digitalFAQ.com Forum

digitalFAQ.com Forum (http://www.digitalfaq.com/forum/)
-   Capture, Record, Transfer (http://www.digitalfaq.com/forum/video-capture/)
-   -   Capture codecs, bit depth, and software? (http://www.digitalfaq.com/forum/video-capture/10884-capture-codecs-bit.html)

Lee B 08-12-2020 10:11 AM

Capture codecs, bit depth, and software?
 
OK, I've finally got my S-VHS rig ordered and will soon be ready to capture.

Originally I was going to get the Pinnacle 710-USB capture device (thread here), but in the end I opted for the cheaper Hauppage USB-Live2 (which I have now ordered) which gives me the extra money to buy a frame TBC (which Lord Smurf has very kindly supplied me with). I also have a Panasonic MR-ES15 line TBC.

So now I just have a few questions about codecs and software… If anyone has any thoughts on any of these, please let me know.

1. Software
I'm planning to use VirtualDub to capture, is that recommended, or should I stick to Hauppauge's bundled software?

2. Bit depth
I would love to capture in 10 bit or higher (for vastly better color correcting results). I'm sure that will not be possible but thought I would ask if there's any way?

3. Capture video codec
Which is better for lossless capture: Lagarith or Huffyuv? By 'better' I mean reliable/efficient. File size is irrelevant as I won't be keeping the lossless files.

4. Capture audio codec
Same question again for capturing lossless audio? I love FLAC and always use that, but not sure if it's best with video?

5. Final render codec
While I will be keeping the audio lossless, I'm not planning on keeping the video lossless as the files will come to about 5tb and that's too big. I'm therefore planning to convert the video to a very high quality codec. I was thinking Grass Valley HQX, but I've also considered Cineform or H.265. Any thoughts?

Compatibility with current technology isn't particularly important to me, but what does matter is having a very high quality format that will last long into the future, for my posterity to be able to watch. This makes me wonder if H.265 would be best?

latreche34 08-12-2020 01:26 PM

Use vdub, AmarecTV or CaptureFlux not the supplied software.
Your capture device's ADC may not support 10bit.
I use HuffYUV just a personal preference.
FLAC? If size doesn't matter as you said why don't you just capture in PCM 48/16 (or 24bit if the card supports it).
H.264 can have audio compatibility issues for anything over 192Kbps. If you want future proof strategy keep the lossless files and encode to whatever future format might be.

Eric-Jan 08-12-2020 01:42 PM

Very high quality ? we're talking about VHS ? and audio if these tapes have a HiFi stereo track,(if it's analog audio/linear track, you can't do much about that anyway) then that's the least of your problem, just try something, what suits you best with your hardware, the different options will put more or less strain on your hardware you've got, a codec like h.265 will not add extra quality, only smaller files, for your end result.
Some hardware (CPU/GPU) support h.265 already so it doesn't put strain on your OS
FLAC audio codec is a complete waste of time for VHS sound, and makes your files only big in the end, you can check audio quality with a spectrum analyzer, you will notice that quality does not come above 16KHz, most recordings will even be much lower, so forget FLAC, and do some basic research first, that saves you money and time for your captures

traal 08-12-2020 05:23 PM

HuffYUV uses less CPU than Lagarith so it will be less likely to drop frames with slower CPUs.

VHS Hi-Fi audio goes up to around 20-22+ kHz, so 48 kHz is an appropriate sample rate. Just capture to uncompressed PCM to avoid taxing the CPU.

For the final render codec, keep it lossless (perhaps ffv1 422p) so you can easily rerender to the codec of the day (H.265, VP9, etc.) without making the quality worse. I also like to keep the original capture files (an 8TB drive is $130 so a 30GB capture file costs about 49 cents) and VirtualDub's .vdscript file.

Lee B 08-14-2020 08:01 AM

Thanks for the answers, guys.

I'm definitely going to stick with Vdub + HuffYUV, based on your answers.

Quote:

Originally Posted by latreche34 (Post 70726)
If you want future proof strategy keep the lossless files and encode to whatever future format might be.

