Missing pixels with s-video vs. firewire capture Sony DCR-TRV460

[theseeker2]

When I capture a Hi8 tape shot in 16:9 aspect ratio via s-video on my Sony DCR-TRV460, everything looks great except pixels are missing on the left and right sides versus capturing with firewire.

Here's an imgsli for easy comparison: https://imgsli.com/MjQwNTgz.

Notice the top left of the thumb cut-off in the s-video version on the left, and the lower balcony continuing on further in the firewire capture on the right.

Here's the firewire capture:


[Attachment 77070 - Click to enlarge]

And here's the same frame with s-video capture:


[Attachment 77069 - Click to enlarge]

I've tried capturing with the TBC setting in the DCR-TRV460 menu turned on and off, and it doesn't make a darn difference. The pixels are cut off on the left and right no matter what.

I've Googled my brains out for the last few hours trying to find another soul who has encountered this issue, but so far it seems to be just me!

Is there any way to avoid this? I do notice that the firewire capture has some kind of noise on the left side, but not on the right, and it's clear that pixels are being cut off on both sides of the s-video capture.

[Bwaak]

I would not care much about a couple of pixels on this or that side, since 3-5% is supposed to be hidden by a TV edges anyway.

But wow, what a difference. Unless DV is just a single field, how blocky does it look. And the levels are raised compared to SVideo, I noticed the same effect too with my Digital8 camcorder. Also, there is bluish artefact on the left edge. Huh.

[theseeker2]

Agreed about the missing pixels not being as big of a deal as the crap picture! But still wondering how the firewire picks up those "extra" pixels and the s-video doesn't. These are both captured interlaced but I screenshotted the deinterlaced playback.

[Alwyn]

...

[dellsam34]

It's your capture device, Some capture devices are known to have out of calibration raster and they don't respect the standard, Firewire usually conforms to rec.601 standards, Are you using an easycap by any chance or one of the modern capture devices?

[theseeker2]

You should have asked. The DV transfer is with a Panny GS25 (SD MiniDV format).

[Attachment 77071 - Click to enlarge]

So I guess it's a standard issue among multiple devices and formats. Good to know and thanks for sharing!

[theseeker2]

It's your capture device, Some capture devices are known to have out of calibration raster and they don't respect the standard, Firewire usually conforms to rec.601 standards, Are you using an easycap by any chance or one of the modern capture devices?

I'm using an ATI 600 USB on Windows XP. Maybe I need different drivers?

[lordsmurf]

This is a terrible comparison, and zero findings can be made.

- one image is lighter, the other darker --- use a proc amp
- other factors, like lack of frame TBC, can weigh on the captures
- the "missing" data is edge data, and it appears affected by chroma offset in the "more", but cropped to full actual image on the "less"

I actually find the values of the ATI 600 USB to be closer to reality here in the first post example. If you want "more detail" (false name here), then just lighten the image.

[theseeker2]

This is a terrible comparison, and zero findings can be made.

- one image is lighter, the other darker --- use a proc amp
- other factors, like lack of frame TBC, can weigh on the captures
- the "missing" data is edge data, and it appears affected by chroma offset in the "more", but cropped to full actual image on the "less"

I actually find the values of the ATI 600 USB to be closer to reality here in the first post example. If you want "more detail" (false name here), then just lighten the image.

For the S-Video capture, I used the histogram in VirtualDub to set the brightness and contrast such that the values were mostly in the non-red range. For this tape, that meant brightness=100 and contrast=32. It does seem a bit dark, but if I bump up even to brightness=101, then I get a little red on the far right, and at brightness=102 I get lots of red on the far right. Should I go into the red a bit for this tape? There is another banister in the far background that is just barely visible in the S-Video version that is quite visible in the Firewire capture.

For the firewire capture, how would I adjust the brightness and/or contrast since it's a direct capture into WinDV with the Firewire cable? Is there a proc amp type adjustment on the camera itself? Or is there a proc amp that takes firewire input? It does seem overly bright.

[Skiller]

With few, usually older (XP era) capture devices a capture width of 704 is the only correct one. If unknowingly you chose 720 nevertheless, you silently get a software based resize from 704 to 720 by the driver.

If done correctly, a properly captured 704 frame will have the exact same proportions when centered over a 720 firewire capture, minus the missing 16 pixels at the sides (which are often black anyways).


Edit: Looking closely at the OP's screenshots, this is apparently not the case here (proportions do line up), but possibly the cause for the discrepancy Alwyn is having.

[Alwyn]

For the S-Video capture, I used the histogram in VirtualDub to set the brightness and contrast such that the values were mostly in the non-red range. For this tape, that meant brightness=100 and contrast=32. It does seem a bit dark, but if I bump up even to brightness=101, then I get a little red on the far right, and at brightness=102 I get lots of red on the far right.

