non-square pixels and pixel aspect ratio

[korben_dallas]

Been doing some light research lately on various codecs, and I've come across something that is not making sense to me: non-square pixels.

The pixel aspect ratio is different than the display aspect ratio.

Can someone put this into layman's terms for me?

[stanger89]

DVDs, 720x480, ie 1.5:1 pixel aspect ratio, with a flag that says if they represent 1.33:1 or 1.78:1 image.

Alternative 2: Some plasmas are 16:9 (1.78:1) but have a resolution of 1024x768.

Alternative 3: Front projector with an anamorphic lens, 2.35:1 DAR, with a resolution of 1280x720.

[blade]

Originally Posted by stanger89

DVDs, 720x480, ie 1.5:1 pixel aspect ratio, with a flag that says if they represent 1.33:1 or 1.78:1 image.

That one confuses me. I understand with DVDs, but what about recorded content?

If I set the Nvidia decoder to content default I get a very narrow image. I'm assuming 1.33:1. Using anamorphic/raw the image is 1.5:1, which is what live TV through my TV's tuner is. Which one is maintaining the true aspect ratio of the recording? I thought 1.33:1 did, but it's not the ratio either of my 16*9 TV's use for 4*3 content.

[Narflex]

This is all handled automatically in V5.1. All you need to specify is the display aspect ratio of your monitor/TV.

[blade]

Wow, I can't wait for v5.1. From all the hints it seems like there are going to be a lot of improvements.

[sleonard]

v5.1? What v5.1?

Scott

[stanger89]

Originally Posted by Narflex

This is all handled automatically in V5.1. All you need to specify is the display aspect ratio of your monitor/TV.

I don't think that's quite what blade meant (sounds like a decoder issue), but since you let the cat out of the bag (pun not intended )...

5.1 has an option to set the Display Aspect Ratio, and will then adjust the size/ratio of video so that it is correctly displayed.

For example if you've got one of the 1024x768 16:9 plasmas, currently Sage assumes it's a 4:3 display and will letterbox 16:9 content (incorrectly). In 5.1 you set the Display Aspect Ratio to 1.78, and Sage will correctly display 16:9 content without bars and will pillarbox 4:3 content.

Originally Posted by blade

That one confuses me. I understand with DVDs, but what about recorded content?

All, recorded content is 4:3 (1.33:1), regardless of resolution. I run everything with "content default" set.

I guess that leads to some questions:
What resolution are you running?
What aspect ratio setting are you using in Sage?
What renderer?

[blade]

I'm using anamorphic/raw in the Nvidia decoder, source in Sage, VMR9, and have my TV set to use standard 4*3.

That leaves me with 3" black bars on each side in Sage. Viewing live TV through the TV tuner I have 4.75" black bars; however the TV masks the edges to hide the edge of the video because the station logo is the same distance from the edge of the screen and all other measurements I've taken are pretty close.

If I use content default in the nvidia decoder and in Sage I have 5.5" black bars on each side, which is a good bit narrower than live TV.

Is it possible my TV doesn't show true 4*3 ratio with live TV and is displaying it as 1.5 like anamorphic/raw does in the decoder? Or am I missing something?

[korben_dallas]

Originally Posted by Narflex

This is all handled automatically in V5.1. All you need to specify is the display aspect ratio of your monitor/TV.

Is the SageTV application aware of non-square pixel content, or is this something that is abstracted by the capture device and its drivers?

What pixel aspect ratio is content which is captured via the composite or s-video input, say from a video camera? How about DV through firewire?

[stanger89]

Originally Posted by blade

Is it possible my TV doesn't show true 4*3 ratio with live TV and is displaying it as 1.5 like anamorphic/raw does in the decoder? Or am I missing something?

Actually, it's probably overscan. Without knowing how big your TV is, I'm not sure if 2" is a lot or not, but I'd guess the whole picture is bigger through the TV than Sage.

You can always adjust the vertical/horizontal zoom in Sage to tweak the display if they're not right.

Originally Posted by korben dallas

Is the SageTV application aware of non-square pixel content, or is this something that is abstracted by the capture device and its drivers?

This is a playback thing, not a capture thing, captures are handled the same.

What pixel aspect ratio is content which is captured via the composite or s-video input, say from a video camera? How about DV through firewire?

I believe NTSC is generally 0.9:1 (ie 720x480 is 1.33:1 instead of 1.5).