Quote:

Originally Posted by traal (Post 70733)
For the final render codec, keep it lossless (perhaps ffv1 422p) so you can easily rerender to the codec of the day

Thanks guys, but that's not the kind of future-proofing I want. I'm not interested in compatibility within my own lifetime. What I'm looking for is a format people can work with long after I'm dead! (As a Christian, I believe I'm going to be resurrected at some point in the future so I'll want to access my videos hundreds of years from now.)

By the way, I assume '422p' is some quality setting, not picture resolution!

Audio

Quote:

Originally Posted by latreche34 (Post 70726)
FLAC? If size doesn't matter as you said why don't you just capture in PCM 48/16 (or 24bit if the card supports it).

You're right, FLAC just adds unnecessary CPU demand. I'll just capture it as PCM and compress it later, thanks.

Quote:

Originally Posted by latreche34 (Post 70726)
H.264 can have audio compatibility issues for anything over 192Kbps

That's interesting. I assumed video and audio would be totally independent. But you're saying I can't just pair up any video with any audio?

Quote:

Originally Posted by Eric-Jan (Post 70727)
FLAC audio codec is a complete waste of time for VHS sound, and makes your files only big in the end

Fair point. FLAC will be about 60gb total for all my videos, which may be pointless if the original audio is poor. I'll have a listen to the quality of the PCM and decide whether lossy compression may be better.

Quote:

Originally Posted by Eric-Jan (Post 70727)
Very high quality ? we're talking about VHS ?

To clarify, I didn't mean a great looking image, I meant a very good reproduction of the original source, including the tiniest speckles of noise, and details beyond visible perception.

traal 08-14-2020 08:47 PM

If you want details beyond visible perception, then you'll need to go lossless, because lossy formats such as H.264 achieve their high compression ratios by removing details beyond visible perception.

lordsmurf 08-14-2020 11:03 PM

Quote:

Originally Posted by Lee B (Post 70722)
1. Software
I'm planning to use VirtualDub to capture, is that recommended, or should I stick to Hauppauge's bundled software?

Yes, VirtualDub.
No, not WinTV, or any other bundled junk included.

Quote:

2. Bit depth
I would love to capture in 10 bit or higher (for vastly better color correcting results). I'm sure that will not be possible but thought I would ask if there's any way?
Consumer analogs like VHS, Video8, Hi8, Betamax -- and semi-pro S-VHS -- all have a near equivalent of 6-bit dithered color. 8-bit is fine. 10-bit and higher does nothing but bloats files with zero quality gain. This is one of those dumb things that you see these days, not much different from wanting 1080p or even 4K for VHS.

Quote:

3. Capture video codec
Which is better for lossless capture: Lagarith or Huffyuv? By 'better' I mean reliable/efficient. File size is irrelevant as I won't be keeping the lossless files.
Lagarith requires more CPU. So Huffyuv is more stable.

Quote:

4. Capture audio codec
Same question again for capturing lossless audio? I love FLAC and always use that, but not sure if it's best with video?
Just capture PCM, no compression., not even lossless.

Quote:

5. Final render codec
While I will be keeping the audio lossless, I'm not planning on keeping the video lossless as the files will come to about 5tb and that's too big. I'm therefore planning to convert the video to a very high quality codec.
Broadcast birate MPEG2 @ 4:2:2 for video, low GOP, with high bitrate AC3 (Dolby).

Quote:

I was thinking Grass Valley HQX, but I've also considered Cineform or H.265. Any thoughts?
Compatibility with current technology isn't particularly important to me, but what does matter is having a very high quality format that will last long into the future, for my posterity to be able to watch. This makes me wonder if H.265 would be best?
I'd strongly suggested against something proprietary like HQX or whatever; I already dealth with that PITA at my studio gig some years ago, funky nuisance now-unsupported formats. Back then, I hated Flash, but they insisted on FLV streams in addition to the MP4 -- what a waste of time (and their money).

I firmly believe that H.265 is a failed format, and will forever be relegated to niche usage. Divx/Xvid was superior to MPEG-2 in some ways, but where is now? Gone. MPEG had a delicate balance of compression time, decompression power, etc -- something that was improved on with H.264. But H.265 doesn't have that balance, it's beast of CPU requirements. H.265 is HEVC, aka High Efficiency Video Codec/Coding, but in a grander picture it's really not efficient whatsoever.

