JVC 30K/40K component output frequency response

[dr1394]

I've been waiting for Greg Rogers to clear up this issue up with Joe Kane, but
I guess it's never going to happen. The 1080i frequency response measurements
published in the August 2003 issue of WSR are not correct.

The frequency burst patterns on the 1080i version of DVE are attenuated
themselves. 100% high frequency bursts are very difficult to encode, and
it appears that the MPEG-2 encoder used to create DVE turned on an adaptive
pre-filter while encoding the burst patterns.

The net result is that the published curves are showing a combination of the
deck's component output response and the response of the encoder pre-filter.
If the encoder pre-filter is subtracted from the result, the component outputs
are actually 8 dB down at 30 MHz (and the Samsung T-165 outputs are nearly flat
to 30 MHz).

The 720p DVE tape does not have this problem. The 720p curve is correct and in
fact, probably reflects the 1080i performance.

I'll post some graphs that show what's on the tape shortly.

Ron

[John Mason]

That'll be interesting to see, Ron, since the plot (http://www.avsforum.com/avs-vb/showthread.php?s=&threadid=369786) in question, at 30 Mhz, appears at what must be the nearly invisible level. So how does all this relate to Joe Kane's published comments last summer that typical telecines on 1080/24p master tapes are <1300 pixels resolvable resolution (dBs down not provided), which sspears here wrote was confirmed with spectrum analyzer tests? Also, somewhere here I believe Glimmie indicated ~1300 was typical, too. -- John

[dr1394]

The 1,2,4,6,8,10 MHz burst.

Ron

[dr1394]

The 10,12,14,16,18,20 MHz burst.

Ron

[dr1394]

The 20,22,24,26,28,30 MHz burst.

Ron

[uzun]

So in fact the 40k's output is down about 8DB at 30Mhz, rather than what was published? That's still quite a drop though, isn't it? Watching the resolution test pattern off the JVC 40k on my Sony KDP-65WS550, it seems the pattern is clearer using the firewire output into the Sony, rather than using the component outputs on the 40k. Does this indicate that the Sony's internal D/A conversion is flatter than the JVC's?

[jamoka]

"it appears that the MPEG-2 encoder used to create DVE turned on an adaptive pre-filter while encoding the burst patterns."

I'll bet they're going to correct that and send out a replacement tape to the people who bought the 1080i version. Has anyone received a notice yet?

Joe

[John Mason]

Nice to have some new data, but it's difficult relating the 3 separate graphs of frequency bursts above to the approximation of the 30k's 1080i graph, component out, in the Aug. '03 WSR issue shown here (http://www.avsforum.com/avs-vb/showthread.php?s=&threadid=369786).

Can't follow the X axis of the 3 separate charts, and how it might relate to the 0 dB point of the WSR chart. Also, thought the highest frequency bursts on the 1080/24p master weren't transferred to the 1080i DVE test tapes since the assumption was they would be too weak. (Rogers writes that 37.088 MHz is needed for 1920 and 1280 pixels.) Yet the 20-30 MHz burst chart above shows a presumably 30-MHz burst between 1680 and 1920 pixels on the horizontal axis. The WSR review chart shows a single plot point at ~30 Mhz and indicates that's ~1553 pixels(at a staggering -25 dB). -- John

[John Mason]

Originally posted by dr1394
The frequency burst patterns on the 1080i version of DVE are attenuated
themselves. 100% high frequency bursts are very difficult to encode, and
it appears that the MPEG-2 encoder used to create DVE turned on an adaptive pre-filter while encoding the burst patterns.

If this happens with test data (automatic adaptive filter activation), how is it avoided with normal motion video? Do all MPEG-2 encoders have this feature, which apparently differs greatly from normally 'tossing out' higher frequencies to achieve compression? Is it related to the great compression change from ~270 Mbps video payload on a HD D5 master (~5:1 compression) to ~23 Mbps video payload on a D-Theater tape? -- John

[dr1394]

The charts are showing the same thing you would see on a waveform analyzer
or oscilloscope. The x-axis is just the pixel position in the raster (0 is the
left border of the screen and 1920 is the right border) and the y-axis values
are the luma levels from one line of video. The frequency bursts are sine
waves, which can clearly be seen on the 1 MHz burst.

To calculate how many dB down a burst is, take it's peak to peak value and
divide that by 219 (235 - 16) which is the value of a full amplitude burst.
Then take the log*20. For instance, for the 30 MHz burst which has a
peak to peak amplitude of about 30 (between 110 and 140):

30/219 = 0.13698 and 20 log 0.13698 = -17.266, so the 30 MHz burst is about
-17 dB on the tape.

Ron

[dr1394]

Adaptive filters are used to mitigate poor encoding. When the encoder detects
that it's doing a poor job (the average quantization level of encoded frames
is high), it will cut in a filter to help itself out.

Burst patterns are very difficult to encode, so the encoder automatically
switches in a filter to make them easier to encode.

On normal material, the same thing can happen. But it's less noticable because
hard to encode sequences usually have so much going on that making it softer
isn't easily detected by the human visual system. It's a trade-off between
having the scene horribly blocky or softer and less blocky.

Ron

[John Mason]

Originally posted by dr1394
To calculate how many dB down a burst is, take it's peak to peak value and
divide that by 219 (235 - 16) which is the value of a full amplitude burst.
Then take the log*20. For instance, for the 30 MHz burst which has a
peak to peak amplitude of about 30 (between 110 and 140):

30/219 = 0.13698 and 20 log 0.13698 = -17.266, so the 30 MHz burst is about
-17 dB on the tape.

Whew. Thanks Ron. I had thoughts of transferring this data to a similar approximation (http://www.avsforum.com/avs-vb/showthread.php?s=&threadid=369786) of WSR's chart, but think I'll take a rain check. :). (Anyone care to work from a 'quote' copy of that linked graph to create a similar chart? An Excel sheet would simply number crunching.)

So I assume your plots are using a JVC 30k or 40k deck and recording test signals on it yourself rather than involving the DVE test tape. Still unclear about the 30 MHz plot, though, since I understand 1920 pixels requires 37 MHz. Your calculation above of -17 dB for 30 MHz [initial post says -8 dB) seems like it would also be "in the mud", although it's not the -25 dB WSR's chart indicates. Your chart shows a 30-Mhz burst between 1680 and 1920 pixels, while the WSR chart 4 (approximation linked above) indicates 30 Mhz is 1553 pixels.-- John

[dr1394]

John,

You're still a bit confused. The charts are what's on DVE. They were generated
directly from the 1394 MPEG-2 video bitstream (remember, I'm a 1394 developer).
When Greg Rogers did his measurements, he assumed that the burst
patterns on DVE were delivering full amplitude for all frequencies including
30 MHz. However, in reality the bursts are attenuated. The 30 MHz burst is
-17 dB down on the tape. So when he measured -25 dB at 30 MHz, -17 dB is from
the tape itself which means that -8 dB is the actual response of the component
output.

You're also confused on pixel position. I'll post another pic later that should
clear it up for you.

Ron

[dr1394]

Here's the 1 thru 10 MHz chart superimposed on the video pattern. Note how
the light and dark areas match the pixel levels plotted on the chart.

Ron

[Oliver Klohs]

Hi,

I plotted the measurements of Ron against Greg Rogers' measurements in the 40000 test and it appears that the JVC 30000/40000 Greg measured are looking good out to 22/24MHZ with a 1 to 2 db boost to frequencies between 10 and 20/22 MHz. From there on they gradually drop off to -7/-8 db at 30 MHz. The Samsung component output is flat out to 30 MHz (+/- 0,5db) - wow !

Ron, thanks for those measurements - shows that with hardware like the Samsung we are prepared for much more to come from HD :)

Now it would be interesting to feed the Samsung into the HD scalers that are available and see what they do with the 1080i component signal when they output 1080p.

Oliver

[John Mason]

Thanks, Ron, for taking the time to clear things up. The 1-10-MHz superimposed makes it clear for those bursts. But the 30-MHz burst in relation to the x-axis pixel count is still puzzling. If you drop a vertical line from the start of the 30-MHz burst it falls just before (one mark) the 1680-pixel readout on the X axis. If that burst start point represents 1553 pixels then the 30-MHz burst corresponds to Chart 4 in the WSR article (the approximated chart sublinked above), which shows the pixel count and the frequencies stacked together along the x axis. (Perhaps the width of the 30-Mhz burst, which seems to cover the area between ~1680 and nearly 1920 pixels is what's confusing.) -- John

[John Mason]

Originally posted by Oliver Klohs
I plotted the measurements of Ron against Greg Rogers' measurements in the 40000 test and it appears that the JVC 30000/40000 he measured are looking good out to 22/24MHZ with a 1 to 2 db boost to frequencies between 10 and 20/22 MHz. From there on they gradually drop off to -7/-8 db at 30 MHz. The Samsung component output is flat out to 30 MHz (+/- 0,5db) - wow !
R
Flatter sure sounds better, Oliver. That approximation chart (http://www.avsforum.com/avs-vb/showthread.php?s=&threadid=369786), posted in the test forum to avoid too much column width here, is for the 30k machine, component out, in the Aug. '03 Widescreen Review. If you have the time, and care to post your plot, please feel free to 'quote' copy it and plug in your values and modify it for another post. With the ASCII plotting, just remember to avoid carriage returns in the graph area, and include the "code" and "/code" (with [] instead of quotes) before and after the ASCII plot.

Now, with Ron's info, perhaps it'll be possible to attempt a rough comparison (frequency/pixel-wise) between what's being delivered on D-Theater tapes at 23 Mbps and what folks are seeing from 1080i OTA and via satellite (<17 Mbps). Earlier here (http://www.avsforum.com/avs-vb/showthread.php?postid=3422866#post3422866) one observer estimated his D-Theater images weren't very different from 1080i viewed via DirecTV. Opinions sure seem to vary. Perhaps to be valid, though, you'd need an all-digital path from a 30k/40k to a digital display with adequate resolution, such as a 1366X788 plasma (owned by that observer). Take that back. For the 30 MHz readout you'd need ~1553 pixels, so perhaps a rare Toshiba 1920X1080 LCoS RPTV would do, even though its final LCD stage is analog, not digital. Guess some 1080p ((1920X1080) plasmas are on the way. -- John

[Oliver Klohs]

Hi,

I have put my calculations in a chart that I hope is sufficiently easy to read.
This is the first time I attach a file, so please bear with me :)

Oliver

[umr]

Thanks for the posts guys. This has been very interesting.

[HDHTPC]

Does anyone have any idea how this compares to attenuation at the higher frequencies of other equipment?

In particular things like the RGB outputs on DTC100, Dish 6000 and HiPix card?

Regarding the HiPix card - I removed the final stage capacitors and diodes on mine which produces a more detailed image. I think some of these devices intentionally attenuate analog outputs so that they can still pass FCC emissions with poorly shielded cables attached.

The attenuation at higher frequencies is actually a saving grace if you are watching an overcompressed, but not pre-filtered enough signal. With my modded HiPix card to a Sony GDM-FW900 compression blockiness is painfully obvious. My other unmodified HiPix card doesn't show the edges of the blocks quite as sharply.

(Thanks again, Ron, for your sharing of interesting data with the rest of us!)

[Oliver Klohs]

HDHTPC,

I think this would have to be evaluated on a per-device-basis, which unfortunately nobody has done as of now, I fear.

BTW you can reduce the visibility of the artifacts you describe by viewing with less ft-lamberts (around 10 I'd say) on a front projection device, obviously seating distance also is a factor.

Oliver

[HDHTPC]

FPs can also be set slightly out of focus to intentionally blur the image.

I use CRT DirectViews for all my viewing, so I don't have that option.

Maybe we could use a "signal" blur circuit with a big gnob. On overcompressed material, you could set it to start rolling off the high frequency signals. If you had a nice, criisp full rate signal you could back off and let the high freqs through again. Sounds like the JVC decks are already doing a bit of blurring for you...

