Topic: Videoclock not being enabled for DVD (Read 3865 times)

Jong · « **on:** January 24, 2013, 10:12:24 am »

See this link to the MadVR thread (discussion with Madshi is only over about half a page). It seems that Videoclock is currently disabled for DVD!

I think this has been the case for a while. When diagnosing my FFW/REW problem I noticed the MadVR reported rate of frame repeats was much higher for DVD than playing the VOB direct. Now, from my discussion with Madshi, it appears clear that Videoclock is simply doing nothing for DVD.

Matt · « **Reply #1 on:** January 24, 2013, 02:12:24 pm »

It seems to be working here. I can watch the clock adapt slightly with Audio Path.

What do you see in Audio Path? What's the playback frame rate and display frame rate? You're not bitstreaming are you?

Jong · « **Reply #2 on:** January 24, 2013, 02:53:40 pm »

Hi Matt. There is certainly something odd going on.

No bitstreaming.

A PAL DVD @50Hz, which is actually 50.001Hz, shows that it repeats frames every 15 mins. MadVR shows 0.0000% clock deviation, yet if I play a VOB from the same film directly it shows .0022% clock deviation and a repeat rate in days. Unfortunately the error playing a R2 DVD @50Hz is so small that the audio path would not show whether it was working or not.

A R1 DVD played @50Hz still shows "tempo at +0.0000x for Videoclock, which doesn't look right.

So the fact that Videoclock appears at all suggests in is kinda enabled, but it isn't working.

jmone · « **Reply #3 on:** January 24, 2013, 03:27:08 pm »

Sounds like it is working to me.... just it does not have to do much in this case. I bet if you turn it off you will not see the "Tempo" in the Audio Path.

Jong · « **Reply #4 on:** January 24, 2013, 03:43:15 pm »

It's on, but not working. As I said, a VOB from the same movie has frame repeats in days, but when played as a DVD it is every 15 mins, which as Madshi has calculated, see thread, is exactly what you'd expect if Videoclock was off.

Matt · « **Reply #5 on:** January 24, 2013, 03:54:36 pm »

I don't have a 50Hz display, so it's hard for me to test the exact scenario.

I suppose one possibility is that we're not getting high-resolution refresh information from madVR, so using 50.0 instead of whatever it actually is.

Does the madVR OSD show something other than 50.0?

Jong · « **Reply #6 on:** January 24, 2013, 04:41:01 pm »

Thanks Matt. As I reported in the MadVR thread, the refresh rate is reported in the OSD as 50.0011. Enough for only 15 minute accuracy, which is exactly what I am seeing. When I play the .vob of the same movie directly I get many hours to days of accuracy reported by MadVR and a clock deviation of 0.0022% instead of 0%, exactly as would be expected if Videoclock was working correctly (0.0011/50 = 0.0022%). I don't know if MadVR would even see the difference between when a VOB is playing and when a DVD of the same movie is. Even if it could it would seem odd for it to pass a different value for refresh rate for the two, yet show the same rate in its OSD.

This is exactly the kind of small correction of clock error Videoclock should eat for breakfast. Is there anyway I can gather more useful info/logs for you?

Jong · « **Reply #7 on:** January 24, 2013, 05:07:40 pm »

Here is a log of playback. Started the log, played one video then quit MC. Maybe it helps? my ignorant eye could not find what MadVR reported to MC or any references to Videoclock or tempo adjustments in the log, so it may be no use, but just in case!

Sandy B Ridge · « **Reply #8 on:** January 24, 2013, 05:16:46 pm »

If videoclock is on, I wouldn't put too much credance to what MadVR reports on the OSD. With some blurays (23.976) playing at 24.000 MadVRs OSD would suggest a frame drop every couple of minutes for me. However, it was quite obvious that this number would go up as the film played, visibly there were no drops and no drops were reported by MadVR either.
Is it just the estimate of 'frame drops per minute' that bothers, or are you actually seeing drops/stutters and the actual drop counter going up? With Videoclock enabled I'd trust your eyes more than the estimate.

SBR

madshi · « **Reply #9 on:** January 24, 2013, 05:25:15 pm »

My understanding is that the madVR measured refresh rate is always the same for Jong, namely 50.0011, with both DVD and VOB playback. I don't think it is likely that madVR is not reporting this properly to VideoClock. I believe the problem is likely to be somewhere else...