Quote:

Originally Posted by Eric-Jan (Post 70727)
FLAC audio codec is a complete waste of time for VHS sound, and makes your files only big in the end

My bigger concern is compatibility on decode/playback and muxing options. Lossless audio is really not in use whatsoever in video. Either do PCM uncompressed, or go high bitrate "lossy" (but no really lssy at all at reasonably high bitrates).

Quote:

Originally Posted by traal (Post 70733)
For the final render codec, keep it lossless (perhaps ffv1 422p)

I don't much care for FFV1, but 4:2:2 is an important aspect to keep, and I'm glad to see that others are taking heed of this. It's the 4:2:0 compressed (or worse 4:1:1) that really tanks quality of master files. For compressed viewing, great, 4:2:0 all the way. But not archives.

Quote:

Originally Posted by Lee B (Post 70762)
Thanks guys, but that's not the kind of future-proofing I want. I'm not interested in compatibility within my own lifetime. What I'm looking for is a format people can work with long after I'm dead!

Then you need to stick to mainline formats. Currently, that means either MPEG-2 or H.264, and nothing else. No frilly offbeat formats, nothing that "may" catch on (aka "the next big thing!"). Follow the majority to have something that most people are willing to still develop legacy products for. As analogy, make vinyl records, not sound cones or reel-to-reels.

Quote:

By the way, I assume '422p' is some quality setting, not picture resolution!
4:2:2

Quote:

That's interesting. I assumed video and audio would be totally independent. But you're saying I can't just pair up any video with any audio?
Correct, must be muxable (multiplex-able).

Quote:

Originally Posted by traal (Post 70785)
If you want details beyond visible perception, then you'll need to go lossless, because lossy formats such as H.264 achieve their high compression ratios by removing details beyond visible perception.

Everything from remove, to complete cover in a blur (lack of bitrate).

Lee B 08-15-2020 12:28 AM

Wow, that is why you're the "Lord"!

Stunning answer! I've learned so much! So glad I asked these questions now!

Quote:

Originally Posted by lordsmurf (Post 70788)
Consumer analogs like VHS, Video8, Hi8, Betamax -- and semi-pro S-VHS -- all have a near equivalent of 6-bit dithered color.

:ohmy: Gosh! I had no idea this was the case! I've always just assumed that analog = continuous. Starting to realise how little I know about video!

And "H.265 inferior to H.264"!? :ohmy: There was I, foolishly thinking that 265 was "one better" because it's a higher number! Thanks so much for the warning sir! :)

I just have one more question…

When I was digging out my tapes recently I was pleasantly surprised to find that the quality still looked good, in fact, better than I remembered.

But I found one tape that seemed completely blank. Nothing on it but noise. It was labeled as having home movies on it, but I guess it's possible the tape was misplaced or mis-labeled, or even somehow recorded over.

But what I'd like to know is… is there any possibility that while all my tapes aged well, this one tape has somehow massively degraded to the point that it seems to be blank? Is that even possible?

Hushpower 08-15-2020 12:42 AM

Lagarith: I have a 10 year-old i5 750 (12gb of RAM) and have just captured 50 hours of VHS through an ES-15 and didn't drop one frame. CPU was never more than 35% and averaged about 25% with 197 processes running under Windows 10. I don't think dropped frames due to slow CPUs are an issue these days.

lordsmurf 08-15-2020 12:44 AM

On the topic of bit depth...
If you want to read the tech details on color-under: https://en.wikipedia.org/wiki/Heterodyne
And that's a brief version.
Which explains why the digital "bit" equivalent is so low. Essentially, color bandwidth is compressed and truncated. That's why it's sub-8bit (but with posterization, so 6-bit dithered). And why 4:2:0 is acceptable (though it does compress more). Note that s-video vs. composite also has some affect on color, but also noting that excess negativity is usually a statement on the device and not the carrier itself.

Memory betrays us. You never know if it was labeled with the intention of being used (and never was), or something else that resulted in a blank video.

