a file after several minutes.

2. I used this file with the convolver for Foobar mentioned several posts ago. I noticed a difference, but, obviously, since I didn't start from a measurement of my system, it wasn't necessarily an improvement in sound quality.

3. I have measurements of my main system taken with ETF. I can and did export the impulse response (a standard ETF option). It's a .pcm file. But, what do I do with it? I tried typing "drc rightmain.pcm" but DRC didn't seem to like that command.

TIA.

Brad

[Jones_Rush]

3-way, you should edit Normal.drc, find "BCInFile =" and change the name of the file to the name of your file. Then you could run "drc normal.drc" and everything will work fine (it just that in the normal.drc file are all the settings, without it, DRC has no clue what you want it to do) Also, it is preferable to use 32bit pcm files (I'm not sure if it's a must though, maybe DRC can work with 16bit pcm files as well), I hope ETF did this for you.

btw, what kind of a mike you have ?.

[3-way]

Thanks, Yoniza. Makes sense.

I have the mic and pre-amp offered by ETF. It's not as full-featured as the Behringer in the link you posted (not even close, actually) and it's much more expensive ($250 for mic and pre-amp), but, I'm happy with it. It comes with a calibration file. ETF is really a fantastic program, especially if you purchase the add-ons (psychoacoustic, etc.)

[K-Wood]

3-way:
If you get this working, can you please post a quick how-to? I have ETF and ETF's microphone, but its been awhile since I played around with it. I was using it primarily to program a Behringer Feedback Pro to eliminate a bass mode in my room. I would love to be able to use a DRC file for the same purpose.
Keep us posted!
Thanks,
Ken

[3-way]

Will do.

[Jones_Rush]

3-way, I don't need any of the features on the UB802, the only thing they can do is degrade the flatness of the frequency reponse, nothing more. It just that Denis heard that it suppose to be a good pre amp, with a pretty clean signal. I've checked a bit about it, and a lot of people are very satisfied with it. I might go for it with the ECM8000.

Anyway, the mic + pre amp you have should be very good. I'm very interested to if you'll hear tonality differences with the corrected signal.

Denis told me that this shouldn't happen, and that some of the people who discovered tonality differences with DRC, found that the tonality that they got using the uncorrected signal with headphones, was like the tonality with the corrected signal and speakers. Meaning, they have got accustomed to the not-authentic sound in their room, to the point they couldn't recognize the authentic reproduction when they heard it.

Don't keep us waiting long... :-)

[3-way]

I'm not sure that this is exactly what you're referring to, but some audiophile friends of mine are very anti-Tact because they believe it adds a tell-tale coloration to the sound. They've heard it on a number of systems and just don't like it. I haven't heard Tact gear often enough to notice.

[ElvisIncognito]

Originally posted by Phat Phreddy
I have spoken with Don a few times and pointed him to AVS digital room correction threads in the past... While friendly and informative he does not want to open up the work he has done on speaker correction as he feels he may incorporate it into a commercial project in the future... I, too, have spoken w/Don, in an effort to get him involved in the PC-PrePro project. He graciously declined. Eventually, I would hope that if enough people continue to try to persuade him, he may decide to help us all out.

XLNT thread, Jones - keep up the good work!

(BTW- Don't look now, Jones, but Yoniza seems to be editing your posts! :eek: ;))

[kromkamp]

I tried to play with this on the weekend. I think I will be waiting for your how-to guide, Jones_Rush :)

I had two problems:

1)When I tried to de-convolve the recorded MLS stream in Aurora, the result did not look like an Impluse response (ie. spike then decay) it was a constant 'noise'.

2)DRC craps out with the settings file specified on that other how-to website. It didnt like one of the parameters (forget which one)

Anyways, this is also meant as a bump because I'd like to check this stuff out...

Andy K.

[RayL Jr.]

Originally posted by 3-way
I'm not sure that this is exactly what you're referring to, but some audiophile friends of mine are very anti-Tact because they believe it adds a tell-tale coloration to the sound. They've heard it on a number of systems and just don't like it. I haven't heard Tact gear often enough to notice. I'm ALL FOR their digital amps ('cept for the price, waiting for the 2nd price drop still). As for their RCS or DRC, I'm with your audiophile friends. Instead of coloration, I'd call it a "veil" if set up in a large "anechoic" room... Maybe that's why Tact recommends removing room treatments for their RCS. Just read the 1st review at - http://www.audioreview.com/Amplifiers/TACT,Millennium/PRD_116186_1583crx.aspx

You could save the 4-5 grand on a big room addition or expansion. I used to have a modest system in a 28’x82’ basement with 8-9’ ceiling (exposed joists). With just bare concrete walls and area rugs (basement mostly empty), my 2-bookshelf Epicure speakers were impossible to locate with eyes closed. Sound was “very addicting” too, but more of a sensation than an addiction. My system sounded better than the monster bouncing rubber Polks playing at an audiophile shop near Stonybrook LI. Maybe that could’ve used the RCS. Or an SCS (speaker correction system).

I could see it doing something in a small or "bad" acoustics room, as Yoniza pointed out. And I think Yoniza will have a lot more money in the bank for later with his solution in this thread... Plus it's a HTPC "upgradable" solution.

<<Flame proofing off - hot dogs, marshmallows and sticks at the ready... :)>>

[Bob Sorel]

BTW- Don't look now, Jones, but Yoniza seems to be editing your posts!

Yup, we will be looking into this. No one should have the ability to edit others' posts unless they have moderator privileges, or have otherwise hacked the board. If either Jones_Rush or Yoniza can explain this, I suggest that they contact one of the mods really soon to clear this up.

[Phat Phreddy]

Yozina has been banned.

EDIT ::: See the link in Jones's Sig... Makes some sense though violates the ToS of AVS.

[Jones_Rush]

About Yoniza,
Relax guys, I didn't hack the boards, it was all done with the help of David Bott and Alan Gouger. They already know the whole story, and were kind enough to help.

kromkamp,
I hope you can wait for the guide. I also had some inconsistencies with deconvolving the MLS signal, this is why in the guide I'll show a way of using a long sine sweep as a test tone (instead of MLS), which you don't need to deconvolve using the same feature. According to Denis, a long sine sweep (let's say from 10hz to 21000hz) will give better accuracy then MLS. So it's a win-win situation.

Flame proofing off - hot dogs, marshmallows and sticks at the ready...
RayL,
Take the hot-dogs back to the fridge. We're on the same side of the fence. I too have fears regarding a possible sonic signature that I will get from DRC. Denis believes that a discernable sonic signature can only come from the recording hardware (mike, pre amp), and this is why I'm taking my time deciding which hardware to order. It can make or brake DRC for me, since I HATE any kind of coloration with my sound.

Anyhow, you need to understand that your room already gives you, today, a really big sonic signature on your sound. I guess in order to make a right decision, one would need to hear his system with and without DRC, and then decide which sonic signature is less offensive for him.

When I asked Denis why doesn't he create a correction only for the ~300hz range and down (and by that eliminating any higher frequency sonic signature), what I understood is, that apart from the technical difficulties (of connecting the corrected and un-corrected frequency range), which shouldn't be THAT much of a problem, considering what he did with DRC, the big reason is that he simply don't find DRC (the whole chain, that is, including the recording hardware that he uses) to have an offensive sonic signature on the sound. I hope he's right.

[dwk123]

I've been away for a while, but things don't look to be too stale.

Originally posted by Jones_Rush


I also had some inconsistencies with deconvolving the MLS signal, this is why in the guide I'll show a way of using a long sine sweep as a test tone (instead of MLS), which you don't need to deconvolve using the same feature. According to Denis, a long sine sweep (let's say from 10hz to 21000hz) will give better accuracy then MLS. So it's a win-win situation.


I've experienced the same thing - I've never been completely happy with MLS signals, even with LAUD. I find they do 'ok' for short impulses, but that might just be the limited resolution inherent in the signal.
I'm also using the logarithmic swept-sine approach (do a web search for Farina for the original paper) and so far I find it to work pretty well. Plus, it's pretty easy to code, although you do need to do a long convolution in the approach I'm using (no problem with BruteFIR).


When I asked Denis why doesn't he create a correction only for the ~300hz range and down (and by that eliminating any higher frequency sonic signature), what I understood is, that apart from the technical difficulties (of connecting the corrected and un-corrected frequency range), which shouldn't be THAT much of a problem, considering what he did with DRC, the big reason is that he simply don't find DRC (the whole chain, that is, including the recording hardware that he uses) to have an offensive sonic signature on the sound. I hope he's right.

I've thought about this as well, since I'm not sure that DRC style correction will be able to properly deal with my full-range dipoles and the reflections they generate. There are a couple ways to aproach it, but it's not straightforward if you are trying to fit into a straight 2-channel convolver. The most direct being a simple linear-phase low-pass/high-pass pair applied to the filter and a dirac impulse which are then summed to create the final filter.
In my case, since I'm doing both xover and room correction, I'm planning on experimenting with only applying the correction to the sub (<80-ish) or the sub and bass (<300-ish) channels to see how it compares with full-range correction. I can see how there could be blending problems, but it's worth a shot IMHO.

What I think is needed to spur this on to a wider audience is a graphical 'filter workbench' where you can
a) capture room impulse responses
b) see the impulses in question (raw, smoothed, corrected etc)
c) semi-interactively alter parameters and see and listen to the results
d) (ideal) switch back and forth between several setups - uncorrected, one or more different corrected responses etc.

Having done many many change params/restart convolution/listen etc cycles, it's really tough to do in an offline/batch mode. There are a couple open-source sound editors that look to be decent candidates for a starting point, but unfortunately I've never progressed past the initial hacking point.

[dkan24]

I have been reading this post with great interest, but have some basic questions before I move on to the difficult ones.

From my understanding, you attach a mic (and preamp) to your soundcard first. You then play a test signal and record it? Is this test signal the only sound you will have to record?

Where do you position the mic? In the sweet spot or all over your room?

Once you get the measurements, a file is created. You then play this file back using a plugin inside winamp such as reverb along with the song? Where does Cooledit fit into this?

The main question I have is - do you have to do this for each song file? Or can one measurment fit all? If the former, it would not be worth it since most people have thousands of songs. How close are we to having real-time music correction?

I am sure that I will have more questions after these are answered, but right now I don't even understand the basics. I do know this: if RCS was to become a serious thing on the PC (and no reason it shouldn't with the horsepower we have), then an HTPC would move into a class all by itself, or at least up there with TACT as super high end. I only head the TACT system once and thought it was unbelievable. It was not as good as a great all-tube setup, but costs considerably less than the best all-tube setups. And an HTPC acting as a TACT would cost 1/10 of that.

[Jones_Rush]

From my understanding, you attach a mic (and preamp) to your soundcard first. You then play a test signal and record it? Is this test signal the only sound you will have to record?
Yes.

Where do you position the mic? In the sweet spot or all over your room?
You position it at the sweet spot, and this is the biggest limitation of room correction, it works great for just one place.

Once you get the measurements, a file is created. You then play this file back using a plugin inside winamp such as reverb along with the song?
You first need to process this "file" in several ways (and this is where Cool Edit plays a role), and then what you said is correct.

The main question I have is - do you have to do this for each song file? Or can one measurment fit all?
The latter is true. The Winamp (or Foobar2000, etc.) plugin works in real time.

[dkan24]

That seems pretty sweet then. So the setup seems pretty easy. How much work is done in Cooledit?

My room is terrible. I live in a studio, and there is all sort of echos and stuff. It sounds better now that I got an area rug, but still pretty bad. I would probably take 2 measures - one for my couch and another for me bed, and use the appropriate one.

On a side note- if you only need to use the mic once, maybe we should look into splitting a group mic and all sharing it.

[Jones_Rush]

On a side note- if you only need to use the mic once, maybe we should look into splitting a group mic and all sharing it.
That's a good idea, only thing is I live outside the U.S, too bad for me.
I also thought of renting a real high end measurement mike, for just one day. If the cost for one day isn't high, I might go for it. All I need to do is find a place who rent this stuff.

[dkan24]

please keep us informed of how the new measurements sound. Also, if you are still considering writing the guide, that would be great. Based on your radio shack measurments, I am moved enough to want to try it. with you guide, i would probably just do it. This is incredible news for the htpc if it works correctly. I have been dreming about a TACT ever since I heard one. To add this to our arsenal, the htpc would be hard to beat!

[tbshooter]

Do you do a filter for each speaker or it doesn't matter?
It would be nice if you can get a B&K reference mic to calibrate.

[Jones_Rush]

Do you do a filter for each speaker or it doesn't matter?
Yes, you must since the speakers are located at different places in your room. At the end you'll have only one file, but inside there would be actually two filters, each for every channel.

[tbshooter]

That makes sense but what about interactions between the speakers?

[Jones_Rush]

tbshooter, I've asked Denis that same question a couple of weeks ago, so I'll just copy&paste:

One more thing Denis, after reading the DRC procedure that Patrick made, I couldn't help noticing a weird thing: The equalization is done for each speaker individually, and that's all!. There is never a stage in the process where both speakers are measured together, to check for their summed response (which might lead to cancellations and reinforcements all around the spectrum). Am I missing something ?.

No, you're missing nothing, but exactly the same problem exists even when the speakers are not equalized. This is called "interaural crosstalk" and is one of the major problems of basic stereo, and get even worse with typical multichannel systems. It can be partially solved using crosstalk cancellation filters, but a system with crosstalk cancelled out is no longer compatible with basic stereo, which requires interaural crosstalk to work, even if interaural crosstalk cause many spectral problems. Just as an example headphones have no interaural crosstalk and usually have horrible imaging performances, with sound coming just from inside your head.

There's another important problem with DRC: it doesn't compensate for level differences or time misalignments between the channels, a thing that the Tact software does. This parameters anyway can be corrected before (time misalignment) and after (level mismatch) channel equalization using standard mesurement methods. Time misalignment can be corrected with a simple meter measuring the exact distance of the loudspeakers from the the listening position and making it the same. Level mismatches can be corrected after channel equalization either with a simple measure or using the level hints supplied by DRC if the measures for both channels were done without changing the volume and using a measuring system which doesn't normalize the impulse response.

This is really important, because after equalization the two channels will start to sound almost identical and differences in the distance down to few centimeters, and level mismatches between the two channels down to 0.2 dB, become easily audible (I can see Elvis smiling). In my system changing the level of one channel of something like 0.2 dB causes the center image (playing a mono male, or better female, voice) to shift something like 5 cm. So it's rather easy to compensate for channel level mismatch just by ear, without further measurements or computations on the levels of the corrected and uncorrected impulse response.

[tbshooter]

Thanks. I would be curious how it would sound if you average the 2 filters.

[Jones_Rush]

Good news. The Foobar2000 convolve filter seem to be able to work with Kernel Streaming output. Furthermore, I checked today (using headphones) a convolve filter that someone did with the ECM8000 mike, for his room, and the result is light years ahead of my Radio Shack SPL meter. The sound is clean, the voices are natural, and sounds very similar to result I get without the filter. I can't still tell if there is an offensive sonic signature or not, since I can hear his room (or kind of an inverted state of his room, more accurately).

[ElvisIncognito]

So, that's how his room would sound in an inverse parallel universe, then...

[dkan24]

Originally posted by ElvisIncognito
So, that's how his room would sound in an inverse parallel universe, then...

And in the inverse parallel of that through headphones.


Jones_Rush:

I just downloaded foobar2000 and think it is hysterical. I see that you can add on many components. I am assuming skins as well although I kind of like the original skin!

I am going to search, but if possible can you post links to all the plugins one would need to play back that same file you have.

thanks,
Dan

[Mooneyass]

Jones_Rush,

Have you checked your PM?

Wes

[Jones_Rush]

Update:
I've ordered the ECM8000 mic + UB802 pre-amp (I've settled on this pre-amp, since it has very low noise and non linear distortion (these factors can not be corrected by software, so it is important that the pre-amp will do good in this area) and its frequency response deviations can be corrected by software, so no worry there). Total price paid: $90 (w/o shipping).
I believe that the package will arrive sometime next week.

[Jones_Rush]

I want to make something clear right at the beginning. The next post you are going to read, is based upon my personal subjective opinion, which has been conceived after listening to DRC, with my own personal recording/playback equipment, and more important, in my own unique room (it is nothing special, but it will be different than yours. It is a pretty small room, untreated acoustically, and the only thing which might shed some light about its nature, is the fact that Denis Sbargion, the creator of DRC, after looking at some plot measures, which were recorded at my listening spot, said it looks to him like a typical room, with the typical acoustic problems, ie. not too good, and not too bad either). So, don't come blaming me if you get different results than what I get. After reading this important message, let's move on.



As some of you might know, I had a lot of concerns about this whole DRC project, especially because I tried it with the Radio Shack SPL meter, and knew I couldn't live with results even slightly similar to what I got, in regard to the higher frequencies.

I received the ECM8000 mike, and UB802 mic preamp (which turned out to be one hell of a mic pre amp. At first I was sure I'll need to correct its frequency response, but after you'll take a look at the file attached below, you'll understand why I didn't bother, at the advice of Denis, of course).

I spent the last couple of days testing the equipment, measuring, A/B testing with headphones, speakers, and doing everything I can do in order to find out the real story about DRC.

Making a long story short (you really don't want to hear about all the near-disappointments I had in the last couple of days), today, about 3 hours ago, after making a tweak which was very successful on my system (going from 16 bit filters, to 32 bit filters), I came to the next definite conclusion: For me, DRC is truly the first "Killer App" for the computer, audio wise. By "Killer App" I mean: an application I MUST have, an application that after I used it once (in the proper way), there is no way in hell I can go back, never, ever. From now on, I will always use a room correction software, and since I don't have $3000 for a TacT Audio device, the HTPC just became mandatory for 2 channel audio listening.

DRC astounded me today, in a way I really didn't believe could be possible (after the last two days of trial and error, at least). I wanted one thing from DRC, I wanted to be able to hear a huge improvement in sound (at least as great as when I moved from a several hundred $ speakers to a several thousands pair), without having the feeling I'm missing anything else in the frequency spectrum. Now, as a wannabe audiophile, I won't say DRC doesn't change the high frequency sound in any way. Yes, under absolute terms, it does. What I am saying, is that I don't care. I really don't care anymore. If this will make me lose any audiophile dignity, I also don't care. I simply don't give a damn about what DRC slightly does to the high frequency spectrum. All I know is that I don't remember myself enjoying my music to a greater extent, in my room. And to be honest, I never heard an overall better artificial reproduction of music in my life (then again, my experience with audio systems is definitely not as great as the one some of the people here have, and to tell the truth, all of the rooms that I heard music in, were never perfectly treated acoustically).