Brightness controls the overall position of the histogram so yes, if you increase the brightness, the right edge will go "more red". In that case, just lower the contrast a bit, which controls only the right edge.

The analogue capture looks good to me; I wouldn't recapture: I'd adjust the levels with Colormill in VDub. With it, you can boost only the darks and leave the brighter areas alone. If you simply use the histogram, you'll brighten the whole screen up.


[Attachment 77084 - Click to enlarge]

As for the DV capture, definitely worse in this case. DV struggles with low light and AFAIK there isn't any way of adjusting the levels until post-capture processing.

[theseeker2]

With few, usually older (XP era) capture devices a capture width of 704 is the only correct one. If unknowingly you chose 720 nevertheless, you silently get a software based resize from 704 to 720 by the driver.

If done correctly, a properly captured 704 frame will have the exact same proportions when centered over a 720 firewire capture, minus the missing 16 pixels at the sides (which are often black anyways).


Edit: Looking closely at the OP's screenshots, this is apparently not the case here (proportions do line up), but possibly the cause for the discrepancy Alwyn is having.

Everything I've ever read about the ATI USB 600 says that 720x480 is the native capture resolution, so your original post had me worried! Glad to see your edit about the proportions lining up.

I did have an issue with the ATI AIW 7500 AGP card and 704 vs 720, so I went back to my ATI USB 600 so I didn't have to worry about silent software resizing issues.

[theseeker2]

For the S-Video capture, I used the histogram in VirtualDub to set the brightness and contrast such that the values were mostly in the non-red range. For this tape, that meant brightness=100 and contrast=32. It does seem a bit dark, but if I bump up even to brightness=101, then I get a little red on the far right, and at brightness=102 I get lots of red on the far right.

Brightness controls the overall position of the histogram so yes, if you increase the brightness, the right edge will go "more red". In that case, just lower the contrast a bit, which controls only the right edge.

The analogue capture looks good to me; I wouldn't recapture: I'd adjust the levels with Colormill in VDub. With it, you can boost only the darks and leave the brighter areas alone. If you simply use the histogram, you'll brighten the whole screen up.


[Attachment 77084 - Click to enlarge]

As for the DV capture, definitely worse in this case. DV struggles with low light and AFAIK there isn't any way of adjusting the levels until post-capture processing.

Thanks for the analysis, Alwyn! I really appreciate the info.

[theseeker2]

I just did another s-video vs firewire capture comparison with another tape, this one shot in the default 4:3 instead of the "16:9 Full" mode from the previous tape that I posted pictures of. There are still more pixels on the left and right in the firewire than the s-video, but there's no blue artifact on the left, although the overall picture quality is still quite blocky and grainy.

It's too bad I'm losing pixels on the sides with the s-video capture, but I'll take it over the firewire ick capture.

[Bwaak]

I am really surprised DV looks that bad. When you say it is a screenshot what do you mean? Is it a Print Screen sort of thing or did you save a frame from VLC or Splash or whatever media player you use?

[oln]

I saw the same symptom in a post about a PAL d8 camcorder about a different issue recently, wasn't sure if it was their capture device that was doing it or not but maybe it is the case that the digital8 camcorders does crop the image a bit on the left side on the analog outputs even with analog tapes. It is a common thing with analog outputs on dv camcorders and dvd players to have 8 px blanking but seems it may be a thing here too. Don't remember it being present on the tbc/dnr equipped hi8 models but not sure. None of my d8 camcorders are in a functional state so can't investigate further

Most TBCs (both internal and external units) also do this since they tend to somewhat adhere to the standards for the blanking area when generating the output from the internal digital component signal which is a little annoying in case of tapes with video in this area. Sometimes also using very sharp borders which can result in some ringing artifacts but since it was all in the overscan area on a crt it wasn't a big concern when these devices were relevant.

[dellsam34]

It can't be the camcorder, it's the same output from TBC that goes to both the DAC and the DV codec chip, If DV was the one that is cropped it would make more sense since it has undergone one processing step into DV, This is not the case here, It's the capture card.

[theseeker2]

I am really surprised DV looks that bad. When you say it is a screenshot what do you mean? Is it a Print Screen sort of thing or did you save a frame from VLC or Splash or whatever media player you use?

I captured it with print screen, but I also just tried saving the frame in VLC which is what I use and it looked the same. It is bad! Is it possible that my firewire cable is defective? I would think if so that it just wouldn't work at all.

I'm capturing the firewire using WinDV which doesn't really have any settings other than type 1 or type 2, and I'm using type2.