[John Mason]

Originally posted by HDHTPC
Does anyone have any idea how this compares to attenuation at the higher frequencies of other equipment?

In particular things like the RGB outputs on DTC100, Dish 6000 and HiPix card?

HDNet's Tuesday 8 am ET test patterns from DirecTV and on some cable systems give you a rough idea. In testing the 1920X1080-pixel Toshiba LCoS RPTV, Gary Merson reported he was able to resolve ~1920 lines from the resolution pattern. (You multiply the chart number readings by 1.78 for lines/picture width.) It's rough, IMO, because the 'merge point' for the HDNet resolution wedge--where lines are no longer distinctly resolvable--is often vague.

Also, I understand that compression demands on test patterns (resolution wedges) and full-motion detailed video differ greatly. A simple resolution wedge--correct me if I'm wrong, Ron--doesn't require the compression, and subsequent 'tossing out' of higher frequencies, that detailed motion video does. But as Ron illustrated in starting this thread, multibursts used for testing can trigger a prefilter in MPEG-2 encoders and diminish higher frequencies to simplify MPEG-2 compression.

What about measuring final high-frequency content piped to your display from HD programs? Earlier sspears wrote (http://www.avsforum.com/avs-vb/showthread.php?postid=2498224#post2498224) that a spectrum analyzer confirmed 720p-advocate Joe Kane's comments last summer that telecined movies on 1080i master tapes provide <1300 pixel resolvable detail, making them about the same as the 1280X720p format for resolvable horizontal detail (according to Kane). Recall Glimmie confirmed that <1300 resolvable pixels on a HD D5 (Panasonic pro deck) for telecined motion video was typical. So it appears this <1300-pixel result is similar to the charts recorded above, which of course depends on the slope of frequency plot; that is, whether the signal amplitudes near and above ~1300 are so low (-dBs) they don't convey HD details all that well. Here's a personal-computer-based 40-Mhz (initially) spectrum analyzer (http://www.eet.com/sys/news/OEG20040219S0040) built with off-the-shelf parts, created for radio astronomy, that might serve for such measurements instead of Tektronix or similar equipment. No doubt this gear is still too costly for most video hobbyists. -- John

[Wendell R. Breland]

Originally posted by dr1394
Here's the 1 thru 10 MHz chart superimposed on the video pattern.

Ron, What hardware and/or software are you using to generate the charts you posted? (And thanks for the work and posting of the pics)

[dr1394]

Originally posted by Wendell R. Breland
Ron, What hardware and/or software are you using to generate the charts you posted? (And thanks for the work and posting of the pics)
First of all, I'm both a IEEE1394 and an MPEG-2 encoder engineer so I have
all kinds of special tools. Here's the tool chain:

1) I have a copy of 1080i DVE that I play on a JVC 30K.

2) I capture the 1394 bitstream on my 1394 development platform. I can
emulate just about anything on this platform including another 5C enabled
D-VHS deck. I don't have any tape, hard drive or optical media to capture to,
so I'm limited to capturing 16 Megabytes of bitstream in memory.

3) The development platform runs VxWorks, so I just FTP the memory buffer
as a binary file over to a Sun Ultra 2 workstation running Unix (Solaris).

4) On the Sun, I demux the captured Transport Stream with a home brew demuxer
and then use an in-house MPEG-2 video bitstream analyzer. With this tool,
I can select a frame and dump the YCbCr values to a file.

5) With a little C program, I pick out the Y values for one line of video
and pipe that to a text file.

6) Finally, I use an old (but sweet) Unix shareware program called "xvgr"
to graph the Y values. For the pic with the graph over the video, I used
Photoshop to cut and paste. For the video, I can either process the YCbCr
file (4:2:0 YCbCr -> 4:2:2 YCbCr -> Solaris .ppm format -> .tga format ->
Photoshop) using a C program, Solaris yuvtoppm and another shareware program
called "xv" or I can use a reference software MPEG-2 decoder to go right to
.tga format and import to Photoshop.