My advice to you is, if your room isn't acoustically treated at all (like mine), you must do yourself a favor and check out DRC, you must!, just go and get (purchase if you need) all the necessary equipment. Even if you are hardcore audiophiles, I'm very sure you'll never go back after trying it.

Now, if your room is already acoustically treated to some extent, then I'll still suggest you try it if you have all the necessary equipment, but if you don't, it's your call. The before and after result you'll get will be less astounding than mine, so I can't be a judge and say if it'll worth it for you, or not.

[Jones_Rush]

For comparison, without the UB802.

[Jones_Rush]

From the listening spot (higher frequency is harder to judge, but will be posted later).

BEFORE DRC:

[Jones_Rush]

AFTER DRC:

[3-way]

That's amazing! Did it have any effect on your impulse response?

[Jones_Rush]

That's amazing! Did it have any effect on your impulse response?
Of course. It corrected it, well, at least for the sweet spot :-)

Btw, I promised a step by step guide if I'll like it, and since I more than like it, you can expect it within 10 days from today.
Oh and I want to thank Denis Sbragion on this occasion. Denis, you single handedly turned out the HTPC to be all I ever dreamed of (ok, not "all", but definitely "most". I still have some other dreams...). Thank you. And the fact that you did everything for free, at your spare time... I'm speechless.

[Mooneyass]

Wow.

Eagerly awaiting the guide..........

How sensitive is it to you moving your head? Are we talking inches or feet?

Wes

[K-Wood]

That's really amazing, Jones_Rush. I've never come close to that with a Behringer Feedback Pro (talk about using a sledgehammer to kill a fly).

I've already invested in the ETF software and microphone and have a Delta 410 card, so I'm interested in what I can acheive using DRC. I also eagerly await your step-by-step guide!

Question however: are they are other playback software programs that can accept the DRC plugins, other than Winamp and foobar2000? For example, will it work with J. River's Media Center?

Thanks,

- Ken

[bblue]

A point which cannot be over-emphasized is that you CANNOT make up for an acoustically corrupt room with EQ (DRC or any other) alone.

Every excited node, every reflective surface (walls usually) adds a particular color to the sound at its effective frequency range. Highly excited room nodes resonate strongly and will extend the time at which that frequency is present (shown by poor impulse response). Compensating electrically can change the amplitude of a frequency, but not the sound of it modified by the room.

For higher frequency sounds, left-right reflections between parallel walls at the listening position (for example) will blur the imaging making for much less precise localization in the sound stage and masked detail. Many other sources of reflection and blurring can be in effect at the same time.

These types of room problems must be optimized as much as possible BEFORE doing DRC. You cannot undo reflections, resonant nodes, poor positioning, etc., electrically after the fact. You CAN make the frequencies flat, but they are still resonating, or echoing just like before.

ETF's Demo Room section goes into some excellent detail about positioning and does contain some information about absorbers in Part 4. If you're really serious about your listening environment, explore your acoustic solutions first and then attack it electrically.

--Bill

[Jones_Rush]

How sensitive is it to you moving your head? Are we talking inches or feet?
Wes,
I really haven't done much testing in this regard (ie. positioning the mic at several locations and measuring). All I can say for now, is that when I sit on the couch, I can comfortably move as naturally as I always do, without ANY perceived change in sound. If I go far though (say 2-3 feet to one side, the sound I hear is of course much worse than what I would have gotten without correction. Room Correction is only for one place, that's always been the limitation. One heavenly spot, and with a bit of luck, it would be you who is going to occupy it, and not your wife).

I've never come close to that with a Behringer Feedback Pro
K-Wood,
I too have the BFD, and it's the only reason my system sounded decent before DRC came along. I used to route the bass below 80hz to the sub, and correct the sub using the BFD. Of course, the BFD does not correct the time domain, like DRC does, but still, it was a nice interim solution.

are they are other playback software programs that can accept the DRC plugins, other than Winamp and foobar2000? For example, will it work with J. River's Media Center?
DRC does not have a plugin, it only creates the correction filter, that a convolver plugin can use. Any player which has a concolver plugin will work, though as I recently found, you might need a plugin which can work with a 32 bit filter. At least with Foobar2000, going 32 bit made all the difference for me. You should check if the J. River's Media Center has a convolver plugin.

A point which cannot be over-emphasized is that you CANNOT make up for an acoustically corrupt room with EQ (DRC or any other) alone.
Bill, I agree that getting a perfect acoustically un-corrupt room is not possible using DRC alone, but DRC can do much more than a simple EQ can (which can only correct the frequency domain, but not the time domain). Look at the pictures I've included. Of course, it's not perfect, but let's agree it's much better than without (I'll include a waterfall plot later, so you'll see things are also improving there, but not as drastically as with the frequency domain, that's for sure).

Every excited node, every reflective surface (walls usually) adds a particular color to the sound at its effective frequency range. Highly excited room nodes resonate strongly and will extend the time at which that frequency is present (shown by poor impulse response). Compensating electrically can change the amplitude of a frequency, but not the sound of it modified by the room.
Bill, what you say is 100% correct regarding regular EQ. DRC, on the other hand, does not work like that. DRC tries to correct the time domain, in addition to the frequency domain, meaning, it will deal with reflections, with varying degrees of success, of course.

These types of room problems must be optimized as much as possible BEFORE doing DRC. You cannot undo reflections, resonant nodes, poor positioning, etc., electrically after the fact.
I completely agree with what you say about trying to make the room acoustically better before using DRC, this is true. This will also make the room listenable from more spots. But, and it's a big "but", in home instalations (in contrast to professional studio installations), you can't always do what you wish to your room, since there are many other things which come in to play (as you know). In these cases, the difference between a DRC'ed room and a non DRC'ed room, will be staggering. This is exactly the situation in my room.

You CAN make the frequencies flat, but they are still resonating, or echoing just like before
I'll have to disagree about this. The resonance and echoing will not be as severe as before, that's the "magic" of DRC, as I've said earlier. Time domain corrections.

I'll sum it up by saying: the better your room will be before DRC, the better it will be after it, BOTH frequency AND time domain wise.

[dwk123]

first, Jones_Rush congrats on getting DRC working. I've had some pretty amazing results, but my situation is in such a state of flux that I don't really have any clear A/B comparisons.
However I have to say :

*WHAT WERE YOU THINKING!?!? 16 bit filters!!?!? 32 is definitely the minimum, and I actually typically use 64 bit floats, just to be sure.


Originally posted by bblue

A point which cannot be over-emphasized is that you CANNOT make up for an acoustically corrupt room with EQ (DRC or any other) alone.

Every excited node, every reflective surface (walls usually) adds a particular color to the sound at its effective frequency range. Highly excited room nodes resonate strongly and will extend the time at which that frequency is present (shown by poor impulse response). Compensating electrically can change the amplitude of a frequency, but not the sound of it modified by the room.

For higher frequency sounds, left-right reflections between parallel walls at the listening position (for example) will blur the imaging making for much less precise localization in the sound stage and masked detail. Many other sources of reflection and blurring can be in effect at the same time.

These types of room problems must be optimized as much as possible BEFORE doing DRC. You cannot undo reflections, resonant nodes, poor positioning, etc., electrically after the fact. You CAN make the frequencies flat, but they are still resonating, or echoing just like before.


Jones already addressed some of your points Bill, but to emphasize:
You are right that passive treatments should be used to get the worst of the problems tamed. In particular, high-requency correction is really not possible in any meaningful way aside from equalizing the initial arrival. Flutter or slap echo in the mid/high spectrum is not really treatable electronically.

However, DRC is a pretty well thought-out time domain approach to room correction, NOT simply a parametric eq that flattens steady-state response. It is definately along the lines of a SigTech or Tact unit, although nobody would claim that it's at quite that level yet. Reflections, modes etc are all treated to some degree by DRC, whereas they obviously aren't in typical parametric/graphical approaches.

Ultimately, though, the value of DRC is that it does a spectacular job at handling bass problems which are very very hard to treat in typical domestic environments. Simple absorption techniques can handle the worst of the high-frequency challenges, but except for the few that can accomodate big tube traps or diaphragm absorbers, most folks have to take what they get in the bass.

[Eiffel]

Awsome! I can't wait to read your 'how to' post.

I've done quite a bit of work getting my 2 channel system to sound good in my room (ETF5 with calibrated mike, diffusors, absorbers, etc), and I am eager to improve it (an maybe have better results with my HTPC system ;) )

Eiffel

[3-way]

Jones_Rush --

Are you using Foobar2000? If so, please note that the latest version of Foobar2000 (v0.7 RC7) requires a version of foo_convolve that has been recompiled. The .dll in the link that was originally posted in this thread (http://www.hydrogenaudio.org/index.php?showtopic=10611&st=0&) no longer works. The proper .dll can now be found at: http://www.saunalahti.fi/~cse/foobar2000/ . While there, users may also want to download the ASIO output .dll.

Thought this might be helpful in your guide. Seeing your results yesterday made me want to try to figure DRC out on my own before your guide comes out (kind of like reading the book before seeing the movie).

Now if we could only get J. River to implement a convolver...

brad

[jriver]

Originally posted by 3-way

Now if we could only get J. River to implement a convolver...

brad
Brad,
Would our DirectX plug-in help?

http://www.musicex.com/cgi-bin/yabb/YaBB.cgi?board=MediaCenter;action=display;num=1062073747

Jim

[3-way]

Unfortunately, I don't think so. I did a search for a DirectX convolver plugin and came up with nothing. Unless someone else reading this thread knows of a DirectX convolver?

[ramick]

Jones_Rush,
Did you increase your volume control or any other EQ type settings from the first graph to the second (corrected) graph, ie. are you pushing more Watts in corrected mode?

Are you using LeAmpII for these? What is your audio equipment list?

[salsbst]

Jim, actually I believe that the DirectX plugin would help greatly, as someone found a wrapper that can wrap VST plugins (at least I think that's what they're called) as DirectShow filters.... but IIRC somebody tried to do so for the purposes of the MC DirectX host and found that it didn't work... darn I can't find the thread.... but talk of the VST/DirectX wrapper started in this thread here: http://www.avsforum.com/avs-vb/showthread.php?s=&postid=2474742&highlight=directx+AND+vst#post2474742.

Regards,
Stuart

[KikeG]

DRC can, theorically, make the sound at you room be the same as it if was played with flat speakers at an anechoic room. The problem is that, that kind of sound is not very natural (zero reverberation at room, zero room coloration), and the sound was not recorded with that kind of neutrality in mind. The ideal would be to reconstruct at you room the sound that the mastering engineers had at the recording studio. For that you would need an impulse recorded at there :)

BTW: foobar2000 does all internal processing in 64-bit float, whether you use 16-bit or 32-bit impulse files. In theory, using 32-bit files vs. 16-bit files shouldn't make much of a difference, since Revo recodings can't go much beyond 16-bit actual resolution due to line-in noise floor (not to talk about mic and room actual noise floor).

And BTW: As some of you said, the 'magic' of DRC is that, in theory, it can solve all room resonance and echo problems, because it works at time domain, not just frequency domain. However, it can't solve speaker distortion problems. Also, as noted, speaker crosstalk is not properly corrected using just a two-channel correction system.

[Mooneyass]

I think Jones_Rush has been very clear that this isnt the be all and end all to 2 channel audio perfection, just a very cheap way to get drastic improvements for us geeks who love computers and 2 channel stuff.

This is great stuff, cant wait to try it out.

Wes

[Jones_Rush]

Jones_Rush,
Did you increase your volume control or any other EQ type settings from the first graph to the second (corrected) graph, ie. are you pushing more Watts in corrected mode?
Yes. Of course. In order to fill room dips, you have to sacrifice Watts RMS. You can control how much dip filling you want to allow DRC to use. As default, it is allowed to use ~6 dB for dip filling. Also, when you actually convolve the filter with the media files, clipping will occur if you'll try to run the filter at -0.0 dB (since it is a VERY complex filter). For me, it takes about -3 dB to get rid of clipping completely. so, ~3 dB + ~6 dB, we're talking about 10 dB of reduced dynamic range. That's the price you're going to pay for room correction.

Anyway, for me this is not too much of a problem. I drive my whole sound system from the Revolution 7.1 (without a pre amp, and I have power amps without volume control), so without DRC, my volume control settings were ~ -15 dB (for a loud presentation at the listening spot), and now with DRC, it's about -5 dB.

I use a Carver 170Wx2 power amp, and my speakers have a sensitivity of 85 dB.


Btw,
The Step by Step guide is coming along quite nicely. I'll try to finish it on this weekend, but no promises.

[bblue]

dwk123 responded after jones_rush regarding my points about fixing the room first:Jones already addressed some of your points Bill, but to emphasize:
You are right that passive treatments should be used to get the worst of the problems tamed. In particular, high-requency correction is really not possible in any meaningful way aside from equalizing the initial arrival. Flutter or slap echo in the mid/high spectrum is not really treatable electronically.

However, DRC is a pretty well thought-out time domain approach to room correction, NOT simply a parametric eq that flattens steady-state response. It is definately along the lines of a SigTech or Tact unit, although nobody would claim that it's at quite that level yet. Reflections, modes etc are all treated to some degree by DRC, whereas they obviously aren't in typical parametric/graphical approaches.

Ultimately, though, the value of DRC is that it does a spectacular job at handling bass problems which are very very hard to treat in typical domestic environments. Simple absorption techniques can handle the worst of the high-frequency challenges, but except for the few that can accomodate big tube traps or diaphragm absorbers, most folks have to take what they get in the bass.Then KikeG writes:DRC can, theorically, make the sound at you room be the same as it if was played with flat speakers at an anechoic room. The problem is that, that kind of sound is not very natural (zero reverberation at room, zero room coloration), and the sound was not recorded with that kind of neutrality in mind. The ideal would be to reconstruct at you room the sound that the mastering engineers had at the recording studio. For that you would need an impulse recorded at there
Let's clear up some misconceptions. For sure DRC and other products that use controlled phase variable width eq, and separately posses the ability to (smartly) shift certain bands of frequencies forward or back in time are highly effective and superior in correcting certain types of problems in rooms. It's pretty amazing, actually. But moving time arrival around can only affect where room nodes occur, it cannot eliminate them. Eq can only help tailor the response of the speakers to the room, and raise and lower bands of frequencies regardless of why they're that way.

But these techniques cannot in any way make up for coloration in a room *caused* by nodes and reflections. Jones made the point that if you use DRC to soften a hot node (assuming it can't do any more with it in time domain) it will magically not resonate as much. In fact it will resonate just the same proportionate to the level of that frequency. While the frequency amplitude in the room will now be flat, it will have a different decay that adjacent frequencies. That will give it a particular color that simply will not go away without room corrections acoustically. In fact sometimes that particular frequency will be lowered a db or two to help hide that effect.

Reflections can only be corrected in the time domain if the reflection is the only source of the sound. Almost always the reflection is *in addition* to the original source of the sound and will arrive at the listeners position slightly delayed and from a different direction. This absolutely cannot be corrected electrically.

'Slap echo' that which is produced by sound bouncing back and forth between side walls, the ceiling to the floor, the front wall to the back wall, or any combination of the above, can not be eliminated electrically.

There's been some testing with identifying these delayed sounds and outputting an identical version of the sound out of phase so it cancels. That works to a limited degree (especially in low frequency ranges) in headphones but not in a room. In a room you're simply adding more contamination to the original.

In no way, can an arbitrary room be made to sound like an anechoic chamber with electrical processing alone. 'Taint possible, no way no how. An anechoic chamber is a room without relections and resonances, and without acoustic leakage from the outside. The only way a room can have no reflections and resonances is to be designed that way. It cannot be coerced after the fact, electrically.

And for the record (no pun), many recording studio control rooms are far from ideal acoustically. Mastering rooms are frequently even worse. Of course, some are pretty good, but they all have their own sound and the one of the roles of a good engineer is to know the room, know how it's wrong or influences the balance, and mix accordingly. It's a black art, to be sure.

Anyway, I'm not trying to pop your bubbles. But if you really want some good sound, identify problems in the room and fix them first. For higher frequencies it could be as simple as a few small sound absorbers strategically placed. For lower frequencies it is more difficult, but speaker placement, including speaker height (and/or distance off the floor), your listening position, positions of other furniture as well as the addition of corner traps (tube traps or other type of resonator/absorbers) will go a long long way to some major improvements of your listening environment. THEN tweak it with DRC or equivalent. You won't regret it.

--Bill

[K-Wood]

You guys are perilously close to the limits of my technical understanding here (which is why I should probably wait for Jones_Rush's guide), but has anyone been able to determine whether a convolver exists that will work with Media Center's DirectX plugin? A thread on the MC Interact forum mentioned Sonic Foundry's Acoustic Mirror convolver, but it appears that you have to buy the full version of Sound Forge to get it ($399 retail), because it is not sold separately anymore. Would the foobar convolver work with MC's DirectX plugin? And assuming there is an affordable convolver that would work, how easy is it to use?

Thanks,
Ken

[Iceman]

Bill,

It seems like you have not fully understood the underlying principles of digital room correction. Every practical aspect of room influence on the signal can be corrected with DRC. The degree to which you choose to do so is mainly due to practical considerations. However, in most cases the room response is measured with an omnidirectional microphone so head/ear-related directionality is not factored in. This problem is negligible for low frequencies in any case. Further, there is nothing theoretically preventing the measurement of the room response with an artificial head coupled with inter-aural crosstalk cancellation for a full correction, i.e. simulated anechoic environment for one specific listener location.

[Iceman]

In addition to Jones_Rush's success with the theoretically preferable FIR approach, I would like to strike a blow for standard, IIR-type, digital equalizers with parametric filters (Behringer etc). It may be interesting for audiophiles unable to go Jones_Rush's route and is way better than its reputation.

I noted there was some discussion on recursive filters (IIR) in the beginning of this thread. These are the digital counterparts to analog filters and will have a minimum phase characteristic, i.e. the phase response is directly linked to the magnitude response.

An ideal room would have well-defined reflections purely of minimum phase type so in this case an IIR filter equipped device would be able to correct fully, both in the frequency and time domains. Unfortunately, the walls of real rooms do not reflect perfectly and instead produce frequency dependent reflection ratios and worse, also introduce frequency dependent time delays. This means that a minimum phase type device will be able to fully correct frequency response magnitude deviations but only about half of the phase/temporal deviations for a normal room. Reverberation time (RT60) is normally cut in half with proper use of a digital eq. In other words, a somewhat too reverberant listening room will sound very dry and quick after eq. Naturally, the correction of loudspeaker temporal imperfections will suffer from the same limitations as room correction. For instance, non-linear crossover-region phase response cannot be corrected this way as it is not of minimum phase type.

