timing.info too short, timeline inaccurate

[sub]

Here is a patch your can try, which derives the seconds from the system time rather than the up-time counter.

[mvallevand]

Can this be made configurable and off by default? I envision all kinds of problems in recordings whenever Windows update to Internet time.

Martin

[johnsonx42]

I'm not sure I see an avenue for a problem here. If the change when windows syncs to "internet time" is fairly small (like less than 10 seconds or so), it probably won't produce a big enough error in the timing.info data to be noticable; if the change is much bigger, then you have a clock problem, and ought to be syncing to the internet more often in which case you'll be happy NPVR is using the system clock for the timing data.

[mvallevand]

It introduces a potential to jump backwards. Hard to consider that worth any risk for a problem traceable to just one users new motherboard. Padding deals with minor time variances comskip is not as forgiving. That being said I now rely on xbmc because it has beautiful comskip.

Martin

[Reddwarf]

It introduces a potential to jump backwards. Hard to consider that worth any risk for a problem traceable to just one users new motherboard. Padding deals with minor time variances comskip is not as forgiving. That being said I now rely on xbmc because it has beautiful comskip.

Martin

The strange thing is that npvr skips really well on some channels while xbmc skips pretty fine on (almost) all channels. I wonder why?

[mvallevand]

In my case it's simple since I use an analog HDPVR I have a nice h/w clock souce for my streams. For broadcast streams you might need timing.info for accurate seeking to deal with timing errors/

Martin

[Reddwarf]

In my case it's simple since I use an analog HDPVR I have a nice h/w clock souce for my streams. For broadcast streams you might need timing.info for accurate seeking to deal with timing errors/

Martin

I do have timing info on all my recordings, but some channels, like Investigation Discovery, is spot on with comskip skips, but FOX and most others are several seconds out. Could it have something to do with the frequency of I/B frames?

[mvallevand]

NextPVR skips the same way so my best guess for variance by channels for comskip would be tied to how comskip itself is detecting commercial transitions

Martin

[Kiwi]

This does seem to be a motherboard timer issue. Have you tried changing the the battery to a new one?

Anyway, drifting 5s per 30 mins is too high for the motherboard.

[johnsonx42]

No, the battery has nothing to do with it. The battery only gets used when the system is disconnected from power. I've done some research on problems like this, found it could be some sort of driver related issue as well. I found a guy who's system kept perfect time with Windows XP, but the clock only advanced at 1/3 speed on Windows 7. That's a pretty extreme case, but it seems having a slow clock is not extremely rare. I'll have to look into this further, as I would like to solve it if possible.

In other news, with the new patch, things are working pretty well. Recordings made so far have very reasonable timeline lengths, and the comskip jumps I've seen so far have been very accurate.