SAR , PAR , DAR - some practical questions

[robw1947]

I try to understand how a video file is displayed at a PC monitor and a LCD TV.

The file is 720x576 with SAR = 5:4 , PAR = 1,422 , DAR = 16:9 (G-spot)

Opening this file in Handbrake with Preset 576p25, this program shows:
Anamorphic Automatic
Pixel aspect 64 : 45 (1,422)
Display size : 1024.

I think, I understand this : 720 x 1,422 = 1023,84

So the pixel is a rectangle with a ratio 1,422 : 1 -> .

When I modify in Handbrake Anamorphic Automatic -> None, then Pixel aspect = 1:1 , Aspect Ratio = 15,96 : 1 , Display size = 720 and Storage size = 720 x 406.
Now the pixel is square 1:1.
After the Encoding I have a new file of 720 x 406.

First question:
I suppose, that the number of horizontal pixels is real 720, but the number of vertical pixels, 576 , 406 or ?????


I also understand, that a monitor have square pixels and a TV has rectangle pixels.
My monitor and TV are FullHD (1920x1080).

Playing both files with VLC software at my PC, I see at my monitor the same video images and I suppose at my TV also.
So the software of VLC (?) and of my TV do upscaling depending on the number of pixels in the file (horizontal , vertical) and the values of SAR , PAR , DAR.

Please some simple explanation if possible.

[davexnet]

Some TV's may ignore any non square pixel setting within the file and just use the SAR which would probably be wrong. You'll have to check it on you own tv.
You can use mediainfo to examine the file details

[DB83]

Once you have square pixels (PAR = 1:1) DAR becomes irrelevant and a player will upscale in direct proportion based on the number of horizontal pixels and vertical pixels and as our friend above states not all players respect any actual DAR on non-physical media.

The numbers you quote are correct and like many other things there are other ways to arrive at the same result. 576 vertical pixels is based on the PAL standard. So 576/9*16 = 1024. So you could keep 576 pixels and handbrake should still give you 1024 horizontal. It is just a matter of convenience that it chooses to retain the horizontal width of 720 (also based on the PAL standard) thus 720/16*9 = 406 (rounded)


And it is my understanding that digital tvs also have square not non-square pixels.

[Cornucopia]

Since the vast majority of modern digital TVs are HD or better, they all use square pixels.

Especially WRT interlaced-sourced material, it makes much more sense to keep the vertical resolution the same and adjust the horizontal.

[SOAPBOX RANT]
There is no such thing as "SAR" (aka Storage aspect ratio). That is just a clumsy shortcut some non-video tech person came up with and it has gotten way more traction than it should.
There is PAR or SAR (aka Pixel Aspect Ratio, Sample Aspect Ratio), the shape of the pixel/sample area. Then there is DAR or FAR (aka Display AR, Frame AR), the shape of the overall image (including any encoded black bars).
Then there is Horizontal and Vertical resolution.
Only those figures are ever stored in the files. Never "SAR".
All the formulas work better when working directly with H & V.
...another reason why Gspot should be deprecated.
Please, do not perpetuate this form of quasi-information.
[/END SOAPBOX RANT]


Scott

[davexnet]

I meant SAR as a shortcut for the actual resolution,
Eg. 720x576

[Ennio]

DAR 16:9 means that the video is meant to be displayed in a 16:9 aspect ratio. Whatever the resolution is. Doesn't get more simple than that.
Do realize that pixels in a videofile have no form or dimensions. PAR and DAR are metadata and tell how video should be interpreted/handled.

As for hardware pixels, like Scott says, it's all square pixels now.
In the early days of HD and 'HD-ready' I do remember Sharp introduced a specific type of TV. Oddly, the display had the same amount of horizontal and vertical pixels, but the pixels themselves had 16:9 H/V dimensions.

As to your question to understand how in this light video is displayed on a monitor or TV, usually on a pc the settings of playback-software determine if a video is displayed in the right aspect ratio. As such, a lot can go wrong if settings are not right.
On a standalone TV, the aspect-ratio and/or zoom settings would dictate the view. So, just as with playback-software, also here wrong settings can totally bugger up the picture.

Some TV's may ignore any non square pixel setting within the file and just use the SAR which would probably be wrong.

