Color issue with TBC CDM-660?
 

Hello!

I have a CDM-660 TBC, bought from Lord Smurf. My setup is:

Panasonic NV-HS1000 -> CDM-660 -> AJA KONA LHe. All connections over S-Video.

When I put the CDM-660 in the chain, I have a strange color issue. The colors get greenish/yellowish. I also believe to see the image is a little more blurry than without TBC, but that may be an effect of the color problem.

Here is a frame with TBC:

Attachment 13787

And here is the same frame without TBC:

Attachment 13788

Is this normal for CDM-660 or does my unit have an issue?

If it is normal: as it is essentially a clone of the green AVT-8710, would I have to expect the same effect from a genuine green AVT-8710? And what about a TBC-1000?

Thanks for help!

Neither image is very good. It appears to be the tape source.

What I see is:
- without-TBC image is blue (cyan) shifted, very overexposed with illegal values (beyond 16-235)
- with-TBC truncates down, and overcorrects WB a bit yellow

This is "essentially" a clone of the AVT-8710, but not 100% identical. Not a rebadge, not a more direct clone like the 1T-TBC. It's PAL only, as NTSC has cyclical interlacing flaws.

The TBC appears fine here, simply reacting badly to sources. Timebase correctiob can alter colours, as a sideeffect of correction.

This unit isn't blurry. It is a tiny bit softer, as it does internally process composite, even though connected svideo. But slight softness isn't blurriness, completely different concepts.

Remember, this is a budget TBC, less than half the cost of most others.

It also needs to be said that this is a very specific version of the CDM-660, from a very specific timeframe. Like other Cypress units, flaws chipsets infest production, so you don't want just any random CDM-660. The cheap units that exist are not what you want.

Thanks for the answer.
It's a tape which was recorded in 1994 and stored in bad conditions since then, so yes, the tape source is not ideal...

So, if I understand correctly, the TBC gets confused because the VCR outputs a signal which is out of bounds (overexposed, illegal values)? If yes, is there something I can do to improve the signal quality before the TBC? Would a proc-amp be helpful?

Anything before the TBC can affect the TBC. So here, it's the tape, and the VCR.

- So tape not great, as mentioned.
- What VCR? Some VCRs don't play well, don't out put the signal well -- that that includes good models that have deteriorated due to age, lack of maintenance, or abuse.

Yes, proc amp would help. (I want to re-test my extra SignVideo, see if it does PAL.)

I have tested it with a Panasonic NV-HS1000 and also a JVC HR-S9600, the yellow shift appears with both VCRs.

By the way, the Panasonic also adds some slight yellow tint when Line TBC is on.

If that had been an AG-1980, I'd say caps were failing. But being a PAL Panasonic, I'm more led to believe it's definitely the tape timing (time base) as originally suspected.

Essentially, welcome to video. Stuff like this is the norm, not the exception. Very often, correcting one aspect slightly (sometimes rarely massively) worsen another. So you have to compensate with more devices. That's not your or the equipment, as everything you're using is known to be quality, that's just the nature of consumer analog videotape formats like VHS.

Had it been random not-suggested gear, then we'd be looking at the gear to trace the fault. I know the TBC is fine, I tested it myself. And your PAL Panasonic and JVC S-VHS decks are likely fine. So we can rule those out.

What's left is the tape, and the wires. Is this all wired s-video? Have you tried different cables?

Yes, everything is wired S-Video. The cables are new "Contrik" (Neutrik) high quality cables.

For the connection between TBC and capture card, I have only one S-Video to 2x BNC cable, so I can't check another one right now.

For the connection between VCR and TBC, I have either "Contrik" high quality 1 m cable or cheap "Value" 5 m cable. Interchanging them does not seem to change the color issue.

So you're not using s-video cables, but BNC>svideo cables? Those often have issues, almost always cheaply made.

That being a standard converter make sure you are not converting to NTSC and capturing it, Make sure it is set to output PAL.

Well, the AJA KONA LHe breakout cable has only BNC inputs. So the cable from TBC to capture card has to be a S-Video to BNC cable. I hope it's a good quality cable: Contrik Y/C-S4P-2BJ-0.3

To rule out this cable as the source of error I hooked up a cheap USB capture stick (Roxio something...) which has a S-Video input, so I can use only S-Video to S-Video cables. I can see the same yellow shift when TBC is connected (see attachments). I'm quite confident the S-Video to BNC cable is ok.