So for those having the opportunity to do FIR correction, go ahead and don't look back. The rewards are nothing less than breathtaking. Others wanting to do do e.g. multichannel correction, for some reason are hesitant of playing with software or are having compatibility problems, use parametric digital eq for full frequency response magnitude correction and decent phase/temporal correction.

Don't forget that room correction is best suited (and most needed) for lower frequencies, below some 300-500 Hz. Correction of higher frequencies is of course possible, but above this frequency region one has to be careful not to introduce more problems than are solved by the correction process.

[JimmyMack]

Originally posted by Iceman
Bill,

It seems like you have not fully understood the underlying principles of digital room correction. Every practical aspect of room influence on the signal can be corrected with DRC. ... , i.e. simulated anechoic environment for one specific listener location.

I remember reading a Stereophile review of the original Tact RCS 2.0 two channel DRC system. Initially he refused to remove all of his room treatments as the designer recommended but after using the system and being continually being amazed at it's abilities, he eventually got rid of them. He ended his review suggesting that if he kept the unit, he's not sure if he would ever return to the room treatments again! I think that's a great testimony to the usefullness of DRC.

Incidently, he also mentioned the phenom KikeG is alluding to, namely, that a completely flat response in generally undesireable. When he dialed in a totally flat target impulse, the sound was quite bland. He said many people think they want a flat response, but if they heard it would be dissapointed. Great article. It's online here : http://www.stereophile.com/showarchives.cgi?437

[SFJoe]

Wow; glad I finally found this thread. I've been searching for a while to find out who has done any work with PC based digital room correction. I noticed that my end goal is a bit different. In my case, I'm not aiming to correct in real time. Instead, I'm planning on ripping my source disc, apply the room correction filters, and buring the disc back to a blank CD. The idea would be to get the impulse responses for different areas (living room, bedroom, car) and being able to create a separate corrected disc for each area that can be played back in a normal CD player, eliminating the need for the PC. Plus you really wouldn't really need a powerful CPU since nothing is done in real time.

Anyone try this yet?

-joe

[Iceman]

Jimmy,

I have heard two-channel as well as multi-channel systems in semianechoic and anechoic rooms as well as with full DRC and it definitely sounds different, hyper-detailed and hyper-dry with normal recordings. It is not really "correct" either as it is not what the sound engineer heard while recording/mixing/mastering the disk as KikeG noted earlier. All things considered, I believe many listeners in a good listening room would be most satisfied with DRC below 300 Hz or so but with DLC (Digital Loudspeaker Correction) for the entire frequency range.


Joe,

Off-line preprocessing is quite easily done with e.g. Matlab. I have tried this myself for fun. Without optimized convolution algorithms it takes a lot of time even on a fast PC, though. I know there is faster software around so your proposed method should be viable.

[Jones_Rush]

Anyone try this yet?
Joe,
It's really not a problem. The only difference is that you'll need to make the frequency response measure, in your car. Other than that, making the convolution off line and burning the result on a CD is a no prob. You can use Cool Edit pro for that.

I have heard two-channel as well as multi-channel systems in semianechoic and anechoic rooms as well as with full DRC and it definitely sounds different, hyper-detailed and hyper-dry with normal recordings. It is not really "correct" either as it is not what the sound engineer heard while recording/mixing/mastering the disk as KikeG noted earlier.
I have to agree, at least partly.
I didn't have the chance to hear too many CDs with DRC yet. But I can attest that the recording quality is going to count now, more than ever. With flat frequency response (or at least what you're getting with DRC), excellent recordings sound simply unbelievable, yet mediocre recordings usually seem a bit too bright and the bass is weak. Anyhow, once I'll have time I'll start testing different target curves other than flat, for all those mediocre recordings.

The Guide is about 80% done now. Do you want it to be able to fit on an A4 page ? (it's not going to be a post), I thought about making it compatible with 1152x864 resolution (minimum), but if you want to be able to print it easily, I'll back off to 1024x768, and make it a bit less wide.

[3-way]

I'm interested in using the psychoacoustic curve in ETF to shape the response after the application of DRC so that the DRC output file is not flat, but matches the algorithm of the psychoacoustic feature of ETF. I think the way it would work is this:

1. I take a reading using ETF.
2. ETF produces the impulse, amplitude v. frequency response curves, etc.
3. I apply the psychoacoustic response feature in ETF, which alters the amplitude v. frequency response curve.
4. The ETF file is exported as a .pcm file, complete with the psychoacoustic response curve applied.
5. I insert the .pcm file into the normal.drc file and run DRC.
6. If the response curve in DRC is set to flat, the psychoacoustic feature applied in ETF will control the DRC output file, and another measurement taken in ETF with DRC applied should show a curve (rather than a flat line) that looks like the ETF psychoacoustic curve.

Is this correct?

[tbshooter]

To complete the correction, one need to do a hearing test and compensate for one's hearing loss. :p
Make the guide printable please.

[RayL Jr.]

WHAT IF YOU LIVE IN A VAN DOWN BY THE RIVER?
http://chrisfarley.netmegs.com/snl/images/mattfoley/MF-7.jpg

[Jones_Rush]

Regarding ETF. It is indeed a great software, but it cost money. The guide is built in a way that you will be able to evaluate DRC without spending a single buck for software. This is why in the guide, I'm using Cool Edit Pro 2.1 (Demo version is enough), which has tons of features, but it is a bit more complicated than ETF, at least for achieving our goal (don't worry, the guide will make things very simple). So, in order to use the guide, you'll need Cool Edit Pro 2.1 (2.0 should work fine too). Recently Adobe purchased the rights for Cool Edit (it is now called Adobe Audition), and as far as I know they still don't offer a demo version. But you can still get the original Cool Edit from several sites.

You can get Cool Edit Pro 2.0 from here (http://www.softnews.ro/public/cat/11/1/4/11-1-4-6.shtml). Anyway, download it, but don't install it yet. Wait for the guide to come out first, since after installation, you only have 21 days before the Demo becomes obsolete for our purposes.

[Mastiff]

Jones_Rush, you sure edit your posts a lot! :D I have read the full thread and wondered about something: Is there any way to utilize this to adjust regular equalizers (12 and 15 bands) in a home theater? I want to have a more even frequency response at least for my center and front speakers, and DRC is unfortunately not possible since I use SPDIF pass through to play my DVDs.

And another thing: Wouldn't it be possible to use ZoomPlayer for playback of this? There are many filters that work with it, and I think it can do any DirectX filter (but I may of course be wrong). :cool:

[Jones_Rush]

Mastiff,
In order to use DRC for DVD playback, you need to add a convolver filter, to come after the Audio Decoder filter. The convolver filter input pin, should be compatible with the Audio Decoder output pin. Then the convolver needs to connect to the Audio renderer filter. Theoretically, you can DRC your entire 5.1 system this way. All you need is the right convolver filter. Unfortunately, no one took the time to write one, yet.

Stay tuned for an edit :-)

[Mastiff]

You mean that it could be used on SPDIF pass through as well? That would really be cool! :cool:

[Jones_Rush]

You mean that it could be used on SPDIF pass through as well? That would really be cool!
Only if you have an AC3 encoder (like the nForce). Anyway, we still can't even DRC two channels from the 5.1 mix. We don't even have that filter.

[nubz69]

"In no way, can an arbitrary room be made to sound like an anechoic chamber with electrical processing alone. 'Taint possible, no way no how."

YET!! You never know what the future may hold.

" And for the record (no pun), many recording studio control rooms are far from ideal acoustically. Mastering rooms are frequently even worse."

What are you talking about? I have yet to go into a studio that was not designed with a good control room. Mastering rooms are even more accousticly designed and treated then studio control rooms. Both of these are usualy way better then 99% of what you find in peoples homes. Only the small 1% that has a dedicaded listening room with full accoustical treatment can really compeate. There are a few small studios and home studios that can't afford a custom built control room but I have never seen a contol room without any accoustical treatment.

[Robert2413]

A couple of random comments on this thread:

FIR filters are not necessarily linear phase; it's just easy to do for FIRs thanks to commonly available algorithms to determine the coefficients. It's easy to make an minimum-phase FIR filter (just put all of the zeros inside the unit circle in the z-plane), or to make a non-minimum-phase FIR filter that is not linear phase. (Virtually all linear-phase filters are non-minimum-phase; the exception is the so-called "gaussian" and "bessel" lowpass filters.)

If a filter is linear-phase, that means it has constant group delay (not "no group delay"). All such filters have symmetrical impulse responses, which means equal pre- and post-echo. Because the temporal masking of the ear is asymetrical in time (pre-echoes are less well-masked than post-echoes), linear phase equalization may not be desirable. This is particularly true if the frequency response irregularity being equalized is itself minimum phase, because then a filter with complementary magnitude response will automatically equalize the phase distortion as well.

Group delay only has physical meaning when it is constant with frequency. It is the negative derivative of the phase vs. frequency, so, when not constant, it does not necessarily have anything to do with physical time delay. Indeed, virtually any minimum phase highpass filter will have a negative group delay over some frequency band, but this does not mean that the filter is non-causal (i.e., the output emerges before the input is received).

IIR filters can be non-minimum-phase and can be equalized to be arbirarily close to constant delay. In fact, one can readily make a graphic equalizer with group-delay-corrected IIR filters that has negligible phase distortion. The impulse response of this equalization may well have less pre-echo than a linear-phase FIR filter bank having no phase distortion at any frequency.

There is no difference between equalizing in the "frequency domain" and the "time domain," as these are duals of each other -- they are just two ways of expressing the same linear operations on a signal. (This assumes that one knows the magnitude and phase response at all frequencies).

Finally, I have to agree with the posters who emphasize that room correction is crucial prior to equalization. Room reflections cause comb filtering in three dimensions, and a one-dimensional correction filter can only correct this comb filtering at one point in space. Above 500 Hz or so, moving the measurement microphone just a small amount will significantly change the amount of reinforcement and null occurring at any frequency. Therefore, the advice of doing full magnitude and phase correction at low frequencies and magnitude correction of the loudspeaker response alone at high frequencies is probably wise.

[Jones_Rush]

Robert, thanks for the informative post.

Therefore, the advice of doing full magnitude and phase correction at low frequencies and magnitude correction of the loudspeaker response alone at high frequencies is probably wise.
I think that I've already talked about it somowhere in this thread.
Anyway, I think what you ask is, at least theoretically, possible with DRC.

Here take a look at this correspondant between Denis and I:

Is it possible to make DRC correct only a limited region of the frequency bandwidth? (for example, only from 20hz to 200hz).

Actually no, usually there's no need to do it. Anyway, you can reduce the correction on higher frequency a lot using a low MPWindowExponent and EPWindowExponent, something like 0.5 or less. This way above a given frequency, depending on the exponent used, DRC starts to correct just the direct sound. Correction is not disabled, but anyway much reduced.


Correcting only the direct sound, is in another words, correcting for the speaker anomalies only.

[Iceman]

Robert,

Good points. It is of course theoretically possible to produce any type of phase response with a FIR filter as well as produce an almost phase distortion free IIR filter. However, in practice, all equalizers with IIR filters I am aware of produce a minimum phase response. Conversely, to the best of my knowledge, all full DRC hardware and software solutions use FIR filters.

[bblue]

bblue: "And for the record (no pun), many recording studio control rooms are far from ideal acoustically. Mastering rooms are frequently even worse."

nubz69: What are you talking about? I have yet to go into a studio that was not designed with a good control room. Mastering rooms are even more accousticly designed and treated then studio control rooms. Both of these are usualy way better then 99% of what you find in peoples homes. Only the small 1% that has a dedicaded listening room with full accoustical treatment can really compeate. There are a few small studios and home studios that can't afford a custom built control room but I have never seen a contol room without any accoustical treatment.Did I say without any acoustic treatment? And we're not comparing studios to peoples homes. Those are two completely different worlds.

If a typical audiophile or musician walked into a studio or mastering facility they would undoubtedly be quite impressed. If a professional in the industry, such as an engineer who works in these facilities for a living (such as I have), walked into the same room they would likely have quite a different opinion.

These rooms are most always designed professionally (except for home brew studios), but that doesn't mean that they all sound good and it doesn't mean that home brew studios automatically sound bad. There are many tradeoffs, many different solutions to an acoustic problem and an 'acceptable' margin of error which varies. Many folks, including some engineers, do not yet appreciate the profound effect that the choice of wiring, amplifiers, speakers have on the room sound apart from but in combination with the acoustics.

In reality, from studio to studio, mastering room to mastering room, there is a wide variance of standards which result in the wide sonic range of recordings on the market. Why are some recordings bass-heavy? Why are others very top-heavy? Why are some recordings super clean and pristine sounding while others are practically unlistenable due to distortion, excessive eq or other processing? Because studios are different, engineers interpret rooms differently, and the goal of the recording is different from artist to artist, producer to producer. Ultimately, they all depend on the sound of the room to form the sound of the mix, and that is a moving target.


And for the Flat fire: Many folks confuse the term Flat and the sound it implies, as how a typical home audio system sounds when its tone controls are set flat (no eq, no loudness). Therefore it's an undesirable goal of room balancing whether it be acoustically, with eq, or with DRC. That isn't how flat sounds with good equipment and a balanced room, and it's not how an anechoic chamber sounds with good quality speakers and associated equipment, and a well recorded source.

The goal is to make the speakers/room as uniform and colorless as possible. That is flat. That is the only way you'll hear what is really on your source material the way it was recorded (more or less). Once you get your room to that point, you still may not like what you hear. A lot of that can be traced to the speakers, amp, preamp, playback source, or even cabling that is being used in the system. Not to mention the type of music you prefer listening to (different recording standards).

If you want to hear nothing more or less than what is on the original, you want a flat colorless (acoustically) room with plenty of amp headroom and good sized quality speakers (general terms here). But once achieved, that goal can also be used as a starting point to further modify the characteristics (acoustically or electrically) to tailor a sound you like or relate to. The resulting sound is almost never flat in a consumer environment. What you like to hear and what is recorded are frequently two different things.

--Bill

[KikeG]

In any mid-quality audio setup, say 99% or more of colorations are just due to speakers & room interactions.

It's pretty easy and common to have a source, amp and cabling, that as a whole, have a frequency response desviation from flat of less than +-0.5 dB from 20 Hz to 20 KHz. It's impossible, quite by far (without DRC, that is), to get a room and speakers that can achieve such a degree of flatness.

Speakers, be alone or interacting with a room, always produce, in comparison with the former, huge frequency response deviations from flat.

On other side, colorless = flat, and the only true flat rooms are anechoic chambers, and it's pretty difficult to build a true anechoic chamber. Anything else is going to be colored, in one or another way. It has been found that rooms or recincts whose reververation time decays with frequency (on in other words, that emphasize somewhat lower frequencies) are usually the most pleasant sounding, and that's how theaters, concert halls and mastering rooms, for example, are treated to sound like.

[Jones_Rush]

The guide will be finished this Tuesday, or Wednesday (max). It will be ~ 1mb. At the beginning I'm not sure I'll find a place to host it, so those who want it, should PM me their email address (and make sure it has 1mb of free space).

[ElvisIncognito]

Originally posted by KikeG
In any mid-quality audio setup, say 99% or more of colorations are just due to speakers & room interactions.Right... That's why a $20K system sounds SO much better than a $5K system - even when they're both in the same room. Enjoy that $500 Denon receiver, then - I've heard (much) better.

We're going to have to start a "Hall of Shame" for you, man. I do believe yours will put Jones's to shame...

[KikeG]

Originally posted by ElvisIncognito
Right... That's why a $20K system sounds SO much better than a $5K system - even when they're both in the same room. Enjoy that $500 Denon receiver, then - I've heard (much) better.

Red herring. Two systems can have both good frequency response and still one have other better properties and sound better than the other. I didn't say otherwise, so you better kept silent.

Or, you could just try to prove me wrong in what I said, instead of just trying to discredit me basing on something I didn't say. And if I had said it, on something more solid than your usual arguments.

We're going to have to start a "Hall of Shame" for you, man. I do believe yours will put Jones's to shame...

Please refrain for such statements or I will have to reply in a similar fashion, and this will probably end into another flamewar. I demand some moderation here. I think I've written enough at these forums (not to talk about other forums) so that I've shown that I do know at least something about actual audio.

[Mooneyass]

Ya ya... Jones_Rush, I'll host the file. You've got PM

Wes

[Jones_Rush]

Thanks!.

[Mooneyass]

I keep meaning to ask, what kind of horsepower does this convolve filter require? I have a PII-450 laying around that I'd like drop my Delta 410 into and turn it into my new CD player complete with dedicated 15" LCD (also currently unused).

Wes

[Jones_Rush]

Well, when running Foobar2000's convolution plug-in, with normal filter accuracy and a 32bit filter (probably what you'll use too), there is a 40% CPU usage on my Athlon 1.1Ghz, so I guess PII-450 might be a bit slow, but it still might be enough.

[Robert2413]

Originally posted by Iceman
Robert,

Good points. It is of course theoretically possible to produce any type of phase response with a FIR filter as well as produce an almost phase distortion free IIR filter. However, in practice, all equalizers with IIR filters I am aware of produce a minimum phase response. Conversely, to the best of my knowledge, all full DRC hardware and software solutions use FIR filters.

I know of several multiband audio processors for broadcast use that employ phase corrected IIR filterbanks, essentially as "program-adaptive graphic equalizers." The filterbanks are generally realized as subtractive trees consisting of phase-corrected lowpass filters and matching delays. There are several advantages to this topology:

1) Selective FIR filters with center frequencies that a small fraction of the sampling frequency require very large numbers of taps to achieve selectivity. While the expense of this can be overcome by several techniques (multirate structures; FFT/process/IFFT), all of the operation-saving techniques of which I am aware add large amounts of delay and are therefore unsuitable for real time applications.

2) Because the phase correction of the IIR lowpass filters is typically applied only to the frequency range occupied by the passband (and perhaps the first part of the transition region), the overall delay is substantially lower (sometimes by a factor of two or more) than that produced for a phase-linear FIR filter of equivalent selectivity but having linear phase at all frequencies. This reduces the overall delay of the IIR filterbank even by comparison to an FIR filterbank that does not use operation-saving techniques such as those listed in (1). (Techniques requiring aliasing cancelation cannot be used when the bands are summed with non-equal gains, as they ordinarily are in equalizers.)

3) The IIR filterbank is very straightforward, and can be used for equalization without any concern for aliasing cancelation, which crops up in some operation-saving low-delay FIR filterbank algorithms (like those used in perceptual codecs). The IIR filterbank also operates at the sample frequency without a need for large numbers of operations (although double precision arithmetic or error feedback may be required for the low frequency filters).

