Unstable Signal Amplitude

[signal]

I've really been going crazy over trying to figure this one out.

I was capturing from LD and noticed that the signal wasn't as clean as I thought it should be.

I hooked up a signal generator to get some stable colorbars and noticed the overall amplitude of the video signal fluctuates at random.

I'm seeing this in capture mode by using VDub and Trevlac's color tools. To set this up:

1. Use Preview mode
2. Preview acceleration mode on (using both fields)
3. Make sure RGB Filtering is selected
4. Add Trevlac's colortools filter
5. Configure the filter (use the histogram with no history)

At first I thought it maybe related to how Trevlac's colortools was sampling the live input but capturing a lossless clip shows the fluctuations as well.

I checked the video source with an oscilloscope pass through to a monitor and the signal is rock solid. Connecting up to the capture card isn't affecting the signal either.

Just in case I tried a TBC with a freeze frame of the color bars and with two other capture cards (on the same system - not all installed at the same time) and get the same results.

So now I'm in a quandary if the fault is with my motherboard/psu or the design/components used on the capture cards (or both).

I don't expect perfection on the ADC's but at least something more stable than this.

Board: ASUS P5WD2-E
PSU: Enermax Liberty 650W

Capture Board #1: Osprey 210
Capture Board #2: ATI TV Wonder Pro
Capture Board #3: Matrox APVe

Here's a sample video you can use with color tools (and no, the colors levels weren't calibrated yet).

junk.avi (Huffyuv 33MB) (http://www.pinesalad.net/example/junk.avi)

Has anyone else experienced this?

[DrP]

You've got some pretty major noise going on in there. It could be coming from lots of places, even your TBC. Lots of low end PC capture cards aren't all that flash when it comes to proper design and can be a source of noise themselves. Even the PSU in your PC could be a potential cause for what's in that clip.

Since you say all three of those capture devices produce a similar quality and since Matrox products used to be fairly good WRT to noise (haven't used a low end Matrox capture device for a long time, so maybe things have changed) you can probably eliminate the capture device as the noise source.

[signal]

Yea, it's one of those unfortunate things where it could be coming from many sources.

The sample avi was captured "raw" without anything inline.

But I think I found it is more than likely the PSU.
The MB could still come into play here so I'm not holding my breath.

While I had the scope out I hooked up to one of the 3.3v rails on the motherboard psu connector and can see a lot of noise with 40-60mv random spikes that seem to correlate to the amplitude issues. The 5v rail is just as bad.

I'll swap out the PSU tomorrow with a different model and see if that "cures" it.

[signal]

I tested a different PSU on a different MB and did not see the same issues on the scope.

Tried that PSU on the ASUS board and the same issue appears.
Might as well have flipped a coin I guess.

Going to try and isolate a few more components to see if they aren't part of the problem. Will also try the capture cards on the other system, but it's an nvidia 680i which has it's own issues.

Failing that, I guess I'm looking for a new motherboard that doesn't cause interference on the voltage lines.

[FlimsyFeet]

Interesting. Do you think it's a faulty motherboard or a design flaw with that model?

[DrP]

Have you thought about tacking some additional bypass caps on the supply rails to the slot you have the card in - assuming the mobo isn't under warranty that is.

[signal]

Interesting. Do you think it's a faulty motherboard or a design flaw with that model?

Not sure. There are a few threads around that attribute some problems wth video capture noise to particular "noisy" motherboard models and chipsets.

The board wasn't around that long and was replaced by ASUS with a revision to handle Core Duo's (already a 975x chipset, but needed to be revised to handle them). I suspect they may have worked some of these kinks out on the newer boards.

Have you thought about tacking some additional bypass caps on the supply rails to the slot you have the card in - assuming the mobo isn't under warranty that is.

I was thinking along those lines but by modding the capture card itself. Seems this should be considered in the design of the card to make sure there is clean voltage on the card but can see where they think a computer PSU/MB should already be doing that.

What value of the caps to use, time, soldering capabilities, and the possibility of damaging something has me looking at another board.

In the end the board is old enough and isn't Core Duo capable that it's a good excuse to get another one rather that fight it. Problem with that is it could be rolling the dice again

[signal]

BTW - I'm still curious what others see with Color Tools on capture (described in my first post). If anything, tested with a decent freeze frame if colorbars aren't available.

I have my doubts that a new MB will cure everything.

Thanks

[DrP]

Test Pattern (http://www.users.on.net/~alfalfa/mpg/testpattern.m2v)* - Produced via svideo out of a STB into a Hauppauge PVR150 then recompressed to ~2Mbit/sec with HCENC.

