Trying to teach benefits of lossless?

[BarryTheCrab]

Forum,
I have spreading the gospel of lossless and am going back and forth with a fairly smart fellow. Can you
verify (or vilify) the following quotes from my esteemed associate?

There are different levels of compression and different results. Starting from a limited bandwidth analog signal, IMO there is little to loose when working at highest quality compressed workflow. But that's just me. I disagree, I hate compression for capture.

The video input ADC device will limit a lot of the bandwidth and have a lot to say about the dynamic range of the video signal (color is amplitude modulated).
Firstly not capturing from S-Video it will apply the comb filter and kill a lot of Luma bandwidth.
Secondly not all the video ADC are created equal, personally I found that the ADC in a Digital8 camera (S-Video input) is very good. Seems OK

Note that the chroma signal is recorded with the same reduced bandwidth in both 8mm and Hi8 formats (only the luminance FM modulation was increased from 1.2 to 2.0MHz). Chroma AM modulation results in a resolution (for both formats) of only around 30 lines horizontally. Compare with about 400-420 lines on luminance channel for Hi8.
So the 4:1:1 NTSC rates applied by DV format are more than double of what Hi8/8 can provide and more than appropriate for capture. Is this true? I thought DV was to be avoided if possible and an S-Video capture would be superior for editing.
PS: The 4:2:0 by PAL is a bad compromise solution IMO... like always in PAL the regulators tried to screw consumers and provide sub-par video recording compared to their over-taxed "pro" market.

[keaton]

Not an expert, but since nobody else has chimed in yet, I can at least say a bit having read a lot on the forum and being a lossless capture guy myself.

It's rare to change somebody's mind on things. Perhaps being a champion of or a promoter of lossless is a better approach. If somebody has it set in their mind and is resolute in their belief, probably not much of a point to argue. I would say promoting lossless to those who have an open mind and/or are new and want to understand why they should consider lossless is a much more productive endeavor.

This forum taught me that if I'm going to be serious about restoration of video, then you should start with lossless. Furthermore, you should try to get the best colorspace you can to start with, i.e. 4:2:2 better than 4:2:0 or 4:1:1, because if you are doing color correction, you want to have as much color resolution as you can so you are not being any more destructive to the original than you have to. I've read MPEG can do 4:2:2, but I think the standard is 4:2:0. I think having the video in a lossless format is also essential to do as little harm as possible with other types of restoration.

Having come from scanning and restoration of photos or camera negatives and using Photoshop to clean and color correct, I used a lossless format like TIFF for my images before compressing the final result to a JPEG. If you work on JPEGs in Photoshop and apply curves, filters, etc. You will likely have a lesser quality JPEG in the end, than if you stayed lossless until the end. Why? Because the operations are destructive when it is not a lossless image being worked on. When it is lossless, this issue is minimized or removed. For image scanning, there are arguments about say 8-bit SRGB Color Space vs 10-bit AdobeRGB Color Space. So there's always the theory of it, versus the actual results one can measure with their eyes. So, like 4:2:2 vs 4:2:0, perhaps it's what one can observe, and it may be case by case with the video/image or the person.

Also, it may boil down to what one can notice. Some of us are much more detail oriented or particular about things. Those that are not, may not be able to tell as much difference.

No big technical arguments I have. I understand the theory of it, and have seen the benefits of using lossless over compressed. If someone doesn't see those results, then they are going to be quite difficult to convince.

Keep trying to spread the good word. Try to educate one individual at a time. But first, evaluate their openness to the subject matter.

[traal]

Quote:

Originally Posted by BarryTheCrab

Starting from a limited bandwidth analog signal, IMO there is little to loose when working at highest quality compressed workflow.

That's true, you don't lose much, unless you need to do any kind of restoration, in which case you need all the information you can get! A highly compressed video for example will toss away information in dark shadows because you can't see it anyway, but if you then try to brighten the image, you will start to see digital "macroblocks" caused by the video compression. Even grain and sensor noise are more visually pleasing because they can easily be ignored, or filtered out in software. You don't want any digital artifacts messing up an already imperfect video, so you really need to start with non-lossy video to get the best results.

Quote:

Originally Posted by BarryTheCrab

Firstly not capturing from S-Video it will apply the comb filter and kill a lot of Luma bandwidth.

That's true. You should capture from S-Video.

Quote:

Originally Posted by BarryTheCrab

I found that the ADC in a Digital8 camera (S-Video input) is very good.

For NTSC, maybe. For PAL, no, it clips reds.

Quote:

Originally Posted by BarryTheCrab

So the 4:1:1 NTSC rates applied by DV format are more than double of what Hi8/8 can provide and more than appropriate for capture. Is this true?

Perhaps, but 4:1:1 YV12 isn't supported as well in software as 4:2:2 YUV. And anyway, if you're using a lossless, compressed format such as HuffYUV, the storage savings will be minimal because, as your associate said, "the 4:1:1 NTSC rates...are more than double of what Hi8/8 can provide and more than appropriate for capture." Right?

[hodgey]

Quote:

Originally Posted by BarryTheCrab

There are different levels of compression and different results. Starting from a limited bandwidth analog signal, IMO there is little to loose when working at highest quality compressed workflow. But that's just me.

