Aspect Ratios 101

[wcndave2]

I have read a lot of posts about this and came to a conclusion as to how it worked, however after using that to transfer a VHS I did some measurements on an item on the screen which I "believe" to be a circle, however it's not proven... which made me question my conclusions.

So I read various articles again, and came up with another possibility...

Here are my workings, can anyone confirm if they are correct or not?

Conclusion 1

• Captured VHS is 720x576
• PAR is 16:15 so it's stretched to 768x576 = 4:3
• The left and right 8px are not meant to be part of the picture - only the middle 704 should contain the image
• Even inside that "margin", there's black borders and head noise
• CRTs used to overscan by more than the 8 each side, resulting in a clean picture

Conclusion 2

• Captured VHS is 720x576
• First we need to discard left/right 8px to get to 704x576
• PAR is 12:11 so it's stretched to 768x576 = 4:3
• Even inside that "margin", there's black borders and head noise
• CRTs used to overscan resulting in a clean picture

I guess the only difference between the two, is do we discard the 16px of width and then stretch to 4:3, or stretch the 720 as it is?

After that we can crop (in the same 4:3 ratio of course), and/or letterbox to remove/hide noise as we like - the AR will be correct.

I thought it was option 2 based on reading https://en.wikipedia.org/wiki/Pixel_aspect_ratio

However here's why I started to doubt myself:

I took a capture and played in VLC, took a screengrab of a frame with a round logo on it with the VLC aspect ration set to default. Then set the AR to 4:3 and took another. Then I cropped to 704, and stretched to 768 and took another.

The first screenshot was clearly taller than wide as expected.
The second one looks spot on
The third is wider than it is tall...

Any confirmation of which approach is correct would be much appreciated!

Thanks, Dave

[Sharc]

Captured PAL VHS is always PAR=12/11 (in very good and practical approximation), and the 4:3 frame is within ~704x576 (unless your capture is flawed and violates the Rec.601 standard).
So yes crop total 16 pixels (left+right) from the 720 frame and display the remaining 704x576 as 4:3 (means stretched to 768x576). If you want to hide the head switching noise at the bottom just mask it (means crop it and replace it by black borders, or encode the whatever cropped picture using SAR=12/11 in x264).

If you would want to keep the 720x576 without cropping, the PAR of the captured PAL VHS is still 12/11, so the 720x576 needs to be stretched to 786x576 which is slightly wider than 4:3, namely ~1.36 rather than 1.333.... for preserving the circles undistorted.

[pandy]

To address your questions you need to provide information about sampling rate (pixel clock) frequency.

[CaptureCraft]

PAL, 4:3
702x575 -> 1150/1053 (for "diagonal" CRT lines, not sure if one can really make it)
702x576 -> 128/117 (for "horizontal" lines, like from CCD)
704x576 -> 12/11 (MPEG-2 compatible with "macroblocks" of 16x16 pixels)
720x576 -> 16/15 (for digital sources authored after circa year 2000)

@wcndave2. So, for you it is really a choice between 702x576 and 704x576.
If your codec is mpeg-2 based and requires 16x16 pixels macro-blocks, then you need to use 704x576.
Otherwise, you can use 702x576.
Either way, you crop even number of pixels from the sides so you would have the desired width, and then you apply the corresponding ratio.
The best way to handle the head switching noise is to mask it with black rectangle, don't crop it.

[pandy]

PAL, 4:3
702x575 -> 1150/1053 (for "diagonal" CRT lines, not sure if one can really make it)
702x576 -> 128/117 (for "horizontal" lines, like from CCD)
704x576 -> 12/11 (MPEG-2 compatible with "macroblocks" of 16x16 pixels)
720x576 -> 16/15 (for digital sources authored after circa year 2000)

So which aspect ratio is for 706.25x576 ?

[wcndave2]

So yes crop total 16 pixels (left+right) from the 720 frame and display the remaining 704x576 as 4:3

This is what I had originally though, however see my further information below

To address your questions you need to provide information about sampling rate (pixel clock) frequency.

Sorry, don't know what that means.

