Elecard Buffer Analyser and x264

[Sagittaire]

I actually make a guide for undestanding VBV to compressionist with x264 (and other codec too).

When I make encoding with x264 Elecard Analyser say that Initial CPB delay is only 0.001 sec. With this bad value Analyser find heavy underflow problem.

With x264 vbv-init=0.9 and buffer=30000 maxrate=40000, delay should be very higher. I think that Analyser can't read correctly initial buffer for x264. Anyway Analyser work good for other H264 encoder like Ateme, Elecard and Mainconcept.

Bug for x264? bug for Elecard Buffer Analyser? How corrige my x264 stream for indicate correct initial delay in HRD information?

[Blue_MiSfit]

Did you add nal-hrd information with the --nal-hrd parameter (set to either cbr or vbr of course)?

[spawnbsd]

He must have, as other wise Elecard won't display any stats unless the stream has HRD.

I think the problem likely has something to do with using such a small CPB delay (0.75 seconds). In my experience I tend to get these sorts of errors if I use less than a 1 second CPB delay in x264, if I use a 1, 2 or 4 second CPB delay, Elecard never gives any errors.

[sneaker_ger]

Originally Posted by spawnbsd

I think the problem likely has something to do with using such a small CPB delay (0.75 seconds). In my experience I tend to get these sorts of errors if I use less than a 1 second CPB delay in x264, if I use a 1, 2 or 4 second CPB delay, Elecard never gives any errors.

According to the x264 wiki, --vbv-init means initial fill = vbv-init * vbv-bufsize for vbv-inits < 1, but initial fill = vbv-init (in kilobits) for vbv-inits >= 1.

[spawnbsd]

Originally Posted by sneaker_ger

According to the x264 wiki, --vbv-init means initial fill = vbv-init * vbv-bufsize for vbv-inits < 1, but initial fill = vbv-init (in kilobits) for vbv-inits >= 1.

Sorry, but I'm not quite sure how this relates to his question, much less my answer.

Since he's using a CPB delay of 0.75 (buffer/maxrate), and vbv-init of 0.9, his initial CPB removal delay *should* be 0.675 seconds.

[Lyris]

I noticed this too, and was concerned, but didn't see any playback problems on devices (BD set top players).

Is it possible to test with any other analysers?

[sneaker_ger]

Originally Posted by spawnbsd

Since he's using a CPB delay of 0.75 (buffer/maxrate), and vbv-init of 0.9, his initial CPB removal delay *should* be 0.675 seconds.

You said you've had problems with delays >= 1 second, so I wanted to make sure that you are aware of the differences in --vbv-unit, thinking it might help finding the cause.