BrightEye 75 first lossless capture

[WaxCyl]

Here's a quote from the BE75 manual:

Quote:

ITU-R 601
This is the principal standard for standard definition component digital video. It
defines the luminance and color difference coding system that is also referred to
as 4:2:2. The standard applies to both PAL and NTSC derived signals. They both
will result in an image that contains 720 pixels horizontally, with 486 vertical
pixels in NTSC, and 576 vertically in PAL. Both systems use a sample clock rate
of 27 Mhz, and are serialized at 270 Mb/s.

I was a bit surprised to find that the NTSC SDI vertical resolution was 486, but then I am in PAL land.

I prefer to de-interlace and upscale my home videos as I believe QTGMC and NNEDI3 are superior to the hardware processing built into TVs. By retaining the 50Hz QTGMC output ensures no loss of any temporal information at all and increases the efficiency of post noise reduction. I use NNEDI3 to upscale to 720P50. Actually, I upscale to 960 x720 and add borders to make it 1280x720. This can be put onto a Bluray disc using lossless H264.
Because I am doing a fair amount of video processing I think this more than justifies retaining 10bit resolution up until final encoding. VapourSynth uses all 16bit internal processing by default for a very good reason; retaining the signal quality. High bit depth is also used in commercial fields for the same reason.
Arguably 10bit may be overkill for you Latreche34 as you have elected to capture as-is with no de-interlacing. Please note: retaining 10bit video will compress more efficiently than 8bit (contrary to Lord Smurf)

[latreche34]

Yes I kind of figured that out, 10bit option is still on the table but I will use it for important stuff where quality matters.

[dpalomaki]

480 vs. 486.

DV chose 480 because it is MOD 8 which digital likes, saves space, and the 6 lost lines are in the overscan area anyway and with most analog sources don't contain any watchable information (and are well outside the nominal action safe area). 486 is only mod 2 which fits interlaced.

I've seen some schematics from 1990's vintage gear with internal digital processing that used 6-bit A/D for the chroma signal read off tape. After A/D they used 8-bit processing and then 6-bit D/A for final analog output C.

[latreche34]

Yes I do get rid of the 6 extra lines but after capturing unfortunately since it is hard coded in the BE75 chips. It is part of the conversion script to mp4 though so it doesn't require an extra step to get rid of those lines.

[hodgey]

Quote:

Originally Posted by latreche34

Anyway, I captured a full tape and I'm doing the conversion to MP4 using this commend while cropping off the extra 6 lines:

ffmpeg -i Input.avi -filter:v "crop=720:480:0:0" -aspect 4:3 -b:a 192k Output.mp4

Maybe you just left out the other details there, but if not: If you just do this the mp4 will be encoded as progressive and not interlaced. Also I think it will use whatever color space the input is in, which may not be what you want.

You can encode interlaced h.264 like so:

ffmpeg -i Input.avi -filter:v "crop=720:480:0:0" -c:v libx264 -flags +ildct+ilme  -top 1  -aspect 4:3 -b:a 192k Output.mp4

Change to:
-top 0 if it's bottom field first (like DV).

If it's a final output file you may also you may also want to specify

-pix_fmt yuv420p

before the output file to get the common 4:2:0 chroma. Otherwise you may get a 4:2:2 or 4:4:4, possibly with 10-bit which is not as widely supported.

[latreche34]

I progressed little bit since that post, I've been using the following ever since, I will give yours a try:

ffmpeg -i "input.avi" -vf "crop=w=720:h=480:x=0:y=0,scale=w=-1:h=-1:interl=1,format=yuv420p,setsar=sar=10/11" -flags +ildct+ilme -c:v libx264 -crf 18 -x264opts bff=1:colorprim=smpte170m:transfer=smpte170m:colormatrix=smpte170m:force-cfr -c:a aac -b:a 192k "output.mp4"

Could you examine it and let me know which is better? I got it from one of VH members after posting for help there.

[traal]

Quote:

Originally Posted by sanlyn

you used zero signal input level control

The video seems dark, and the histogram reflects this, but is there a more objective way to determine that the signal input level was too low, after the fact?

More generally, how would you determine whether a VHS capture was a good one, in case someone chooses to send their tapes out instead of procuring all the necessary equipment to do it themselves? Starting with the codec they used (HuffYUV vs. DV vs. MPEG-2 or H.264). What are some things to look out for?

[lordsmurf]

Quote:

Originally Posted by traal

The video seems dark, and the histogram reflects this, but is there a more objective way to determine that the signal input level was too low, after the fact?

No, there is not. Sometimes it's easy to point out color flaws without realizing the condition of the ingest (the source tape). But it's also why proc amp can be an important part of workflow. This is a variable that must be accounted for in capture conversation that include coloring/IRE issues. Tapes are often all over the place. I've been accused of "bad levels" before, but the person didn't yet understand the tape was atrocious. The bad levels were actually "good levels" after the proc amp.

[hodgey]

Quote:

Originally Posted by latreche34

I progressed little bit since that post, I've been using the following ever since, I will give yours a try:

ffmpeg -i "input.avi" -vf "crop=w=720:h=480:x=0:y=0,scale=w=-1:h=-1:interl=1,format=yuv420p,setsar=sar=10/11" -flags +ildct+ilme -c:v libx264 -crf 18 -x264opts bff=1:colorprim=smpte170m:transfer=smpte170m:colormatrix=smpte170m:force-cfr -c:a aac -b:a 192k "output.mp4"

Could you examine it and let me know which is better? I got it from one of VH members after posting for help there.

