[TESTING] Linux 4.9 kernel and improved video stack for Vero 4K / 4K +

Testing
227

This update solved none of my issues with the March 25 4.9 kernel update.

1- After update on reboot I only have 720p or 480p screen resolutions.
2- No sound from AAC files
3- no sound from AC3 encoded video files (DTS encoded files do have sound, but the soundbar doesn’t register them as DTS encoded)

Logs attached in which I tried to play an AC3 encoded video file (Ask This Old House), a DTS encoded video file (Tomrrowland), and an AAC music file (Your Love is My Drug). I made no changes to any of my settings yet, so this is a “clean” upgrade.

https://paste.osmc.tv/qubezagubo

I then unplugged the HDMI cable and plugged it back in, and after that problems #1 and #3 are gone. I also checked to make sure the audio volume is all the way up on Kodi (using the OSMC remote) and that it isn’t muted. The sound volume indicator shows a full circle, and there is no red audio icon with an X through it. So I’m relatively confident the sound is “on.” Despite that, the only way to fix #2 is to set Audio>Output Configuration to FIXED. But then the soundbar no longer registers DTS passthrough (although I do get 5.1 surround from what I can tell).

If I reboot I once again only have 720p and 480p resolutions, but at least AC3 and DTS files work (AAC sound still only work with FIXED output configuration as above). Unplugging/replugging the HDMI cable gets my 1080p again, although the second time I had to go into the settings, change it to 720p and then change it back to 1080p. Otherwise Kodi said it was outputing 1080p, but the TV was saying it was getting 720p.

Here’s a set of logs after unplugging/replugging the HDMI that at least shows the EDID was parsed properly. I didn’t have debug logging on in Kodi and didn’t retry any of the other stuff.

https://paste.osmc.tv/odorunaquy

I’m going to limp along with this config for now and see if there are things I can do to help you all figure out what’s causing all this with the 4.9 kernel (I have zero problems in the 3.x kernel).

228

Thanks for the feedback. We haven’t come up with a fix for the EDID problem yet. But one of the devs can reproduce it on his LG 32LB580V even without an AVR/soundbar in the way.

Could you please try loading an EDID from file so we can see what happens when you do have the right EDID. Grab this https://collab.osmc.tv/s/iTqWyv3dsLGaWre and put it in /home/osmc then you can go:

echo load pkscout.bin | sudo tee /sys/class/amhdmitx/amhdmitx0/edid

and re-start Kodi with Power->Exit.

Then does everything work @pkscout ?

Oh, and does AAC work as expected when it;s a video sound track rather than an audio file (it does here, but I don’t have any AAC audio files).

229

Hi, after installing the 4.9 kernel (coming from 3.14) connman does not see my 5GHz wlan anymore - no problems with 2.4 GHz. crda is installed (REGDOMAIN=DE). Pls let me know if you have an idea.

230

Yes, after loading that file Kodi starts at 1080p. I even did a full reboot and it correctly came up at 1080p.

I had to pull a video file with AAC audio and load it (I mostly use Kodi for recorded TV and movie’s I’ve bought and ripped, and all those have AC3 of some variant). On my system the video with AAC audio also produces no sound, and the same “fix” resolves that (changing the output configuration to FIXED). I also looked through my library for some MP3 files, and those are exhibiting the same issue.

And just to try and eliminate things, I copied one of the MP3 files to the local OSMC volume (all my media are on a NFS volume connected using Kodi’s internal NFS client) and then played it from there using the File Manager. Still no sound.

Oh, and it’s probably worth noting: when I had this problem with the March 25 test kernel, I renamed my .kodi folder and started fresh. Same problems. So I’m hesitant to point at the Kodi configuration as an issue.

EDIT: I plugged the Vero 4K+ directly into the TV the AAC audio now works. Oddly, if I plug the soundbar back in, AAC will continue to play through the TV speakers (as long as I have them on). Still no idea what in the 4.9 kernel would be causing this, but I’ll see if I can reset the soundbar.

2nd EDIT: I set the speaker output to 2.0 and now with FIXED set I get nothing from AAC or MP3 files (or even the Kodi GUI sounds). If I set the speaker output to 5.1 and leave it on FIXED I once again get AAC and MP3 playback using the soundbar. My head hurts. I quit. :wink:

3rd EDIT: Resetting the soundbar had no affect on the problems.

231

Well i must say:

My previous reported issues are all gone.

  • The m2ts issue [seeking correct refresh rate and black screen in between] is gone. 4k and 1080p both.
  • Switching between several files [ mkv, mp4, m2ts]
  • no sad faces anymore in between.
  • The lg chess also plays perfect 4k60 without refresh rate seeking.
  • Also the 1080p files with 30fps still works great.

Then i never played 4k iso before, but dts-ces iso 4k played flawlessly

To be fair, rebooted after update twice, but did not remove power or disconnect hdmi yet.

Thanks you all who worked on this update!!!

232

I do beg your pardon, @tanio99 (and @sam_nazarko) - 3D m2ts files are working with a 60Hz desktop. They take longer to start than a standard film does; adding on extra time for my video processor to resync with my TV, it takes a good 16 seconds before I see a picture and I wasn’t waiting long enough! :man_facepalming:

1 Like
234

Just tried playing a bit of the original Planet Earth series. It’s a 1080p/25 video, stored as 1080i/50; VC-1 codec and stored as a blu ray folder image rather than MKV.