PAL, 4:3
702x575 -> 1150/1053 (for "diagonal" CRT lines, not sure if one can really make it)
702x576 -> 128/117 (for "horizontal" lines, like from CCD)
704x576 -> 12/11 (MPEG-2 compatible with "macroblocks" of 16x16 pixels)
720x576 -> 16/15 (for digital sources authored after circa year 2000)

@wcndave2. So, for you it is really a choice between 702x576 and 704x576.
If your codec is mpeg-2 based and requires 16x16 pixels macro-blocks, then you need to use 704x576.
Otherwise, you can use 702x576.
Either way, you crop even number of pixels from the sides so you would have the desired width, and then you apply the corresponding ratio.
The best way to handle the head switching noise is to mask it with black rectangle, don't crop it.

More detailed - but matches what I had read regarding 702


However I did another experiment.

I decided to go into photoshop and be a bit more accurate, rather than drawing circles.

I took a capture from a DVD which I also have on VHS.

I then took my VHS and took the 720x576 capture cut it down to 704 and then stretched it to 768x576
I then took the original 720 and stretched it to 768

I overlaid all 3 in photoshop with different transparencies so I could see them all in place.

The 704 stretched to 768 was clearly too wide compared to the DVD screen.
The 720 stretched to 768 was a perfect match.

Unless the DVD is not showing a correct AR, something is not matching the advice you're all giving, which was what I had first assumed to be correct from my reading.

Can you think of any reason why this would occur?

I am capturing in 601 using amarectv FWIW.

[CaptureCraft]

You can't have 706.25.
Your analog video data is 702 pixels. If you take wider, you just include black borders.
Your true PAR (pixel aspect ratio) is always 128/117.

[pandy]

To address your questions you need to provide information about sampling rate (pixel clock) frequency.

Sorry, don't know what that means.

This is simple - active video line in PAL is 52.3uS long (64.0uS - 11.7uS [line blanking interval]) - so you need to use number of pixels with specific sampling rate. Maximum number of active pixels (active video) is 706.25 if you are using sampling rate 13.5MHz (ITU601/656) - if you are using different sampling rate hen number of pixels will change - active video duration is constant - other numbers are related to it.

[CaptureCraft]

So yes crop total 16 pixels (left+right) from the 720 frame and display the remaining 704x576 as 4:3

This is what I had originally though, however see my further information below

To address your questions you need to provide information about sampling rate (pixel clock) frequency.

Sorry, don't know what that means.

PAL, 4:3
702x575 -> 1150/1053 (for "diagonal" CRT lines, not sure if one can really make it)
702x576 -> 128/117 (for "horizontal" lines, like from CCD)
704x576 -> 12/11 (MPEG-2 compatible with "macroblocks" of 16x16 pixels)
720x576 -> 16/15 (for digital sources authored after circa year 2000)

@wcndave2. So, for you it is really a choice between 702x576 and 704x576.
If your codec is mpeg-2 based and requires 16x16 pixels macro-blocks, then you need to use 704x576.
Otherwise, you can use 702x576.
Either way, you crop even number of pixels from the sides so you would have the desired width, and then you apply the corresponding ratio.
The best way to handle the head switching noise is to mask it with black rectangle, don't crop it.

More detailed - but matches what I had read regarding 702


However I did another experiment.

I decided to go into photoshop and be a bit more accurate, rather than drawing circles.

I took a capture from a DVD which I also have on VHS.

I then took my VHS and took the 720x576 capture cut it down to 704 and then stretched it to 768x576
I then took the original 720 and stretched it to 768

I overlaid all 3 in photoshop with different transparencies so I could see them all in place.

The 704 stretched to 768 was clearly too wide compared to the DVD screen.
The 720 stretched to 768 was a perfect match.

Unless the DVD is not showing a correct AR, something is not matching the advice you're all giving, which was what I had first assumed to be correct from my reading.

Can you think of any reason why this would occur?

I am capturing in 601 using amarectv FWIW.

DVD is completely different story. There are a lot of tricks. Pure digital way uses 720 pixels, which directly transform to 768 pixels (4x3).
Unless you want to dig really deep, you just have to accept that the true 4:3 source for analog video is 702 pixels. So, the ratio between 702 and 768 is 128/117.

[Sharc]

DVDs are inconsistent. Some early 4:3 DVDs respected Rec.601, means the PAR was ~12/11 and typically showed small left and right borders. Later DVDs violated Rec.601 and put the 4:3 picture into 720x576 frame, means PAR 16/15.
DVD's actually don't even specify the PAR. They just specify the DAR (DisplayAspectRatio), like 4:3 or 16:9 how it should be viewed
All this is confusing at the beginning - and ambiguous or even plain wrong, I agree, but the error is within 2.3%.

[CaptureCraft]

To address your questions you need to provide information about sampling rate (pixel clock) frequency.

Sorry, don't know what that means.

This is simple - active video line in PAL is 52.3uS long (64.0uS - 11.7uS [line blanking interval]) - so you need to use number of pixels with specific sampling rate. Maximum number of active pixels (active video) is 706.25 if you are using sampling rate 13.5MHz (ITU601/656) - if you are using different sampling rate hen number of pixels will change - active video duration is constant - other numbers are related to it.

Not quite so. The total visible line is 53.3µs (720 pixels). Active video is 51.95µs, which is 702 pixels.

[wcndave2]

To address your questions you need to provide information about sampling rate (pixel clock) frequency.

Sorry, don't know what that means.

This is simple - active video line in PAL is 52.3uS long (64.0uS - 11.7uS [line blanking interval]) - so you need to use number of pixels with specific sampling rate. Maximum number of active pixels (active video) is 706.25 if you are using sampling rate 13.5MHz (ITU601/656) - if you are using different sampling rate hen number of pixels will change - active video duration is constant - other numbers are related to it.

Simple for you, I still don't know what that means. I don't think I need to.

[wcndave2]

DVD is completely different story. There are a lot of tricks. Pure digital way uses 720 pixels, which directly transform to 768 pixels (4x3).
Unless you want to dig really deep, you just have to accept that the true 4:3 source for analog video is 702 pixels. So, the ratio between 702 and 768 is 128/117.

I have seen numbers like 59:54, 16:15, 12:11, however not 127:117 before...

From what I understand, 702 is correct, however a lot of the time, due to the blocks etc, 704 needs to be used instead, and the AR error is tiny.

So I have been doing the right things it seems, however on visual inspection I do feel it's just a bit wide. The camera adds 10 lbs!

I am making one more "approximating error" in the process, which is that even with the 704, there are black bars on the sides, and of course the head noise.

Instead of just covering up and having lots of empty space, I have cropped a further 14 from sides and 12 from top/bottom which is as close to 12/11 it's possible to get.

According to calculations, the final AR is therefore off by 0.0967% which I think I can handle.

When I toggle between the original, the 720->768 and the 704->768 one can see the AR changes.

When I have 690x564 and compare, apart from losing the bars/noise and therefore zooming in, you can't see the tiny AR change by eye.

[Sharc]

....Active video is 51.95µs, which is 702 pixels.

Some (TV-) players have a problem with 702 width, therefore use 704 to be on the safe - means most compliant - side.

I have seen numbers like 59:54, 16:15, 12:11, however not 127:117 before...

128/117 does not correctly take the top and bottom half-scanlines into account. You have to dive deep into the subject though, for understanding all of the nitpicking .....

And PAR 16/15 is definitly wrong for standard PAL VHS captures.

[pandy]

I don't think I need to.

Well - isn't this is what you are asking for? Aspect ratio 704/720/768 pixels?

In analog domain they share same display aspect ratio (described by display horizontal and vertical diameters)

768 is not ITU601/656 compliant - this is pre-DTV times when 4xFsc ( Fsc is chrominance subcarrier frequency) was used as a sampling rate (4*4433618.75Hz=17.734475MHz) and as such it should be not mixed with ITU.

[CaptureCraft]

....Active video is 51.95µs, which is 702 pixels.

Some (TV-) players have a problem with 702 width, therefore use 704 to be on the safe - means most compliant - side.

127/117 does not correctly take the top and bottom half-scanlines into account.

There won't be any 702 pixels after stretching them to 768.
And how exactly you can take these half-scanlines when in reality they are shown as full lines anyway?

[pandy]

Not quite so. The total visible line is 53.3µs (720 pixels). Active video is 51.95µs, which is 702 pixels.

Are you able provide source for your claim? I use ITU-R BT.1700 as a reference.

[Sharc]

....Active video is 51.95µs, which is 702 pixels.

Some (TV-) players have a problem with 702 width, therefore use 704 to be on the safe - means most compliant - side.

127/117 does not correctly take the top and bottom half-scanlines into account.

There won't be any 702 pixels after stretching them to 768.
And how exactly you can take these half-scanlines when in reality they are shown as full lines anyway?

This has been discussed and argued thoroughly many times in the past (years ago), here and over at doom9. Just lazy to find it. Anyway we are splitting hairs.

[wcndave2]

....Active video is 51.95µs, which is 702 pixels.