Ron

[Wendell R. Breland]

Originally posted by dr1394
First of all, I'm both a IEEE1394 and an MPEG-2 encoder engineer so I have
all kinds of special tools.

Ron, Thanks for the process info. Now more questions.

Do yo have a MyHD MDP-120 DTV tuner card?

The KTech Model VSB-FRQ-200 Converter can input SMPTE310M, DVB-ASI and 8VSB. It can output SMPTE310M, DVB-ASI and IEEE1394.

Do you know of a commercial device that can input IEEE1394 and output SMPTE310M and/or DVB-ASI? RGBHV and/or YPbPr??

[PoolShark]

I hope you guys don't mind I have some "how to" questions, but this is the only thread that came up when I searched for JVC 40000.

I have a Samsung Sir T165 and would like to buy the JVC 40k to record OTA HD.

Is this easy to do?

Which box controls the programming settings?

Which box outputs the recorded HD to the tv? How (DVI, component)?

Is it dependable?

Is there any downfall to this set up?


Thanks!!!!!!!

:cool:

[jimg]

Originally posted by PoolShark
I hope you guys don't mind I have some "how to" questions, but this is the only thread that came up when I searched for JVC 40000.

I have a Samsung Sir T165 and would like to buy the JVC 40k to record OTA HD.

Is this easy to do?

Which box controls the programming settings?

Which box outputs the recorded HD to the tv? How (DVI, component)?

Is it dependable?

Is there any downfall to this set up?


Thanks!!!!!!!

:cool:

Quick notes... (so you can do searches to find out more):

The JVC has built-in MPEG decoders so you can use it's outputs directly.
The T165 can decode the firewire output of the JVC and display via its outputs.

For the firewire connection, you just need a 4-pin to 4-pin firewire cable.

Power both off, connect the cable, then power both up.
They will handshake automatically. The T165 will identify the JVC 40K as I-1.
Press the firewire button on the remote, it will bring up the firewire menu. Verify you can power up and down the JVC. That's it.

T165 timers are "scheduled events", off the main menu... very straight-forward.

Caveats for timing: (There's an extensive T165 thread on issues... do the search).
1) Always power the JVC on and off from this menu. The T165 gets confused easily.
2) The T165 must be off before it can do a timer event.
3) The T165's timer uses the PSIP of both the last-tuned-to channel as well as the desired channel. i.e. the clock will be synchronized to the last tuned-to-channel when powered off, and will resync when it wakes up for the timer event. And if the PSIP is wrong, your recording won't start/stop as you programmed. Best to tune to the target channel, check the difference between the channel's time and the correct time, then set the scheduled recording accordingly (i..e. if channel is 10 minutes late, set the start/stop times 10 minutes later so it matches).

Other issues....
The T165 contols but doesn't know how to do "daily" "M-F", etc. See other thread about
using an external timer.

Sometimes subchannels cannot be timer-recorded. They'll work fine manually however, so an external timer-capable programmable "remote" can be programmed to do the recordings.

[PoolShark]

Thanks for the input Jimg,

Unfortunately I think the timer on the Samsung 165 is going to be a deal killer because the times here in Dallas are way off, so if it does not have a manual setting for the 165 it will always miss my programs.

I called Crutchfield today and they said the JVC 40k would control the 165 and turn on and tune in from a menu on the 40k. I guess that guy was wrong.

I guess I could do it manually but I would prefer an automated system.

I guess I'll wait some more. :mad:



I need a HD Stand Alone Tivo!!!!!!!!!!!!!!!!!!!!!

[Troy LaMont]

Ron,

I've been following this thread and I've yet so summise the effect of the drop in frequency response in lay terms.

For me, what does this mean? Lower resolution, worse picture quality or both?

Thanks.

Troy

[HDHTPC]

Generally speaking it is a slight "softening" of the picture you are watching.

If the material is "over-compressed" and subject to frequent DCT artifacts,
than it might actually make the picture look better... But it limits your ability
to view fine details in the highest quality recordings.