[Iceman]

Robert,

I was referring to consumer equipment in my statement above, but it was interesting reading nonetheless. Actually, I have used similar techniques to those you describe in 1) for non real time filtering purposes.

Another perhaps interesting point is that DRC can be much closer to true absence of reflections (for a small space surrounding the microphone position) than anechoic chambers, especially for frequencies below some 70-100 Hz depending on the chamber. This is due to the absorbing wedges producing a rather sharp loss of absorption when their length is shorter than 1/4 of the sound wavelength. Further, anechoic chambers are not totally anechoic even for higher frequencies where there may be small, but not insignificant, reflections from the wedges and/or floor structures, ventilation openings, lighting installations etc.

[Mooneyass]

OK, I guess I'll spend $70 on a new mobo seeing as I have an Athlon 1.2 sitting in my parts bin too. Gotta use some of this junk....

Wes

[3-way]

There are now a number of people who have responded to this thread who seem to have a solid understanding of DSP and experience working with audio DSP. Is there sufficient interest and knowledge to create a C or C++ based program to implement crossovers, loudspeaker EQ, time alignment and room correction? Something that could run either on the PC directly, or could be downloaded to a DSP evaluation board? A similar commercial product is about to be released (for about $3,500 + $750 for the program needed to set it up) (see http://www.clarityeq.com/ ).

As a starting point, I'd like to develop something like the following:
1. Vanderkooy filters (high order, linear phase crossover filters, as discussed in various papers by Lipshitz and Vanderkooy, published in the Journal of the AES in the early- to mid- 1980s) for 2-5 drivers per side.
2. DRC or a DRC-derivative applied to each driver just beyond the frequency band used.
3. Able to be loaded onto a DSP evaluation board (for more processing power, allowing for the processing of rear and surround channels, as well as ambiophonics). Also, the DSP boards could be placed after a pre / pro so that volume for all channels would be controlled with a single volume control, and gain for each driver could be set within the DSP's GUI.
4. DSP board should have AES3, spdif and balanced XLR inputs and high quality ADCs and DACs with balanced XLR and single-ended outputs.
5. A nice GUI setup wizard and control so that the final program is easily accessible to all.

I've just started looking into this more seriously over the past couple of weeks, so, what I proposed may be way off base. I'm not speaking from authority or experience, just as someone who *really* would like to see such a program mature. I don't think that the hardware side of things should be too difficult, and the algorithms for the Vanderkooy filters exist (I can provide them), but I don't know enough about the underlying glue that holds everything together to know whether this is possible.

Any comments?

Brad

[K-Wood]

I'd think that there would be a small, but significant, market for what you describe. By analogy to the Holo3D card, I would suspect that people might be willing to pay up to $800 to $900 for such a card. Of course, a software application that used existing hardware (Delta 410) and made adjustments only below 300Hz would have a much bigger market.

[Jones_Rush]

The guide will be released tomorrow afternoon.
Warm up your mikes...

[Robert2413]

Originally posted by 3-way

[snip]
As a starting point, I'd like to develop something like the following:
1. Vanderkooy filters (high order, linear phase crossover filters, as discussed in various papers by Lipshitz and Vanderkooy, published in the Journal of the AES in the early- to mid- 1980s) for 2-5 drivers per side.

Brad

Mind you, I'm not volunteering (as I have a Real Job(TM) to keep me busy), but I'd just like to comment that one really doesn't have to limit oneself to the Vanderkooy filters (I'm relying on a vague recollection of the paper, which I read when it first came out). One can use virtually any lowpass filter -- Butterworth, Chebychev, elliptical, etc. -- with a relatively flat passband and then compute an allpass phase corrector for the filter using any of a number of filter design programs. If you're lucky, the resulting passband delay will be precisely n times the sample period, where n is an integer. However, nature seldom works that way, so an additional allpass section to pad the delay to the nearest integer sample is often required. (Alternatively, one can oversample to increase the fineness of the delay quantization, but the allpass usually works pretty well and has to avantage of being much more finely adjustable.)

As described in one of my posts above, one makes filterbanks by taking the output of such a phase-corrected lowpass filter and subtracting it from a delayed version of its input to get a voltage-complementary highpass output. This operation can be repeated as many times as necessary to create all of the bands in the filterbank by further dividing the output of either the highpass or lowpass filter (depending on whether you start the filterbank with the lowest or highest frequency). Of course, you also have to add compensating delays in the branches of the tree that is produced, making the delays between the input and all of the outputs the same. These trees are strictly voltage-complementary (that is, the sum of the outputs is unity with no magnitude or phase distortion even if the phase correction of the lowpass filters is imperfect). However, such crossovers are not power-complementary, and one has to think carefully about the implications of this if the loudspeaker is to be listened to under farfield conditions, where the reverberant field tends to randomize phase. Of course, DRC will not work very well (if at all) in the far field, so this may be a moot point in this project.

(The only crossover of which I am aware that is simultaneously power- and voltage-complementary is a simple first-order analog crossover. Does anyone know of any others?)

[0db]

Originally posted by ElvisIncognito
Right... That's why a $20K system sounds SO much better than a $5K system - even when they're both in the same room. Enjoy that $500 Denon receiver, then - I've heard (much) better.


Um... he did say "speakers AND room interactions."

Unless you're working near maximum output levels for each system and the expensive system happens to have loads more available power, I honestly doubt you'd hear a "SO much better" difference between a 20k and 5k system running the SAME speakers in the same room. Some difference, yes... but I think it's pretty realistic to say that speakers and listening environment make up the vast majority of the sonic coloration of your output.

[bblue]

Unless you're working near maximum output levels for each system and the expensive system happens to have loads more available power, I honestly doubt you'd hear a "SO much better" difference between a 20k and 5k system running the SAME speakers in the same room. Some difference, yes... but I think it's pretty realistic to say that speakers and listening environment make up the vast majority of the sonic coloration of your output.Hmmm. Have you actually heard good high-end equipment (and comparable speakers)? If you really have heard a great system (let's say one with higher grade Krell, Levinson, Rowland or equivalent power amps, pre-amps and/or processors), high grade cabling (not the likes of Monster) and a well designed set of speakers (Meridian, B&W, WATT, and many others), good source material and a not-completely-awful room, you wouldn't even notice the room for some time -- you'd be that blown away by the rest of it. Oh, and this is at any output level -- it's not unique to extreme high power listening, or minimum/maximum output levels, though you will consistently find that high grade higher *current* amps take the lead on sonic quality even if that capacity is not used.

Of course room improvements will almost always improve things if done acoustically. Electronically as well if the electronics of the gear processing the signal don't destroy it sonically before or after it gets to the actual filtering (which is quite frequently the case, by the way).

You would need to maintain a pure digital path, no computer based reclocking or resampling, no DA and AD conversions, etc.

If the audio was being generated in the same computer as the filtering (such as ripped wav files or better), you would need to carefully select playback codecs/filters, the player itself, maintain an all digital unconverted path, and exit the system digitally into a very high grade DA converter to hand off to the rest of the audio gear. It is not something you could even approximate with a winamp plugin (because of winamp, not the plug-in necessarily).

--Bill

[hyslopc]

What's specifically being discussed here is the *relative* importance of each component compared to the rest of the components in the listening chain. I agree with many others who has posted here saying speakers are #1, room is #2 (assuming we're talking typical listening rooms - obviously nothing will sound good in a small concrete basement). The rest of the components are important, too - no-one's doubting that. It's just that they're not as important as the room is.

[KikeG]

bblue:

foobar2000 plus the convolution plugin, with Win98 waveout, or WinXP KS or ASIO or kmixer-immune drivers, would accomplish perfectly the player part according to your requirements. And it's free!

[hyslopc]

Unfortunately foobar2000 won't do it for me: I've already got hundreds of GB of music in an MC9 database. Is anyone looking seriously at getting a convolver plug-in working with MC9?

[3-way]

Robert2413 --

I think that the method you described in your last post is what Weiss Engineering describes in the manual for their eq1-lp product (pages 10-11 at http://www.weiss.ch/eq1/EQ1-LP_Manual_V5_0%20E.PDF ). Is this correct?

Would you recommend using only IIR filters for the crossovers, speaker EQ, room EQ and "music" EQ? Or is there a place for FIR filters? Also, could you recommend a book or articles re audio dsp?

To respond to other posts, there's a saying I've heard among long-time audiophiles re the importance of various parts of the audio system: "The room is 80%, the speakers are 19% and the rest of the system is 1%." This may be a bit of an exaggeration, but I would agree that someone spending even $200 on a pair of cables for a system that is in a room without treatment, that has never done any testing on speaker placement, etc. is focusing his efforts / money in the wrong place. Standing waves and early reflections will effect cheap speakers and $100,000+ speakers. Better equipment will give you more headroom, dynamic range, etc., but a great DAC will never eliminate standing waves or early reflections.

brad

[Jones_Rush]

I'm really sorry to disappoint, but the guide won't be ready for today.
Don't get me wrong, everything is finished, except for one important thing: Safety measures!.

It might seem ridicules to postpone the guide for this reason, and I thought so too after Denis told me this yesterday, but believe me, after almost losing my entire sound system today, + getting 220V shock to my body (my neck still hurts!), after running for the electricity outlet, in order to disconnect my amp, in order to stop my near-exploding speaker to fulfil his sine-sweep death wish, I can't stress the importance of safety enough.

DRC, unlike ETF, can not work well with a 5 seconds MLS test which produce a 60 dB S/N measure. It needs a ~ 90 dB S/N recording, so a long log sine sweep must be used, at pretty moderate levels (about 90 dB SPL, at the listening spot). I'm still in the process of writing a fail safe procedure, in order to get the Sine Sweep measure with the least risk possible for the speakers. During this process, I've ruined a pair of headphones, and was THIS close to lose my entire system today, when my left speaker, which was supposed to be muted (Damn Revo driver!!!), suddenly bursed with 100 dB SPL of static noise, in the middle of a multimeter voltage check of the right channel.

Anyway, I'll do my best to complete everything for tomorrow. Sorry, but I don't want to feel responsible for your damages.

[Robert2413]

Originally posted by 3-way
Robert2413 --

I think that the method you described in your last post is what Weiss Engineering describes in the manual for their eq1-lp product (pages 10-11 at http://www.weiss.ch/eq1/EQ1-LP_Manual_V5_0%20E.PDF ). Is this correct?

Not quite. The Weiss uses a trick where they pass the signal through the filter in normal order and then in reverse order. This doubles the magnitude response (on a dB scale) but cancels out the phase shift. In order to have this work in less than infinite time, the signal has to be divided into frames and then spliced back together (an issue not mentioned in the Weiss manual).

The Weiss manual is incorrect in claiming that it is impractical to phase-equalize IIR filters by cascading them with allpass filters. This is actually very common, although usually limited to phase-equalizing the filter in its passband and perhaps in the first part of the stopband

Would you recommend using only IIR filters for the crossovers, speaker EQ, room EQ and "music" EQ? Or is there a place for FIR filters? Also, could you recommend a book or articles re audio dsp?

brad

I think there is a place for FIR filters, and I have seen programs where the user draws a target frequency response on-screen and the computer synthesizes an FIR filter that closely approximates it.

When one is trying to match such a target response, one is doing what is called an "optimization" in mathematical terms. The optimizations required to approximate FIR filters are numerically stable and often involve solving linear equations, which is a very well understood procedure. Optimizations that fit IIR filters to measured or arbitrarily specified responses usually involve solving nonlinear equations where numerical stability is much harder to achieve globally.

That being said, I prefer second-order IIR filters for "musical" EQ because there are a small number of familiar parameters to manually adjust (center frequency, bandwidth, and amount of peak boost or cut). It's really a matter of preference and what one has trained one's ears on.

A useful audio signal processing book is "Digital Audio Signal Processing" by Udo Zoelzer (Wiley, ISBN 0 471 97226 6). This text assumes that one has already worked through an introductory course in discrete-time signal processing using a textbook like Oppenheim & Schafer, plus a thorough understanding of z transforms.

[Jones_Rush]

This text assumes that one has already worked through an introductory course in discrete-time signal processing using a textbook like Oppenheim & Schafer, plus a thorough understanding of z transforms.
LOL.

[hyslopc]

There seem to be three conversations going on here:

1) A how-to guide to get a PC to implement audio DRC
2) A theoretical discussion on audio filters
3) A debate as to how important DRC is relative to other factors

Perhaps we should split this thread into three, so that each of these conversations can get coverage on their own? Personally I only have interest in #1 (above).

[KikeG]

Jones: for increasing SNR you could just use a longer MLS measurement. On the other side, using an IRS signal instead of MLS, the measurement will be also more inmune to speaker distortion. Take a look at Aurora help.

[Iceman]

If a 5 s MLS measurement produces 60 dB S/N, one would need 5000 s for 90 dB S/N, perhaps a bit impractical. I believe a log sine sweep is more useable here, especially as it makes speaker non-linearities less of a problem.

[Myren]

nuts, gotta go run off to class. on page 6, this looks awsome, cant wait to keep reading.

how much potential is there for multi channel (see: quadrophonic and 5.1)?

(dont most MAudio's have builtin phantom power? *cross fingers for my 410 saving the day*

Myren

[Jones_Rush]

Ok, I'll do it like this:
I'll do my best to complete the safty procedure until tomorrow. If I wont finish it by tomorrow afternoon (eastern), I'll send the guide anyway, to anyone who will request it via PM. But, anyone who decides to follow the guide, before the safety procedure is complete, should have a really, REALLY good idea at what he's doing, in regard to speaker protection.

As a general rule of thumb, if your speaker's sensitivity is rated around 90 dB, and your tweeter has a *continuous* (not peak) power handling of at least 10 Watts @8 Ohm (real world), you should be completely on the safe side, and you should have no problem to achieve a 90 dB SPL of the log sine sweep, at the listening spot, without risk.

[KikeG]

Jones, are you sure you need such high measurement SNR?

I have been doing some synthetic tests with noisy MLS signals with CoolEdit Pro 1.2a, and used a 5-second MLS signal with a noise floor of -40 dB RMS (around 34 dB SNR), and obtained from there an impulse response and corresponding inverse filter. The impulse response had a SNR of 87 dB. Trimming the inverse filter impulse response so that nasty impulsive garbage before and after main impulse is removed, I had a inverse impulse that worked well down to 20-25 Hz. Convolving that with a 16-bit RMAA test signal, I achieved a noise floor level of -85 dB and dynamic range of around 95 dB.

With a longer MLS I could improve results even further. For example, using 40 sec. MLS instead of 5 sec, SNR of the impulse response was of 100 dB. With a better SNR measurement than 34 dB I used, results can be improved too.

[bblue]

KikeG: Thanks, yes you are correct. I've had foobar2000 sitting here to be installed but had not played with it yet in my audio editing/listening equipment environment. Did that last night, and it does seem to do a very good job just passing a basic signal (44.1 or 48khz wav file) to SPDIF out of my Delta 1010 card. It's amazing how many players get even that much wrong.

Kernel threading mode doesn't seem to work but I haven't investigated if that's a compatibility issue with the card drivers, or what.

With any such player I need it to be able to process a bit for bit representation of the source. The source, the playback or the editing of the source through Cool Edit Pro, when fed to my Merdian 518 digital processor on SPDIF out, need to be sonically identical. Foobar2000 does seem to accomplish that, and in this environment is the only one that does besides WMP.

--Bill

[hyslopc]

bblue: Have you tried MC9 with all DSP settings turned off and output mode of ASIO? I use the Delta 1010's S/PDIF out, and find that in this mode, it gives me a pure 44.1kHz signal.

[bblue]

3-way writes:To respond to other posts, there's a saying I've heard among long-time audiophiles re the importance of various parts of the audio system: "The room is 80%, the speakers are 19% and the rest of the system is 1%." This may be a bit of an exaggeration, but I would agree that someone spending even $200 on a pair of cables for a system that is in a room without treatment, that has never done any testing on speaker placement, etc. is focusing his efforts / money in the wrong place. Standing waves and early reflections will effect cheap speakers and $100,000+ speakers. Better equipment will give you more headroom, dynamic range, etc., but a great DAC will never eliminate standing waves or early reflections....and average equipment will never produce decent sound regardless of how flat the room is.

There's no real audiophile that I've ever spoken to or read about that will entertain the notion that the electronics of a system are of only 1% importance to the resulting sound. That sounds like something 'specification audiophiles' (you know, the ones that either don't hear it, or have never heard it) would conclude or imagine by looking at the specs for each piece of equipment and then drawing a conclusion about how it probably sounds. Doesn't work that way.

I would venture that in terms of percentages, the electronics with speakers is more like 80-85% and the room is the fine-tuned remainder that makes the difference between a damn-good sound and an extraordinary one.

Speakers in particular play an interesting role because they interact with both the electronics as well as the room. But it's truly amazing how good mediocre speakers can sound when connected to a quality system. Uncanny, actually.

Room correction is not a solution. It is an important one of many steps necessary to achieve great sound.

--Bill

[bblue]

hyslopc: No, I don't use MC9, but since it's based on WMP I don't doubt that capability in the least.

As an accessory to evaluate audio content, I have just left WMP at its default settings, but set the MMC panel to use only defined devices for playback, and the defined device is Delta 1010 SPDIF out. This avoids kmixer and allows faithful output of any bitrate file or file format. In the case of MP3 or other similar formats it uses only what I have installed as codecs for DirectX to use, so the path and content is neutral.

Foobar2000 setup is similar except it doesn't need to rely on the Windows defined devices so I just set it to produce the same path. Works very well.

--Bill

[KikeG]

Originally posted by bblue
There's no real audiophile that I've ever spoken to or read about that will entertain the notion that the electronics of a system are of only 1% importance to the resulting sound.

Maybe you should read more from actual audio engineers. Anyway, what I claimed was about "flatness" of the system, not just final sound quality. "Flatness" is just frequency response, and what I said about speakers and room interaction accounting for 99% of flatness (frequency response) problems is evident and very easy to prove.


I would venture that in terms of percentages, the electronics with speakers is more like 80-85% and the room is the fine-tuned remainder that makes the difference between a damn-good sound and an extraordinary one.

No system can sound very well with bad room acoustics, period. Nearfield monitoring can alleviate some of these problems, but in usual hi-fi semi-diffuse-field monitoring, bad acoustics will ruin sound from any equipment.


But it's truly amazing how good mediocre speakers can sound when connected to a quality system. Uncanny, actually.

