On 3/17/06, James Richard Tyrer <[EMAIL PROTECTED]> wrote:
> Timothy Miller wrote:
> > On 3/17/06, James Richard Tyrer <[EMAIL PROTECTED]> wrote:
> >> The extra RAMDAC bits are used for gamma and color correction.  You
> >>  still have only 8 bits per color in memory.
> >
> > Understood.
> >
> >> The Philips 3xDAC has 10 bits per channel, and we are grounding the
> >>  2 LSBs rather than connecting them to the FPGA.
> >
> > As I mentioned in another email, we simply haven't got the spare pins
> >  for it.
>
> Perhaps using LVDS driver chips (also to be un-populated on the standard
> board) could be use to free up some pins.  These would need to be
> mounted very close to the FPGA
>
> Short search found Fairchild FIN1102 rated at 800 MB/s 3.3v supply would
> accept 2.5v LVC input levels with no extra parts unless you wanted a
> series termination resistor.  Priced @ $1.20 (1K).  Only two channel so
> you would need 18 of them.  Doesn't sound like a great idea, but it
> would work.

Are you suggesting inserting an extra device in between the FPGA and
the DAC?  For a high-speed signal?

And regarding using the LVDS pins, it's all or nothing.  For
high-speed signals, you really want point-to-point.  We're doing badly
enough routing them to two places.  Also, you seem to be suggesting
that we canibalize pins from the high-speed head to improve the
low-speed head.  What if you wanted to use both at the same time?  Or
are you saying you've identified some totally unused LVDS pins?

>
> > If it becomes REALLY important, we can do temporal modulation.
>
> I don't like this idea.  Just MHO.

Me neither.

> > Tech Source and plenty of others have been doing it for ages with
> > medical LCD displays at 60Hz.
>
> Interesting but ... I think that gamma correction/color matching
> software works by loading a palate.

And the same would be the case here.  Rather than just spitting the 10
bits out, we'd run it through some extra logic.  Nevertheless, I agree
that it's a bad idea.

> > One of the design requirements is that OGA should be very good for
> > 90% or more of desktop users.  Very few are going to be overly
> > concerned about gamma correction at all, let alone how many bits of
> > precision on the DAC.
>
> Yes, you are correct on this.  The only real market for fine control of
> gamma and color is pre-press color work.  Perhaps most of those people
> still use Macs. :-)

Yes, but wouldn't we like them to all move over to Linux?  :)

> > I'm not, by any means, saying you don't have a valid point.  But I am
> >  trying to determine how critical this is for most users and see how
> >  we could possibly solve the problem given the hardware constraints
> > we have.
>
> OTOH, if this hardware were available, then Linux color matching
> software could easily be changed to support it.  Gamma correction
> doesn't really work well with only 8 bits.

Oh, I know.  You end up seeing bad contour lines in darker pixels. 
It's bad enough WITHOUT gamma correction throwing away some of your
luma levels.
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to