The difference between DVD and VOB playback for Jong seems to be the audio clock deviation. If VideoClock is doing it's job, it should notice that the refresh rate is too high and should consequently speed up the audio master clock, too, by the same factor. If VideoClock is doing that, madVR's "deviation" measurement should show that. According to Jong this does happen with VOB playback. There audio clock deviation is measured by madVR with 0.0022% which is exactly what is needed to "fix" the too high video refresh rate. However, with DVD playback it seems that VideoClock is not speeding the audio clock up at all. madVR measures a audio clock deviation of zero. As a result madVR has to repeat frames.

@SBR, the latest madVR version v0.85.8 has a new "repeated frames" counter in the OSD which counts repeated frames due to a mismatch between movie framerate and video refresh rate. And this counter actually goes up for Jong (it does, doesn't in, Jong?), in accordance to what the "frame repeats per minute" estimate suggests. So there is a real problem here.

Matt · « **Reply #10 on:** January 24, 2013, 05:27:48 pm »

Please provide another log after testing the next build.

In particular, we'll want to look at these:
CVideoRefreshRateThread
COutputPluginAudioRenderer::SetVideoFrameRate

I'm guessing the audio renderer doesn't have the right (or any) input frame rate for some reason.

Jong · « **Reply #11 on:** January 24, 2013, 05:35:23 pm »

Quote from: madshi on January 24, 2013, 05:25:15 pm

My understanding is that the madVR measured refresh rate is always the same for Jong, namely 50.0011, with both DVD and VOB playback. I don't think it is likely that madVR is not reporting this properly to VideoClock. I believe the problem is likely to be somewhere else...

The difference between DVD and VOB playback for Jong seems to be the audio clock deviation. If VideoClock is doing it's job, it should notice that the refresh rate is too high and should consequently speed up the audio master clock, too, by the same factor. If VideoClock is doing that, madVR's "deviation" measurement should show that. According to Jong this does happen with VOB playback. There audio clock deviation is measured by madVR with 0.0022% which is exactly what is needed to "fix" the too high video refresh rate. However, with DVD playback it seems that VideoClock is not speeding the audio clock up at all. madVR measures a audio clock deviation of zero. As a result madVR has to repeat frames.

@SBR, the latest madVR version v0.85.8 has a new "repeated frames" counter in the OSD which counts repeated frames due to a mismatch between movie framerate and video refresh rate. And this counter actually goes up for Jong (it does, doesn't in, Jong?), in accordance to what the "frame repeats per minute" estimate suggests. So there is a real problem here.

Thanks Madshi, yes I see the repeated frame counter go up almost exactly once every 15 mins. by the way I have tested with both the version of MadvR bundled with .120 and with the latest with repeated frame counter and both show the same clock deviation problems and an estimated 15 mins between frame repeats for DVD only - not a VOB of the same movie.

Jong · « **Reply #12 on:** January 24, 2013, 06:01:04 pm »

Matt, tried to do this, but as reported in .121 thread I can't play anything with that version. It's also midnight here, so unfortunately time for bed unless another build goes up in the next 10 mins! np if it doesn't I'll pick it up tomorrow.

Jong · « **Reply #13 on:** January 25, 2013, 07:32:04 am »

Matt, more evidence if its needed:

1. R1 VOB (copied from DVD below) played @60Hz (not 59.940). Audio path shows a tempo adjustment of 0.1%, as expected. MadVR also shows this as the "Clock Deviation".

2. R1 DVD with 60Hz refresh rate. With the DVD there is no Videoclock correction for the .1% NTSC speedup. MadVR also reports no Clock Deviation. Videoclock is enabled but not working.

This should be easy for you to test yourself.

Mike Noe · « **Reply #14 on:** January 25, 2013, 07:49:26 am »

Jong, just curious, what happens if you turn deinterlacing off (ctrl+shift+alt+d)....

Jong · « **Reply #15 on:** January 25, 2013, 08:02:02 am »

Mike, just tried it. Same problem. only change is the rate of frame repeats changes to 33 secs instead of 16 due to halving of the frame rate.

Jong · « **Reply #16 on:** January 25, 2013, 09:46:55 am »

Matt, using my 60Hz R1 test I can see that Videoclock works using RO Std. So it is an ROHQ problem. Specifically I think it is linked to LAV as I can replace MadVR with EVR, using custom mode based on ROHQ, and still audio path shows the tempo as +1.0000x.

Ps. I am assuming you no longer need a log from me as you should be able to reproduce using 60 Hz at your end, but if I am wrong, or if you are unable to reproduce, of course let me know.

Matt · « **Reply #17 on:** January 25, 2013, 09:47:50 am »

Quote from: Jong on January 25, 2013, 09:46:55 am

Ps. I am assuming you no longer need a log from me as you should be able to reproduce using 60 Hz at your end, but if I am wrong, or if you are unable to reproduce, of course let me know.

I would like to see a log showing a case where you don't believe it's working properly.

Jong · « **Reply #18 on:** January 25, 2013, 09:50:59 am »

OK Matt. No problem. It will be on its way in a few mins. But please look at those screen shots above, I think they show clearly it is not working here. of course there could be another setting at the root of the problem. It would be interesting to to see if you get the same results as me with an R1 DVD @(true)60Hz.

Jong · « **Reply #19 on:** January 25, 2013, 09:57:54 am »

Here it is:

Matt · « **Reply #20 on:** January 25, 2013, 10:16:53 am »

Well, this is why it's not working:
0026738: 6732: Playback: COutputPluginAudioRenderer::SetVideoFrameRate: Frame rate: 0 fps

I think this means the input pin to the video renderer isn't providing a frame rate for DVD when we call ConnectionMediaType(...) at graph building time.

Maybe DVD is special and doesn't have a real frame rate until the graph is running? Maybe we need to continually monitor the refresh rate of the connection to the video renderer?

Hendrik · « **Reply #21 on:** January 25, 2013, 10:20:41 am »

I think the DVD Navigator refreshes the media type shortly after the graph was built with the correct information.

madshi · « **Reply #22 on:** January 25, 2013, 10:22:03 am »

Aren't there even DVDs where the menu has a different framerate than the movie? Not sure, to be honest.

Matt · « **Reply #23 on:** January 25, 2013, 10:23:21 am »

Do you think the audio renderer should periodically query the connection of the video renderer to keep the frame rate updated? How often?

The audio renderer already holds a reference to the video renderer to support the custom madVR interfaces, so this wouldn't be a big deal.

Jong · « **Reply #24 on:** January 25, 2013, 10:29:00 am »

Guys, it's interesting (to me anyway!) that RO Std works OK. There must be something different in how/when info is provided when using LAV. And notice the problem happens even if custom mode based on ROHQ replaces MadVR with EVR.

Quote from: madshi on January 25, 2013, 10:22:03 am

Aren't there even DVDs where the menu has a different framerate than the movie? Not sure, to be honest.

Yes, some (probably a lot) have interlaced 30fps menus and telecined movies, but I'm guessing Matt at least checks this every time a new graph is built, right Matt?

madshi · « **Reply #25 on:** January 25, 2013, 10:31:39 am »

I'm wondering if there shouldn't be a VideoClock mode in cases when the source framerate isn't known? E.g. if the refresh rate is 50.50Hz, it is pretty clear that it's 1% too high. So in this case VideoClock could make a %1 adjustment to the audio speed, too, without having to know the source framerate. Ok, this way you can't make 23.976fps content play smoothly with a 24.000Hz refresh rate. But I would say if you don't know the source framerate, a good workaround would be to guess that it is likely to be one of the usual framerates which is the nearest to the measured refresh rate. E.g. if the refresh rate is 50.50Hz, guess the source framerate to be 50fps. If the refresh rate is 23.99Hz, guess the source framerate to be 24.000fps. If the refresh rate is 59.941, guess the source framerate to be 59.940. If the refresh rate is 47.998, guess the source framerate to be 48.000, etc. Of course that would just be a workaround for situations where (for whatever reason) you don't know the correct source framerate.

I don't see a problem if you poll for the framerate once every e.g. second or so. But there might be situations where the pin info will never contain a valid framerate.

Hendrik · « **Reply #26 on:** January 25, 2013, 11:15:46 am »

Quote from: Jong on January 25, 2013, 10:29:00 am

Yes, some (probably a lot) have interlaced 30fps menus and telecined movies, but I'm guessing Matt at least checks this every time a new graph is built, right Matt?

But the DVD Navigator doesnt build a new graph when you switch from menu to playback.

Jong · « **Reply #27 on:** January 25, 2013, 11:20:58 am »

Matt · « **Reply #28 on:** January 28, 2013, 01:33:27 pm »

Next build:
Changed: VideoClock works better with videos that start with an unknown frame rate or change frame rates during playback.

INTERACT FORUM

Author Topic: Videoclock not being enabled for DVD (Read 3865 times)