References & b-pyramid?

[rwill]

Originally Posted by benwaggoner

MPEG-2 had a serious fall off in reference quality, which is why such short GOPs were used. A big part of that is MPEG-2 used straight up floating-point DCT, not an integer iDCT which was standard by VC-1/H.264. Different decoders would do the DCT in different precision, so the longer the GOP, the bigger the gap between how different devices would decode the same frame! 14 frames of IbbP kept the max P-frames per GOP very low.

Well, IEEE 1180-1990 specified the precision requirements for the IDCT. Its just that no one really cared as can be seen on older DVDs. The inaccuracies creeping in were also not really visible on CRTs, but are quite noticeable on flat monitors. There is even a test stream in the official (?) Mpeg-2 conformance package that tests IDCT accuracy over around 200 inter pictures. Mpeg-2 is fine really, its just that some people screwed up.

Oh and I think its 15 frames keyint, not 14.

[benwaggoner]

Originally Posted by rwill

Well, IEEE 1180-1990 specified the precision requirements for the IDCT. It's just that no one really cared as can be seen on older DVDs. The inaccuracies creeping in were also not really visible on CRTs, but are quite noticeable on flat monitors. There is even a test stream in the official (?) Mpeg-2 conformance package that tests IDCT accuracy over around 200 inter pictures. Mpeg-2 is fine really, it's just that some people screwed up.

Yeah, if everyone had followed the IEEE spec it would have been fine. But once there's an installed base with an incompatibility, it's really hard to roll back from there. Like with AV1's Film Grain Synthesis. Most devices do it well know, but some early ones don't so I don't think anyone is actually using FGS. To use it one would need a way to filter out nonconformist devices and send them something else instead.

Oh and I think it's 15 frames keyint, not 14.

It was 14 relative to 24p, which was "fake" 24p with field repeat tags for DVD. HD-DVD was, IIRC, 24p with frame repeat tags to pad it out to 29.976, which was easily reversible.

[HD MOVIE SOURCE]

Originally Posted by benwaggoner

I believe 12 was the default, but the DVD and HD-DVD specs specified a maximum GOP duration of 0.6 seconds. Those extra two frames did help reduce ABR. And it was Open GOP, thank goodness, to reduce keyframe strobing.
Only IDR frames are independently decodable, so are the best ones for random access. Some decoders do random access by following the reference hierarchy and only decoding the frames needed to decode a given frame. Decoding just P-frames is a lot lighter weight intermediate.


On Blu-ray you can use a single-slice 2 second GOP if encoding to Level 2.0 15 Mbps peak. That said, it's not like commercial Blu-rays are hurting for space. The 50 GB was designed around MPEG-2, and no one had a clear idea of how GOOD/efficient H.264 would get over the years.
The shorter GOPs were important for optical media to make random access less painful. The latency of a random access read of an optical desk makes even a 5400 rmp HD seem incredibly spritely.


MPEG-2 had a serious fall off in reference quality, which is why such short GOPs were used. A big part of that is MPEG-2 used straight up floating-point DCT, not an integer iDCT which was standard by VC-1/H.264. Different decoders would do the DCT in different precision, so the longer the GOP, the bigger the gap between how different devices would decode the same frame! 14 frames of IbbP kept the max P-frames per GOP very low.


Please report your results!

--radl=? What value do you set it to? Can you set it to the same value as the number of bframes? I will try what is suggested in x265 documentation of 3 with an open-gop, which I believe you can use together with this setting.
If it can reference outside of the GOP, should I also increase my rc-lookahead value to 25? Because this makes it so you can look into the first frame of the next GOP right?

I wonder if --radl is compatible with 4K Disc? I believe open-gop is, so it would be interesting if this could work.

[HD MOVIE SOURCE]

Tested --radl=3 last night, and now instead of having every GOP ending with a pframe, they now end with bframes. I think every single GOP ended with a frame. This means that it had to have looked into the next GOP to gather information right?

I didn't haven to change the rc-lookahead either, I kept it at 24, but I may check to see if a further lookahead can now improve things.

[benwaggoner]

Originally Posted by HD MOVIE SOURCE

Tested --radl=3 last night, and now instead of having every GOP ending with a pframe, they now end with bframes. I think every single GOP ended with a frame. This means that it had to have looked into the next GOP to gather information right?

I didn't haven to change the rc-lookahead either, I kept it at 24, but I may check to see if a further lookahead can now improve things.

--radl 2 or 3 is generally good. Ideally it would be adaptive, but it isn't.
It's not that the GOP is ending with a B-frame, but that GOPs are starting with B-frames that are only forward references.

[jpsdr]

Originally Posted by HD MOVIE SOURCE

--radl=? I will try what is suggested in x265 documentation of 3 with an open-gop

Where did you read that ? The documentation says it requires glosed gop...

[benwaggoner]

Originally Posted by jpsdr

Where did you read that ? The documentation says it requires glosed gop...

Yeah. --radl is essentially the replacement for Open GOP when Closed GOP is required for adaptive streaming.