Original MPEG2 source (http://www.users.on.net/~alfalfa/mpg/source.mpg)* - that was fed to the STB to produce the clip above.

*clips no longer available.

[signal]

Thanks for the samples DrP.

Using VDUB mod I can see the source had very minimal variations in amplitude and the capture wasn't radically affected.
Looks like it was filtered somewhat as the center spike seen in the original is forced to black in the captured.

I'm still hammering away at this with no conclusive answers.
I've replaced the motherboard and PSU (which I wanted to do anyway) and results are still bad.

So I've been starting over and from scratch working up from power source to video source again.

[signal]

After four days I feel like I'm trying to reinvent the light bulb or something.

I just can't nail down where this interference is coming from.

I've gone through the power from outlet to pci slot.
Checked all ground potentials, used a physically separate power circuit with nothing else on it, emi/rfi filters, ac voltages, etc.

Tried shielding the card, changed cables, used an isolation transformer on the video, different motherboard, PSU, case.

Tried 4 different video sources (5th if you count the signal generator) and now 5 capture devices.

Even internally generated video from the video card exhibits the same symptoms when trying to capture it. Capturing with an ADVC110 is even bad.

Taking a break I looked up how 3 of my various Conexant based cards are designed and noted how the Conexant datasheets offer ways to reduce noise in the design. The Osprey kind of follows this advice by having a fairly decent analog voltage section but then ruins it by not following Conexant's advice on not connecting a couple of the analog voltage pins to the right voltage bus and how the analog grounds are handled (they tie everything to the digital grounds).

Regardless, it might have helped but as this problem is widespread on my setup the problem is still in my court.

[signal]

I have a JVC monitor that has a passthru and tried something.

I hookup the video source to the JVC and it looks great.
As soon as I plug it into the capture card, a dark band appears between green and purple. This is the same as when the video source is plugged directly to the capture card.

What causes this?

Here's a current snapshot and video.

Latest Capture (snapshot below taken from this):
cb2osp-c1.avi.avi (Huffyuv 26MB) (http://www.pinesalad.net/example/cb2osp-c1.avi)

Snapshot:
http://www.pinesalad.net/examples/ospreycb1ur0.png


Any more ideas?

[DrP]

Capturing with an ADVC110 is even bad.

Woah Nelly! If you are still getting the interference shown in your initial clip while grabbing the LD output via an ADVC110, then the problem is most likely not your PC itself, even if the ADVC110 is being run with power supplied from the PC via the firewire cable. If the ADVC110 is being run with an external power supply, then that goes doubly.

That dark band is most likely a combination of your cards (rather poor) comb filter and a few other factors.

[signal]

Capturing with an ADVC110 is even bad.Woah Nelly! If you are still getting the interference shown in your initial clip..

I should have clarified that better. It isn't as bad as the inital clip, but still seeing noise and ghosting.

Here's a sample of the avisynth colorbars being fed from a Matrox APVe to the ADVC.

m2advc-1.avi (DV 17MB) (http://www.pinesalad.net/example/m2advc-1.avi)

Snapshot from above:
http://www.pinesalad.net/example/m2advc-1.png

...then the problem is most likely not your PC itself, even if the ADVC110 is being run with power supplied from the PC via the firewire cable. If the ADVC110 is being run with an external power supply, then that goes doubly.

That dark band is most likely a combination of your cards (rather poor) comb filter and a few other factors.

I tried both an external power supply with 4 wire firewire (no power) but the grounds are still there to the PC.

More than happy to admit it's something external to the PC (i.e. the video sources themselves), but can't seem to nail that down either.

Everything looks great going to the JVC or other monitor but seems to get dragged down as soon as the signal is in a path with the capture card(s). This is the same when the output of the APVe is put through the JVC monitor to the ADVC110 (a test I descirbed in my last post).

And unless I'm doing this wrong, I can't find a ground potential difference between the shield on the video source cable to the ground on the PC.
I'm using a Fluke 77 to measure voltage across the two and using the scope to see if any voltage is on the ground.

I see what you mean about the comb filter and going to try some of the chroma settings in the BT8x8 tweaker with the Osprey to see if that helps.

[DrP]

While there is a bit of noise in your last sample avi, it's pretty darned good and I honestly doubt you'll get much better in terms of noise without stepping up to some fairly expensive equipment (for the source and acquistion devices).

The variation in the video seems to be +/- 2 at most when viewing the picture as RGB which is fairly quiet IMO.

[signal]

I guess I've been staring at the histograms too much in that the variations seem similar to that of the last osprey clip.

Any suggestions for the ghosting?