VHS to digital - help

[LMotlow]

You don't have to upload such long clips. Shorter clips with a mix of bright, dark, highly saturated colors, motion and stills would be better.

@freee right, I'm glad jagabo clarified that. Also, I didn't mean that you're forced to post an entire capture.

You can cut a sample from your DV-AVI without otherwise modifying or changing anything using VirtualDub.
Open the DV capture directly in VirtualDub.
Use the navigation and cut/edit tools at the bottom of the VDub window to edit your sample segment.
In VDub's top menu panel, click "Video..." -> "direct stream copy"
The click "Save as AVI..."

[freee]

You don't have to upload such long clips. Shorter clips with a mix of bright, dark, highly saturated colors, motion and stills would be better.

@freee right, I'm glad jagabo clarified that. Also, I didn't mean that you're forced to post an entire capture.

You can cut a sample from your DV-AVI without otherwise modifying or changing anything using VirtualDub.
Open the DV capture directly in VirtualDub.
Use the navigation and cut/edit tools at the bottom of the VDub window to edit your sample segment.
In VDub's top menu panel, click "Video..." -> "direct stream copy"
The click "Save as AVI..."

I have attached DV-AVI capture files.

[jagabo]

I don't see any dot crawl artifacts in the composite cap -- one of the biggest problems with composite. But there's not a lot of high saturation color (where dot crawl shows up) in the clip. The clip still has blown out brights. You need to see if you can adjust the capture device's proc amp settings to reduce that. I also saw a suspiciously large amount of macroblocking -- especially visible in the foliage behind the butterfly. Is your capture device really capturing DV? Or is it capturing MPEG video and converting to DV?

[freee]

I don't see any dot crawl artifacts in the composite cap -- one of the biggest problems with composite. But there's not a lot of high saturation color (where dot crawl shows up) in the clip. The clip still has blown out brights. You need to see if you can adjust the capture device's proc amp settings to reduce that. I also saw a suspiciously large amount of macroblocking -- especially visible in the foliage behind the butterfly. Is your capture device really capturing DV? Or is it capturing MPEG video and converting to DV?

I cannot adjust anything on the device as I am using a DV camera (analog pass-through functionality). It should be caputring DV, not MPEG and then converting to DV.

I have two TV cards that I can use for analog capture (Phillips saa7134 and one bt878). Perhaps I should try them and adjust the settings as per your advice. Is the Phillips saa7134 better than bt878 chipset?

I have just realised that s-video capture did not upload at all. I hope it will be attached now.

[Skiller]

Had a look at both samples. Can confirm what jagabo wrote: In the Composite capture the highlights are clipped. Interestingly both captures have their hot peak at 252, which is kinda odd – I would have expected the Composite capture to clip at 255, but instead it clips at 252 and has almost nothing beyond 252 even though there should be something like the S-Video capture clearly demonstrates.

I do not see any obvious MPEG compression artefacts. It's good old DV after all.


Edit: Looking at the samples again today, I realized the Composite capture has it's line jitter nicely stabilized by the camcorder while the S-Video capture, which was first digitized by the Yamaha, does not. Reason is the Yamaha does not stabilize line jitter as it seems, and after the video has been digitized once the camcorder cannot correct the original tape jitter anymore.
It's quite obvious with the top part of the video in the butterfly scene. In the S-Video cap it looks like jello while the Composite cap is stable. That's bad news because it means both capture routes have serious flaws – just different ones.

[LMotlow]

Capturing as was done e evarlier, the levels problems is more complicated than one would think. Below, frame 174 of the s-video cap (the composite levels are worse), the YUV histogram shows raised black levels and clear clipping at the right end (brights). Borders and head switching nosie are cropped off to prevent throwing off the dark (left) side of the histograms. The RGb histogram (right-hand histogram) shows brightness clipping early, with chroma climbing up the right-hand wall (chroma clipping).



Below, the same frame with everything brought toward the left-hand (darker) side using YUV contrast, offset, and slightly lower gamma. In the YUV histogram the bright cutoff is clearly visible and doesn't change, meaning that brights above a certain point are simply gone, nothing left to retrieve. In RGB the clipping points for both brightness and color are clearly seen and won't change shape. Clipping limits the dynamic range that can be achieved -- you can "stretch" the dark end and the bright end more if you want, but the high end and specular highlights will still be clipped and the image will look flat and washed out. Small "gaps" in the histogram indicate missing detail pixels, so further stretching will cause banding effects in large clear areas like big expanses of sky, water, snow, etc. Even with correction, the camera's auto gain during the pan creates problems of its own.


Below, the unadjusted butterfly shot is a little better off, but still some mild clipping in the bright wings. Again, black levels are off and high gamma makes the image look thin and pale. Can probably be repaired, but it'll be tricky when gamma and midtones are twisted up during capture.

[jagabo]

I thought I posted another response in this tread yesterday but it was in the wrong thread:

https://forum.videohelp.com/threads/368611-Edge-noise-when-TBC-is-active?p=2359321&view...=1#post2359321

[freee]

jagabo

I thought I posted another response in this tread yesterday but it was in the wrong thread:

https://forum.videohelp.com/threads/368611-Edge-noise-when-TBC-is-active?p=2359321&view...=1#post2359321

I experimented with the capture cards last night. The Phillips based card (Compro T300) appears to be superior to BT848 card, however I am getting some kind of noise with Compro T300. It apears as short horizontal lines that appear randomly throughout the capture (static like). I don't know how to remove it or how to call it. I will try different drivers but I am not sure if it will work. With Compro T300, I have full access to the driver in VDub but BT848 card's proc amp settings are not available (I am using a universal WDM driver that has been circulating in the Internet for 10 years and nothing new seems to be available). No horizontal lines issue with BT848. I will post examples tonight. I have only one PCI slot in my PC and I have to keep swapping the cards all the time.