So uncanny I can't believe it, sorry. Speakers are the most imperfect devices of the whole audio chain, in every respect. Room and speakers play a huge role, in *whole* final sound quality. Any people that really knows something about acoustics, speakers and electronics knows this.

[hyslopc]

No, I don't use MC9, but since it's based on WMP I don't doubt that capability in the least. To my knowledge (long-time user and beta-tester) MC9's audio isn't based on WMP, and never was. Their first implementation of video playback was based on WMP, but they subsequently wrote their own code to handle video before MC9 was released.

the defined device is Delta 1010 SPDIF out. This avoids kmixer I think you better check on that. I don't see how that avoids kmixer at all. The only way that I'm aware of to avoid the kmixer are with specially-written sound-card drivers (which to my knowledge are not available for the 1010) or via ASIO. So to my knowledge, the only way to get bit-perfect audio with the Delta 1010 is to use the ASIO driver.

[KikeG]

kernel streaming works too for avoiding kmixer. I don't know if MC9 can use it, though.

[bblue]

KigeG writes:Maybe you should read more from actual audio engineers. Anyway, what I claimed was about "flatness" of the system, not just final sound quality. "Flatness" is just frequency response, and what I said about speakers and room interaction accounting for 99% of flatness (frequency response) problems is evident and very easy to prove.Well I wasn't responding to your post, but I don't disagree with your statements about flatness. Only that flatness is an all-encompassing measurement that somehow quantifies how the system will sound. It does not. Flatness does not equate to overall system sound, only static frequency response.

Read more from audio engineers? How about I just talk to them? Most of my friends are audio engineers and audiophiles and active in that field. I was an audio engineer (recording) for about 10 years, and active in the field semiprofessionally for as many. I've designed and modified a variety of high-end amplifiers, preamplifiers, and speakers, and have designed and modified recording and listening rooms in studios and homes. Not to mention 100's (1000's?) of hours at the console.

So yeah, I guess I could read more, but I don't speak without the experience to back it up, nor do I express an opinion on something I haven't heard or measured myself.

--Bill

[bblue]

hyslopc: I know little about MC9, so my comment about their internal use of the WMP engine is based only on their web page text. I don't know myself.

On the kmixer issue, the card in this particular machine is the Delta 66 not the 1010 -- that's in a different machine. Driver-wise I believe they are the same. I was told by a Midiman tech that in order to use SPDIF out on any of the Delta cards the clock has to be directly derived from the source, natively, and it cannot go through kmixer.

I never thought too much more about it because the Delta clock (and therefore SPDIF output) is always at the clock of my source data, which as far as I know could only be if kmixer was not involved. Comparing 44k/48k wave source, one input directly to my digital processor, and the other originating on the computer to the digital processor, the resulting audio is identical between inputs and when mixed 1 for 1.

Wave files played out SPDIF to DAT and then back to wave have the same sample count. I haven't pursued it beyond that, though.

--Bill

[KikeG]

I agree that flat response is not all, there are other things that can make into an equipment sounding better than other even when both are equally flat.

About the rest... Well, I'm an audio engineer too, studied engineering at university several years ago. I've worked as acoustic engineer (not now), and I've read from several actual engineers that work and know about audio. I'm afraid that they and me disagree from some of the things you say, sorry. On the other side, it's quite usual that audiophiles don't agree with audio engineers. I haven't heard many audio engineers talking about expensive cables and such...

[KikeG]

about kmixer: my Audiophile 2496, which is a Delta card and uses Delta drivers, is not inmune to kmixer bit-mangling even at the digital output. Ony ks and ASIO avoid kmixer. I've verified it too.

[occammd]

I'm an audiophile newbie, the only nice thing I own are Grado SR325 headphones and the Grado amp. I understand the reason for performing these types of calibrations, but can someone explain to me the real benefit of doing this type of calibration and installation as opposed to open-air headphones w/a subwoofer. My living room is large (20x45x30high) and I was thinking about setting up a sound system using headphones like the Grado's instead of a 5.1 speaker system for actual movie watching. Obviously it would be best to have an HT, but it just isn't practical.

I understand the previous arguments and actually just finished testing the group delay on a radar FIR mixer/filter, but I was curious about the application side of your calibration.

Thanks,
Ray

[Jones_Rush]

Damn it!, this guide is a monster!, I can't seem to finish it.
Don't worry though, I'm not going to sleep today until the guide is finshed and sent to all the people whose PMs I received. Even if it means staying awake until the morning. I really hope it won't come to that though...

[Jones_Rush]

Holy cow!, it's 6:15 AM here, and I'm finally done!.
The guide is now being sent to the people who PM'ed me.
It will be posted tomorrow, after Denis will take a look at what I did today.

[kyrill]

Thanks Jones, for the trouble and enthousiasm you took.
It awakens in me a long forgotten dream to EQ my room.
I'll await your guide:)

[mfombellida]

An interesting link:
http://www2.gol.com/users/pcazeles/

Cheers,
Michel

[Jones_Rush]

An interesting link:
http://www2.gol.com/users/pcazeles/

Yes, Michel, this is Patrick's outdated guide to DRC (it is already mentioned in this long thread, somewhere). Anyway, I hope you'll find my guide, a bit more, in-depth ;-).

The guide will be posted later today.

[Jones_Rush]

That's it, the guide was sent to everyone who PM'ed me from the start.
Wes, if you can host it now, it would be great.

Everyone, please don't forget to post your impressions after you get results, and remember, this guide will get you started, but in order to exploit DRC to its fullest (ie. fully tailor it for your room and likings), you'll have to take a look at the readme file which is attached to the DRC program.

[Mooneyass]

Done.

http://www.mooneyass.com/DRC/DRC.html

Wes

[Jones_Rush]

Thank you very much Wes!.

[gazzagazza]

Re room eq, this came through in a TI newsletter this morning...

Texas Instruments and Audyssey Laboratories Create "Sweet Spots" for All Listeners in the Home Theater
Delivering the highest quality sound for high volume home audio, home theater in a box (HTiB) products and professional sound systems, Texas Instruments Incorporated (TI) (NYSE: TXN) and Audyssey Laboratories today announced the patent-pending MultEQ(tm) and MultEQ LE(tm) technologies that are the first to automatically calibrate entire audio systems for multiple listeners in a room by eliminating frequency response distortions inherent in most acoustical environments. Developed by Audyssey exclusively for use with TI's digital signal processors (DSP), these technologies will be demonstrated during the CEDIA EXPO (September 4 - 6, 2003) in TI's demo suite at the Indianapolis Marriott Downtown.

[nubz69]

Can someone post some before and after waterfall and FR graphs on this? I need to reread this whole thread because I don't know how this process works. If someone could explain the mechanics in laymans terms that would be great.

[Myren]

on the texas insturments whoobajoob....

"for multiple listeners"....

*scratch head*

[Jones_Rush]

"for multiple listeners"....
Must be doing averaging. I can't see how that will sound anywhere as nice as one spot correction though.

[3-way]

I put together a filter with DRC using ETF and Jones_Rush's guide. I used ETF because I wanted to use the PSD/Sweep signal and the psychoacoustic add-on features of ETF. I can't say whether I have an optimized filter, though I suspect that I don't. There are a couple of variations to the Guide when using ETF. If people are interested, I can post the steps that worked for me.

I can't say for certain that I have an optimized filter because I can't think of a way to test it. The PSD/Sweep signal is created by measuring background noise first and then creating a signal that is supposed to take the background noise out of the measurement (important when you're living in an urban environment, like New York). So, the signal is unique. In order to test the filter accurately, I would need to play the PSD/Sweep signal through the filter and then apply the Psychoacoustic feature to the FR graph. As far as I know, ETF does not export the PSD/Sweep file and it doesn't have a convolver, so, I can't play the signal through the DRC filter to make an "after" graph. I think that if I can get the PSD/Sweep signal recorded onto a CD, I could play that through a CD player and use ETF to measure. I will write the author of ETF to ask where the PSD/Sweep file is saved.

Two other downsides to using ETF:
1. The impulse .pcm that is exported is 16 bits, instead of 32, although you can choose 44.1 or 48khz.
2. I think that the psychoacoustic curve in ETF produces a result which is more desriable than a flat response. I think that if the psychoacoustic response is applied to the FR graph before the .pcm file is exported, it changes the .pcm file so that flattening the resulting graph actually gives you the correct psychoacoustic response. The problem here is that, according to my reading of the DRC Readme file, DRC no longer aims for a flat response -- Denis wrote that he has tried to implement some psychoacoustic improvements in later versions. It would be nice if the psychoacoustics Denis added to DRC could be overcome by changing a value in the normal.drc file. Maybe it can be? I haven't looked into this carefully enough, yet.

The problems I see with using DRC in general are:
1. It only works for two channels at a time.
2. It only works for files played through a program with a convolver, like Foobar.

These aren't problems with the program itself - DRC seems to do a very good job at doing exactly what it's supposed to do, but, what about DVD playback, or TV? In order to use DRC, I'm cancelling the EQ on my Driverack 260. So, if I want to watch TV or a DVD, I need to change the settings on the Driverack and lose the benefits of DRC. As far as I know, there are no multi-channel convolvers.

I think (and I'm hoping someone will correct me here) that in order for multi-channel digital crossovers and digital room correction on the PC to become a reality, we need a program that gives us access to audio routing on the PC (something like WaveWarp or MSP/MAX, though neither provides a total solution at this point). Audio routing would allow us to capture a stream either from any program or from the soundcard inputs and process it (surround sound decoding, room correction and XO) before it leaves the PC.

I ordered a program called Soundeasy (www.interdomain.net.au/~bodzio/) that looks like it does digital crossovers, as well as DRC-like room correction. I've never seen it in use, but, from the user's manual online, it looks like it too, will only work for stereo. It may be that multiple PCs running Soundeasy could each accept two channels from a surround processor (pre-/ pro-) apply crossover and room correction and pass the signal to the amps. Aside from the cost of having three separate computers for 6 channels, you've got the headaches associated with proper time alignment and maintaining three computers that all must be working simultaneously.

[Jones_Rush]

1. The impulse .pcm that is exported is 16 bits, instead of 32, although you can choose 44.1 or 48khz.
If what I remember is correct, Denis told me that this is not good. The processing need to be done at 32bit domain, otherwise you'll have better chances of screwing the impulse response.

2. I think that the psychoacoustic curve in ETF produces a result which is more desriable than a flat response. I think that if the psychoacoustic response is applied to the FR graph before the .pcm file is exported, it changes the .pcm file so that flattening the resulting graph actually gives you the correct psychoacoustic response. The problem here is that, according to my reading of the DRC Readme file, DRC no longer aims for a flat response -- Denis wrote that he has tried to implement some psychoacoustic improvements in later versions. It would be nice if the psychoacoustics Denis added to DRC could be overcome by changing a value in the normal.drc file. Maybe it can be? I haven't looked into this carefully enough, yet.
The psychoacoustic table is a file named "bk-2.txt" which store the target curve (unless you mean something else he did). You can put any value you wish there. I think that Denis did a good job, attenuating the higher frequencies, otherwise it sounds too bright.

Here is bk-2.txt:
0 -20
10 -10
20 0
400 0
800 -1
1600 -2
3200 -3
6400 -4
12800 -5
20000 -6
21500 -10
22050 -20

Regarding the rest of what you said, well, while it is true that certain compromises needs to be done in order to work with DRC, I trully believe, that once implemented correctly, DRC more than compensate its user, for all the inconvenience. I also think that before you try and make things yourself (using ETF and such), you should first follow the guide precisely, hear the results, and only then, after you'll hear the result of the same process that I hear, go and test different ways of achieving the goal. It just that there are too many places you can go wrong, and if you did, it's a shame.

[dwk123]

Originally posted by 3-way

The problems I see with using DRC in general are:
1. It only works for two channels at a time.
2. It only works for files played through a program with a convolver, like Foobar.

These aren't problems with the program itself - DRC seems to do a very good job at doing exactly what it's supposed to do, but, what about DVD playback, or TV? In order to use DRC, I'm cancelling the EQ on my Driverack 260. So, if I want to watch TV or a DVD, I need to change the settings on the Driverack and lose the benefits of DRC. As far as I know, there are no multi-channel convolvers.

I think (and I'm hoping someone will correct me here) that in order for multi-channel digital crossovers and digital room correction on the PC to become a reality, we need a program that gives us access to audio routing on the PC (something like WaveWarp or MSP/MAX, though neither provides a total solution at this point). Audio routing would allow us to capture a stream either from any program or from the soundcard inputs and process it (surround sound decoding, room correction and XO) before it leaves the PC.


Well, the obvious answer is to put your convolution engine into a DirectShow filter so that it's installable in any filter graph. To be honest, I'm somewhat baffled as to why nobody has done this. Convolution isn't *that* hard, and multi-channel convolution isn't really any different than two channel. I guess it's possible that there are some limits/constraints with DS that I"m not aware of that would make this difficult, although I'd be a bit surprised.


I ordered a program called Soundeasy (www.interdomain.net.au/~bodzio/) that looks like it does digital crossovers, as well as DRC-like room correction. I've never seen it in use, but, from the user's manual online, it looks like it too, will only work for stereo.
[QUOTE]

Hmm, I though SoundEasy was basically a speaker design program. Their digital xover was mostly designed for previewing an analog xover topology before building. It does look like they've added general digital eq, although I doubt it'll be particularly convenient to use this as your 'media player'.

[QUOTE]
It may be that multiple PCs running Soundeasy could each accept two channels from a surround processor (pre-/ pro-) apply crossover and room correction and pass the signal to the amps. Aside from the cost of having three separate computers for 6 channels, you've got the headaches associated with proper time alignment and maintaining three computers that all must be working simultaneously.

Well, IMHO now you're being a bit sily. One computer running Linux will get you everything you want and more - *provided* you're willing to live with running analog inputs into the Linux box. As I've indicated repeatedly, I think Jack + BruteFIR is pretty much the ideal environment for doing this. Obviously most folks want a Windows based solution, but if nothing else it serves as an example of how things could operate. The big prolem with Linux is the lack of native multichannel decoders: AC3 only - no DTS, DPL2 etc, which makes a purely digital path a problem.

Assuming my system stabilizes soon, one area I'd like to look at is a digital link via Firewire from a Windows box over to the Linux box. With a suitable DS Audio Renderer, you could capture the decoded multichannel audio and transmit it digitally over to the Linux box. This would be a relatively long-ish term project, though, given the relative immaturity of firewire audio.

[3-way]

dwk123--

Using 2-3 computers to process is a bit silly and using Soundeasy to play back media would likely be inconvenient / undesirable. I do think that Soundeasy can be and has been used permanently, as both xo and eq for two channel. I know it's not an ideal solution, but, I thought it would be worth the price to take a closer look at it. I build speakers, too, so the additional speaker design tools might be interesting.

I hadn't thought about using a Windows box to control audio and then outputting to a Linux box for processing, though. What about passing the stream between the two boxes via AES3 or ADAT? Would it be possible to decode on the PC, send the signals out through a Lynx with the AES-16 card and then into the Linux box? As I understand it, there are no Linux drivers for the Lynx Two cards, unless this has changed? Is there a way to send the signals between the two boxes via ADAT or AES3, instead of Firewire?

[3-way]

Also, doesn't Xine do DTS decoding via Open Sound System?

[RayL Jr.]

Besides the DRC in this thread, I also really like the idea of digital crossover filters, phase control (as well as digital equalization and volume/attenuation). A lot of this stuff could also apply to people who don't want DRC, but the things I just mentioned.

It could also help make the HTPC an "all in one" solution for all kinds of speaker and amp setups. :D I'll try to start a separate thread on this...

[dwk123]

Originally posted by 3-way
dwk123--

I hadn't thought about using a Windows box to control audio and then outputting to a Linux box for processing, though. What about passing the stream between the two boxes via AES3 or ADAT? Would it be possible to decode on the PC, send the signals out through a Lynx with the AES-16 card and then into the Linux box? As I understand it, there are no Linux drivers for the Lynx Two cards, unless this has changed? Is there a way to send the signals between the two boxes via ADAT or AES3, instead of Firewire?

Hmmm, you know - I hadn't even thought of ADAT, but of the various ways to connect digitally, ADAT is probably the closest to viable *now*. AES3 or SPDIF would requrie 4 separate cables, and the only box I've ever heard of that *might* allow this under Linux is the Hoontech DSP24 with the DMIII box (you're right - no Lynx support under ALSA, and likely never will be). Firewire has promise, but would require some pretty significant development effort - Linux has basic support for the underlying protocol, but that's about it. I haven't looked into how to get any sound out of a Windows box by raw firewire either.

ADAT would probably work, but the potential problems I see of the top of my head are:
1) Getting your dvd software to use the ADAT outs - might not be a problem at all, but I simply don't know
2) 48kHz only. Not a *huge* problem since you could use spdif for 96kHz stereo.
3) cost. this might not be horrible if the Terratec EWS88D is supported under Linux, since they run $200/card. I *think* this uses the Envy24 chip, which would be great. Otherwise, you're probably looking at an RME card.
4) On the Linux side, if you send things back out ADAT to an external DAC, then we're probably done. If you wanted to use another card (like my Delta 1010) then you have clock sync issues unless you explicitly sync the clocks. I *think* running an spdif cable in addition to the ADAT link and slaving the output card off the spdif would allow this to work. You'd need some magic in your .asoundrc to trick Jack/Alsa into making both cards look like a single interface, but that already works 'in theory'.

Interesting idea, though. I may just have to look into this a bit. Two cards for $400 might be an acceptable price if I 'knew' it would work with my existing Delta 1010.

Of course, you could just get a Revo and run the analog outs into a Delta 1010 or something which would probably be *almost* undetectable on the vast majority of DVD/multichannel stuff, but an all-digital link is way sexier.

[dwk123]

Originally posted by 3-way
Also, doesn't Xine do DTS decoding via Open Sound System?

SPDIF pass-through only, as far as I know. I know that an awful lot of Windows codecs have been 'wrapped' with Wine code to use them directly from Xine/Mplayer etc. However, unless I've missed something none of the surround formats have been done. It's on my list to look into this, but as always real-life gets in the way. If it's simply that nobody has gotten around to it, then it might be 'relatively straightforward' to implement. There could easily be reasons why complex multichannel DirectShow filters haven't been wrapped, though.

[3-way]

What kind of control does one have over the crossover algorithms in BruteFIR? Is it possible to design a filter in something like MATLAB and export it into BruteFIR? If not, what types and slopes are available in BruteFIR? Would a 2.8Ghz P4 with 1 gig of fast RAM be capable of processing:
1. 4-way front speakers
2. 2-way surrounds
3. 3-way rears
and DRC on each channel?