Same goes for mediaplayers. Even some of today's "modern" ones still struggle with anamorphic. Where the zoom settings of an ancient TV can come to rescue

[hello_hello]

[SOAPBOX RANT]
There is no such thing as "SAR" (aka Storage aspect ratio). That is just a clumsy shortcut some non-video tech person came up with and it has gotten way more traction than it should.
There is PAR or SAR (aka Pixel Aspect Ratio, Sample Aspect Ratio), the shape of the pixel/sample area. Then there is DAR or FAR (aka Display AR, Frame AR), the shape of the overall image (including any encoded black bars).
Then there is Horizontal and Vertical resolution.
Only those figures are ever stored in the files. Never "SAR".
All the formulas work better when working directly with H & V.
...another reason why Gspot should be deprecated.
Please, do not perpetuate this form of quasi-information.
[/END SOAPBOX RANT]

Scott

How did you end up with such a bee in your bonnet over "storage aspect ratio"?
"Frame aspect ratio" makes some sense, even if it doesn't seem to say anything about how much of the storage aspect ratio constitutes a frame, and it does sound like a clumsy shortcut a non-video tech person came up with because they were pissed that sample aspect ratio gained enough traction to make SAR ambiguous.

[Sharc]

It would be helpful if people would write out (just once in brackets, for example) what they understand with the acronyms they are using, just to prevent misunderstandings and confusion in the discussion. SAR, PAR, DAR, MAR, CRF, CFR, FPS, fps, CBR, VBR, ABR to name a few which are often getting mixed and mingled or are subject to personal interpretation.

[lollo]

It would be helpful if people would write out what they understand with the acronyms they are using

I agree! Also the software should do it, IIRC ffmpeg, while encoding, calls SAR what we normally call PAR "Pixel Aspect Ratio"

[Ennio]

It would be helpful if people would write out what they understand with the acronyms they are using

I agree! ...IIRC ffmpeg... ...SAR

lollo (laughing out loud, like outrageously)

[Sharc]

[QUOTE=lollo;2648830]

... IIRC ffmpeg, while encoding, calls SAR what we normally call PAR "Pixel Aspect Ratio"

The question here is just who are "we". SAR (SampleAspectRatio) has been introduced by international standards like H.264, H.265 where PAR (PixelAspectRatio) has been dropped. Justification was - if I got it right - that the SAR (SampleAspectRatio) describes the sampling process as horizontal vs. vertical spacing of the samples, and a sample does not have a dimension. Hence the encoders x264 and x265 are using the --sar parameter as well.
Not to argue what is better or more correct. Just pleading for clarification what one understands when using "well known" but possibly ambiguous acronyms.

[lollo]

Not to argue what is better or more correct.

Yes. That's why I support your idea "write out (just once in brackets, for example) what they understand with the acronyms they are using" also for the report of the software. SAR is an acronym for Sample Aspect Ratio but also for Storage Aspect Ratio

[dellsam34]

Yes the SAR (sample or sampling aspect ratio) is the number of pixels the frame is sampled at, Alone it doesn't say anything about how those pixels will be displayed until a PAR flag (pixel aspect ratio) is provided, Or a DAR (display aspect ratio) is provided. If the PAR is provided the DAR is ignored as far as I know.

There are a lot of options and configurations based on what software and standard used, Rule of thumb, never down scale in any direction, Always up scale to avoid loosing resolution. Use the PAR flag whenever possible to keep the pixel count intact, if not use a resize scheme that gives you the correct 4:3 DAR if the pixels were considered square, In your case resize from 720x576 to 768x576 for a SD frame, or 1440x1080 for a HD frame. If the correct DAR is 16:9 not 4:3, resize to 1024x576 for a SD frame or 1920x1080 for a HD frame.

[Cornucopia]

In most of the official tech docs, the original designations were PAR and DAR (at least up through the MPEG2 era). Those migrated around the time of MPEG4 to SAR (sample) and FAR, respectively, for reasons that are in line with what Sharc mentioned.

How I got that "bee in my bonnet", is pretty clearly laid out historically here on these forums, considering all the misuse and confusion about them I've seen and needed to correct over the years. This site has been quite good at drumming into users the basic formula of "Bitrate = Filesize * RunningTime". Too bad they cannot seem to do it as well with "DAR = PAR * H / V" (or alternately using new terms "FAR = SAR * H / V").

For example, I've seen users floundering trying to do a resize given only certain variables including Storage AR, Yet using the proper formula, it is a breeze to drop in DAR1=PAR1*H1/V1 and DAR2=PAR2*H2/V2 and solve using equivalents (lets say DAR1=DAR2). Plus, if one stops relying on the shortcut, the bigger picture -pardon the pun- is easier to understand.

What I don't understand is why folks keep falling back on that habit of using Storage AR, when it is ultimately less helpful and certainly is not backed by any engineering standards.

In order to avoid SAR/SAR confusion, I will probably be retaining DAR & PAR in my formulae, because that is still quite clear what is refered to.

@robw1947, I apologize if my rant has sidetracked your topic.

Scott

[dellsam34]

Even in some scripts it says setsar which clearly refers to the Pixel AR such as "setsar=SAR=10/11", Yes very confusing for a lot of people. I agree SAR as in Storage AR should be dropped and replaced by H/V resolution.