This one is probably better, provided the field order is correct. It takes into account a few more things.

[WaxCyl]

Quote:

Originally Posted by latreche34

I progressed little bit since that post, I've been using the following ever since, I will give yours a try:

ffmpeg -i "input.avi" -vf "crop=w=720:h=480:x=0:y=0,scale=w=-1:h=-1:interl=1,format=yuv420p,setsar=sar=10/11" -flags +ildct+ilme -c:v libx264 -crf 18 -x264opts bff=1:colorprim=smpte170m:transfer=smpte170m:colormatrix=smpte170m:force-cfr -c:a aac -b:a 192k "output.mp4"

Could you examine it and let me know which is better? I got it from one of VH members after posting for help there.

You are encoding with BFF. Why is this?

[latreche34]

Quote:

Originally Posted by WaxCyl

You are encoding with BFF. Why is this?

I don't know what it means and I don't know why, The whole commend was written for me based on a lossless sample I've provided over at VH. Is there a better way?

[traal]

Quote:

Originally Posted by WaxCyl

Arguably 10bit may be overkill for you Latreche34 as you have elected to capture as-is with no de-interlacing. Please note: retaining 10bit video will compress more efficiently than 8bit (contrary to Lord Smurf)

If you use the capture card's built-in proc amp to compress the histogram into the safe zone for capturing, then after capturing decompress the histogram for RGB output, won't this fragment the histogram and cause banding/posterization unless you captured at more bits than your final output file supports?

[lordsmurf]

Quote:

Originally Posted by WaxCyl

Arguably 10bit may be overkill for you Latreche34 as you have elected to capture as-is with no de-interlacing. Please note: retaining 10bit video will compress more efficiently than 8bit (contrary to Lord Smurf)

- What does deinterlacing have to do with bit depth?
- Why would you deinterlace on capture?
- Why do you think 10-bit makes a difference when a VHS/S-VHS/Video8/Hi8 source is closest 6-bit dithered at most? Not even 8-bit.

[latreche34]

By the same argument VHS is only 30 lines of chroma so it doesn't deserve 4:2:2 capture, But I think both of you have valid points and it should be a middle ground somewhere for the most part, Going extreme is overkill and doing a sloppy job is not worth the effort. But at the end of the day it is the importance of the material what decides in which direction someone can take the task.

[lordsmurf]

Quote:

Originally Posted by latreche34

By the same argument VHS is only 30 lines of chroma so it doesn't deserve 4:2:2 capture

But 4:1:1 is not adequate to acquire the color fidelity, and 4:2:0 is right at/under. VHS is probably a different sort of x:x:x ratio, but not one I care to measure for.

[latreche34]

I totally agree.

[WaxCyl]

Quote:

Originally Posted by latreche34

I don't know what it means and I don't know why, The whole commend was written for me based on a lossless sample I've provided over at VH. Is there a better way?

I do apologise Latreche34, my mistake. I checked google, I did not know this NTSC is BFF or lower field first! I thought you had it wrong...
The BE75 output for me at least is TFF - top field first (PAL)

[WaxCyl]

Quote:

Originally Posted by lordsmurf

- What does deinterlacing have to do with bit depth?
- Why would you deinterlace on capture?
- Why do you think 10-bit makes a difference when a VHS/S-VHS/Video8/Hi8 source is closest 6-bit dithered at most? Not even 8-bit.

Deinterlacing has nothing to do with bitdepth. I have talked about maintaining high bit depth throughout my processing chain (which includes deinterlacing). Pinterf over at Doom9 Forum has gone to an awful lot of trouble to enable high bit depth support for all the important AVIsynth plugins. Why would I convert back to 8bit just to do deinterlacing? Any filtration that uses Fourier Transformation may well benefit from high bit depth (eg noise reduction)

I have never deinterlaced on capture, why would anyone do this?

My Panasonic VCR (NV-HS1000) uses an 8 bit D-A. I can provide the CCT diagram and IC data sheet if you would like. The chip is a Hitachi HA-19211BNT. It has 4 X 6bit DAs internally (=24bit) with 8bit output. The chip is Y only. C goes to another chip which may well be 6 bit as claimed above .

Quote:

Originally Posted by dpalomaki

I've seen some schematics from 1990's vintage gear with internal digital processing that used 6-bit A/D for the CHROMA signal read off tape.

[WaxCyl]

Quote:

Originally Posted by latreche34

By the same argument VHS is only 30 lines of chroma so it doesn't deserve 4:2:2 capture, But I think both of you have valid points and it should be a middle ground somewhere for the most part, Going extreme is overkill and doing a sloppy job is not worth the effort. But at the end of the day it is the importance of the material what decides in which direction someone can take the task.

.
I believe you are correct about the chroma resolution, but I think it is only for the horizontal direction. Vertically the chroma resolution is half the Y resolution. Capturing at 4:2:2 IS overkill but guarantees no resolution loss!

[WaxCyl]

Quote:

Originally Posted by traal

If you use the capture card's built-in proc amp to compress the histogram into the safe zone for capturing, then after capturing decompress the histogram for RGB output, won't this fragment the histogram and cause banding/posterization unless you captured at more bits than your final output file supports?

The BE75 does have basic proc amp controls and pedestal adjustment etc. This processing is all internally performed at 12 bit resolution. There are number of circumstances where higher bit depth has big benefits. Colour space conversion ( to/from RGB etc) would be just one of them. 10 bit is really worth it! 1024 shades of grey= no banding (or at least no de-banding)