LMotlow

Below, the same frame with everything brought toward the left-hand (darker) side using YUV contrast, offset, and slightly lower gamma. In the YUV histogram the bright cutoff is clearly visible and doesn't change, meaning that brights above a certain point are simply gone, nothing left to retrieve. In RGB the clipping points for both brightness and color are clearly seen and won't change shape. Clipping limits the dynamic range that can be achieved -- you can "stretch" the dark end and the bright end more if you want, but the high end and specular highlights will still be clipped and the image will look flat and washed out. Small "gaps" in the histogram indicate missing detail pixels, so further stretching will cause banding effects in large clear areas like big expanses of sky, water, snow, etc. Even with correction, the camera's auto gain during the pan creates problems of its own.
Attachment 28850

Perhaps a graduated ND filter or a polarizing filter would have helped on a day like that. It was a very hot summer day in Europe preceeded by a weeek of hot weather without a drop of rain. In the light of your assessment, would you recommend that I use the s-video converter in my Yamaha receiver rather than the composite input?

[Skiller]

With the two options we have examined yet it's "pick your poison". With the Yamaha (S-Video) route you have much less clipped highlights but there's no line jitter stabilization. The Composite capture has quite severe clipping but nicely straightened out lines (no jello effect in the top part of the butterfly scene).

[LMotlow]

Perhaps a graduated ND filter or a polarizing filter would have helped on a day like that. It was a very hot summer day in Europe preceeded by a weeek of hot weather without a drop of rain. In the light of your assessment, would you recommend that I use the s-video converter in my Yamaha receiver rather than the composite input?

Exposure problems aside on that one shot -- which sounds a bit odd because the camera's AGC was obviously enabled -- the buitterfly shot was made in subdued or overcast light but still looks overexposed. The type of cable used isn't the problem. The problem stems from not controlling levels during capture. Levels mismatch between source and capture chain is common with all capturing situations. Aside from buying a proc amp -- good ones are not cheap, and cheap ones will ruin your videos -- brightness and contrast can be controlled with the capture device software or with VirtualDub Capture's hook-in to graphic device image controls. Someone more familiar with capturing DV in the manner you're using will have to advise how to control certain parameters.

It's also possible that your captures demonstrate what the source really looks like.

[jagabo]

You can bring the bright level down post-capture and that will get the pixels above 235 down below 235 to conform with the spec. But that won't help with all the pixels that should have been brighter than 255 but are crushed at 255 because you can't have values greater than 255 in 8 bit per channel video. Those have to be adjusted with the video proc amp before capture.

[freee]

It's also possible that your captures demonstrate what the source really looks like

.
Spot on conclusion, this is not just a possibility but the fact. I am getting an impression that no marked improvements can be achieved.

[jagabo]

It's also possible that your captures demonstrate what the source really looks like

.
Spot on conclusion...

No doubt the signal on the tape is hot, but both videos are hard clipped at Y=252 (with just a little DCT ringing artifacts above that) even though they have very obvious differences in levels (ie, the composite cap is overall brighter than the s-video cap). I suspect you'll be able to get more bright detail using a capture device with a video proc amp. If you capture like you are now you will NEVER get those bright details back.

You said the Philips capture device has a functioning proc amp? Can you bring the brightness down using that? Adjust the proc amp to see if you can get more detail in those blown out skies.

[freee]

It's also possible that your captures demonstrate what the source really looks like

.
Spot on conclusion...

No doubt the signal on the tape is hot, but both videos are hard clipped at Y=252 (with just a little DCT ringing artifacts above that) even though they have very obvious differences in levels (ie, the composite cap is overall brighter than the s-video cap). I suspect you'll be able to get more bright detail using a capture device with a video proc amp. If you capture like you are now you will NEVER get those bright details back.

You said the Philips capture device has a functioning proc amp? Can you bring the brightness down using that? Adjust the proc amp to see if you can get more detail in those blown out skies.

I can definatelly bring the brightness down with the Phillips capture card. However, the card (or the driver) introduces those horizontal lines during the capture. I will post the results tonight when I get home.

[jagabo]

I can definatelly bring the brightness down with the Phillips capture card.

Yes, but is there any more detail in the bright areas? If the tape was saturated you will not be able to get any more detail. If it was just the DV camcorder that was overwhelmed you may be able to get more detail.

[freee]

What is causing these lines to appear? It only happens on the Phillips based card and only when capturing at 768x576. 720x576 looks good.

[Skiller]

IRQ conflict or just too much data for the controller. You probably do have other extension cards in use as well, right?

But capturing at 768x576 is pointless anyway. The sampling clock for the digitization of the analog signal is fixed to ITU-R BT.601 standard (maybe not with Bt8x8 cards and custom drivers, but certainly with SAA7134) so capturing at 768x576 will just resize the already digitized video to that.

The only correct capture res for your SAA7134 card is 704x576 (not 720). Comparing it to your 720x576 camcorder capture you can tell by the missing and smaller black sides that it is capturing the active 52 µs window of the analog signal which equals 704 pixels.

[jagabo]

What is causing these lines to appear? It only happens on the Phillips based card and only when capturing at 768x576. 720x576 looks good.

That sounds like a software problem. Keep in mind that VHS only has a resolution of about 350x576. So capturing at 704 or 720 is already way above what's on the tap. 768 isn't going to give you any more detail.