Uncompressed 8-bit vs 10-bit MOV to Huffyuv AVI?

[metaleonid]

Hello,

I have been playing with Black Magic Design Intensity Shuttle for Thunderbolt. Will probably use workflow:
LD player Pioneer Elite LD-S2 Composite out -> Panasonic DMR-ES25 Component -> Black Magic Design Intensity Shuttle for Thunderbolt.

BM has 3 options: 8 bit uncompressed MOV, 10 bit uncompressed MOV and DPX.

I have no idea what DPX is? I was planning on recording in either 8 bit or 10 bit uncompressed MOV and then convert it to Huffyuv AVI using AVISynth QTInput.

First off all for the source such as LD is 8 bit uncompressed MOV good enough? Or would 10 bit be more preferred?
Second question is: am I losing the quality if I perform conversion to Huffyuv AVI using QTInput? I don't want to keep uncompressed since it takes too much disk space. But I don't want to be losing video quality either.
Third - what the hell is DPX? Just sequence of images?

Thanks.

[msgohan]

Why not HDMI?

BlackMagic 10-bit uses the v210 pixel format, which isn't supported by Huffyuv.

DPX is used for post production work: http://en.wikipedia.org/wiki/DPX

[metaleonid]

Thank you. I couldn't make HDMI work yet. BM doesn't recognize the signal. I have ordered this http://www.amazon.com/gp/product/B00...?ie=UTF8&psc=1 as suggested. I will see if it's going to recognize HDMI.

I will go with 8 bit after reading this:

http://www.digitalpreservation.gov/f...53.shtml#notes

--Leonid

[lordsmurf]

Quote:

Originally Posted by metaleonid

I will go with 8 bit after reading this:
http://www.digitalpreservation.gov/f...53.shtml#notes
--Leonid

Bumping this post, because it's still relevant. Again, seeing a lot of questions lately, on this very topic.

And I wanted to quote that site:

Quote:

8- and 10-bit sampling. In principle, 10-bit encoding is superior in clarity to 8-bit, due to the reduction in tonal contouring and other artifacts. Some specialists argue, however, that there is no benefit for certain classes of material. One university expert wrote, "We digitize Betacam SX tape to 8-bit UYVY but Digibeta to 10-bit V210 because these selections align with the nature of the data that is actually sent out over SDI from these tapes. . . . SDI is 10-bit data, but when I piped the SDI video data from an SX tape to a binary display I could see the 9th and 10th bits were always zero. Thus by taking only the first 8 bits I could get all meaningful data . . . . I have about 3,000 SX hours to preserve and choosing 8-bit instead of 10 saves me about 90 TB of storage" (private communication).

In my opinion, for all analog formats, 8-bit is plenty fine, and 10-bit is overkill. I think 10-bit is best for the ITU-R Recommendation BT.709 color space, as used on modern formats.

[dpalomaki]

8- vs. 10-bit would depend on what one intends to do with the captured material, and to a certain extent the image content.

8-bit is fine for most purposes - the human eye is only good for about 7-bits of gray scale. 8-bits should look like a smooth gradient, 6-bits can look banded, and 7-bits will depend on how good the viewer's eyes are (and the display of course).

However, if one intends to do a significant amount of editing, especially large amounts of color correction, brightness/contrast adjustment, or filtering then 10-bit may (its not assured) give a nicer end result due to reduction of truncation artifacts. (6 dB of gain applied to an 8-bit gray scale will brings it down to 7-bit size steps.) It will be more apparent in images with large areas of relatively uniform brightness; e.g., sky, flat walls, etc.

[BmacSWA]

This is an older thread but a frequently asked question. I've been through all the discussions on 1's and 0's and file size and what consumer VHS was recorded on anyway. I've gone back and forth wrestling on whether to capture 8bit or 10bit old VHS video and I decided on 10. I did a small capture of both 8 and 10 bit video coming from the same tape through the same hardware only changing the capture format between 8 bit and 10 bit. You decide for yourself with the attachments. The 10 bit frozen on the same frame on the same monitor both in VLC looks better. Whys and wherefores aside, how can you choose 8 bit if file size isn't an issue?

B

...of course the .png file doesn't completely show the difference

[latreche34]

Analog digitizing over HDMI is only preferred if it adheres to rec.601 for SD materials such as some DVD recorders from the era of SD, none of those amazon analog to HDMI dongles qualifies obviously, Not only the conversion but the ingest device has to adhere to the same standard as well, A lot of those cheap HDMI to USB dongles don't, They resize, de-interlace, change other video properties, So a brand name product with good reputation is recommended.