[theseeker2]

I saw the same symptom in a post about a PAL d8 camcorder about a different issue recently, wasn't sure if it was their capture device that was doing it or not but maybe it is the case that the digital8 camcorders does crop the image a bit on the left side on the analog outputs even with analog tapes. It is a common thing with analog outputs on dv camcorders and dvd players to have 8 px blanking but seems it may be a thing here too. Don't remember it being present on the tbc/dnr equipped hi8 models but not sure. None of my d8 camcorders are in a functional state so can't investigate further

Most TBCs (both internal and external units) also do this since they tend to somewhat adhere to the standards for the blanking area when generating the output from the internal digital component signal which is a little annoying in case of tapes with video in this area. Sometimes also using very sharp borders which can result in some ringing artifacts but since it was all in the overscan area on a crt it wasn't a big concern when these devices were relevant.

That was an interesting thread, thanks for the link. In their case, pixels on the left were cut off, but not on the right. In my case, pixels on both sides are cut off. Of course, theirs was PAL and mine is NTSC, so maybe that accounts for the difference.

[theseeker2]

It can't be the camcorder, it's the same output from TBC that goes to both the DAC and the DV codec chip, If DV was the one that is cropped it would make more sense since it has undergone one processing step into DV, This is not the case here, It's the capture card.

Maybe there's some kind of driver issue? I just checked and my driver is from 8/31/2007 and is version 5.7.831.0.

I'm capturing with VirtualDub using the settings found this guide: https://www.digitalfaq.com/forum/video-capture/7427-capturing-virtualdub-settings.html.

[poisondeathray]

The DV transfer is with a Panny GS25 (SD MiniDV format).

Your screenshot is quite different than the OP's. All borders are affected in yours. Similar to a zoom . Unless you or the software cropped/resized /did something else for that screenshot

[Bwaak]

I captured it with print screen, but I also just tried saving the frame in VLC which is what I use and it looked the same.

Can you share a small DV-AVI clip with this particular shot that you compared above?

[Skiller]

If it's really caused by the capture device, the blanking would have to be about the same with any signal source connected.
Try capturing anything else you have at hand as a test. VHS, DVD, game consoles, other camcorders...


(My educated guess: It's the DCR-TRV460 adding more blanking on it's analog outputs.)

[Alwyn]

Your screenshot is quite different than the OP's. All borders are affected in yours.

Fair call. Here is a completely unretouched comparison, DV verses analogue GV-USB2. The pillarboxing (if you could call it that) of the S-Video version is pretty much blanking the area shown by the DV version.

[poisondeathray]

Your screenshot is quite different than the OP's. All borders are affected in yours.

Fair call, same result though. Here is a completely unretouched comparison, DV verses analogue GV-USB2. The pillarboxing (if you could call it that) of the S-Video version is pretty much blanking the area shown by the DV version.

Not the "same" result in terms of the previous screenshot. The 2nd screenshot is more like what I would expect . The top and bottom are a lot closer . Compare that to your 1st screenshot , where the top and bottom are cropped off and it gives the appearance of zoom in or smaller FOV. The 2nd screenshot is more similar case to the OP's - notice the OP's top and bottom are very close . My point is your 1st screenshot is misleading

<deleted by request>

[Skiller]

The pillarboxing (if you could call it that) of the S-Video version is pretty much blanking the area shown by the DV version.

Which is perfectly normal and to be expected. The DV frame carries a longer scanline than what the length of the active picture in an analog video signal shall be. To put it into numbers, the DV frame carries a length of 53.333 µs (microseconds) worth of picture in one scanline, but only 52 µs of those are needed (in 50 Hz systems) for outputting S-Video.

[Alwyn]

My point is your 1st screenshot is misleading

OK ok. To remove any doubt, I've deleted it, and I would appreciate you doing the same.

[Sharc]

Maybe it is simply a "feature" of the particular capture device to crop a few pixels left and right into the picture, considering that the left and right borders of standard analog captures are often fuzzy and discolored.

[Alwyn]

Maybe it is simply a "feature" of the particular capture device to crop a few pixels left and right into the picture

This is occurring with the GV-USB2 as well. See post #24.

[Sharc]

Also there are differences with consumer DV videocams as far as I remember from previous tests:
1st generation followed Rec.601 as per the SONY DVCAM specs, means active picture is ~702....704 wide, PAR (PixelAspectRatio) =12/11 (for 4:3 PAL). Some just filling left and right to 720 with picture content, making the overall DAR 1.36 (aka 15:11). 16:9 - if available - was 16:9 in a 4:3 frame by masking top and bottom.
Later generation(s) supporting 4:3 and 16:9 violated Rec.601, making the active picture 720 wide with a 'generic' PAR of 16/15 for 4:3 PAL and 64/45 for 16:9 PAL.

Capture devices for analog capturing (cards, dongles) are supposed to follow Rec.601, means to sample the analog video luma signal at 13.5MHz with an active picture width of ~702...704 pixels, PAR=12/11 for PAL (in very good practical approximation).

Sorry if going off topic as it doesn't explain the extra cropping/masking of the s-video (which has been my original intention) .....