Natural degaussing is unlikely, but it may have been subjected to a strong magnetic field in past decades. For example, right next to a speaker, for an extended time (years). What you must always do is list possible hypothesis, NEVER jump to conclusions (as that's how myths start), and then test those hypothesis (aka the scientific method). Stuff like "tape fade" is nonsense, not scientific whatsoever, but is still often parroted online.

Quote:

Originally Posted by Hushpower (Post 70790)
Lagarith: I have a 10 year-old i5 750 (12gb of RAM) and have just captured 50 hours of VHS through an ES-15 and didn't drop one frame. CPU was never more than 35% and averaged about 25% with 197 processes running under Windows 10. I don't think dropped frames due to slow CPUs are an issue these days.

My tablet drops frames on Lagarith. It's an i7 fanless. Not that I capture to my tablet at all, but I do use it for Win10 card testing. Note that it's not your average tablet. IPS matte, Wacom, etc. Not many tablets cost $1k+

Lee B 08-16-2020 07:51 AM

Thanks again for all the info.

Color-under is over my head, but I'll take your word for it!

Quote:

Originally Posted by lordsmurf (Post 70791)
Natural degaussing is unlikely, but it may have been subjected to a strong magnetic field in past decades.

No, all my tapes have always been kept stacked together (vertically) in plastic containers, so if one had been near a strong magnetic field, they all would have.

Quote:

Originally Posted by lordsmurf (Post 70791)
You never know if it was labeled with the intention of being used (and never was)

That's a good hypothesis, I hadn't thought of that! :-)


Quote:

Originally Posted by lordsmurf (Post 70791)
Stuff like "tape fade" is nonsense, not scientific whatsoever, but is still often parroted online.

Gosh, I had no idea that was a myth. I can't tell you how many times I've seen TV shows and articles talking about the phenomenon!

A quick Google search led me to this:
https://www.refinery29.com/en-us/201...-deteriorating

Which led to this:
https://www.clir.org/pubs/reports/pub54/2what_wrong/

Presumably, these kind of articles make you cringe?

lordsmurf 08-16-2020 08:32 PM

The nonsense term "tape fade" doesn't appear in either link. That's the asinine idea that the image "fades" or otherwise "degrades" (less sharp, color loss, etc), which is not possible. The signal is either there, or it isn't. The quality is 100% about the VCR and transfer method being used. In other words, "tape fade" is user error, the person is using cheap garbage hardware.

However, the Refinery29 article refers to the bogus date range of "15-20 years" for degradation. The actual range is 35-65 years, which is why we're starting to see more and more early 80s tapes have issues. That article author has credits for all sorts of random nonsense, and is extremely likely not qualified to write on this topic.

The NPR article quotes a person who has fear without merit, nor a scientific basis in fact.

That Refinery29 article further also leads to a CNET article giving horrible transfer advice. The entire Refinery29 article appears to be a sham "article" meant to drive visitors to the single transfer business mentioned at the bottom.

Be smarter.

I'd bet most of those people will have zero understanding of basic concepts like overscan.

Lee B 08-17-2020 05:50 AM

Thanks for the clarification. In a world so full of misinformation and clueless 'experts', I'm glad I found you!

Subarit 01-10-2021 02:54 PM

Quote:

Originally Posted by lordsmurf (Post 70788)
Consumer analogs like VHS, Video8, Hi8, Betamax -- and semi-pro S-VHS -- all have a near equivalent of 6-bit dithered color. 8-bit is fine. 10-bit and higher does nothing but bloats files with zero quality gain. This is one of those dumb things that you see these days, not much different from wanting 1080p or even 4K for VHS.

Further to the colour depth issue:

https://www.scart.be/?q=en/content/s...hiving-masters

Quote. "The higher the bit depth used to digitise video the better the digital sample depth, ... For maximal quality retention of the original source a 10 bits digital sample is required. This is true for any videotape formats, even U-matic, Hi8 or VHS. In fact, retaining the maximum chrominance and luminance information from these formats might be even more important than for high-quality standards such as Betacam SP or 2" Quad tapes. The same goes for an already poor analogue source, for which any type of compression will only make the low quality of the image worse." Unquote.

The article is entitled 'A short guide to choosing a digital format for video archiving masters' and is a compilation of works by some industry experts. The statement seems to be based on a presentation by George Blood Audio and Video (TX) dated 2011 (no.8 in reference notes).