Yes, the TBC is set on PAL output mode (double checked).

What software are you using to control the Aja card? I know Jwillis messed with those cards and I think he mentioned that if used with the right software it has a frame synchronizer, So you could capture directly to it, What problems you had that led you to using a frame TBC? do you have audio/video sync problems?

I am using Aja's own software "AJA Machina".

I don't have audio/video sync problems anymore. In fact, I had a/v sync issues before, with the cheap USB capture stick, but they are gone with AJA KONA. I think they react differently on bad source frames. The USB capture device seems just dropping them (with each skipped frame, the audio sync offset increased).

With AJA KONA, the audio always stays in sync with video, but on the other hand, bad frames on tape introduce bad artifacts in the video file. I see some (or parts of) old frames flashing up here and then. For me, it looks like the KONA takes whatever was still in its RAM buffer at the place were the new frame should be. This creates "jumps" in the video file. I have attached a short example where you can see 4 of such "flashes" (frames 11, 17, 66 and 72). For example, frame 17 appears to be a copy of frame 2. Or in frame 66, a part of frame 53 or 52 is blended over the frame 66 (a single field?). With other (also bad, but not as bad) tapes, some old frames were copied over the actual frame, but with a spatial offset, e.g. the lower left corner of an old frame displayed on the upper right corner of the actual frame.

This is why I decided to buy a frame TBC. I wanted to pass a clean signal to the AJA Kona. And in fact, the flashing artifacts are gone with the frame TBC, so besides the color shift and the slight softness, the video file looks good. But as Lord Smurf writes, correcting one issue can introduce a new one. This is now my case. And I'm trying to resolve this new problem...

If there's a possibility to make the Kona handle bad frames in a more graceful way (e.g. take the last good frame, not one which is 15 frames behind in time), that would be very nice, so I would get rid of the color problem. But I guess that's not possible...

I just tried two other tapes. The yellow cast also appears with them when passed through the TBC. So it seems not only my first tape being affected. See the attached frames. In the "tape 02" examples, the effect is very pronounced for my eyes.

In "tape 03" it looks much better, though there's still a slight yellow shift (or isn't there?), but the colors look more natural for me. But there, the TBC seems to make a strong exposure/saturation correction.

I didn't expect a frame TBC to change the color/exposure/saturation so much.

Not every sample frame there is shifting yellow.

For example, tape 3, the original has a purple/brown muddy casting. The post-TBC version is more properly lit, colors pulled out, scene lit better.

In tape 2, your source is blue/cyan tinted, and the black is gray and muddy. The post-TBC version is a tad overdone, now has a sunburn, colors a tad too saturated now. And the saturation pulls out a bit more chroma noise as a result. Neither source nor post-TBC are perfect here, either proc amp and post-capture color correction is needed. If it were my footage, I'd have a proc amp somewhere between the extremes. I'd put back saturation, yet adjust contrast/black down only slightly.

Tape 1 is yellow due to intensifying green.

Overall, this is unexpected behavior. I don't have a huge collection of PAL tapes, as I do for NTSC tapes. In all of my testing, color was never augmented, and that included some homemade LP mode PAL tapes from Australia. Using JVC S-VHS decks with line TBC, all HR series. (NTSC, however, is a problem with these units, both the good and bad version. It created interlacing artifacts, and cyclical jumping.)

I've had a few of these over the years, and previous buyers never saw this (or at least never said anything). So I still wonder if other factors are in play. Capture card, VCR, wiring, even power and power grids. Negative workflow interactions happen, which is a main reason I provide whole workflow (VCR>TBC>capture), though mostly only NTSC. Workflow problems are probably the most frustrating issues I've run into over the years, especially after major gear moves.

Proc amp is the fix here.

Just in case you're doubting this unit, follow the money. It cost less than half of most TBCs. So you can add a great proc amp, and still be below costs of the alternatives. Too many people fail to understand video, and look for a single piece to resolve all problems. Doesn't work that way. Compensate, compensate, compensate, until quality exists, and output is satisfactory. And after adding proc amp, you should be fine.