[aramkolt]

Quote:

Originally Posted by latreche34

Analog digitizing over HDMI is only preferred if it adheres to rec.601 for SD materials such as some DVD recorders from the era of SD, none of those amazon analog to HDMI dongles qualifies obviously, Not only the conversion but the ingest device has to adhere to the same standard as well, A lot of those cheap HDMI to USB dongles don't, They resize, de-interlace, change other video properties, So a brand name product with good reputation is recommended.

Is there a certain test pattern or way to tell if Rec601 is being appropriately utilized? I'd like to include something like that in my testing of various capture devices.

As for the 8 bit vs 10 bit discussion, depends on the content I think whether you can really tell. In scenes where there is low contrast such as sky or generally very dark scenes, 10 bit should be visibly better because the similar brightnesses on screen near each other aren't "rounded" to the same values creating a blocky/band like appearance. That said, the "recommended" TBCs all do this same 8 bit rounding anyway, so if you're using something like an AVT-8710 or TBC-1000 in the chain, you might as well capture in 8 bit. AVIsynth at least used to downsample to 8 bit before doing most filters anyway, but I'm not sure if that is still the case. Still wouldn't hurt to have a 10 bit master/archived copy if nothing in your chain is limiting you to 8 bit.

[latreche34]

Studio test equipment can tell the difference but who has them, This is why one should stick with brand name devices.

[lordsmurf]

Even 10-bit gear isn't guaranteed to do anything, especially when the source is vastly below even 8-bit. The processing noise involved tends to be greater than any bit depth. For too many years, decades actually, bit depth has been a marketing ploy, and not something with actual translation. I've seen this in cameras and software since the 1990s.

Worse yet, many times, 8-bit is intentionally made to look worse. Again, processing.

[latreche34]

Just to clarify, my above post refers to telling the difference between rec.601 compliant hardware and cheap non standard crap using test equipment, I wasn't referring to 8bit vs 10bit.

[lordsmurf]

Quote:

Originally Posted by latreche34

Just to clarify, my above post refers to telling the difference between rec.601 compliant hardware and cheap non standard crap using test equipment, I wasn't referring to 8bit vs 10bit.

And I was mostly replying to @aramkolt, but didn't quote the message.

[Gary34]

Quote:

The processing noise involved tends to be greater than any bit depth.

What exactly do you mean here? Are you saying the noise from capture dithers the gradient?

[lordsmurf]

Quote:

Originally Posted by Gary34

What exactly do you mean here? Are you saying the noise from capture dithers the gradient?

It can, yes. The palette may be 10-bit, but the data isn't necessarily so. And that can include the item itself processing the video/photo. Consider cameras, and megapixels, not too different from bit depth. The palette may be 50 megabit, but the optical resolve may not be more than 5 megapixels. This problem predates digital.

[Gary34]

I think I got what you’re saying. Not just capture noise but also post processing and compression smooth out the banding and you don’t take out all of the noise in post. You leave some noise so you don’t get a plastic look. You do take out most noise because it affects the compression algorithm’s ability to use other frames for reference. So really comparing raw unprocessed captures of 8 and 10 bit doesn’t really matter much to the quality of what you end up with.

The shuttles also captures 6 extra lines that are not part of the usable picture. Those lines are full of noise or black and it’s an extra 100 MB/second vs Huffy 8 bit.

[aramkolt]

Ah yes, the 486 line argument haha. Technically full frame is 486 lines of visible data I believe. Most capture cards just decide to capture 480 of them and it can vary how many top or bottom lines are left out. In a perfect world, you'd want the bottom 6 lines to be left out since those are often head switching noise anyway.

AJA and Blackmagic typically do capture all 486 of those lines if capturing as ProRes. Most others limit you to the 480 lines.

The "head switching noise" part often can be completely eliminated too with head switching point adjustment - but the problem is that it's usually different for every tape, so changing that over and over isn't exactly recommended. The exception might be playing a tape that was recorded on the same VCR you are playing it back on - you might not see any head switching noise at the bottom at all in that case with no adjustment. Haven't tried that myself. I have tried the head switch point adjust and that does work.

[lordsmurf]

Quote:

Originally Posted by Gary34

I think I got what you’re saying. Not just capture noise but also post processing and compression smooth out the banding and you don’t take out all of the noise in post. You leave some noise so you don’t get a plastic look. You do take out most noise because it affects the compression algorithm’s ability to use other frames for reference. So really comparing raw unprocessed captures of 8 and 10 bit doesn’t really matter much to the quality of what you end up with.

As hodgey/oln likes to point out (and for obvious reasons, as that project's main dev), there is no image on the tape, only RF signals. Sort of like The Matrix, "there is no spoon".

But to me, that's philosophical hooey. I'm more pragmatic. There is/was an intended source image, and it's the job of the VCR/gear to extract it.

When you start to exceed the specs of the source -- 10-bit, HDMI/component, etc -- you introduce upscale issues. Not "upscale" as in higher resolution, but upscale by dictionary definition (ie, "increase size of something").

Tossing out bigger numbers, for the sake of bigger numbers, is called measurebating, a term coined by Ken Rockwell (photo/camera gear reviewer) about 20 years ago.

The inverse is also true. When you create something that does a task well (ie, capture 10-bit), and then "also create" (or have a feature that "also does") another task, it tends to do it quite poorly. Because, duh, that's not what it was designed to do. So you may seen a "10-bit" item do bad at 8-bit, or you may even see a dedicated sibling product line that does it badly because it's based on the 10-bit logic/designs.

Or the inverse, bad 10-bit extracted from 8-bit.

Sometimes it all sucks, the company just makes crap.

Quote:

Originally Posted by aramkolt

Ah yes, the 486 line argument haha. Technically full frame is 486 lines of visible data I believe. Most capture cards just decide to capture 480 of them and it can vary how many top or bottom lines are left out. In a perfect world, you'd want the bottom 6 lines to be left out since those are often head switching noise anyway.
AJA and Blackmagic typically do capture all 486 of those lines if capturing as ProRes. Most others limit you to the 480 lines.

Correct.

Quote:

The "head switching noise" part often can be completely eliminated too with head switching point adjustment - but the problem is that it's usually different for every tape, so changing that over and over isn't exactly recommended. The exception might be playing a tape that was recorded on the same VCR you are playing it back on - you might not see any head switching noise at the bottom at all in that case with no adjustment. Haven't tried that myself. I have tried the head switch point adjust and that does work.

I've only done this with rare footage. In fact, I have some tapes I need to re-capture, and try to remove the head switching. In general, not worth it, and you will just damage the VCR as tolerances start to slip non-tight. This isn't what it VCRs were design for.

[Gary34]

Quote:

Most capture cards just decide to capture 480 of them and it can vary how many top or bottom lines are left out.

These cards were made when MPEG 2 was popular and people were making DVDs. With 480 it’s divisible by 16 so a macroblock will go into it evenly then with 720 your player gets rid of the 8 pixels cropped on each side if you are making a DVD. 480 is better optimized for lossy compression then 486. Doesn’t matter with us though because we aren’t using macroblocks until the final encode.

[latreche34]

Be carful when changing the head switch point for tapes with HiFi audio, You may introduce buzzing in the audio at certain points of the tape, You may not hear it right away until the VCR reaches a weak signal spot on the tape and boom you have that buzzing, When adjusting make sure to allow some headroom, don't stop turning the stator right after the buzzing disappears, give it an extra fraction of a degree. If the tape has linear track mono only this would not matter.

Back in the day when overscan was a thing, this did not matter, now we are looking at every bit of image to recover, But you don't have to do it for every tape, do it strictly for important or rare tapes.

[lordsmurf]

Quote:

Originally Posted by latreche34

Be carful when changing the head switch point for tapes with HiFi audio, You may introduce buzzing in the audio at certain points of the tape, You may not hear it right away until the VCR reaches a weak signal spot on the tape and boom you have that buzzing, When adjusting make sure to allow some headroom, don't stop turning the stator right after the buzzing disappears, give it an extra fraction of a degree. If the tape has linear track mono only this would not matter.

Back in the day when overscan was a thing, this did not matter, now we are looking at every bit of image to recover, But you don't have to do it for every tape, do it strictly for important or rare tapes.

For important tapes, where I'd go to obscene extra effort of tweaking heads to remove a few more pixels of overscan, I'd easily capture the audio separately as well. In fact, odds are high that I'd capture video on a certain JVC, and capture the audio on a certain Panasonic (or another JVC, or even a Sharp VHS).

In general, yes, be careful for sure. This isn't a process that you'd want to overdo, or do as a standard.