Why is VHS 720x480 and interlaced?
 

As much googling and deep diving as I have done, there are still some things I don't understand/would like confirmation on before I start doing my capturing

First, what I have:
- JVC HR-S9600U
- Pinnacle 510 USB
- Windows 10 PC

- VirtualDub 1.9.11 w/ Filters downloaded from this forum
- HuffYUV v2.1.1 CCESP Patch v0.2.2

- 12-ish VHS tapes that need to be archived, with all footage in 4:3 format

- and my exporting goal is to have these available online

First are the questions I was hoping to get confirmation on:
1. HuffYUV is a lossless codec. This means I do not need to set a bitrate for capture, correct?

2. Under "Raw Audio Format", VirtualDub doesn't give me the the option to choose between compressed/uncompressed audio, but it does allow me to choose between the sample rates (32, 44.1, and 48 kHz). Obviously I will choose 48kHz, but is it safe to assume that these are all uncompressed options?

3. I found this post: https://www.digitalfaq.com/forum/vid...o-640x480.html
My understanding is that if I were to put my videos straight on DVD, I would just burn the 720x480 version and it would automatically be 4:3 on my TV, is this correct?

And these are the questions I just don't understand:
1. WHY do we capture at 29.97 fps when analog video was filmed at 59.94 fps? I find it so confusing, I read here on the forum that you are supposed to take your 29.97i footage and de-interlace it into a 59.94p video file using either VirtualDub or QTGMC inside AviSynth.
a. First, why not just record the capture in 59.94? My thinking is that it's because 29.97i is what's needed to create 59.94p, meaning it doubles when de-interlaced, is this correct?
b. Given that, does de-interlacing always double the framerate? What do I do if I want to de-interlace but keep my framerate at 29.97?
c. What do I do if the footage on my VHS tape isn't 59.94? For example, I have one with a TV show recorded at 23.976 fps, do I still capture at 29.97?
d. Lastly, is it worth it to use AviSynth over VDub or is VDub enough? I read that VDub has motion issues and that QTGMC is the best

2. Continuing from #3 from the above section: When exporting for the internet, however, I need to convert the rectangular pixels to square, myself.
a. First, why did rectangular pixels ever exist?
b. More importantly though, what is the best way to do this? Again VirtualDub has a resize filter, but I imagine AviSynth does this too, so again, is there a reason to use one over the other? I also found this really helpful forum post: https://forum.videohelp.com/threads/...eeps-On-Giving

3. Hybrid is the current top pick for encoding software at the moment, correct?
4. Color correction is just if I want to, right? HuffYUV is lossless, so I imagine that when I capture, the colors aren't going to turn into LOG footage and get all flat and contrast-less.
5. I imagine the workflow looks something like this (Plz tell me if I am missing something, did something wrong):
a. Capture AVI in VDub, 48 kHz audio, 29.97i, HuffYUV
b. In VDub, remove black bars, convert rectangular pixels to square
c. De-interlace in AviSynth
d. Color grade also in AviSynth if I want to
e. Encode into h.265 or h.265 in hybrid
Is this correct-ish? Seems like a lot of back and forth, I imagine I could consolidate it in some ways, I'd love to hear what your workflow is!

Thanks in advance! I know this is a bit of a long one but I tried to wait until I had a good amount of questions before I posted.

You lack frame TBC. That will likely be a problem. But for 12 tapes, you can gamble, cross your fingers.

Most people don't even know what "CCESP Patch" is for anymore. A long-gone program, CinemaCraft Encoder (CCE), had issues. CCE wasn't fixed because most CCE users were just pirating the software (namely the SP version). The irony is that CCE was a noisy MPEG encoder, not worth the $2k whatsoever, nor even using for free. So these pirate users patched Huffyuv, instead of getting CCE to bugfix it. However, it also introduced issues, and should not be used now.

FYI, MainConcept toppled CCE, Canopus Procoder, Cleaner, and several others in the 00s. MC won the payware MPEG war, and still exists. ffmpeg won the freeware MPEG war.

Correct. Only confirm the default colorspace used, YUY2(YUV) and not RGB.

Correct, no options, use 48kHz.

Correct, as-is captures are authored 720x480 (if captured 720x480), which corrects by 8 pixel padding.crop offset, and the viewing pixel AR from rectangle to square.

That's wrong. The video is not 59.94 fps.

It's 50% of 59.94 fps. For convenience, let's round to 60 fps. Each half second is a frame. But that's not accurate. Each half second is half a frame. Not as in top/bottom, but a comb-tooth interlace. It's temporally offset. Each portion of the image represents a different moment in time. No true 59.94fps exists, and deinterlacing to 59.94fps is essentially creating a half frame from nothingness (actually neighbor frames, with some % of interpolation/guess/"AI").

