[vdr] Any really working HD video output systems for VDR?

Georg Acher acher at in.tum.de
Wed Jun 10 00:50:42 CEST 2009


On Tue, Jun 09, 2009 at 11:44:29AM +0200, Mattia Rossi wrote:
 
> Hi Georg,
> sorry if that came out as 'they are hiding something', I was really
> thinking more in terms of 'there are so many variables that the best
> combination is probably achieved in an environment where the user has
> much less control as we have in a custom vdr compiled environment,
> without even taking into account the endless combinations of display
> devices that people will want to connect to the eHd'
> 
> That said, I really really would like a different option than 'don't use
> the scaler on the eHd-HDMI output, it's crap', is there a way to change
> the scaling algorithm it uses ?
> 
> Is there any other setting we can work on regarding the eHd outputs,
> even if they are not listed in the reelbox-3 setup ?

Not really. You can either use upscaling without deinterlacing (and get
tearing for real interlaced moving signals) or use the chips "deinterlacer".
That is effectively independent scaling of both fields, so the field order
is correct, but fine details show horizontal ghost lines as the
interpolation doesn't know anything about the full context. It depends a bit
on the display how visible these lines can be seen, some displays seem to
have their own filtering and hide them better.

There is a third possibility that is currently not settable: Blending. That
mixes both fields together. So still pictures are sharp and movements are
"only" blurred and not teared. I can enable it in a "experimental"
linux.bin.

BTW: We chose the DeCypher despite "a real" deinterlacer was missing at the
time, because we were told that it will be enabled later. The rest is
history...
 
> For example, I am using the coax digital output on the eHd mini-din,
> instead of the pass through HDMI (the HDMI cable goes straight to the
> VPR, the coax cable goes to my main amp), I had to change a couple of
> lines in the reelbox-3 plugin in order to allow for AC3 to pass through
> the digital output. The option was there, it just wasn't compiled in
> because of REELBOXLITE ifdefs

That was because the MiniPCI-Card for the Lite used a DeCypher-revision with
a dead SPDIF-port. So we need to feed the AC3 SPDIF output over the main
sound card. The AVG-card has a working SPDIF, so the mainboard sound is used
only for PCM (if enabled).
 
> What I would like to achieve is decent quality (either scaled or
> unscaled) SD MPEG-2 output, and by decent I mean at least at the same
> quality of the MPEG-2 output of a FF card without having to use a 2Keuro
> deinterlacer. Please note that I'm talking about MPEG processing
> quality, not still image quality 

You can output 576i over HDMI and let the display do the deinterlacing. But
there is a dependency with the analog output that allows only 576p when the
analog port should emit 576i (CVBS etc.). For 576i over HDMI you need to
switch the analog port completely off.

> I understand that this is not easy to fine tune because of all the
> different variables and because of all the 'not straightforward' test
> environments every single vdr user will have set up.
> 
> At the moment my vdr 1.7.x install is really basic (vdr, extensions
> patch, reelbox3 filebrowser xinemediaplayer and menuorg plugins) and I
> am completely satisfied by the stability (both of vdr and of the eHd) ,
> I even don't notice the problems with the recordings handling (the
> single frame forward in a recording isn't really working, it looks like
> it is skipping from i-frame to i-frame (doesnt'do that when going
> backwards, it is the opposite behaviour as with the ff card, where
> forward skipping works, while backward skipping goes from i_frame to
> i-frame ..)

I guess the vdr 1.7 has it's own intelligence about frame-precise playback
for TS that doesn't match the requirements of the plugin... It took us quite
a while to make it work almost as responsive as on the FF, maybe there are
some dependencies with our "old" 1.4. But up to now, I had no time to try
1.7...

-- 
         Georg Acher, acher at in.tum.de         
         http://www.lrr.in.tum.de/~acher
         "Oh no, not again !" The bowl of petunias          



More information about the vdr mailing list