Proc amps are great, I often suggest adding one anyway. :)

Ok, thanks. Well, so I think I'll have to find a PAL Proc Amp with S-Video in/out. Do you have any recommendations?

In my opinion that frame jump is a software/driver issue, Do you have another computer to try with? Are you on a Mac? If not try AmarecTV.

For now, I don't have another computer to try with, unfortunately. I am running the AJA Kona LHe on Windows 7.

I don't know if AJA would agree on this being a software/driver issue, as it works well with clean signals. I already captured a whole tape without a single bad frame.

AmarecTV seems to work with DirectShow devices. If I'm not wrong, the Kona LHe does not work with DirectShow (LHe Plus would, but not the version without plus: https://www.aja.com/support/directshow)

Do you think the jumping frames would not occur with the mac software/driver for AJA Kona LHe?

While waiting to find a suitable proc amp I did some further "research". I tried to find out what the CDM-660 is doing to the color.

For that, I grabbed frames from different tapes in a lossless YUV format (V210, yuv422p, 10 bits per channel). For each tape, I made sure to capture the very same frame once with and once without tbc. Then, I aligned them (image registration) as good as possible, so that every pixel of the frame without tbc corresponds to the same pixel in the frame with tbc. Then i cropped them (because the tbc adds a black border) and scaled them down, so I have two 10bit YUV images which should differ only in luma and chroma (well, close to. the tbc also seems to do some other manipulation, but let's say close enough).

Then I created some 2D-plots where I plotted the Y, U and V value "with tbc" vs "without tbc" for each pixel. For instance, if the U value for a pixel is 0.6 without tbc and 0.65 with tbc, a point will be plotted at (X: 0.6, Y: 0.65) on the U plot. If there were no change in a channel, all the points should all lie on the red line f(x)=x.

tape 02:
Attachment 13836

tape 02b:
Attachment 13837

tape 03:
Attachment 13838

tape 04:
Attachment 13839

What I believe to see on the plots:

• The Y (luma) channel gets processed linearly (e.g. like f(x) = a*x, like gain), but with a different factor depending on the tape (or scene). sometimes the gain is >1, sometimes <1.
• The U and V (chroma) channels seem also to be processed linearly, but always in the same direction (chroma gain >1).

It seems that the chroma gain is constantly too high on all my tests (3 different tapes, different sources).

Then, I manually searched for a function to correct the U and V channels, so that the points lie on the red line. And I found: u_corr = (u-0.02)*0.72 and v_corr = v*0.7 for all 3 tapes (and correct the y channel for each tape independently). Yes, I added a small offset to U for better accuracy. This is what I get:

tape 02:
Attachment 13840

tape 02b:
Attachment 13841

tape 03:
Attachment 13842

tape 04:
Attachment 13843

So, basically, the color shift seems more or less constant and thus can be corrected in the same way for all tapes.

What is strange, though, is that when the Line TBC of the VCR is switched on, i have to correct less, i.e. the tbc is more accurate. Still tinted, but slightly less. In the tape 01 example, I can't record without line tbc because of bad tape. I can then correct the color with u_corr = (u-0.01)*0.9 and v_corr = (v-0.01)*0.9:

tape 01:
Attachment 13844

So the chroma gain of the tbc seems to be dependent on whether the line tbc is on or off... Or maybe it's just the one tape in bad conditions. I'll have to make some more tests.

But anyway, my tbc seems to push on the chroma gain a little too hard. Does anyone know if there is a calibration possibility for the CDM-660? Is there a potentiometer inside to adjust the chroma gain? Can this be due to ageing?

I have no idea what those graphs are supposed to show. :question:

It may still be workflow conflicts, devices don't play nice. For example, VCR output can cause TBC to overcorrect. The line TBC altering this output raises an eyebrow.

Aging is unlikely.

However, values changes are possible, due to shipping, namely changes in elevation. This has happened to me with every move, and sometimes happens during shipping. There's about 15 electrolyte caps in this unit, and one could have shifted. There's no adjustments inside the unit.

Ok, thanks a lot. Anyway, apart from the small color shift the tbc works very well.