FYI, if you see anybody claim that "only" 29.97 is "throwing away data", they're wrong. It's not accurate. Some motion data may, or may not, be lost. But for certain new data is created, and that new data isn't always good data. Which is why some video fights being deinterlaced to 59.94, even with complex Avisynth/QTGMC scripting.

There's no "supposed to", but rather the option to. And sometimes that's the right option, and sometimes the wrong option. It depends on factors. For example, some video will not cleanly and easily deinterlaced to 59.94 without heavy scripting. And then not every viewing situation will require 59.94.

Bad question due to bad/wrong assumption. See answer above.

It's sort of "doubling", but not really, because 59.94 never existed, and 29.97 also isn't the true nature or the interlaced source. When you deinterlace to 29.97, you can (but not necessarily!) lose some motion data, but then it tends to be retain in-frame data (as it's not interpolating as much).

It's never 23.976, nor 59.94.

23.976 is non-analog, only film is 24. To show 24fps film on analog medium, it has pulldown added, time converted to 23.976 progressive, with frames added, to achieve 29.97 for analog TV. VHS copies are baked-in at 29.97, and any attempt to remove pulldown (IVTC) still results in various analog artifacts. Unlike DVDs, which is when/where most people learned about 23.976.

For what? Deinterlace?

Never VirtualDub, old deinterlace filters/methods. Only use VirtualDub for capturing, certain specific filters.

Always QTGMC in Avisynth. Nothing better/newer has existed for about 15 years now, and may never. It's quite good at what it does, and is fast when using the 64-bit version (not quite realtime).

Also crop 8 from each side, then resize. Otherwise AR may be wrong.

Note: Analog is analog, but "pixels" refers to digital.

It's storage.

Why did books store pages back to back, when you could not read them that way? It's sort of a daft question, when you think of it in those terms.

And it's mostly trivia for users. For example, why is grass green? Does the yard man really need to know why it's green, or just that it need to be cut? There are reasons to everything, but does it really matter? When converting videos, do you need to know why, or just understand that it is what it is?

In actuality, it's 480i video, and the "720" is the measurement of the other axis. And there are many numbers, with multiple AR. 352, 480, 544, 640, 704, 720, etc. Each has a reason.

For resize, bicubic is fine. Each resize methods has pro/cons (lots of cons for some!), but bicubic tends to balance sharpness with artifacts. VirtualDub and Avisynth (and others) all have this basic resize method. Something like Lanczos tends to overshapen/ring, while bilinear tends to ghost/ring. Lots of newbies tend to gravitate to Lanczos, and butcher videos in the process. Decade-old posts proclaim it as "best" back when it was the new and shiny thing.

Correct, sort of. Hybrid is just a really advanced GUI/frontend, the underlying encoders are CLI open-source. So it's software (and excellent!), but technically not "encoding software".

Color correction can also be chroma noise reduction (CCD filter in VirtualDub). Tweaking colors can be quite optional, but it really depends on how off/bad it is.

Crop 8px (the "black bars" you refer to?), mask overscan by 2 pixel increments (bottom of 8-12)..

Color in Avisynth is not fun. Instead, try VirtualDub ColorMill filter, or NLEs (Premiere, FCP, Vegas, whatever).

Convert everything to a new lossless after VirtualDub/Avisynth/whatever. Then convert that to H.264/5, in MKV/MP4 container.

With video, if you do 2x as much in one step, it often takes 3x+ as long. More is not better. It all goes faster if you do several steps here, several there, several here again. Newbies have this idea of "I can do it all in one pass", and they are sadly mistaken. It always backfires quite terribly.

You asked good questions. I've not seen some of those in a while. :congrats:

I'm only going to address the title of the thread since LS covered all the questions, NTSC analog tape formats including VHS are not 720x480, They are 525 scan lines with no horizontal pixel structure, Out of 525 scan lines there are about 476-486 scan lines that are used for active image depends on the content, The rest are used for signal timing and other data.

The 720x480 (486 pro gear) came about a decade or two later after the birth of the first video tape format, When conversion of analog to digital standards were in the works, It was calculated that around 704 samples per video scan line was more than enough to preserve the analog signal in a digital form, The choice was precisely 720 samples or pixels by today's terminology per scan line, adding 16 extra pixels, 8 pixels per side to accommodate for different cameras and telecine machines' overscan, Vertically the choice was settled on 480 lines for consumer gear since consumer tapes have less active scan lines, pro gear is 486 lines because pro tapes such as Betacam usually contain more active lines than consumer tape formats in my opinion.

Thank you Thank you Thank you for this wealth of information, this has helped tremendously, I've read over this twice though and I still have some followup questions:

So wait, did I download the wrong thing? I just verified, I downloaded the 15.9 KB HuffYUV.rar file that was recommended on this forum, from this link:
https://www.digitalfaq.com/forum/vid...lters-pre.html
When I said I had the CCESP patch, that's just what shows up under "Video > Compression" on virtualdub, so if it's wrong then what do I download?