Some (TV-) players have a problem with 702 width, therefore use 704 to be on the safe - means most compliant - side.

I have seen numbers like 59:54, 16:15, 12:11, however not 127:117 before...

128/117 does not correctly take the top and bottom half-scanlines into account. You have to dive deep into the subject though, for understanding all of the nitpicking .....

And PAR 16/15 is definitly wrong for standard PAL VHS captures.

Yes, I was referring to numbers I have seen around, not that I should use.

[Attachment 93615 - Click to enlarge]

[wcndave2]

I don't think I need to.

Well - isn't this is what you are asking for? Aspect ratio 704/720/768 pixels?

In analog domain they share same display aspect ratio (described by display horizontal and vertical diameters)

768 is not ITU601/656 compliant - this is pre-DTV times when 4xFsc ( Fsc is chrominance subcarrier frequency) was used as a sampling rate (4*4433618.75Hz=17.734475MHz) and as such it should be not mixed with ITU.

I thank you for trying to help, however perhaps if you're a deep expert, it's hard to realise what it's like as an amateur/dabbler.
I want to know if a VHS capture should be cropped to 704 before stretching to 768, or stretched directly from 720.
You then start asking about frequencies and so on, about which I have no idea...
I would like to keep it fairly simple - the idea of having super deep knowledge, or else you're not allowed to play, doesn't fit with me

[Sharc]

Not quite so. The total visible line is 53.3µs (720 pixels). Active video is 51.95µs, which is 702 pixels.

Are you able provide source for your claim? I use ITU-R BT.1700 as a reference.

BT.1700 specifies the parameters. Attached herewith for those interested.

[CaptureCraft]

Not quite so. The total visible line is 53.3µs (720 pixels). Active video is 51.95µs, which is 702 pixels.

Are you able provide source for your claim? I use ITU-R BT.1700 as a reference.

Here is the summary:

|--- Front Porch ---|--- H-Sync ---|---------- Back Porch ----------|================ Active Video ===============|
| ____ 1.65 µs ____ | __ 4.7 µs __ | __________ 5.7 µs ____________ | ________________ 51.95 µs _________________ |
| _________________ | ____________ | [*** Color Burst (2.25µs) ***] | ___________________________________________ |
|___________________|______________|______________ _________________|________________________________ ____________|
\__________________________ Horizontal Blanking (12.05 µs) __________/
\______________________________________________ Total Line (64.0 µs) _____________________________________________/

The vertical black bars we have are portions from the back porch and front porch of the next line.

[Sharc]

I want to know if a VHS capture should be cropped to 704 before stretching to 768, or stretched directly from 720.

Hasn't this been clearly answered in post #2?
.. So yes crop total 16 pixels (left+right) from the 720 frame and display the remaining 704x576 as 4:3 (means stretched to 768x576)

and forget the rest of the "video expert babble" .

[wcndave2]

I want to know if a VHS capture should be cropped to 704 before stretching to 768, or stretched directly from 720.

Hasn't this been clearly answered in post #2?
.. So yes crop total 16 pixels (left+right) from the 720 frame and display the remaining 704x576 as 4:3 (means stretched to 768x576)

and forget the rest of the "video expert babble" .

Yes!! I was just explaining to Pandy why I didn't need to know/understand what they were trying to tell me. They seemed to think I did need to know it in order to have the question answered

Thanks!

[VWestlife]

I keep all uncorrupted video info intact regardless if it's "illegal" or not. Sony never consulted the ITU when they decided to make camcorders which record a full 720-pixel-width frame of video with no black borders. (The American government never busted them for using NTSC-J on U.S.-market equipment, either!)

And even if there are a pixels' worth of border area, so what? It's most likely going to end up pillarboxed on a 16:9 monitor anyway. No sense to crop out the black edges only to add them back upon playback.

[pandy]

I thank you for trying to help, however perhaps if you're a deep expert, it's hard to realise what it's like as an amateur/dabbler.
I want to know if a VHS capture should be cropped to 704 before stretching to 768, or stretched directly from 720.
You then start asking about frequencies and so on, about which I have no idea...
I would like to keep it fairly simple - the idea of having super deep knowledge, or else you're not allowed to play, doesn't fit with me

Thread you started triggered my curiosity - i don't understand why you willing to resize 704 to 768 in first place - secondly 704 can't be properly resized to 768 - maximum active video line duration for PAL system is 52.3uS - if you sample your video with ITU601/656 compliant rate (13.5MHz) this gives you 706 pixels so you need to upsample 704 to something like 766 pixels ((704*768)/706)...isn't better to keep original pixel number and signal aspect ratio properly instead blindly resize one number to another especially that 768 pixels is unique for NTSC systems where active video line duration is 53.055uS and sampled with (NTSC) 4*Fsc you get 758 pixels (so it was rounded to nice binary 768 i.e. 48*16).