If I play it with my Oppo 203, outputting 1080i/50, send that to to my video processor, and tell the processor to deinterlace in film mode and then output 1080p/25, it looks good - motion is very smooth.

However, if I play it on the Vero 4K+ with output set to 1080p/25, it looks terrible -really jerky.

No guarantee this is a 4.9 issue of course.

Does this sound interesting? If not (e.g. if it’s something you already know about), fine; if so, let me know and I’ll see if I can do you a sample video tomorrow.

235

This problem still exists.

236

Well playing 50i as 25p has always been less than ideal. I don’t think that’s peculiar to 4.9.

Out of interest, why convert a 25p stream to 50i?

237

Need to see logs for that.

238

You should probably ask the BBC that, that’s the format the original blu ray is in.

If you want me to guess: maybe at the time the disk was published, all HDTVs sold in Europe could reliably handle a 1080i/50 signal, but not all could process 1080p/25.

Plenty of blu ray players can’t output 1080p/25 either. Even fairly modern players like my Oppo 203 can’t do that. (It can do 2160p/25 but not 1080p/25).

239

Fair enough.

On reflection, this could simply be a symptom of the Vero 4K not being able to deinterlace 1080i correctly in hardware - which, as you say, is not a new problem. Maybe I’ll bring it up again once 4.9 is stable. :slightly_smiling_face:

240

On first look, it appears that HDR10+ reverts back to HDR10 only on the current update.
Will have to grab logs a bit later.

241

If you can double check that, that would be good.

Thanks

Sam

242

I know, frame rate switches take too long. I haven’t had time to look at it yet. My priority is to get everything up and running and performance issues will follow later. Thanks for testing!

1 Like
243

Indeed.

Currently, @tanio99 has worked around some M2TS issues by telling ffmpeg to analyse more of the file that is to be played, but this is not a solution for all files as some do not seem to present enough detail until up to 20+ seconds later. If we did this for every file, we’d 1) have to be sure that 20s is the exact limit; 2) have to suffer a significant startup penalty.

This will be looked at a little bit later, in the same way Ethernet performance will be – there’s still a bit of packet loss that isn’t ideal and is caused by PBL being off.

Cheers

Sam

244

Looks like HDR10+ is indeed being output as HDR10 with the current update, tested with a few known good titles (all of them being UHD MKV remuxes), and the TV is only registering HDR10 output (through a HDR10+ compatible sound bar).

Logs are here.

Sample HDR10+ movie is here.

EDID still reports HDR10plus support however… :thinking:

osmc@osmc:~$ cat /sys/devices/virtual/amhdmitx/amhdmitx0/edid
Rx Manufacturer Name: MEI
Rx Product Code: a296
Rx Serial Number: 01010101
Rx Product Name: Panasonic-TV
Manufacture Week: 0
Manufacture Year: 2019
Physical size(cm): 128 x 72
EDID Version: 1.3
EDID block number: 0x1
blk0 chksum: 0x1c
Source Physical Address[a.b.c.d]: 2.2.0.0
YCC support 0x03, VIC (native 97):
ColorDeepSupport 0x38 10/12/16/Y444 1/1/0/1
97 96 16 31 102 101 5 20 32 33 34 4 19 3 18 7 22 93 94 95 98 99 100 95 94 93 98 353 352 358 357 
Audio {format, channel, freq, cce}
{1, 1, 0x7f, 0x07}
{1, 7, 0x7f, 0x07}
{2, 7, 0x07, 0x50}
{7, 7, 0x1e, 0xc0}
{10, 7, 0x07, 0x03}
{11, 7, 0x7e, 0x01}
{11, 7, 0x7e, 0x03}
{12, 7, 0x7f, 0x03}
{15, 3, 0x04, 0x59}
Speaker Allocation: 0x4f
Vendor: 0x000c03 (HDMI device)
MaxTMDSClock1 300 MHz
Vendor2: 0xc45dd8
MaxTMDSClock2 600 MHz
vLatency:  Invalid/Unknown
aLatency:  Invalid/Unknown
i_vLatency:  Invalid/Unknown
i_aLatency:  Invalid/Unknown
Colorimetry: 0xff
SCDC: 1
RR_Cap: 0
LTE_340M_Scramble: 0
  HDR10plus 1  HDR  DeepColor
checkvalue: 0x1c3a0000

Edit: You should be able to see the exact playback logs if you search for the file name of the sample file. :slight_smile:

250

I have installed last updates for this kernel and I am no longer getting hdr10+

251

A coincidence that this remark should be made when this weekend I noticed how poorly the Vero on 3.14/March OSMC is dealing with 1080/50i material sourced from native 50i video. I was looking at some recently released “classic” Dr Who mkv’s from blu-ray (h264). The studio sequences on the Vero show distinct signs of motion judder when actors move across screen or when the camera pans. The judder is not as bad as the broken VC-1 on the 4.9 kernel, but definitely there. I switched to a Pi and the motion is perfect. Hopefully this can be fixed with 4.9?

252

To be fair, best case scenario there is a 576i signal deinterlaced, upscaled to 1080p and then reinterlaced - and depending on the story in question, the original source may be far more mangled than that. (In some cases, for example, the original 576i source was converted to NTSC and then converted back to PAL again. In others the 576i video was stored on 16mm black and white film at 25fps and the film has distorted with age. In some cases you’ve got colour information from an NTSC conversion and luminance information from 16mm film.) So, all in all, that’s a pretty good torture test for a deinterlacer!

Still, if the Pi gets it right and the Vero doesn’t, that’s interesting.