[mythtv-commits] Ticket #5754: Enhancement to make EIT Active Scanning end after a period, then restart later (a duty cycle)

[MythTV]

#5754: Enhancement to make EIT Active Scanning end after a period, then restart
later (a duty cycle)
----------------------------------+-----------------------------------------
 Reporter:  simonwalls at yahoo.com  |        Owner:  stuarta   
     Type:  enhancement           |       Status:  new       
 Priority:  minor                 |    Milestone:  0.22      
Component:  eit                   |      Version:  0.21-fixes
 Severity:  medium                |   Resolution:            
  Mlocked:  0                     |  
----------------------------------+-----------------------------------------

Comment(by simonwalls at yahoo.com):

 My test had to end after 3 days run because I had to restart the machine
 to fiddle with the HDDs. What I saw until the test ended was this:

 - The guide data indication on Mythweb Status stayed at 8 days throughout,
 indicating that 12 minutes per hour is enough to obtain data for 6
 multiplexes in the UK. Yes, if a last-minute reschedule arrives it may be
 missed, so that could be handled better. This might be by ensuring that
 the Active Scan time(s) immediately precede a recording, rather than being
 asynchronous.

 - The EIT transmission rate is very high, so that with a correctly chosen
 on-time, which looks like it can be <2 minutes per mux, the EPG can be
 kept full. There isn't an indication of how full it is, at present, as the
 Mythweb (from Mythbackend) indication is the time of the last entry in the
 guide. 'Holes in the guide' are not measured at present.

 - The EIT present/following info is transmitted for the tuned mux at least
 every 2 seconds (according to the PDF link I supplied above). This means
 that immediately before a recording, a quick now/next check could be made
 to confirm the intended programme is still about to be transmitted. Using
 the SDT data for the tuned mux would require at least 10 seconds - still
 feasible.

 - 12 minutes (precisely) timed using the system clock is not enough time
 for 2 runs through 6 muxes at 1 minute Active Scan per mux. (Only one and
 a half runs through the set of 6 muxes were made). In my logs, the time
 between "Looking for EIT data on multiplex of channel ... " messages is 1
 min 23 secs whereas max_seconds_per_source is set to 60s in my mythtv-
 setup. There are other tasks going on between these messages, like
 database access etc, and since the EIT process runs at idle priority this
 may explain this.

 - The second period of scanning, when it occurred for a mux, was some 9
 minutes later and resulted in smaller numbers of EIT events (2% to 50%)
 than the first period.

 - The power saving was worthwhile at 11 Watts AC (~8.4Watts DC after PSU
 inefficiencies) which was in a machine with 2 Nova-T PCI cards, and one
 was using Active Scan. If you use mythshutdown, the backend will usually
 not be on for long enough to make worthwhile savings (and you will want
 all the guide data you can get to be gathered during an on-time).

 - I noted some 'cx8802_start_dma() Failed' errors in dmesg, which
 correlated in time with when I started testing patches that closed the DVB
 card. According to postings on the [linux-dvb] list:  "It means the dvb
 sub-driver hasn't claimed the IRQ, probably because the demux was not
 opened (which indirectly causes the subdriver to claim ownership). The
 error is reported because someone asked DMA to start, but no subdriver
 owns the transport bus."
 That's good enough for me. It may mean that the way I closed the DVB
 channel, CloseChannel(), is not good or complete enough and something
 (DMA?) is still set up.
 So, to confirm that it was the patch, I reverted my patch and recompiled,
 the cx8802 messages had disappeared (test duration 1 hour run).


 I am going to have to leave this for some time, as we are expecting a new
 baby in the family!  I will look into, perhaps, a measurement of 'how full
 the guide is' so that a future EIT patch could keep scanning until the
 guide is full, then stop and re-scan when it drops below some limit.  And
 a way is needed for EIT scanning to close the DVB card, without this
 cx8802 error occurring. I am sure Myth does it properly, if the correct
 call is used, the one I picked was insufficient in some way.

 Thanks for all your comments and perhaps we'll speak again in the future.
 This ticket could be closed now, so it could be reopened in the future.

-- 
Ticket URL: <http://svn.mythtv.org/trac/ticket/5754#comment:8>
MythTV <http://www.mythtv.org/>
MythTV