[lollo]

SAR as in Storage AR should be dropped and replaced by H/V resolution.

+1

[Ennio]

Does somebody recognise this?
I remember the early days struggling with these different acronyms and ratio relations. It was the dvd era where everything was anamorphic. Drove me crazy.
"PAR x SAR = DAR" etc. Yeah, great. I had to see for myself. Even in pc times, for me a lot of things come together when I visualise. So I took out pencil & paper and started drawing. After having done the math (no calculator, just fractions) a couple of times with different example-values for pixel amounts and DAR, it all made sense. Once you go back to basics and work things out, you'll never forget.

[lollo]

Yes!
I have an excel spreadsheet with all the math behind PAR, SAR (Storage AR), DAR, SAR (Sample AR), FAR, 576 scanlines or 574+2*1/2=575 scanlines, 702 or 704 active area inside 720 width, ITU-R BT.601-4, MPEG-2, MPEG-4

[Sharc]

...Once you go back to basics and work things out, you'll never forget....

Agree. Most important is to understand the basics, always. Takes some time and efforts though. And only then one will better understand tools and GUIs.

[robw1947]

What I don't understand is the number of vertical pixels, is this a real number?
I started with a file 720x576 with pixel aspect = 1,422 and DAR = 16:9.
So 720 x 1,422 -> 1024 Hor and 16:9 -> 576 Vert.

After my modification in Handbrake Anamorphic Automatic -> None and doing a export / produce, I got a 720x406 file with pixel aspect 1,0 and DAR = 16:9
So, with this file 720 Hor and 406 Vert.

As I understand 1024 is not the number of pixels but 720 streched pixels.

I am afraid, that there is more then just pixels, these pixels with all kinds of info belong to / creates a frame.
And this frame must be upscaled to my FullHD panel in my monitor and TV.

Im am from 1947 and have learned in the late 60 and early 70 the principals of a analog amplifier, radio, TV (B/W), taperecorder.
The TV with interlaced lines.
After my study I got a job in the software brache, administrative processes, so not the technical world.
Until now never thought about the digital method / operation of modern devices.
Perhaps this makes it difficult for me to understand with only my knowledge of the analog devices.

[Sharc]

The PC monitor or digital TV screen has square pixels only. All pixel values are real numbers.

The 720x576 video file has 720 pixels horizontal and 576 pixels vertical dimension. When you watch this with your monitor you would see an anamorphic distorted picture with a width to height ratio of 720/576 = 1.25, means it would appear horizontally squeezed. The 16:9 DAR (DisplayAspectRatio) flag informs the player that it should "stretch" the picture to fit a 16:9 aspect ratio.
This conversion can be done by

either
a) streching the picture horizontally by 1.4222 (=PAR, PixelAspectRatio), and keeping the vertical size untouched, i.e 720x1.4222:576 = 1024:576 = 16:9. So the "new" picture has 1024 real square pixel in horizontal direction and 576 real square pixels in vertical direction, and the aspect ratio is 16:9 as requested.

or
b) leave the horizontal size untouched but shrink the vertical size by 1.42222, i.e. 720 : (576/1.42222) = 720:405 = 16:9. So the "new" picture has 720 real square pixels in horizontal direction and 405 real square pixels in vertical direction, and the aspect ratio is 16:9 as requested.

So both methods a) and b) deliver an undistorted picture with a width to height ratio of 16:9 = DAR (DisplayAspectRatio)

Method b) is apparently what Handbrake does when you request it to convert the anamorphic 720x576 source to non-anamorphic square pixels.

...... is the number of vertical pixels, is this a real number?

All pixels are real numbers as long as the signal stays in the digital domain. It is eventually left to the player or TV to zoom the 1024:576 or 720:405 square pixels to "full screen" by applying a linear scaling.
If however one plays a DVD with a DVD player and connects the (legacy) DVD player to the TV via the SCART, Composite or S-Video interface for example, the DVD player converts the digital information (the "pixels" stored on the DVD) to an anlog signal and presents it as an analog signal to the TV's analog SCART or Composite or S-Video input. The analog video signal has no pixels anymore. It is an analog waveform which the TV processes further on according to its technology (legacy analog CRT TV or current digital TV).

[dellsam34]

What I don't understand is the number of vertical pixels, is this a real number?
I started with a file 720x576 with pixel aspect = 1,422 and DAR = 16:9.
So 720 x 1,422 -> 1024 Hor and 16:9 -> 576 Vert.

I have a feeling that you have 4:3 contents not 16:9, Can you post some screen shots? 720x576 refers to non square pixel 4:3 D1 frame unless it is from a broadcast recording that used the widescreen signaling.