This why i asked for your sampling rate - if you don't know your sampling rate and reproduction sampling rate then perhaps it is better to not resize your source?

[pandy]

Not quite so. The total visible line is 53.3µs (720 pixels). Active video is 51.95µs, which is 702 pixels.

Are you able provide source for your claim? I use ITU-R BT.1700 as a reference.

Here is the summary:

|--- Front Porch ---|--- H-Sync ---|---------- Back Porch ----------|================ Active Video ===============|
| ____ 1.65 µs ____ | __ 4.7 µs __ | __________ 5.7 µs ____________ | ________________ 51.95 µs _________________ |
| _________________ | ____________ | [*** Color Burst (2.25µs) ***] | ___________________________________________ |
|___________________|______________|______________ _________________|________________________________ ____________|
\__________________________ Horizontal Blanking (12.05 µs) __________/
\______________________________________________ Total Line (64.0 µs) _____________________________________________/

The vertical black bars we have are portions from the back porch and front porch of the next line.

This is not i've asked for. I've asked for source of your claim - worldwide accepted standards says: "line-blanking interval is 12uS +0/-0.3uS" so with line period 64 uS and using maximum standard allowed deviation from nominal line blanking interval (0.3uS) this gives us 64-11.7=52.3uS - if you sample this active video duration with 13.5MHz sampling rate then your visible video has 706.05 pixels (truncating to 706).
So i appreciate your drawing skills but please provide source (worldwide standard) - you can check my calculation with ITU-R BT.1700
available here for free: https://www.itu.int/rec/R-REC-BT.1700-0-200502-I/en

[CaptureCraft]

@wcndave2.
The consensus among "experts" is to store your captured video as is, without any alteration and using lossless codec.
Then, you take this video and convert it/crop it/mask it/whatever you want and watch it/send it to your friends and family, etc.
Why is that? Because today you think one way to process it, tomorrow another way, next year you will decide to upscale it using AI or whatever.
If you have your original capture, then you're good to go. But if it is gone, then it is gone...
Just a point of view...

[CaptureCraft]

Not quite so. The total visible line is 53.3µs (720 pixels). Active video is 51.95µs, which is 702 pixels.

Are you able provide source for your claim? I use ITU-R BT.1700 as a reference.

Here is the summary:

|--- Front Porch ---|--- H-Sync ---|---------- Back Porch ----------|================ Active Video ===============|
| ____ 1.65 µs ____ | __ 4.7 µs __ | __________ 5.7 µs ____________ | ________________ 51.95 µs _________________ |
| _________________ | ____________ | [*** Color Burst (2.25µs) ***] | ___________________________________________ |
|___________________|______________|______________ _________________|________________________________ ____________|
\__________________________ Horizontal Blanking (12.05 µs) __________/
\______________________________________________ Total Line (64.0 µs) _____________________________________________/

The vertical black bars we have are portions from the back porch and front porch of the next line.

This is not i've asked for. I've asked for source of your claim - worldwide accepted standards says: "line-blanking interval is 12uS +0/-0.3uS" so with line period 64 uS and using maximum standard allowed deviation from nominal line blanking interval (0.3uS) this gives us 64-11.7=52.3uS - if you sample this active video duration with 13.5MHz sampling rate then your visible video has 706.05 pixels (truncating to 706).
So i appreciate your drawing skills but please provide source (worldwide standard) - you can check my calculation with ITU-R BT.1700
available here for free: https://www.itu.int/rec/R-REC-BT.1700-0-200502-I/en

The source is the same. The difference is you take max intervals, while I take standard/default.

[Sharc]

I keep all uncorrupted video info intact regardless if it's "illegal" or not. Sony never consulted the ITU when they decided to make camcorders which record a full 720-pixel-width frame of video with no black borders.

That was long after VHS times. SONY (not all) adopted the loose DVD practice. The OP asked about analog VHS tape captures, and once people ask it is so simple - but up to everyone - to do it correctly.