[dwk123]

Originally posted by 3-way
What kind of control does one have over the crossover algorithms in BruteFIR? Is it possible to design a filter in something like MATLAB and export it into BruteFIR? If not, what types and slopes are available in BruteFIR?

BruteFIR is nothing more than a convolution engine - it uses FIR filters that you design to do the filtering - it doesn't really have anything 'built in'. This is 'good' in that it'll do any filter topology you can imagine. It's 'bad' in that you have to completely define the filters - it's not just a matter of plugging in "4th order LR @2500Hz'.
Matlab can certainly be used to design the filters - I originally used a MatLab clone called SciLab for some of this, but the measurement import was a bit of a pain.

There are other options for implementing IIR filters if that's whats desired that will fit in easily with the Jack/BrutrFIR approach, and will use significantly less cpu.


Would a 2.8Ghz P4 with 1 gig of fast RAM be capable of processing:
1. 4-way front speakers
2. 2-way surrounds
3. 3-way rears
and DRC on each channel?

I can't say for certain on this one. I _can_ say that you'll be spending a small fortune on hardware to do this if you care about quality at all- that's 18-22 channels (not sure whether you included a 4-way center in there or not). Just the amps will cost a fortune, to say nothing of 18+ channels of decent D/A.
I do know that Anders (who wrote BruteFIR) is using it for Ambiophonics, and is running huge setups - at least that big with very long filters. In the readme he claims 26 channels of 128k tap filters on a 1GHz Athlon. However, he's running with very large buffer sizes which result in 200-200ms delay which obviously won't work in a video system. The CPU requirements rise rapidly when you try to reduce the latency down to the sub-30ms range.

My honest feeling though is that you're getting a bit carried away to consider thinking about that level of a system up front. This is not a 'point and shoot' type of project. If you undertake this in phases - eg start with the front L/R xovers, then bring in the multichannel time alignment, then bring in the room correction etc, then I strongly suspect that you'll be a couple generations of CPU away from 'now' by the time you're close to 'completing' this project.

[3-way]

I'm glad I can input my own filters. It may take some effort, but I'll take power over convenience any day.

The front speakers have been designed over the course of the past 12 months and the drivers are now on order. Drivers from the Phoenix I built 18 months ago will be paired with Mangers and made into rear channels. I have a pair of Mackie HR824's as surrounds right now (internal active XO and amplifiers), but would like to upgrade to a pair of Alix-Anne's from American Power & Light. The whole system has gone through several iterations already, and will no doubt continue to evolve. But, it will likely come together within the course of the next year. It will be a 6.2 system (no center channel).

I was hoping to be able to use the DACs on the Lynx Two B soundcard, rather than external DACs, but, if I need to use Linux, I may have to rethink this... Thanks for the information. I'll have to re-install Linux tonight and take a closer look at BruteFIR.

[Steve Dodds]

Very interesting thread. I've been using digital room correction of various sorts for the past few years with excellent results via TACT and other units.

One thing I noticed that hasn't been brought up is the response of the Behringer ECM mic. While very flat below 10K or so, it does get a bit squirrely above that and has a droop.

If uncorrected, this will mean the software tries to over-correct, making the top a little brighter than it should be.

So it ought to be corrected before doing any measurements. Here's a look at the corrections that are needed:


Cheers

Steve

http://www.doddsy.net/ecm.jpg

[KikeG]

Good to point out. However, I recall some people at the rec.audio.pro newsgroup measuring their ECM8000 units and finding them to be flat within a couple of dBs from 20 Hz to 20 KHz. Maybe it's just unit-to-unit tolerances, but, what did you calibrate your ECM8000 against?

[jimwhite]

Microphone calibration...... now there's a subject I'd love to see some how-to's on..... I remember reading in an AES paper about using the "reciprocity principle" as postulated by Berenack (sp?) in "Acoustics". Basically you used 3 uncalibrated sound transducers (speakers, microphones), one or two of which must be bi-directional, in a round-robbin fashion, and end up with a calibrated microphone! :)

:cool:

[Steve Dodds]

The calibration file is the one from TrueRTA, a very nice RTA program (www.trueaudio.com). I've checked it against ETF and my dbx RTA and it seems about right.

Still, it could be worse. Here's the Radio Shack meter one:

http://www.doddsy.net/rs.jpg


Cheers

Steve

[Jones_Rush]

Steve Dodds,
Yes, I'm aware of the ECM8000 inconsistencies above 10-12khz. Still, for $40, it's pretty good. And as I've been told, inconsistencies above those frequencies are less audible than below (assuming we're not talking about huge differences, and by your graph, it doesn't look like we do).

Anyhow, once I'll finish setting up my listening room completely, I'll rent a better mike, for a day, in order to create some really accurate filters.

[Mooneyass]

Steve

I havent been able to find the cal file at truerta, any chance you can post a direct link or post it here?

Thanks

Wes

[Steve Dodds]

You're correct that higher frequency shifts, especially this high, are often less audible. However there are those who find a truly flat response too bright, and the extra boost from the uncalibrated response will make it just a bit brighter.

Basically the response you posted earlier (about page 4 ;)) will actually be closer to the response in the ECM chart above than completely flat.

Most people prefer a gentle slope downwards from low to high. It may be possible to do this with some kind of shelving filter.

I've attached the calbration file from TrueRTA in .txt format, although I have no idea how to make it work with any the programs you guys are using.

Cheers

Steve

[drewmc]

I have run into a snag with Cool Edit Pro. I installed the program fine and it said I had 21 days left. I transfered the Aurora files to the main directory as in the software notes. After I launched the program I had Aurora, but my trial had expired. Any way of resetting my trial short of nuke and paving my system?

Another issue entirely, I plan to use the system ultimately for EQ'ing a car environment. It is impossible for me to have the speaker lengths exactly the same. It seem I would need to put a delay in prior to recording the sine sweep to equalize the time arrival or offset the waveform in CEP, and I would also need an equal delay on playback. Any Ideas on what would be the best way to go about this?

[Jones_Rush]

I have run into a snag with Cool Edit Pro. I installed the program fine and it said I had 21 days left. I transfered the Aurora files to the main directory as in the software notes. After I launched the program I had Aurora, but my trial had expired. Any way of resetting my trial short of nuke and paving my system?
Have you worked with Cool Edit before on your current WinXP installation ?.
Anyway, PM me, so I'll give you a list of possible solutions.

Another issue entirely, I plan to use the system ultimately for EQ'ing a car environment. It is impossible for me to have the speaker lengths exactly the same. It seem I would need to put a delay in prior to recording the sine sweep to equalize the time arrival or offset the waveform in CEP, and I would also need an equal delay on playback. Any Ideas on what would be the best way to go about this?
I'll check it out.

[Jones_Rush]

You're correct that higher frequency shifts, especially this high, are often less audible. However there are those who find a truly flat response too bright, and the extra boost from the uncalibrated response will make it just a bit brighter.

Basically the response you posted earlier (about page 4 ) will actually be closer to the response in the ECM chart above than completely flat.

Most people prefer a gentle slope downwards from low to high. It may be possible to do this with some kind of shelving filter.
By default, the target curve that DRC uses is this:

0 -20
10 -10
20 0.0
200 0.0
400 -0.5
800 -1.0
1600 -1.5
3200 -2.0
6400 -2.5
12800 -3.0
20000 -3.5
21500 -10
22050 -20

[Steve Dodds]

That makes sense, but don't tell the hi-rez people.

:)

Steve

[Jones_Rush]

Another issue entirely, I plan to use the system ultimately for EQ'ing a car environment. It is impossible for me to have the speaker lengths exactly the same. It seem I would need to put a delay in prior to recording the sine sweep to equalize the time arrival or offset the waveform in CEP, and I would also need an equal delay on playback. Any Ideas on what would be the best way to go about this?

I've asked Denis, and introducing delays in the filters could be done using CoolEdit, adding samples set to 0 at the beginning of the filter that needs to be delayed, and adding them also at the end to the filter that doesn't need to be delayed to keep the filters of the same lenght. The delays introduced is equal to N/F wher N is the number of samples added and F is the sampling frequency. Here the minimal delay is one sample, but probably nobody need less than this. In theory you could introduce even subsample delays, but this is much more difficult to do.

[3-way]

In response to Jones-Rush's post above, I asked Denis how using 16 bit instead of 32 bit impulse response files would affect DRC's results:

Me:
"I really want to use .pcm
>files from ETF, but the .pcm files ETF exports are 16-bit
>integers instead of 32 bit floats. How critical is this?"

Denis:
in my experience, not so critical, at least for DRC. To get the most out of
DRC you need an impulse response with something like 80 dB of S/N,
something 16 bits are able to deal with.


Denis (on ETF's PyschoAcoustic feature):
The PsychoAcustic feature of ETF is not suited for DRC use. DRC has it's
own simplified psychoacutic model. It is probably much simpler than the ETF
one, but it also accounts for some physical limitations which are inherent
to room compensation and which are often more critical than the
psychoacustic model itself, especially at high frequencies.

[RayL Jr.]

I found this at DIY speaker building information - http://home.earthlink.net/~etunstal/diy.htm - "Panasonic makes electret microphone capsules - http://www.panasonic.com/industrial/components/rec_mic.htm - used in the above projects and other mic designs found on the net. Calibrated mics increase the accuracy of your measurements. Calibrated Panasonic mic capsules are available from Kim Girardin. He can be reached at:

Kim Girardin
Wadenhome Sound
1400 Homer Rd. Suite 2
Winona, MN, 55987, USA
507-454-8844
kmgrdn@luminet.net

Right now I have my hands full designing a sealed 18" subwoofer with sealed 3-way or 4-way main speakers (large woofer in it's own box under a good 3-way speaker w/8-10" woofer). Mains down to 20-25Hz, set the xover to 40-50Hz, total bass down to 16Hz (corner placed sub)... :)

This is a similar design - 500 lbs. each, 16Hz - 35kHz -3db (+/- 1dB in midrange), $18,950/pair! http://www.vonschweikert.com/vrline/VR-6W.HTM
http://www.lalena.com/audio/faq/speaker/vr8.gif


PS - A neat trick you can do with this design is implement "full range" center channel and surrounds for hi-rez 5.1/6.0 discs or 5.1 DD, DTS...

[jimwhite]

I bought this Microphone (http://www.linearx.com/products/microphones/m31/M31_1.htm) for doing some RTA, equlization and maybe even some DRC.... it has a single ended output (but on an XLR connector) and my Delta 1010-LT has balanced XLR mic inputs... anybody got a pointer to an inexpensive but good transformer to use?

:cool:

[Jones_Rush]

I wonder,
Ever since the guide came out, I haven't seen any reports of anyone actually using DRC. Now, I've personally emailed the guide to at least 20 people, so I know some people have it.

Is there any reason why you still haven't tried it, or why there are no reports about your DRC experience ?. Is there anything which was unclear in the guide, which makes it impossible to use DRC ?. Or, maybe you tried it, and the results weren't satisfying ?. I would really like to know, so I could improve the guide, or help you get better results.

[KikeG]

You've done a very good job with the guide. I had similar DSP correction on my schedule from before you wrote this guide, but I will try it when I have more time and feel like doing it. I have already a ECM8000 mic, but I still need to build or buy (I think the later) a mic preamp. I want to experiment also with in-ear, self-made, cheap WM-60AY Panasonic mics.

BTW, didn't you still try the kmixer blind test? :D

[Jones_Rush]

I have already a ECM8000 mic, but I still need to build or buy (I think the later) a mic preamp.

Good point. The process does require you to have some hardware tools which are not handy to anyone. Still, they are not THAT rare these days...

BTW, didn't you still try the kmixer blind test?
No I haven't, but I will today...

[K-Wood]

The guide is excellent. I've been in the processing of rebuilding my HTPC, so that has taken precedence. But attempting DRC is next on my list.
Awesome effort with the guide, btw. Many thanks.

[boubou]

Hi Jones Rush, Hi all,
This thread and your guide have drawn our attention in a french forum called homecinema-fr.com
When something is smart AND nearly free, people do listen from far away!

Jones, I 'm wondering if the procedure described in your guide is only aiming at correcting the FREQUENCY response of a system (speakers + room), or if it can correct as well its PHASE response, as a Tact RCS or a Sigtech would do.
According to me, if I take only the frequency response of a system in order to construct a correcting FIR filter, and disregard the phase, I would miss half of the problem! Am I wrong here?
It is not clear in your step-by-step guide whether or not your take care of the PHASE response of the system, for instance by comparing the capture of the sine sweep with the original sweep.
My doubt stems from the fact that you could not use the "genuine" impulse response or MLS signals in the first place (for SNR reasons if I read well).
Thanks for any clarification!
Regards from France,
Boubou.

[Jones_Rush]

Jones, I 'm wondering if the procedure described in your guide is only aiming at correcting the FREQUENCY response of a system (speakers + room), or if it can correct as well its PHASE response, as a Tact RCS or a Sigtech would do.

Yes, it tries to correct for phase, too.

It is not clear in your step-by-step guide whether or not your take care of the PHASE response of the system, for instance by comparing the capture of the sine sweep with the original sweep. My doubt stems from the fact that you could not use the "genuine" impulse response or MLS signals in the first place (for SNR reasons if I read well).

In the guide, at section #3, it is written:
* Be aware that an inverse of the log sine sweep, has been automatically created and copied to your windows clipboard, right after the log sweep was generated. We’ll use it later

Then, in sections 14-16, we Convolve the recorded sine sweep, with the Clipboard. This is how the original log sine sweep is being accounted for. Denis offered to send me some technical papers which explains how this "magic" of convolving the inverse of the log sine sweep with the recorded log sine sweep, creates the impulse response that DRC later work with. To me, the point that it simply "works", was enough :-), at least at the point when I wrote the guide, when I didn't have too much spare time...

I didn't mention this in the guide, since I didn't want it to seem more complicated than it must be.

Sorry for the confusion!.

[KikeG]

The sine sweep technique should obtain same results as a proper MLS measurement, which is the impulse response of the system. This impulse response will contain both the frequency and phase response of the transfer function of the system. This technique is used instead MLS just because if offers a better SNR.

[Jones_Rush]

Just in order to show that DRC indeed is useful in the time domain, here are some images from measures taken at my listening spot.

Both images are in sync, so values from the Y axis can be compared.

Go here (http://www.hometheaterforum.com/htforum/showthread.php?s=&threadid=160078) for the image comparison.

[3-way]

I've been reading about a program called Max/MSP this morning. I haven't looked at it long enough to understand exactly what it does, but, it runs on Windows, and provides filtering and crossover features for what looks to be N channels. There's a convolution engine plug-in, and, it takes VST plug-ins. The convolution plugin looks like it's part of a package called Pluggo, which, for now, is only available on the Mac.... On the upside, Max/MSP is a ReWire application, so, it could be connected to any front-end that supports ReWire, or, maybe it could be connected to others via Virtual Audio Cable.

More info on Max/MSP at www.cycling74.com

I'm hoping this will provide a way to implement DRC on multiple channels and user-definable crossovers on the Windows platform. I'll look at it more carefully this weekend and report back.

[Jones_Rush]

I'm hoping this will provide a way to implement DRC on multiple channels
3-way,
But the multichannel convolution engine will have to come in the form of a direct show filter, in order to work with DVDs. Unless there is a DVD player which can use this convolution engine as an integral DSP.

[3-way]

Jones_Rush--

I don't understand why the convolution engine needs to be a DirectShow filter.

Virtual Audio Cable will let you take audio signals being sent to the soundcard's outputs and reroute them to another program for further processing. I think ReWire works similarly. If so, you should be able to take whatever audio is being output by whatever program you're using (or even take input to the soundcard from an external DVD or CD player) and run it through the Max/MSP convolution plugin and xo -- it would be like taking the output from one program and then running it into the input of another program before sending it to the soundcard's outputs. The only issue I see is possibly latency. I really have no idea how much latency this would introduce.

If I'm missing something, please correct me.

[Jones_Rush]

Virtual Audio Cable will let you take audio signals being sent to the soundcard's outputs and reroute them to another program for further processing. I think ReWire works similarly. If so, you should be able to take whatever audio is being output by whatever program you're using (or even take input to the soundcard from an external DVD or CD player) and run it through the Max/MSP convolution plugin and xo -- it would be like taking the output from one program and then running it into the input of another program before sending it to the soundcard's outputs.

Oh, I didn't know it works like that. In this case, it should be fine.

The only issue I see is possibly latency. I really have no idea how much latency this would introduce.
I'm sure we'll find a way to delay the video stream in comparison to the audio stream, in one way or the other. But, the latency added by the convolution engine needs to be constant for this to work well, otherwise we'll lose sync pretty easily.

[Jones_Rush]

3-Way,
We need to ask them if Max/MSP can take the multichannel convolution plug-in from "pluggo", and use it as its own. I'll send them an email.

[dsinder]

Originally posted by Jones_Rush
I wonder,
Ever since the guide came out, I haven't seen any reports of anyone actually using DRC. Now, I've personally emailed the guide to at least 20 people, so I know some people have it.

Is there any reason why you still haven't tried it, or why there are no reports about your DRC experience ?. Is there anything which was unclear in the guide, which makes it impossible to use DRC ?. Or, maybe you tried it, and the results weren't satisfying ?. I would really like to know, so I could improve the guide, or help you get better results.

I ordered the Mic and Mixer as soon as your guide came out (or maybe just before). I had the order in a few days. It arrived a week ago.

Some background... I had recently installed some DYI acoustic panels made with John Manville type 475 duct board covered with GOM. I had done this primarily to improve my 7.1 channel sound. What I finally did was to cover the entire wall behind my 65" RPTV with the DYI acoustic treatment. I also installed a 6 foot long by 4 foot high strip on both side walls so as to catch L-C-R speaker reflections to the seating area. There was a very marked improvement to the 7.1 sound. I then listened to some 2-channel music and found much improved stereo imaging there too.
The cost for duct board and GOM was about $400.

Next came DCR. I used the guide to make normal filters. The subjective frequency response was MUCH better. The bass boominess was gone. The bass was remarkably tight, distinct, and imaging as pin point. The sound stage spread beyond the sides of the room with pretty good imaging in the middle. The problem with imaging in the middle was that most sound in the area of the RPTV seemed to come from above the RPTV. I guessed that if I covered the face of the RPTV a non-reflective material that might solve that problem. So I had just enough duct board left over to try the experiment. But I was out of GOM. So I covered the duct board on both sides with 1/4" polyester batting and rigged a simple way to hang it over the face of the TV. That indeed solved the "high middle". But things still did not sound quite right. So I remade a filter with a measurement made with the RPTV face covered. That did it. I've now got the best sound reproduced sound I've heard anywhere right in my house. It's actually a little scary at times how good it is. It does have it's draw backs. Primarily, the reproduction is so good that any defect in the recording is revealed. A violinist taps a music stand with a bow and I hear it - right over "there". That can actually detract a bit.