[korben_dallas]

Originally Posted by stanger89

This is a playback thing, not a capture thing, captures are handled the same.. I believe NTSC is generally 0.9:1 (ie 720x480 is 1.33:1 instead of 1.5).

So if I could look at the actual frame, it would be squished or stretched?

[stanger89]

If you look at an actual NTSC frame, it's stretched slightly horizontally. Probably wouldn't even notice without a test pattern.

[korben_dallas]

Originally Posted by stanger89

If you look at an actual NTSC frame, it's stretched slightly horizontally. Probably wouldn't even notice without a test pattern.

So when it is encoded with MPEG-2 by the capture card and stored on the hard drive by SageTV, is it still in a non-square pixel format? Or was that lost when the capture card captured it? Or am I missing something subtle here?

And is it still interlaced or progressive or ?

[stanger89]

Originally Posted by korben_dallas

So when it is encoded with MPEG-2 by the capture card and stored on the hard drive by SageTV, is it still in a non-square pixel format?

I believe so, I'm pretty sure they're recorded a 4:3, 720x480, which is (slightly) non-square.

Or was that lost when the capture card captured it? Or am I missing something subtle here?

You might be thinking of true "anamorphic" content. If you record something that's 16:9 with the capture card, that is lost, because everything is recorded as 4:3.

And is it still interlaced or progressive or ?

Interlaced.

[blade]

Since this thread has shifted a little more towards capture, I have another question. I know mpeg2 can be flagged as either film or video. If I set the Nvidia decoder to automatic everything plays back as video. If I use smart some play back as film and others as video. The problem is I have very poor detection and end up with stuttering and artifacts on some shows while others look great.

Is everything recorded by Sage flagged as video or does it not use a flag at all? I could have sworn at one time using automatic with Sage did both film and video, but I could be mistaken.

There are utilities that allow the flag to be set on previously recorded mpegs and I tried one of them, but it wouldn't work on a program stream. It will only work on an elementary one. If I could figure out a way to get proper detection I'd be very happy.

[stanger89]

Originally Posted by blade

Is everything recorded by Sage flagged as video or does it not use a flag at all?

Everything recorded by the Hauppauge cards (and I imagine others as well) is recorded as video.

I could have sworn at one time using automatic with Sage did both film and video, but I could be mistaken.

I doubt automatic ever did film on recordings, but smart should alternate, because there's a lot of film sourced content on TV.

There are utilities that allow the flag to be set on previously recorded mpegs and I tried one of them, but it wouldn't work on a program stream.

Without re-encoding, that's probably not a good idea, because they truely are video recordings, ie 59.94 fields/sec.

It will only work on an elementary one. If I could figure out a way to get proper detection I'd be very happy.

I have no trouble with using Smart (or IVTC enabled).

[korben_dallas]

As I understand it, the material flag is often wrong, which is why the Nvidia decoder does per-pixel adaptive.

[korben_dallas]

Originally Posted by stanger89

I believe so, I'm pretty sure they're recorded a 4:3, 720x480, which is (slightly) non-square.. interlaced

Do you know if this is also the case with digital formats, like OTA HD?

I've read a few forum posts (in other forums) about ppl working with DV that is both non-square pixel ratio and interlaced, which seemed odd to me that a DV camera would not record in square pixel progressive format.

So I take it that it is best to transport/recapture/store the video as near to its native form as possible?

[stanger89]

Originally Posted by korben_dallas

Do you know if this is also the case with digital formats, like OTA HD?

Sage doesn't do anything special with those. However they're almost always either 1080i60 or 720p60.

Or do you mean square pixels, AFAIK HD is always square. Digital cable can be non-square, last time I recorded digital SD, it was like 528x480 (4:3), and I think HD light on satellite is often 1280x1080 (16:9)

I've read a few forum posts (in other forums) about ppl working with DV that is both non-square pixel ratio and interlaced, which seemed odd to me that a DV camera would not record in square pixel progressive format.

DV is, AFAIK, 720x480 and 4:3, ie non-square.

So I take it that it is best to transport/recapture/store the video as near to its native form as possible?

Generally, as few transformations as possible is optimal.

[blade]

Originally Posted by stanger89

I have no trouble with using Smart (or IVTC enabled).

I have heard the quality of the capture affects how well smart detection works. I guess my analog cable isn't all that clean and causes problems.

Some shows are perfect and others unwatchable with smart. A bios update for my 6600GT did seem to help some though.