I think I found the link to the above on this forum. The statement is in line with the principles professed by the forum reputable members which remind me of doctors’ Hippocratic oath not to do any harm to a patient in the first place.

If I understand it correctly if one had a choice he/she would go for a 10 bit capture device but under current circumstances (lack of 10 bit capture cards on sale) this is not worth bothering. The PCI cards I looked at so far (AVerMedia CE310B, Canopus DV Storm 2) all have 8 bit colour depth except for Blackmagic Intensity Shuttle USB which is 10 bit (they boast in their ads that this is the only such device on the market). I also look at PCI(e) cards AVerMedia Ezmaker SDK Express C725B and Hauppauge ImpactVCB-e as they accept SECAM which I need, but was not able to find any information on their colour depth.

In a discussion http://www.digitalfaq.com/forum/vide...depth-vhs.html LordSmurf mentions that Theatre chips used in ATI AIW cards are 8-bit as well.

Is it correct that if I were to use 10-bit at the capture stage then it would be preferable to preserve 10-bit all thru’ the entire workflow chain? I have already FC-400 TBC and Panasonic DMR-ES10 pass-thru’ ready for my workflow. I have not yet decided which way to go in terms of whether I should use a PC or a notebook based work-flow and thus the choice of a capture device (BM Intensity Shuttle is not compatible with my notebook). Taking into account that FC-400 is 10-bit and ES10 is claimed to be 12-bit (there was a discussion though on videohelp.com ref which converter - AD or DA - uses 12-bit) would it be a good idea to stick to this rate throughout the chain? I am inclined to trust 1) the above mentioned article ref retaining signal quality 2) a detailed description of specs I found on the Inet which says ES10 has a 12-bit ADC 3) my gut feeling that this makes sense.

Maybe forum members can suggest some other 10-bit capture cards apart from BM Intensity Shuttle? It would be a shame to lose an opportunity of having a 10-bit colour depth from capture to the final product.

I understand LordSmurf was skeptical about the colour depth visual differences (10 vs 12) yet as far back as in 2005 (https://forum.videohelp.com/threads/...-ES10-Reviewed) however maybe 8 bit vs 10 bit in the setup configuration matters at the editing stage?

If there is no 10 bit PCI cards and I am forced to choose between AVerMedia Ezmaker SDK Express C725B and Hauppauge ImpactVCB-e (for their SECAM) would Hauppauge be better for capturing purposes (it has Conexant cx23885 chip)?

traal 01-11-2021 12:16 AM

Adjusting curves can result in undesirable posterization, so it's useful to capture at a greater bit depth than what you're targeting for the final product. These 8-bit cards weren't designed with restoration in mind.

I'm experimenting with converting to 16 bits at the beginning of my avisynth script so that all subsequent calculations are 16 bits wide. Processing takes a lot longer but the output has more colors so it seems to be a slight improvement. Still I'd rather start with 10 or 12 bits!

pinto 01-11-2021 08:46 AM

Quote:

Originally Posted by Subarit (Post 74210)
The PCI cards I looked at so far (AVerMedia CE310B(...)) all have 8 bit colour depth (...)

Where did you get these informations from? Look at chipset CX23888 technical specifications :warning:
Quote:

Quality video is accomplished by using 10-bit video ADCs, a full 10-bit video data path, and an
adaptive 3D comb filter.

lordsmurf 01-11-2021 01:04 PM

Quote:

Originally Posted by Subarit (Post 74210)
short-guide-choosing-digital-format-video-archiving-masters

There's nothing short about that guide. :weird:

Whenever I see somebody post a link like that, I already know what I'm in store for: "science dictates blah blah blah". And my response = "how naive, how cute, of that article author". "In theory" is not "in practice". I'm all for science, but it's not as binary or rules-based as some simpletons mind make it out to be. Practical application differs from rigid theory. So while "the rules" may state X, in practice you get X-1 or X+1. With video, a huge variable is the equipment in use. And understand that attempting to find the item that replicates the theory is chasing unicorns.

BTW, that guide suffers from a common overzealousness for then-new formats. The glowing commentary of FFV1 has never materialized, and likely never will.

