KnoppMyth on a 320G drive

Wed Dec 6 03:10:48 UTC 2006

| From: Colin McGregor <colinmc151-bJEeYj9oJeDQT0dZR+AlfA at public.gmane.org>

| --- Merv Curley <mervc-MwcKTmeKVNQ at public.gmane.org> wrote:

If you hvae free PCI slots, keep both tuners in your computer.  That
lets you record two signals at once.  Or watch one live and record
another.

I have 5 tuners in one Myth box (PVR150, PVR250, 3xadaptec)!  The CPU
is an Athlon XP 1700+

My understanding is that thpe PVR software bundled with WinXP MCE does not
support the Adaptec card.  Some PVR-150 models are supported and some
are not.  Of course you can use the software that comes with the card
but I doubt that would support both the cards together.

| > Using the XP Media Center software, displays a
| > shockingly poor quality TV 
| > picture with the Hauppage card. Using the software
| > with the Adaptec  card is 
| > about the same, maybe a bit better. The notes with
| > the WinDVR software say 
| > that it uses software encoding so I assume that it
| > doesn't use the hardware 
| > encoding available on the Adaptec card. Perhaps the
| > the same with the XP 
| > Media Centre and the Hauppage PVR150. Perhaps I just
| > don't understand when 
| > the hardware encoder is used.
| 
| The hardware encoder has a CPU on the tuner card that
| relieves a lot of the demands off the main CPU. When
| doing the above article I didn't want to @#$% my main
| MythTV box, so I set-up a 2nd test PC for installs
| with an AverMedia TV98 tuner card (as the name
| suggests the card is about 8 years old and built
| around a Brooktree BT-878 chip, in other words, old
| and ugly) which does software encoding. My main MythTV
| box and my test box both have a PIII 866 CPU, but big
| difference in CPU usage when watching TV.

That sounds odd.  For recording, the encoder is valuable.  For live
TV that should not be the case.  I've watched TV with at BT848 card in
a CPU slower than 866 (Celeron 300a clocked at 450MHz).

Mind you, I watch live TV using tvtime with the same card in a dual
core Athlon 64 X2 3800+ and the CPU gets pegged.  tvtime is vicious.
xawtv used much less CPU.  (tvtime does some filtering to improve the
picture.)

| The PVR150
| box runs at around 10% CPU when watching TV, the TV98
| based box runs around 90% CPU.

Some of the load is in kernel code, including interrupt handlers.  I'm
not confident that this shows up in normal statistics.  What do you
use to take those readings?

| > I am using a 19" LCD monitor, set to 1280x 1020, is
| > this wrong? Somewhere in 
| > the Knoppmyth musings I saw to set ones monitor to
| > 800x600. I don't have a TV 
| > connected to the video card at this point,  if that

800x600 is the resolution you seem to need on ATI and nVidia cards so
that the TVout works.  If you are not using TVout, those don't matter.
Consider them a hint about the best imaginable resolution from NTSC.

If you use 1280x1024, you will be using something (hardware or software) to
scale the picture.  That can eat processor power (if done in
software).  At that resolution, you will be able to see how crappy
NTSC is.  It won't make the signal crappier, just make the crappiness
more evident.

I don't completely understand this aspect of Myth either.

I have two Myth machines, a master and a slave (not sure that is the
right terminology).  Both run the FE and the BE.  The two backends
know about each other.  Intuitively, I feel I should need only a FE on
the slave machine because I don't wish to record or watch live TV on
that unit.  But that didn't seem to be what was expected.  For one
thing, the FE and BE were in the same packages (as far as I remember).

Note that the version of the FE must match the version of the BE.
Knoppmyth is a version behind these days (or was when I last looked).

I used FC5 for the master and Ubuntu 6.10 for the slave.  Ubuntu
seemed easier but that may only reflect:

(1) I'd done this before

(2) the Ubuntu packages were released later.

(3) the slave was a bit simpler

(4) Fedora's stack-smash protection exposed a Myth bug
    that the developers cannot replicate so I had to
    figure it out.  http://cvs.mythtv.org/trac/ticket/2420

BTW you can perform many Myth functions remotely via a browser -- no
installation required.  MythWeb is pretty good.

If you want multiple machines to work together then you probably
want to use the IP address of the ethernet interface.

If you are only going to have one box, 127.0.0.1 might be slightly
better.  It will not expose the box to the LAN.  It will work whether
or not the LAN is up or is renumbered.

BTW, you must use a "dotted quad" (four numbers separated by dots).
If you use a domain name here, it will silently fail.  I consider this
a bug (actually, several) but the Myth developers don't agree:
  http://svn.mythtv.org/trac/ticket/2469

| By default MythTV stores files in a format that offers
| high quality with LARGE file sizes. There are times
| when you want small file sizes and can live with 3rd
| rate picture quality. The transcode option is about
| converting to a lower quality, small file size.

Transcoding is pretty CPU-intensive.  The order of magnitude was about
1 hour of CPU for transcoding 1 hour of program.  Better to record in
the quality you desire.  Furthermore only certain resolutions are
standard DVD resolutions.  And not the default Myth one!
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists