Problem after upgrade to 1+ serviio .. mkv loading errors

[jhb50]

The Samsung specs http://downloadcenter.samsung.com/content/UM/201112/20111208185713733/Web_GSDVBEUA_Eng.pdf indicate that your TV does not support either of those files natively. While you may have found otherwise for those 2 particular files, the Serviio profile should be left as is in order to ensure all files will play correctly. If bitrate is causing stuttering due to your network capacity or CPU limits, you can set maxVbitrate to a lower value than 20000.

The Samsung specs http://downloadcenter.samsung.com/content/UM/201112/20111208185713733/Web_GSDVBEUA_Eng.pdf indicate that your TV does not support either of those files natively. While you may have found otherwise for those 2 particular files, the Serviio profile should be left as is in order to ensure all files will play correctly. If bitrate is causing stuttering due to your network capacity or CPU limits, you can set maxVbitrate to a lower value than 20000.

limit was not network (cabled gbit,maxed at 100mbit) , but 212+ cpu with transcoding

but its not just these files.. same with a lot of other mkv files as well (and i have a lot) , they all work fine without transcoding including those with 5.1 headers
so seems samsung is less picky than advertised (goes for most scene mkv files as well)

in 2012 there has been a firmware samsung update as well for the tv

[jhb50]

they all work fine without transcoding including those with 5.1 headers

Yes they should per the samsung spec including those with false 5.1 headers, but Serviio must respect the headers in case the file is really a 5.1 which will fail per the specs and my experience. Real 5.1 files are few and far between. Planet Earth is one, but most are mislabeled 4.1's.

[zip]

they all work fine without transcoding including those with 5.1 headers

Yes they should per the samsung spec including those with false 5.1 headers, but Serviio must respect the headers in case the file is really a 5.1 which will fail per the specs and my experience. Real 5.1 files are few and far between. Planet Earth is one, but most are mislabeled 4.1's.

Do you have one that has header > 4.1 and doesn't play? What profile would it fit according to: http://wiki.serviio.org/doku.php?id=get_h264_level

[jhb50]

Do you have one that has header > 4.1 and doesn't play?

Yes #1 and #3 below with transcoding off but that doesn't tell us anything. You need a file with header >4.1 and ref frames below the 4.1 level that does play natively and a file with header < 4.2 and ref frames greater than the 4.1 level that does not play in order to prove the TV only uses ref frames.
#3 and #4 below do that.

First I think the wikipedia table is better than the formula for determining level, so I'll refer to that below as well as the formula. http://en.wikipedia.org/wiki/H.264/MPEG-4_AVC

For 1280 up to 5 ref frames = 3.1, 6 to 9 ref frames = 4.0
For 1920 up to 4 ref frames = 4.0, 5 to 13 ref frames = 5.0 and 14 to 16 ref frames = 5.1

Here are my test files.

#1 5-5 A clip of the original Planet Earth 1920x1080 w/16 ref frames and valid 5.1 header (16=5.1 & 1920x1080x16/256= 129,600 = 5.1)
#2 4-4 Same clip w/3 ref frames and valid 4.1 header (3=4.0 & 1920x1280x3/256= 24,300 = 4.0)
#3 5-4 Copy of #1 (16 ref frames) with header changed to 4.1
#4 4-5 Copy of #2 (3 ref frames) with header changed to 5.1

With transcoding off (note new bug https://bitbucket.org/xnejp03/serviio/i ... il-serviio) the Samsung D will play #2 and #4 but not #1 or #3. This demonstrates that the Samsung D uses resolution and ref frames (either by table or formula)rather than header value to determine playback capability which they document as "up to level 4.1" (ie: 32,768).

With transcoding on, Serviio transcodes #1, #3 and #4 due to caching the higher of header or ref frames. Given the above, it would be appropriate to have Servio only use the ref frames value in determining h264 levels with Samsung C/D/E.

I also found that the level change tool reporting of the "AVC level pattern found at file offset nnn" just reports the stored header value and does not report the real level based on ref frames.

To confirm the above, I tested another file.
#5 5-3 A demo clip 1280x720 w/4 ref frames and incorrect 5.1 header (4=3.1 & 1280x720x4/256=14400 = 3.1)

With transcoding off the Samsung D plays it correctly, again confirming the Samsung D ignores the header and uses ref frames.

[zip]

Great, thanks mate. I'll create a ticket for this. https://bitbucket.org/xnejp03/serviio/i ... istinguish

Btw, what do you mean formula vs table? The formula produces a number which then matches a limit of a level within the table.

[Illico]

I test on my Sony Bravia TV with a following file :
Transformers.HL5.1.ts

Video
Format : AVC
Format profile : High@L5.1
Format settings, ReFrames : 8 frames
Width : 1 920 pixels
Height : 800 pixels

So according to the formula : dpbMbs = (1920* 800 * 8 ) / 256 = 48000 => Should be 5.0 profile level

With TsMuxer tools and change header profile level and create different files:
Transformers.HL5.1.ts => Black screen
Transformers.HL5.0.ts => Black screen
Transformers.HL4.2.ts => Video play with some little blocks artefacts
Transformers.HL4.1.ts => Video play with some little blocks artefacts


I try with another file:
GIJoe.HL4.0.ts

Format : AVC
Format profile : High@L4.0
Format settings, ReFrames : 12 frames
Width : 1 280 pixels
Height : 528 pixels

So according to the formula : dpbMbs = (1280* 528* 12 ) / 256 = 31680 => Should be 4.0 profile level
With TsMuxer tools and change header profile level and create different files:
GIJoe.HL5.1.ts => Black screen
GIJoe.HL5.0.ts => Black screen
GIJoe.HL4.2.ts => Video play fine
GIJoe.HL4.1.ts => Video play fine
GIJoe.HL4.0.ts => Video play fine

Conclusion: Sony devices "buffer" should support base on Reframes an AVC video stream up to 5.0 profile, but the sony firmware ignore stream with AVC Header stream >4.2 profile level

[zip]

Thanks Illico... so for Sony we will have to keep it as it is, ie matching against the higher value from the two methods

[Illico]

Thanks Illico... so for Sony we will have to keep it as it is, ie matching against the higher value from the two methods

Yep

[jhb50]

Btw, what do you mean formula vs table? The formula produces a number which then matches a limit of a level within the table.

I mean the table in wikopedia, which if you look at my #1 file says it is a 5.1 whereas the formula makes it a 5.0

[jhb50]

@ Illico
How could these play?

Transformers.HL4.2.ts => Video play with some little blocks artefacts
Transformers.HL4.1.ts => Video play with some little blocks artefacts

After tsmuxer and changing heading you would still have 8 ref frames so the file is still really a level 5.0 so how can the sony play it with out transcoding?

[zip]

Btw, what do you mean formula vs table? The formula produces a number which then matches a limit of a level within the table.

I mean the table in wikopedia, which if you look at my #1 file says it is a 5.1 whereas the formula makes it a 5.0

You're reading it wrong, 129,600 is more that the max limit of level 5.0 (110,400), therefore it's 5.1

[zip]

@ Illico
How could these play?

Transformers.HL4.2.ts => Video play with some little blocks artefacts
Transformers.HL4.1.ts => Video play with some little blocks artefacts

After tsmuxer and changing heading you would still have 8 ref frames so the file is still really a level 5.0 so how can the sony play it with out transcoding?

I assume the TV has got just about enough RAM to buffer the 8 frames, as it's on the lower scale of level 5.

[jhb50]

You're reading it wrong, 129,600 is more that the max limit of level 5.0 (110,400), therefore it's 5.1

Ooops..misread.. edited my post.. but the table is a lot simpiler.

I assume the TV has got just about enough RAM to buffer the 8 frames, as it's on the lower scale of level 5.

That's a stretch! 1920x1080x8/256=64800 or twice the buffer needed by 4.2

[Illico]

I see that on formula we don't take the video bitrate into account (or framerate)...
For example a 4.2 level device supports max 522240 macroblocks per second for decoding speed.

[zip]

Yes, the formula is from Decoded picture buffering chapter. I don't want to rely too much on the bitrate given by FFmpeg, not sure if it's always right.

[jhb50]

While we are on the topic of level, my Samsung plays up to level 4.1,but the Serviio test parameter says > 4.2. Shouldn't it be > 4.1 for Samsung?

[zip]

While we are on the topic of level, my Samsung plays up to level 4.1,but the Serviio test parameter says > 4.2. Shouldn't it be > 4.1 for Samsung?

probably

[jhb50]

While we are on the topic of level, my Samsung plays up to level 4.1,but the Serviio test parameter says > 4.2. Shouldn't it be > 4.1 for Samsung?

probably

Could you expand?

The comment in the profile says test for > 4.1, but the rule says "level greater than 4.2". So what does the code check for? The formula value? If so which value, >32,768 for 4.1 or >34,816 for 4.2?

[zip]

it checks for 4.2, but it's probably wrong and should be 4.1