Quote:

I think I found the link to the above on this forum. The statement is in line with the principles professed by the forum reputable members which remind me of doctors’ Hippocratic oath not to do any harm to a patient in the first place.
Indeed, do no harm, but inversely don't overdo it (10-bit+ capturing).

The format mentioned -- VHS/S-VHS, U-matic, Hi8/Video8, etc -- all had low color depth. VHS was closest to 6-bit dithered, so 8-bit can perfectly capture the entire color palette.

Quote:

If I understand it correctly if one had a choice he/she would go for a 10 bit capture device
No.

Quote:

In a discussion http://www.digitalfaq.com/forum/vide...depth-vhs.html LordSmurf mentions that Theatre chips used in ATI AIW cards are 8-bit as well.
Yes.

Quote:

Is it correct that if I were to use 10-bit at the capture stage then it would be preferable to preserve 10-bit all thru’ the entire workflow chain?
No.

Quote:

and ES10 is claimed to be 12-bit
No. :laugh:

Panasonic was a chronic BSer with their DVD recorders.
I actually often wonder if the ES10/15 is even 8-bit, due to the posterization. Sure, some chip may be 12-bit, but it doesn't seem to matter whatsoever on the actual output video.

Quote:

I understand LordSmurf was skeptical about the colour depth visual differences (10 vs 12) yet as far back as in 2005 (https://forum.videohelp.com/threads/...-ES10-Reviewed)
Blast from the past, there! gshelley (retired broadcaster) and mazinz (fellow toon collector) also involved. As usual, lots of misunderstandings by others, about the tech, but they saw (or think they saw). That's one of the earlier posts when I posted about the "sync filter" (definition later refined to be a minimalist TBC'ish, as it's a crippled line TBC + non-TBC frame sync). Still recommended ONLY for tearing correction, using it as a budget TBC is done begrudgingly as it has quality-reducing side effects (posterization, aggressive always-on NR, PAL luma, etc).

Quote:

Originally Posted by traal (Post 74219)
These 8-bit cards weren't designed with restoration in mind.

The higher-bit cards weren't either. Most are "measurebating" similar to megapixels on cameras in the 2000s. It's not just the number of bits, but how the bits are used, the quality of the bits. Because bits aren't just bits.

Quote:

I'm experimenting with converting to 16 bits at the beginning of my avisynth script so that all subsequent calculations are 16 bits wide. Processing takes a lot longer but the output has more colors so it seems to be a slight improvement. Still I'd rather start with 10 or 12 bits!
I would imagine a very slight improvement, and one that is easily lost from latter delivery encodes. But it all depends on the sources.

traal 01-11-2021 04:31 PM

Quote:

Originally Posted by lordsmurf (Post 74237)
The format mentioned -- VHS/S-VHS, U-matic, Hi8/Video8, etc -- all had low color depth. VHS was closest to 6-bit dithered, so 8-bit can perfectly capture the entire color palette.

That sounds about right for chroma. I would like more bits for luma.

Subarit 01-12-2021 08:48 AM

2 Attachment(s)
Many thanks, guys. Will continue with my homework on this.

Quote:

Originally Posted by pinto (Post 74224)
Where did you get these informations from? Look at chipset CX23888 technical specifications :warning:

Pinto, I would be grateful for a link to info on Avermedia card chipsets.
The source of my info is Avermedia - see attached PDF file.

In another Avermedia source info on this card colour depth is simply missing:https://storage.avermedia.com/web_re...s_brochure.pdf

I also attach a Word file with some different colour depths test pictures and the link where they were taken from. Probably of some interest. (https://forum.blackmagicdesign.com/v...p?f=3&t=109259).

pinto 01-12-2021 01:40 PM

Quote:

Originally Posted by Subarit (Post 74268)
Pinto, I would be grateful for a link to info on Avermedia card chipsets.
The source of my info is Avermedia - see attached PDF file.

https://pdf1.alldatasheet.com/datash...T/CX23888.html


All times are GMT -5. The time now is 12:35 PM

Site design, images and content © 2002-2021 The Digital FAQ, www.digitalFAQ.com
Forum Software by vBulletin · Copyright © 2021 Jelsoft Enterprises Ltd.