For 5 days straight I never got to bed before 3 AM. (That's late for a 51 year old geezer like me.) I just could not stop listening. I heard new and wondrous things in every recording I listened too. I also discovered that I needed to re-rip my whole CD collection. I thought I had been smart by using the LAME -insane option to compress at 320Kb/sec. But that's not good enough now. So I ripped the whole collection to uncompressed .wav.

Another problem - A couple of days ago I finally got around to watching a DVD. Now what I thought was good 7.1 sound is boomy and muddy. I've gotten used to what proper reproduction should sound like and my 7.1 is still a long ways off!

I want multi-channel DRC!!

Thanks so much Jones for putting this guide together. It's made a world of difference to my music listening.

PS. I also did DRC on a pair of Maggie MG-IIIa speakers in another room. That room has bad acoustics. But it was still better for two channel than my HT room before I added the DYI acoustic treatment to the HT room. The DRC did not bring the Maggies in the untreated room up to nearly the quality of the Diva 6.1s in the treated room with DRC. I guess I can retire the Maggies now and recover a bunch of space.

Dale

[Jones_Rush]

Some background... I had recently installed some DYI acoustic panels made with John Manville type 475 duct board covered with GOM. I had done this primarily to improve my 7.1 channel sound. What I finally did was to cover the entire wall behind my 65" RPTV with the DYI acoustic treatment. I also installed a 6 foot long by 4 foot high strip on both side walls so as to catch L-C-R speaker reflections to the seating area. There was a very marked improvement to the 7.1 sound. I then listened to some 2-channel music and found much improved stereo imaging there too. The cost for duct board and GOM was about $400.

Do you know what is the absorption coefficient for the duct board with the GOM ?.


For 5 days straight I never got to bed before 3 AM. (That's late for a 51 year old geezer like me.) I just could not stop listening. I heard new and wondrous things in every recording I listened too.

Yep, happened to me too.

Another problem - A couple of days ago I finally got around to watching a DVD. Now what I thought was good 7.1 sound is boomy and muddy. I've gotten used to what proper reproduction should sound like and my 7.1 is still a long ways off!

I want multi-channel DRC!!

As you see, we are currently working on finding a way to make this happen too.


Thanks so much Jones for putting this guide together. It's made a world of difference to my music listening.

Thanks!. I think Denis Sbragion should get a bigger thanks though :-), he's the real mastermind behind all of this.


PS. I also did DRC on a pair of Maggie MG-IIIa speakers in another room. That room has bad acoustics. But it was still better for two channel than my HT room before I added the DYI acoustic treatment to the HT room. The DRC did not bring the Maggies in the untreated room up to nearly the quality of the Diva 6.1s in the treated room with DRC. I guess I can retire the Maggies now and recover a bunch of space.

Yes, this is an important issue. While DRC works very well up to 1-2khz, above this, it can't do much more than correct the direct sound, or in other words, if you will, DRC can fix time domain problems, more effectively below 1-2khz. IMO, in order to get the full high quality experience, one should first treat his room for frequencies above 1-2khz. This shouldn't be much of a problems though, since absorbant material which does not need to treat low frequencies (below 500hz), is dirt cheap.

[Jones_Rush]

In the meantime, I got a response from www.cycling74.com.

I've asked them the following question:
Hi,
I would like to know if you have a product which can give me the ability to convolve multichannel sources, in real time, for Windows XP environment.

Your "pluggo" product seems like it can do this, but only for MAC users.

Thank you.

This is their answer:

Hello Jones,

Max/MSP can do this for you. In fact, Max/MSP is the environment in which all the Pluggo plugins are created. So perhaps it is not quite as friendly out of the box, but offers you a great deal more flexibility. I would suggest that you grab the demo, and have a look at the tutorials, and when you're comfortable have a look at the "FFT fun" folder inside the included examples. You'll find some core convolution and phase vocoding things in there which should give you an idea of how we set it up. BTW, we are working on Pluggo for Windows but at this stage have no definite release date.

All the Best


Andrew Pask
--


Cycling '74

Business Office
379A Clementina Street
San Francisco, CA 94103 USA
p: 415-974-1818
f: 415-974-1812

Technical Support:

Windows
p: +1 415 869-3717
e: winsupport@cycling74.com

Macintosh
p: +1 415 974-1811
e: macsupport@cycling74.com

http://www.cycling74.com : software

http://www.cycling74.com/c74 : CD releases

[dsinder]

Here are a couple of links. The GOM should not change it much.

http://www.jm.com/Insulation/Products/DataSheets/PerformanceMaterials/AirHandling/ahs200_superduct.pdf

http://www.jm.com/Insulation/Products/DataSheets/PerformanceMaterials/AirHandling/ahs334_matfaced_microaire.pdf

Compare to:

http://www.jm.com/Insulation/Products/DataSheets/BI_US/RigidSemiRigidFGBoards/hig1214dk_insul-shield.pdf

often used in HT installations.

If you are going to use much, the insul-shield is the way to go. Easier to handle and less expensive. But comes in 4 ft wide by 100 ft roll at about $1/sq. ft. The Duct board is about $40 for a 4x10 sheet. I used 4 boards total. In my case I wanted to experiment with different configurations and do so without messing up my walls too much.

[Jones_Rush]

Thanks for the info, Dale.

[boubou]

Originally posted by Jones_Rush
Yes, it tries to correct for phase, too.



In the guide, at section #3, it is written:


Then, in sections 14-16, we Convolve the recorded sine sweep, with the Clipboard. This is how the original log sine sweep is being accounted for. Denis offered to send me some technical papers which explains how this "magic" of convolving the inverse of the log sine sweep with the recorded log sine sweep, creates the impulse response that DRC later work with. To me, the point that it simply "works", was enough :-), at least at the point when I wrote the guide, when I didn't have too much spare time...

I didn't mention this in the guide, since I didn't want it to seem more complicated than it must be.

Sorry for the confusion!.


Hi again Jones,
I will think about this magic ;)
Thanks a lot for the clarification!!!
Boubou.

[dwk123]

Originally posted by dsinder

PS. I also did DRC on a pair of Maggie MG-IIIa speakers in another room. That room has bad acoustics. But it was still better for two channel than my HT room before I added the DYI acoustic treatment to the HT room. The DRC did not bring the Maggies in the untreated room up to nearly the quality of the Diva 6.1s in the treated room with DRC. I guess I can retire the Maggies now and recover a bunch of space.

Dale

I used DRC on my homebrew Carver Ribbon based dipoles, and also found that the results were questionable. The bass correction was tremendous, but the midrange had problems. I'm fairly certain that the assumptions used within the DRC correction/inversion are thrown off by the dipole ratiation pattern and secondary reflected dipole impulse.

I'm a week or two away from beginning another test cycle with my revamped speakers, and will be looking at this more closely.

[Jones_Rush]

Some more updates:

My question:
Hello,

I have a certain need, and I wonder if you have a product which can fulfil it.

I have a software which have created 8 inverse impulse response filters (using Fir filters), one for each speaker in my HT setup, based on the acoustics in my room (in order to compensate for its anomalies). I need a real-time convolution engine, which can convolve 8 channel inputs (of a 7.1 DVD movie sound, for example), and output it through the 8 channel analog outputs of the sound card. I don't care about latency/delay, because I can delay the video stream too, in order for it to be in sync, with the audio stream.

Do you have a product which can do this ?.

Their response:

Yes, but it isn't released. It is a VST plugin that can do e.g. a 2x5 upmix (10 convolutions) of 3 second impulse responses with 256 pnt latency and 20% CPU load on a 1.8 MHz laptop. It is in many ways a realtime MultiVolver that can do up to 8x8 and your case is 1x8 so no problem. However, as indicated, it isn't released and there are some business decisions to make first. The plugin market is a mess with extremely low prices and still a *lot* of piracy so it may never be
a plugin at all in the end.

Ther are two share or freewere alternatives one is SIR and one is BruteFIR, the first will bog down the machhine and will have to be run in several instance too, the second is made excatly for your purpose but it runs under Linux and if that is not a problem it would fit well. Both of these can be found by Google.

Best,


Bengt-Inge Dalenback
.........................................................
c a t t * Mariagatan 16A * SE-41471 Gothenburg * SWEDEN
http://www.catt.se bid@catt.se phn/fax: +46 31145154

My question:
Hi,
Do you have any plans to port BruteFIR to Windows ?, or, will BruteFIR remain a Linux software, forever ?.

Do you know of any real-time, multichannel convolution engine for Windows platform ?.

Thank you very much.

His response:
I have no current plans to port it. I don't use Windows at work or at home, so I don't even have the platform to develop on. But one should never say never...

> Do you know of any real-time, multichannel convolution engine for Windows platform ?.

No, but there could be, I'm not very familiar with windows software.

/Anders Torger


To be continued...


P.S
I once promised some before/after waterfall plots of the bass region. So here they are: before (http://www.avsforum.com/avs-vb/attachment.php?s=&attachmentid=12731)/after (http://www.avsforum.com/avs-vb/attachment.php?s=&attachmentid=12732).

[Arnie]

It's just occurred to me, would it be possible to use DRC to implement some form of headphone virtualisation?
Sometimes I have to watch TV with headphones on, and even though they are Sennheiser HD-600s, the sound can be a bit 'dry'. Ideally, it would be nice to have a soundfield just like I was listening to speakers in my room (corrected, of course!).
And hopefully the 3D would seem more realistic than the multitude of analogue and virtual headphone DSPs that I've tried.

Cheers
Arnie

[Jones_Rush]

Arnie,
Since headphones do not need time domain corrections, the only thing that DRC could theoretically do, is correct the direct sound, ie. make the response flat to 20khz, if they are not already. I'm not sure this will give you the "wet" sound you're looking for. It just that listening through headphones is a totally different thing than through speakers. With speakers, the waves reach your ears from different angles, while partly merging into each other, plus, the room also gives you an echo from behind, which the headphones do not. In order to get the same experience as with speakers, you'll have to use some kind of an algorithm, like "Aureal" once implemented with their MX sound cards, which will enable you to fool your brain into thinking it is actually listening into speakers which are located far in front of you, while hearing echoes from behind. I don't think such algorithm exist these days in a consumer product software.

[Arnie]

Thanks for the reply. I understand the differences between headphone listening and speakers, but it seems to me that it should be possible to apply the room correction in reverse (assuming the headphones already produce a flat, time-aligned response) to mimic the sound of an uncorrected room. Ideally, just the time component leaving the frequency response reasonably untouched.

Perhaps it would be necessary to use a dummy head/binaural microphone when doing the calibration...

[Jones_Rush]

Still, I don't think it will work. At best, it will ruin your sound, just by a little. You want to add the decay you get in a room, to the headphones. I am not sure you'll like the results. Just for experimentation, maybe you should try to add several milliseconds of reverb to your sound (using Cool Edit, or some other sound editor), and see if you'll like it... (the reverb will have to be added with some fade off slope, to mimic the decay in a room). Again, I don't think it will work, since the reverb will hit your ears from the wrong "angle" (ie. the wrong time cues).

[Jones_Rush]

My question:
Hi Andrew,
I've taken a look at Max/MSP convolution demos. Let me ask you a more specific question: Currently I'm using a stereo convolution plug-in, which takes a .wav song, and convolve it with an inverse impulse response of the acoustics in my room (I have a special software which create the right filters for my room, according to MLS sweeps which are recorded from the listening spots and played
from the speakers). This way, I get room correction for most of the frequency and time anomalies of the sound.

What I need, is to do exactly the same, just with 5.1 multichannel sound. I want to take the decoded 5.1 sound of DVD movies, run it through Max/MSP, so each channel out of the 6 (5 + subwoofer) will get convolved with its own filter, in order to make it fit with the acoustics of my room. So, I need to run 6 independent channels, to be convolved with 6 independent impulse response filters, and then output the result through the 5.1 analog outputs of the sound card, in real time.

Do you understand what I'm trying to accoimplish ?, if so, is it possible with Max/MSP ?.

Thank you.

His answer:

Hi Jones,

Sure you can do this, this is the sort of thing our users love to get involved in. I am quite confident that Max/MSP has the capability to deliver an app for you which can do what you want.

Here are a couple of thoughts about what you are doing.

Firstly, you would need to be sure that the filters you create in Max/MSP exactly represent the filters you have determined match the acoustics of your room . I guess this would involve a fairly technical comparison and a matching of various parameters etc. This would be the fun part.

Secondly, and this is more of a practical concern than a reflection of the capabilities of Max/MSP, there will be a significant amount of latency introduced as a result of all this. Even if we could assume that the signal processing in Max takes place in no time (a reasonably dangerous assumption), you would still be looking at a round trip through your AD convertors, and I wonder if the resulting
latency might cause another, altogether unsatisfactory side effect.

It sounds like it will be an interesting project though.

Good luck!

While I have no clue what was he trying to say in his first point (the filters are a given to Max/MSP, so what's his point ?). His second point shouldn't be a problem, since we can just delay the video stream as well.

There are two other problems though, more serious ones:

1) Max/MSP is WAY complicated, and it will take me at least a week of hard work, to try and figure out how can I do this.

2) Max/MSP is not a shareware, and the Demo version expires after 30 days. The price is set to $495 USD. I don't think that at that price, people will bother.

Overall, it's nice to know that we can, at least theoretically, already achieve the goal of DRC for HT with Max/MSP, but practically, we should look for a different product.

[Arnie]

Originally posted by Jones_Rush
Still, I don't think it will work. At best, it will ruin your sound, just by a little. You want to add the decay you get in a room, to the headphones. I am not sure you'll like the results. Just for experimentation, maybe you should try to add several milliseconds of reverb to your sound (using Cool Edit, or some other sound editor), and see if you'll like it... (the reverb will have to be added with some fade off slope, to mimic the decay in a room). Again, I don't think it will work, since the reverb will hit your ears from the wrong angle.

I've tried playing with CoolEdit, and building various (analogue) crossfeed circuits. The trouble with mimicing a room response is that there would be multiple reflections, each with a different frequency response from all the objects in the room i.e. a lot of work to do manually. And at best it's going to be unlistenable.
Nothing gives that "out of head" experience. This is why I suggested using a dummy head mic, as used in binaural recordings, to make the calibration data - it would take into account the shape of one's ears and the way that the frequency response changes with incident angle, just as it does in a binaural recording.
No matter, the only way to find out is to try it, which is something I will have to tackle when I've some spare time.

Cheers.

A.

[Jones_Rush]

No matter, the only way to find out is to try it, which is something I will have to tackle when I've some spare time.
Even though I don't understand how what you say can work, please keep me posted, if you're successful :-)

[3-way]

I didn't understand his answer, either, but I"m not yet convinced that Max/MSP isn't the right product. If you look in C:\Program Files\Cycling '74\MaxMSP 4.3\examples\fft-fun, there's a PAT file called "Convolution Workshop". It looks like it's a stereo convolution engine. I think it could serve as the basis for an N-channel convolver.

Also, as I understand it, you can create a file in Max/MSP called a Collective that can be run on another computer with only the Max/MSP Runtime Environment (which is free) installed. From the Tutorials:

"You can also give your collective to someone else to use, without worrying whether you’ve included all the necessary files. If the person you give your collective to doesn’t own Max, you can give (but not sell!) them the MaxMSP Runtime application along with your collective. This will allow them to run (but not edit) your program."

So, if one person builds it, he/she can distribute it to everyone else. And, if it can be built w/in the 30-day trial period, no one will pay a penny.

[Arnie]

Originally posted by Jones_Rush
Even though I don't understand how what you say can work, please keep me posted, if you're successful :-)

OK, I'll try to explain as best I can (it's getting late, oh so late).

1. Dummy head microphones are basically two microphones embedded in a human head-shaped object, complete with ears. Thus they can pick up a soundfield as it is picked up by human ears, complete with outer ear and head reflections. They are used to make a class of audio recording that displays very acute 3D properties. Check out some of the recordings at http://www.binaural.com/

2. Thus, if a dummy head is used to record the output of a set of speakers, in a room, it will record the sound as it would be picked up by a human sitting at that position, complete with all room and ear-based reflections.

3. Assuming that a pair of headphones is of a sufficiently neutral nature, any sound that is played through the headphones will be for all intents and purposes, WITHOUT any room and ear reflections.

4. Now, having taken a measurement of how the room affects a sound with the dummy head, we should be able to use that measurement IN REVERSE to effectively ADD the same distortions to the signal as the combination of speakers, room and ears would do.

5. Finally, playing the sound with the distortion ADDED through the (neutral) headphones, should re-create the effects that the room and speakers have on the sound.

Yes? :D

[Jones_Rush]

Arnie,
Just because I hate to surrender easily :-), I'm giving it another shot:

If you take a dummy head, and where the ears are located, you put two microphones which have the same directionality of our ears. Then you put this head where your head is located when you listen to stereo through your speakers, and then play a song though your speakers, while recording it with the dual mikes on the head. I assume that if you then listen to that recording with headphones, you'll hear something which actually resembles what you hear with your speakers. This might get you what you want.

But, if you'll take this same head, even with the same microphones, and just get the impulse response from your listening spot when your speakers are playing an MLS signal, and then try to mimic it with convolution while listening to a song with your headphones, then you won't get same thing as the former situation, not at all. In this case you'll lose all the time cues which makes the sound "wet". You'll lose the time cues which makes your ears differentiate between front and backwards. An impulse response does not carry such information. All you'll get is the frequency response and the decay that your ears hear when the speakers are playing, and I'm afraid this is not what you're looking for. This is exactly what DRC is trying to eliminate in the first place, which is, the unique pattern that each listening spot has regarding frequency response and decay.


The only way to get something which might please you, might be to use a 5.1 headphone simulator (I heard there is such a thing), which using a special algorithm, can create front and back cues. Then, using an ambience recovery algorithm, you'll need to use the semantic "surround" channels in the headphones, to play the ambience back. This might give you something which will more resemble your speaker setup. And in simpler words, you just need an algorithm which can play DPL-II or CS-II through headphones, while maintaining all the necessary front and back cues.

[Jones_Rush]

So, if one person builds it, he/she can distribute it to everyone else. And, if it can be built w/in the 30-day trial period, no one will pay a penny.
Sounds sweet 3-way, touche for your sharp observations ! :-).