As John Mason points out, most telecined films are already resolution limited, so the major impact would be attempting to view high bitrate D-Theater HD-*VIDEO* productions on a very capable display (such as a Sony G90 projector). Even so, people rave about how great the "Beauty of Japan" demo tape looks under such conditions, so we can only image how much better it could be if the component outputs supported higher bandwidth.

[HDHTPC]

It is possible that there are intentional filters on those component outputs that could be removed. I don't have a JVC D-Theater DVHS to look for myself.

[Oliver Klohs]

One could always buy the Samsung receiver and get a flat frequency response without any modifications.

From Ron's measurements it seems that the encoders are the main culprits for the bad frequency responses of HD material originating from movies and it would be interesting to see how actual movie transfers fare, not only the DVE.

[John Mason]

One reason I posted an approximation of G. Rogers' JVC 30k component-out frequency response plot (http://www.avsforum.com/avs-vb/showthread.php?postid=3454358#post3454358) from the Aug. 03 WSR was an attempt at estimating what some may see, fidelity-wise, with other 1080i sources. Ron writes his tests indicate that multiburst test signals triggered prefiltering in the MPEG-2 encoder for the DVE 1080i test tape (from a HD D5 1080/24p master tape). So, a graph of the 30k's component output without this prefiltering attenuation should appear flatter, like the pdf-file plot (http://www.avsforum.com/avs-vb/showthread.php?postid=3454358#post3454358) prepared by Oliver.

It's not clear when--if at all--such excessive prefiltering might be occurring with other broadcast 1080i. As usual, the mix of variables grows. Can an encoder setting, deliberate or otherwise, trigger such automatic encoder prefiltering? Or, as mentioned above (http://www.avsforum.com/avs-vb/showthread.php?postid=3459194#post3459194), if telecined films are typically <1300 pixels in resolvable detail, might the frequency response of typical telecine hardware be the key limitation? Here's a plot (http://www.avsforum.com/avs-vb/showthread.php?postid=3317574#post3317574) of just one data point for a widely used CCD-based telecine (Thomson's Spirit) compared to plots of CCD and CMOS sensors used for HD cameras. (CMOS sensors provide flatter frequency response, and are being used in new HD cameras from Germany's Arri and Canada's Dalsa, as well as 1080/60p camera (http://www.smpte.org/conferences/38synkozlowski.cfm) development .)

Perhaps some comparisons of video-based D-Theater productions, as HDHTPC suggests above (http://www.avsforum.com/avs-vb/showthread.php?postid=3483807#post3483807), will offer clues. They don't involve telecines, and such 1080/60i tapes delivered on HDNet or Discovery HD Theater appear crisper than most telecined productions. They're also 'smoother' when motion is involved since they're captured at 1080/60i (60 fields per second) instead of 24 frames per second with film or 1080/24p tape. Most are originally produced from HDCAM tapes deliberately filtered >1440 pixels. But upcoming movie productions, such as "Collateral" (http://www.hollywoodreporter.com/thr/film/feature_display.jsp?vnu_content_id=2076963) starring Tom Cruise, are mostly created with a camera such as Thomson's Viper not filtered >1440; Sony's new HDCAM SR omits this >1440 filtering, too. So when a "Collateral" version--not telecined--reaches home HD theaters it'll be interesting to see if any added 'wow factor' gets through. -- John

[oleus]

pardon my ignorance....i'm new to all this and was thinking of buying a 40K but i'd be using the component inputs from it to watch recordings. (i'd be recording from my motorola 6208 HD PVR via firewire).

would i be better off with the 30K? are the component outputs on these things faulty?

much of this thread is over my head so bear with me.

thanks

[moggy]

Originally posted by HDHTPC
It is possible that there are intentional filters on those component outputs that could be removed. I don't have a JVC D-Theater DVHS to look for myself.

Some thread cross pollination required:-
Another thread trying to come up with a mod for the JVC 30k.
It appears that modifying some output filtering improves the perceived resolution quite markedly.
Try:-
http://www.avsforum.com/avs-vb/showthread.php?postid=5069056#post5069056