[Sharc]

[QUOTE=dellsam34;2649007]

I have a feeling that you have 4:3 contents not 16:9, Can you post some screen shots? 720x576 refers to non square pixel 4:3 D1 frame unless it is from a broadcast recording that used the widescreen signaling.

720x576 is standard PAL DVD resolution. It applies for 4:3 or 16:9 content, depending on the DAR flag. But it's always recommended to upload a sample to prevent misunderstandings.

[Ennio]

Isn't PAR for 16:9 PAL ( (16x576) : (9x720) ) which simplifies to 64/45 which is approx 1.422? Handbrake would be right here.

Of course bad AR metadata can bugger up things

[Sharc]

Isn't PAR for 16:9 PAL ( (16x576) : (9x720) ) which simplifies to 64/45 which is approx 1.422? Handbrake would be right here.

Yes of course, nothing wrong with it.
And a 720x576 PAL DVD for 4:3 DAR (DisplayAspectRatio) would have a PAR (PixelAspectRatio) of (4x576) : (3x720) = 2304 : 2160 = 16/15=1.06666....
Remember however that a DVD (mpeg2 video) does not explicitly specify its PAR. It only specifies the DAR as either 4:3 or 16:9. The PAR for DVDs can then be calculated as above, but one can't read this number directly from the videostream or from the DVD container.

Edit:
@robw1947: For a more general explanation beyond DVD see the discussion here:
https://forum.videohelp.com/threads/404855-SAR-for-544x576-4-3#post2649073
Note that SAR (Sampling Aspect Ratio) which is used there is basically the same as PAR (Pixel Aspect Ratio) used here.

[robw1947]

I have done research to refresh my knowledge about analog TV and the relation with a DVD player.
PAL : total 625 lines, visible 576.
This explanes the number of 576.
But why 720 pixels?
That question had also someone else.
Answer at that site:
For PAL the allowed bandwidth is defined to be 4.2 - 6 MHz (depending on the PAL variant).
The whole line lasts 64 µs but the visible part is 51.9 µs.
Again the rest is used for the beam in CRT to return from the right to the left.
The bandwidth in the course of the duration of the visible part could allow up to about 622 "pixels" per line.
Note also that the colour information has a lower bandwidth so its horizontal resolution is lower in analogue PAL.

At a Dutch site, I read:
DVB : the horizontal resolution is max 720 pixels and must be dividable by 16 because of the MPEG compression.

The first generation of DVD-payers had a Scart output, being analog.
So, I think, the software in the player did the conversion.

All with all, interesting stuff for me after all these years from analog to digital.
And for me, from a main computer with puch cards to what there is now.

[Skiller]

In my experience if the frame size is 720x576 and the DAR flag is 4:3 – and especially (but not necessarily!) if there is a total of about 16 pixels of black padding on the sides – the SAR/PAR is 12:11 (not 16:15).

By default I would always use 12:11. I would actually even go so far as to advice not to ever use 16:15 unless you know for sure it is correct.
12:11 is correct for basically anything that was produced in the 4:3 aspect ratio era and also most digital originated 4:3 stuff.


Different story if the DAR flag is 16:9. In that case 720x576 = 16:9 turns out to be the right choice in my books.

[Ennio]

Could you please confirm if I understand this side-padding correctly?

In such a case, do you mean that the 704x576 videopart itself is already meant as 4:3 and the padding is just "black fill" for 720x576 delivery?

[Sharc]

In my experience if the frame size is 720x576 and the DAR flag is 4:3 – and especially (but not necessarily!) if there is a total of about 16 pixels of black padding on the sides – the SAR/PAR is 12:11 (not 16:15).

Fully agree IF the DAR flag is 4:3 AND there are small black borders left and right, because the 4:3 picture is then typically represented by the center 704x576 portion of the frame. It's just unfortunate that "modern" DVD players with digital interface usually ignore this and display the entire 720x576 frame (means including the left and right black borders) as 4:3, i.e. the active picture becomes 2.4% horizontally squeezed. Legacy DVD players however with analog SCART interface used to play the picture correctly (at least my old one did). Greetings from the circle test. Modern times.
(The OP however wrote in his first post that his source is DAR flagged as 16:9).

Edit:
Remember the dicussion here:
https://forum.videohelp.com/threads/404626-How-do-i-upscale-PAL#post2646882

[Sharc]

Out of curiosity - or just for fun - I let you guys guess or analyze whether the PAR (pixel aspect ratio) of these 720x576 4:3 PAL DVD clip is 12:11 or 16:15

Frankly I don't know as I didn't find a nice circle for 4-3 (1).m2v, but I would vote for 16:15 for this clip.

And a second example 4-3 (2).m2v with circles where the winner is 12:11 as I see it.