[mythtv-users] Back-toBack recordings from the same source

[Karl Dietz]

On 25.04.2012 00:58, Pieter De Wit wrote:
>
> Is there a way to tell Mythtv to use the same source for back-to-back
> recordings of programs, with overlapping start and finish times ? (what
> a mouth full :) )

You can record overlapping programs from one DVB card with the multirec
feature. By default each DVB cards gets added as 2 virtual tuners.
You can push that up to 5 in mythtv-setup.

> Let see how my ASCII art skills are:
>
>
> ------>time
>
> --->(1)--->(2)--->(3)-->(4)--->(5)--->
>
> 1) Start of "P1" recording
> 2) Real start of "P1"
> 3) Start of "P2" recording
> 4) End of "P1"/Start of "P2" following 1 or two ads (In New Zealand,
> zero ads between programs on some channels, there is title scrolling and
> all that)
> 5) End of "P1" recording
> --> Live goes on
>
> I have found it *very* hard to get the over-recording right (I think
> they call it "island time" :D ). The best way for me is to set the
> pre-start recording time to 120 seconds and over record by 300 seconds.
> The problem this causes is that "all" my tuners are busy (hauppauge Nova
> S Plus)

I run everything with 5 minutes start-early and end-late as default/in
each recording rule. (MythTV has start-early/end-late and pre-/postroll
which are two different things. the former is enforced but the latter
is not due to different use cases)
For some programs I have end-late set to 45 minutes, works very well,
too.

> I am told this card can do up to 8 recordings (via that magic that Myth
> does) but I could have sworn the last time I set it to 8 the quality
> went down (the picture appeared snowy) If the card is happy to handle 8
> decodes, then that solves my problem, but what about people that arn't
> that lucky.

The card should just be pumping bits and not be decoding any streams.
The restriction is likely on the number of different PIDs/streams that
the card can filter out in hardware.
And the GUI setup will only allow you to configure 5 parallel
recordings.

> Here is what I thought could happen:
>
> At 3) - create another ring buffer that "feeds off" the already created
> ring buffer for 1)
> At 4) - release the ring buffer for 1)
> At 5) - release the ring buffer created at 3)

Rewriting the backend to allow overlapping recordings on all card types
or hiding the virtual tuners in a generic 1xMPTS -> NxSPTS filter is
certainly possible and a nice idea, but it will likely require a very
low level rewrite of much recording and scheduler code for little gain.

Regards,
Karl