Average/Peak Bitrate for AVCHD

[kenguru2005]

Hallo,

I am wondering if the specifications of AVCHD standards (actually in v2 28 MBit/s) mean the average bitrate or the peak bitrate?

Greetings

[MrVideo]

AFAIK, average. I personally set the x264 encoder to 20 Mbps for Blu-ray content.

Not worth going any higher.

[Ghitulescu]

Where lies the problem, that the maximal bit rate must be below to 24 Mbps? You can't have an average bitrate topped at 24 by the definition of average ... :)

http://www.avchd-info.org/format/OverviewChart1.jpg

[kenguru2005]

No ... but concerning VBR encoders you can set peak bitrates about 40 MBit, lower bitrates about ie 5 MBit, and mean/average bitrates 24 MBit. Would this blow out the frame of AVCHD specification?

[MrVideo]

No ... but concerning VBR encoders you can set peak bitrates about 40 MBit, lower bitrates about ie 5 MBit, and mean/average bitrates 24 MBit. Would this blow out the frame of AVCHD specification?

I'd have to doublecheck, but I think I set th peak rate to 24 Mbps when I told my script I was encoding for AVCHD.

But since I now only recode for Blu-ray, I don't worry about the numbers :D

[kenguru2005]

The br spec also sets a limit ... perhaps 48 MBit/s, will it be speca conform to have peaks of let´s say 62 MBit/s?

My background:
I´m cutting 1080p50 h264 video footage with 28 MBit/s bitrate. The final m2ts that I produce should be conform to the new AVCHD standards:

http://www.avchd-info.org/format/OverviewChart2.jpg (http://www.avchd-info.org/format/OverviewChart2.jpg/)

Now the question is: am I limited to 28 MBit/s only concering average bitrate of the movie, or am I also limited to 28 MBit/s concerning peak bitrate (ie could I have 40 MBit/s at fast movements if the "over-all-mean-bitrate" of the movie stays 28 MBit/s)?

[Ghitulescu]

The br spec also sets a limit ... perhaps 48 MBit/s, will it be speca conform to have peaks of let´s say 62 MBit/s?

My background:
I´m cutting 1080p50 h264 video footage with 28 MBit/s bitrate. The final m2ts that I produce should be conform to the new AVCHD standards:
Now the question is: am I limited to 28 MBit/s only concering average bitrate of the movie, or am I also limited to 28 MBit/s concerning peak bitrate (ie could I have 40 MBit/s at fast movements if the "over-all-mean-bitrate" of the movie stays 28 MBit/s)?

There's no average bitrate in standards, there are exclusively peak bitrates (some do have also a minimum bitrate).

[MrVideo]

There's no average bitrate in standards, there are exclusively peak bitrates (some do have also a minimum bitrate).

Ah, good to know. I never dug too deep into that, since all my encodes worked :D

BTW, your signature should probably read:

It's not difficult to find the ideal woman. What is difficult is hiding her from the wife.

[kenguru2005]

Ok found some more information in a german board: it is in fact an over all average limit, peaks can be higher as long as the mean bitrate does not exceed those 28 MBit/s.
Checked my Sony Cam: it captures 1080p50 "around" 28 MBit/s reaching higher peaks but most time staying under the limit.

[Ghitulescu]

The reason for imposing peak bitrates is that a particular medium should be able to cope with it, otherwise buffer overflow conditions may occur. That a particular medium, through hand picking, can sustain higher bitrates (like using a class 6 SD card instead of a class 4 one) doesn't mean that there there have been meant average. Then these restrictions are used to design the interfaces and the codecs/algorithms.

Any player would have a problem in correctly displaying a 10 seconds long video sequence of ~48Mbps that comes after 10 seconds of "digitally blank" (so the average will be ~24Mbps). Of course, unless the player has been designed to support ~48Mbps as a peak bitrate.

[kenguru2005]

I don´t think, encoders wouldn´t or shouldn´t produce a bit rate delta of 48 MBit/s (i.e. coming from 0 goin´to 48).
I think they work in an adaptive way, holding bit rate pronounced under the line of max, keeping the diviation small, and "accelerating" in a moderate extent if necessary.

[Ghitulescu]

If they won't, how can they obey the average bitrate? The case is purely fictional to underline the idea.
Ok, taking the "moderate way", that means the peak bitrate would also be more than 48 for this fictional example :) because of the lost bits during the "acceleration" slope.

Another example: I have in school 3 marks of 2 (out of 10, 10 is the best). I badly need an average of 5 not to repeat the year. Well, I need for the next mark to be a 14 (2+2+2+14 all divided by 4, the number of marks). Well, bad news, I can't have a 14. The school standard limits the peak mark at 10 (28Mbps) while requires in the same time an average of at least 5. There's also a lower limit (mark 1, there's no mark 0), similar to the lower limit for bitrate in MPEG-2/DVD standards.

PS: I don't own yet an HD camcorder, so I cannot test myself if the case average bit rate ~ peak bit rate, which would generally mean a sort of CBR (a sort of VBR with an extremely low amplitude), is a reality or a misinterpretation of the values.

See below what I mean
http://www.dvdbeaver.com/film2/DVDReviews48/i_love_you_beth_cooper_blu-ray/bitrate_graph_ILYBC.gif
This is not VBR.
http://www.hitchcockwiki.com/files/captures/5402_bitrate.jpg
This looks more like a VBR.

[mpucoder]

AH, the old maximum versus average bitrate problem. OK, here's what happens. The bitrate stated is the system bitrate, which is a constant rate. You can fill it with data and overhead until it is full or let some bits slip by if not needed. Now, the streams are multiplexed into the system with 2 delays, one to allow the decoder time to decode, and the other to allow for peaks exceeding the system rate. You still cannot exceed the limit, but, due to the delay, the instantaneous bitrate (the number of bits divided by the duration of one frame) can exceed the system bitrate. This is at the cost of the next picture. But this is very acceptable because all encoding schemes use full pictures followed by differential pictures. There is another constrant, too, that is the size of the buffer. This determines the absolute maximum amount of data for one picture.

[kenguru2005]

Thank you, now I got it ....