Additionally, out of curiosity, why don't we use the "No recompression (YUY2)" option under the same menu?

Then what is the true nature of the interlaced source? 29.97i?

Technically, yes, you are correct, but I don't think I correctly communicated my question. Here's some context: Before learning about this forum, I tried capturing with OBS and an old Elgato Game Capture HD I had laying around. This was when I *presumed* my home movies were recorded in 59.94, I could tell because when scrubbing the video frame-by-frame there were no repeated frames. However when I captured the aforementioned TV show at 59.94, there were several repeated frames. After doing some experimentation, I found that when captured at 23.976, there were no repeated frames.

Given that, I do youtube, where sometimes I want the final framerate to be 23.976. So what I really want to know is what do I do/how would the process would change if I wanted my end product to be in 23.976? Would my capture settings differ or would I do this in post?

And lastly, would this process change depending on the source (the TV show recorded at 23.976 vs my home movies recorded at a much higher framerate)?

This is understandable, I was just hoping to get a general framework of what to do in what order. Based on what you said, now I'm thinking the order would be:
a. Capture AVI in VDub, 48 kHz audio, 29.97i, HuffYUV
b. In VDub, remove 8 pixels on left and right, resize, convert rectangular pixels to square using bicubic filter
c. Color grade (if desired) in ColorMill
d. Move to AviSynth and de-interlace using QTGMC
e. Convert to New Lossless (using what, Hybrid? And with what codec, HuffYUV?)
f. Convert to MKV/MP4 using Hybrid for web

Is this a good framework to fall back on?
Thanks again!

For HuffYUV you can use Hofmand's installers for modern versions of Windows

Yes, interlace videos are recorded at 29.97i, but it can have footage that is either truly interlace, as in the footage was shot onto tape (de-interlaced to 59.94 progressive) or instead, like what you've seen from movies and TV shows shot on film, telecined from 23.97 to 29.97 (Deinterlacing that footage results in repeated frames, like what OBS was doing as it was probably deinterlacing on the fly) That kind of footage needs to be detelecined back to it's original frame rate.

Now you might get tapes where some footage was shot at 23.97 fps but also have footage or added effects that are at 29.97 fps, for videos like that were the footage is a mix of film and tape sources, the only option is to deinterlace

FPS can mean two things, Fields per second, or frames per second, In the analog world we use fields per second, NTSC use to be 60 FieldPS, but when the TV color transmission was introduced they reduced the FPS to 59.94 (60 fields every 1001 fields or simply 59.94 Hz) for technical reasons not related to this thread, you can look it up yourself online.

In digital, we refer to frames either as progressive or interlaced, if they are progressive we add P if they are interlaced we add I, For example (let's use 60 for easy typing) if your video has 60 fields it is called 30i, each two fields are considered one interlaced frame, if it contains 30 full progressive frames it called 30p. So where did the 60p come from? two sources: 1- When 30i is deinterlaced into double frame by whatever interpolation scheme used it becomes 60 full progressive frames a second or 60p, 2- Some camcorders has the ability to shoot in 60 progressive frames per second, I believe this was introduced in early 2010 in HD, I remember buying a Panasonic HD camcorder that recorded 60p to AVCHD and had a hard time playing those files on my computer due to lack of support back then.

I went to this URL, downloaded the installer, ran it according to the instructions and it still says I have HuffYUV 2.2.1 CCESP patch 0.2.2 so I suppose I did it correctly

Hm, maybe uninstall it and try again. VirtualDub should report it as 'HuffYUV v2.1.1'

Awesome, that worked! For anyone in the future who needs it, here is how to uninstall HuffYUV on Windows 10:

1. Remove the HUFFYUV.INF in C:\Windows\INF
2. Remove the HUFFYUV.DLL file in C:\Windows\System32
3. Remove the HuffYUV.DLL file in C:\Windows\SysWOW64

Then reinstall using hofmand's msi installer and you have HuffYUV v2.1.1 INSTEAD of the CCESP patch.

Speaking of, shouldn't the hofmand installer be added to this guide? https://www.digitalfaq.com/forum/vid...lters-pre.html

Glad you figures it out. I knew #2 and #3, and was going to post on it this weekend. You beat me to it.

I completely forgot about #1, and that may explain an issue I'm running into on a system. :hmm:

I like to use VirtualDub to determine the values to put into Avisynth, then Avisynth does the color correction and deinterlacing all in the same script. Then later if I find I did something wrong, I fix it in the script and rerun it.

Feel free to copy/paste it into a guide or any sticky posts you have on the forum, thanks for the help!