It depends a bit whats mean by that, with max quality ProRes maybe, less so with DV where the DCT quantization introduces noticeable artifacts, especially on noisy content. It is essentially reducing the video resolution, just in the variety of frequencies that can be represented rather than the amount of x/y pixels. I find the main problem with the DV compression is when it comes to filtering the video afterwards, since the compression artifacts causes extra issues for noise reduction. If you're not doing anything other than playing it back I suppose it may not be as much of an issue.

Quote:

Originally Posted by BarryTheCrab

The video input ADC device will limit a lot of the bandwidth and have a lot to say about the dynamic range of the video signal (color is amplitude modulated).
Firstly not capturing from S-Video it will apply the comb filter and kill a lot of Luma bandwidth.
Secondly not all the video ADC are created equal, personally I found that the ADC in a Digital8 camera (S-Video input) is very good.

The bandwidth of the analog signals are way more limited than a modern Video ADC, the video is sampled by a rate of at least 4 times the color carrier frequency (which is 4.43 MHz in PAL and 3.58 MHz in NTSC), usually a lot much more.

Capturing using the S-Video input on a D8 camera is limiting as you lack any control of the levels (other than possibly in service mode, haven't checked), and you are forced to use DV compression. The ADC itself seems ok, after all the video signal will go through the cameras ADC for timebase correction on playback and it doesn't seem to cause any issues there.

Quote:

Originally Posted by BarryTheCrab

Note that the chroma signal is recorded with the same reduced bandwidth in both 8mm and Hi8 formats (only the luminance FM modulation was increased from 1.2 to 2.0MHz). Chroma AM modulation results in a resolution (for both formats) of only around 30 lines horizontally. Compare with about 400-420 lines on luminance channel for Hi8.
So the 4:1:1 NTSC rates applied by DV format are more than double of what Hi8/8 can provide and more than appropriate for capture. Is this true? I thought DV was to be avoided if possible and an S-Video capture would be superior for editing.
PS: The 4:2:0 by PAL is a bad compromise solution IMO... like always in PAL the regulators tried to screw consumers and provide sub-par video recording compared to their over-taxed "pro" market.

Translating between analog and digital resolution is not very straight forward, I haven't used NTSC much, others on here can probably explain better the issue with 4:1:1, from what I've read on the forums, the horizontal resolution loss is noticeable though, how much I don't know.

4:2:0 is a bit complicated as well, PAL has reduced vertical color resolution due to the format, and additionally the color signal from tapes (in both PAL and NTSC) will go through a comb filter in the playback device (to reduce crosstalk, not as extensive as what will happen in the ADC with a composite signal), even when using S-Video. I.e the color is going to be smeared a bit verically, but I don't know if it's technically 1/2 vertical resolution still.

[lordsmurf]

Quote:

Originally Posted by BarryTheCrab

There are different levels of compression and different results.
, IMO there is little to loose when working at highest quality compressed workflow. But that's just me.

While those statements can be true, it really depends on the compression he refers to. ProRes422, for example, can be fine. Even broadcast depth 15mbps+ 4:2:2 MPEG can be fine. But something like H.264 is not fine, as that is a delivery format.

Quote:

Starting from a limited bandwidth analog signal

However, the analog bandwidth really has nothing to do with digital compression.

Quote:

The video input ADC device will limit a lot of the bandwidth

No.

Quote:

and have a lot to say about the dynamic range of the video signal (color is amplitude modulated).

It can, but I think he's being facetious, seeing as how compression will be the limiter far more than any capture device. The bit depth of the capture device is secondary to the compression.

Quote:

Firstly not capturing from S-Video it will apply the comb filter and kill a lot of Luma bandwidth.

Comb filter doesn't really harm luma bandwidth. The bigger issue is crosstalk artifacts when using s-video (aka separated video; because luma and chroma are separately carried), be it from composite or coax.

Quote:

Secondly not all the video ADC are created equal, personally I found that the ADC in a Digital8 camera (S-Video input) is very good.

He thinks DV compression is good?

Quote:

Note that the chroma signal is recorded with the same reduced bandwidth in both 8mm and Hi8 formats

There is no "8mm" format. When somebody says something like this, I raise an eyebrow, and prepare myself for technobabble. The formats are Video8 (comparable to VHS), Hi8 (comparable to S-VHS), and Digital8 (DV on Video8/Hi8 tape).

Quote:

Chroma AM modulation results in a resolution (for both formats) of only around 30 lines horizontally.

No. Color data is never that low. Video8 is about the same "240 lines" as VHS data, with chroma half-to-quarter that. Hi8, like S-VHS, increased luma resolution only. It's definitely NOT 1/8th color data, otherwise it'd be unwatchable.

Quote:

So the 4:1:1 NTSC rates applied by DV format are more than double of what Hi8/8 can provide and more than appropriate for capture.

No. The bandwidth may be low, but that doesn't equate to the data present. VHS is the closest equivalent to 4:2:2. and the idea that NTSC VHS is 4:1:1 is a myth due to bad understandings of how bandwidth carries data. "VCR Troubleshooting and Repair" by Robert Brenner is a good resource for understanding what all is involved. It's not simple math, where you divide luma by chroma.

Quote:

PS: The 4:2:0 by PAL is a bad compromise solution IMO... like always in PAL the regulators tried to screw consumers and provide sub-par video recording compared to their over-taxed "pro" market.

4:2:0 is a compromise, but not nearly as bad as 4:1:1. The former is alternating halving, similar to 4:2:2, while the latter is simply butchery of the signal, pure quartering (halving the available half of color). So we agree, but for different reasons.