DAC1 / DAC2 for ex137x - ES1370/1/2/3 , CT5880/CT4730(EV1938)

[Pyun YongHyeon]

On Sat, Oct 08, 2005 at 02:06:26AM +0800, Ariff Abdullah wrote:
 > On Fri, 07 Oct 2005 09:42:28 +0200
 > Alexander Leidinger <Alexander at Leidinger.net> wrote:
 > > 
 > > A thought comes into my mind...
 > > 
 > > Stupid as I am sometimes, I thought so far the bad sound quality of my
 > > SoundBlaster PCI (128 or 512 ... damn, I have to look it up at home)
 > > is some kind of hardware problem in my PC... I thought this, because
 > > in Windows it sounds as bad as with FreeBSD. Ok, Windows is bad, but
 > > regarding the multimedia topic I've never seen such bad behavior from
 > > Windows XP. That's the reason for thinking it's a hardware problem.
 > >
 > I have lots of dropouts in Windows, mostly because of VIA
 > motherboards and particularly because of bad driver. It amaze me that
 > this soundcard sound alot better in FreeBSD, with correct tuning and
 > configuration.
 > 
 > > But now I'm not sure anymore. I have PREEMPTION in my kernel. And it
 > > doesn't sound like stuttering. It sounds like distortion. I don't know
 > > how to describe this further. It seems it is related to activity on
 > > the PCI bus.
 > > 
 > > Does this sound like a problem which can be solved with
 > > latency_timer/slave channels?
 > >
 > Most likely.. especially with VIA motherboards. Increasing default
 > buffersize also can help but usually it's a sign of bad PCI timing. I
 > don't have any problem with Intel boards even sharing the interrupt with
 > others.
 > 
 > > I think I have to get some time at Sunday to give it a try...
 > > 
 > You won't regret it. This is one of the cheap / best sounding soundcards
 > ever exist.
 > 
 > http://www.pcavtech.com/soundcards/summary/index.htm
 > 

For es137x, it has no DMA limitation(supports up to 4GB) which means it
could be used on any architectures including sparc64/ppc. But it
supports only little endian format, so kernel endian conversion feeder
is required on big endian systems. This would add unavoidable playback/
capture latency under high system loads. In addition, its complex
sampling register access methods and timing requiried on accessing
CODECs effectivly defeats parallel execution capability of kernel
threads on today's high speed CPUs.

-- 
Regards,
Pyun YongHyeon