If you look in C:\Program Files\Cycling '74\MaxMSP 4.3\examples\fft-fun, there's a PAT file called "Convolution Workshop". It looks like it's a stereo convolution engine. I think it could serve as the basis for an N-channel convolver.
Yes, I've looked at it too. Were you able to convolve anything except the noise they give you there (of drums and that other annoying sound) ?. Also, were you able to make the DSP see the entire 8/6 analog outputs of your sound card ?.

[nubz69]

is there a way to apply an impulse filter in realtime through a sound cards inputs and outputs? I want to have the audio go into the sound card and then be processed and output to the sound cards outputs.

How hard would it be to build an external DSP device that could accept an impulse file so we wouldn't need to use our computers? Is there a way to take multiple impulses and merge/average them so that we can widen the sweet spot? I have an idea on how to do this but I don't have a mic pre I can use (I already have a test mic).

When I was looking at the waterfall plots I couldn't help but notice that your system seemed to exihibite a lot of break up modes/ringing. Why is this? It seems that by using this form of DRC we are ruining our transient responce. I have always belived that transient resoponce is one of the most important aspects of audio reproduction.

[Jones_Rush]

is there a way to apply an impulse filter in realtime through a sound cards inputs and outputs? I want to have the audio go into the sound card and then be processed and output to the sound cards outputs.
This appear to be possible with Max/MSP, but unless your source is digital, you'll add ADC to the chain, is this wise ?. I believe that the DRC filter might be more sensitive to "dirtier" signals, since there is so much processing which is done.

How hard would it be to build an external DSP device that could accept an impulse file so we wouldn't need to use our computers?

Why not use our computer ?. Of course such device could be built, but why spend the time on building a hardware that is already available ?. You can just build a dedicated PC for the process. You know, one with a really tiny case, with quiet fans.

Is there a way to take multiple impulses and merge/average them so that we can widen the sweet spot?
I think you can do this with Cool Edit's Mix/Paste option. An alternative mode will be to go with a softer correction mode. This will widen the sweet spot as well.

When I was looking at the waterfall plots I couldn't help but notice that your system seemed to exihibite a lot of break up modes/ringing. Why is this? It seems that by using this form of DRC we are ruining our transient responce. I have always belived that transient resoponce is one of the most important aspects of audio reproduction.
I'm not completely sure of what you mean. Can you edit the waterfall image I've included, and add circles around the possible problems ?. If it is what I think it is, it is already there, to some degree, in the "before" image (and this just goes to show you the importance of working with clean signals with DRC). But I'm not sure if that's what you mean.

If you look at the buttom of this (http://mail.infotecna.it/drc/) page, you'll see some waterfall images that Denis did. Can you spot the same problem there ?.

Anyway, I did the filter which its waterfall image you see above, a pretty long time ago. I'm afriad I didn't follow my suggestions in the guide back then (since there wasn't a guide, nor I had the knowledge to write a good one, back then), and I used a much weaker signal for the log sine sweep, than what I later recommended in the gude. This might have caused this problem. Once you'll show me excatly what you mean, I'll look at my more current filters.

[Arnie]

Originally posted by Jones_Rush
Arnie,
Just because I hate to surrender easily :-), I'm giving it another shot:

If you take a dummy head, and where the ears are located, you put two microphones which have the same directionality of our ears. Then you put this head where your head is located when you listen to stereo through your speakers, and then play a song though your speakers, while recording it with the dual mikes on the head. I assume that if you then listen to that recording with headphones, you'll hear something which actually resembles what you hear with your speakers. This might get you what you want.
Exactly! But i don't fancy having to re-record all of my music etc. that way. And besides, it would be nice to only use the positive aspects of in-room reproduction rather than the limitations of my speakers...

But, if you'll take this same head, even with the same microphones, and just get the impulse response from your listening spot when your speakers are playing an MLS signal, and then try to mimic it with convolution while listening to a song with your headphones, then you won't get same thing as the former situation, not at all. In this case you'll lose all the time cues which makes the sound "wet". You'll lose the time cues which makes your ears differentiate between front and backwards. An impulse response does not carry such information. All you'll get is the frequency response and the decay that your ears hear when the speakers are playing, and I'm afraid this is not what you're looking for. This is exactly what DRC is trying to eliminate in the first place, which is, the unique pattern that each listening spot has regarding frequency response and decay.

Aha! Now I think I understand. I had assumed that DRC had the ability to fully capture the acoustics of a room, and effectively nullify it so that the sound would be as if played in an anaechoic chamber... ah, a little knowledge...

The only way to get something which might please you, might be to use a 5.1 headphone simulator (I heard there is such a thing), which using a special algorithm, can create front and back cues. Then, using an ambience recovery algorithm, you'll need to use the semantic "surround" channels in the headphones, to play the ambience back. This might give you something which will more resemble your speaker setup. And in simpler words, you just need an algorithm which can play DPL-II or CS-II through headphones, while maintaining all the necessary front and back cues.

Yes, I've spent many hours with WinDVD (+audio packs), and all manner of hardware crossover filters (some of which are in Headwize (http://headwize.com/)), and all of them are lacking in some way. Or perhaps I'm a fussy bugger.

Jones,

Thank you immensely for the time you have taken to explain this to me. I sit at the feet of a fountain of knowledge and unerstanding :D

Arnie.

[3-way]

believe that the DRC filter might be more sensitive to "dirtier" signals, since there is so much processing which is done.

I'm not sure that I understand this one... AFAIK, the DRC filter file is a constant and is applied in the same manner, regardless of the source of the music signal. It applies correction based only on the information it gathered from the impulse file it was fed. Once the correction file is spit out, that's the file that is used to convolve the music signal, regardless of its source. I believe that the processing is the same, regardless of the source, and that the time / resources the convolution engine eats up is determined by the length of the filter DRC creates (the number of taps). More taps would be required for more complete correction (which might be more necessary with a 'dirtier' signal), however. It is possible that if the test signal used to create the impulse file is generated by the computer and then the music source signal comes in through the convolver via ADCs that the quanitzation done by the ADCs is less than perfect. This would effect DRC's ability to correct the music signal, but, I don't believe that it would require more processing.

I've looked into building an external DSP device. I don't think that it would be THAT difficult, though it would be considerably more difficult than creating a convolution engine with Max/MSP, and, it would take me months, if not years, to do it, even with proper tools, like MATLAB. Here, I feel obligated to say that I'm a lawyer, not an engineer, and I have no experience with DSP, other than what I've picked up around the internet and in a couple of books. I've read several posts on DSP boards suggesting that DSPs might be going the way of the dinosaurs because their usefulness is limited given the computing power of modern PCs. The only real advantage that I could see in building an external DSP device to handle convolution and crossover filters *might* be stability. The cost of building a custom external DSP device on our own would be high. Aside from the time involved, the cost of even an evaluation board powerful enough to do several channels of convolution would be $3-400 + you need the box, plus the ADCs and DACs (if you want to replicate the flexibility of a computer with a soundcard). I decided that if DSPs aren't significantly more capable than PCs, it's not worth the money, time and effort that would be required for me to develop an external DSP device.

The MSP Tutorial that comes with the Max/MSP download contains a pretty good introduction to digital audio basics. In addition, there's a great resource on Music DSP available at: http://www-ccrma.stanford.edu/~jos/filters/filters.html

[Jones_Rush]

I'm not sure that I understand this one...

and then the music source signal comes in through the convolver via ADCs that the quanitzation done by the ADCs is less than perfect. This would effect DRC's ability to correct the music signal,

This is what I ment. I said that because the filter does a complicated job (which of course is constant!, and doesn't change according to the signal), then a dirty signal comming from the ADC, might sound bad, after the correction. I see now how my post could have been read like you read it at first. Sorry for that.

[RayL Jr.]

Interesting, I came across this "FM (Filter Maker) is a utility for creating Digital Filter Transfer Functions to send to a DSP based Soundcard Digital Crossover Emulation program. It permits you to cascade standard filters and output at the sample rate and bandwidth of your choice to Rect Linear DSP FRD file format." - http://www.pvconsultants.com/audio/dsp/dplay.htm

From this pretty amazing site - http://www.pvconsultants.com/audio/frdgroup.htm - which I got from here - A summary of recommended speaker design software packages - http://ldsg.snippets.org/appdx-c.php3 - another nice site... ;)

I learned how to design a full range speaker, 20/25Hz-20,000Hz +/- 1-3 dB, 4-way (mid hi and lo drivers), using all 1st order crossovers, L-Pads, R-C (Zobel Network) and R-L-C (series notch filter) network (impedance compensation). Similar to Thiel CS7.2 speakers: specifications 4-way system - http://www.thielaudio.com/THIEL_Web/Pages/cs7_2.html @ $20,000/pair. Only mine uses a sealed box(es) instead of a passive radiator woofer... :)

Now that I have the whole circuit layout I just have to get the drivers and in-box parameters, etc. I’m using Unibox software - Design Your Own Loudspeaker Cabinet - http://www.danbbs.dk/~ko/ubtxt.htm - http://www.danbbs.dk/~ko/ubdwnld.htm Also WinISD - http://www.linearteam.dk/

Now only if I can make cement/ceramic cast half-round enclosures for mains and sub. This stuff is fun, ain't it? :D

[Nik Sinkola]

Hi all,

I have been following this thread for a while and had a thought. Have any of you heard of Creamware? They make Ultra high quality sound cards with DSP processors. The Pulsar II has 6 and the Scope/Fusion has 15. You can add as many cards as you like to get more processing power. They even have a card with only DSPs for expansion.

These cards allow for VST plug-ins and have increadable software. I was thinking that everything that we would like to do might be possible with these cards in real time without taxing the main CPU at all. These cards have been around for a few years and the software is Very mature.

Thay are used for creating software synthesizers, echo, delay, reverb, all kinds of effects. All in real time and all at the same time.

http://www.creamware.de/en/Products/SFP/Pulsar/default.asp

What do ya think?
Nik

[3-way]

I've looked into the Creamware cards several times over the past 18 months. They always look like an appealing solution, but I've never found the plug-in that would do exactly what I want it to do. I even called Creamware Tech Support about 10 months ago to ask whether I could do crossovers with their cards and software. The response was that it could probably be done, but he didn't recommend trying it... If Tech Support is telling you not to use their product, you should probably listen.

I'm sure that the Creamware cards are great for digital editing, but they're just not custom made for signal processing. As with all of the other digital audio editing programs and plugins, I've found that the functionality is never *quite* what I'm looking for. Just this morning I was at B&H here in New York, asking about a program called Waves that does linear phase EQ or another program with a convolver and that would allow me to do high order, linear phase xo + delay per driver. He said that Waves wouldn't quite cut it. I mentioned that I had been looking at Max/MSP and he said that he thought that that was probably the only program that would allow me to all three things mentioned above.

It's going to take some work to put it together, no doubt. I'll probably have to buy the program, and I'll probably have to pay for one-on-one tutorials at Harvestworks (a non-profit group here in New York) in order to put together a high quality program. But, afer looking around for 18+ months for the ideal program, I think the only solution is going to be to build it myself using Max/MSP. Maybe if a 3.0ghz P4 isn't enough muscle to run the program without latency, I'll need to look into Creamware (or the Mackie or TC Electronics equivalents).

[RayL Jr.]

Here's a software package I found with plug-ins, one of them "Room Acoustics Plugin" - http://www.purebits.com/order.html using your sound card. There's also the Liberty Audiosuite (which uses familiar soundcards) or LMS Analyzer from LinearX (expensive).

I want to measure the drivers impedance curve across the frequency band from my amp - http://www.purebits.com/appnote16.html - and don't want to calibrate the TS parameters just from listed specs (even with "matching" drivers). I did find excellent software for that (like Soundeasy v. 8.00, LEAP 5, WinSpeakerz, LspCAD) which I will do box and crossover design and optimization, possibly even room acoustics with.

Acoustic Measurement Software (and hardware, mics...) - http://www.me-au.com/measure.html Although it says "The ME Technologies section of this web site is out of date" - it's a great list (I'm checking newer listed versions, the ones I'm familiar with are top notch)...


PS - Here's 2 more links on measuring TS parameters:

The Subwoofer DIY Page v1.1 Measurements http://www.diysubwoofers.org/measure.htm
How do i measure the TS parameters of a misc driver? http://www.diysubwoofers.org/cgi-bi...p&msg_num=34170 http://www.avsforum.com/avs-vb/showthread.php?s=&postid=2715467#post2715467

[Jones_Rush]

Anyone can take a look at this (http://www.gaips.upv.es/AmbioVolver/AmbioVolver.zip) multichannel convolver, and see why everytime an impulse response is being loaded to it, there is a message: "Division by 0 error".

This convolver can be great for combining DRC with an Ambiophonics (www.ambiophonics.org) setup.

[MHenckel]

Jones - let me congratulate you on a very written guide to DRC. I have already made god use of it.

I have two proposals for additions that i do myself on as part of setting up the filters;

First proposal:
Step 17 in the guide - just before saving the IR of the speaker/room you need to delete everything up to the main spike - signals prior to the main spike is 2, 3, 4rd etc hamonics distortion IR. I am new to this forum so no link provided but do a google search for;

"Log sine sweep NPL-workshop-2002_files"

And you will see were i am comming from.

No need to let DRC work on the distortion products !

Also the noise floor prior to the main spike will give some indication of the S/N of the particular measurement - just as a check.

Second proposal:
Since you refer to the Behringer ECM8000( i have one for myself) you should now that it is not totally flat( approx -4 dB @ 20 KHz). I can provide a generic correction file for this mic if needed. If you use this Mic and follow the guide DRC willl actually try to compensate also for the Mic´s frequency response. So prior to saving the IR you need to do an invers filter to compensate for the mic. Use FFT filters in Cool Edit for that. Also as sanity check you should advice of the option cool edit has for showing the frequency response of the measured system ( Alt+Z)

I have very good experiences with using DRC on Mid-Fi systems and even on Low-Fi(PC speakers) and can only say that it makes a lot of improvement in my set-ups, but it is not a high end system i have and the room is very problematic in the low end.


For Digital X-over which also has been discussed in this thread i belive that we all should head over to hydrogen audio and try to influence the author of the foo bar convolution plug in to make it into a multi channel convolution engine - Foobar already supports multichannel sound - so it should be "easy"


Best regards

Morten

[Jones_Rush]

First proposal:
Step 17 in the guide - just before saving the IR of the speaker/room you need to delete everything up to the main spike - signals prior to the main spike is 2, 3, 4rd etc hamonics distortion IR.
I thought DRC knows how to handle this by itself... I'll check with Denis as update the guide if necessary. Thanks!.

Second proposal:
Since you refer to the Behringer ECM8000( i have one for myself) you should now that it is not totally flat( approx -4 dB @ 20 KHz). I can provide a generic correction file for this mic if needed. If you use this Mic and follow the guide DRC willl actually try to compensate also for the Mic´s frequency response. So prior to saving the IR you need to do an invers filter to compensate for the mic. Use FFT filters in Cool Edit for that. Also as sanity check you should advice of the option cool edit has for showing the frequency response of the measured system ( Alt+Z)
Yes, I am aware of this problem. It already has been mentioned in this VERY long thread a couple of times.

As of Version 2.3.0, DRC has a mic calibration file. So no need to do it with Cool Edit. I will update the guide for this feature.

For Digital X-over which also has been discussed in this thread i belive that we all should head over to hydrogen audio and try to influence the author of the foo bar convolution plug in to make it into a multi channel convolution engine - Foobar already supports multichannel sound - so it should be "easy"
Not only for Digital X-over. A real-time multichannel convolution for windows is something that I'm searching like crazy. As of today, there are none. Not even one. Such a convolver, if implemented right, will be able to DRC a 5.1/7.1 signal, which is crucial for HT. It will also be able to convolve the reverb channels of the Ambiophonics setup, which I'm experimenting with these days, with great success.

Morten, do you already have the email address of the guy who wrote the foobar convolution engine ?.

Thanks for your constructive comments! :-)

Edit,
Btw Morten, I think that Digital-XO is possible today with DRC, but I'm not sure with what level of success. Actually, anyone who used DRC with a sub, like me, is already doing a digital-XO. All you need to do is power all the drivers of your speaker (I assume from different amps ?), and then go into SpectraLab (or any other spectrum analyzer software) and using a noise (I'm not positive which will be the best though) watch the 1/3 band bars with infinite averaging, and using the volume control of the amps which drive the drivers, try to calibrate for the flattest response possible. Then, just go into Cool Edit and follow the guide as usuall. You'll get both Digital-XO and Room Correction at the same time.

The only problem I could see, is how to split the signal which suppose to drive all the amps which drive the same speaker, without losing sound quality. OTOH, allowing each driver its own sound card channel is also kind of a waste in my eyes.

[ewildgoose]

Well, I'm keen to give this all a go.... I have the Behringer mic and pre on order, and it should arrive in a week or so's time.

Now the main issue is that I have only linux here, so I'm interested in feedback from anyone who has done the sine-sweep under linux...? Any pointers as to software and perhaps some notes as to how you went about it?

Thanks all. And thanks for the guide as well - extremely detailed! Well done!

Ed W

[dsinder]

Even if one had a multi-channel convolver it seems to me there is another problem to be overcome. One has to get ahold of each of the channels to be convolved. So if you are playing a Dolby Digital or DTS title you need a DD or DTS decoder in the software. It would then be nice to put it back into a DD or DTS stream to send it to your receiver/pre.

Last night I decided to install the MS DirectX 9 SDK and the Windows Media Player SDK and look at using BruteFIR to make a plugin to WMP 9. The plugin itself does not seem to be too difficult. A bigger challenge is to transplant the guts of BruteFIR.

Am I missing something about getting each of the channels?

[Jones_Rush]

Am I missing something about getting each of the channels?
You're not missing anything. There are two possible options to do that, and both talks about decoding DD/Dts in software:
1) Implementing the multi channel convolution engine inside a directshow filter. This way you can use programs like ZoomPlayer.

2) Creating a plug-in which can work with anything which your sound card produce. Kind of a last station before the sound leaves the sound card. We can use Max/MSP to acomplish that, but this software is more like a computer language by itself. Learning it is a pita.

Anyway, if you have an nForce board, you can always encode back to Dolby Digital, but quality will suffer.

[ewildgoose]

If you have certain cards such as the RME 9632 or any of their DIGI stuff, then I think you can get all your channels out digitally and individually.

So presumably as long as you can decode in software, then convolve, you can then sent each channel individually to your amp. The RME cards synchronise precisely across channels, but there is some jitter introduced by the digital connection on the way to the amp, so whether everything would remain accurately clocked is anyones guess...

You could also get 7 analogue channels out of most of these cards as well. (A few people have claimed that the RME 96/8 dac's sound nicer than a Lexicon MC12 for example...)