ice1712 and jack

ice1712 and jack

[mike start]

I've been having difficulty getting an ews88MT (ice1712 chipset) to work 
in full duplex using the jack audio server (jackd).  Capture only mode 
works fine, even down to 64 samples latency, Playback only mode works 
fine down to 64 samples latency.  Duplex mode causes 'xruns' at even 
2048 samples latency.  Looking at the interrupt handler, if I add some 
diagnostics to tell me the difference between the capture and playback 
DMA counters timing out I see what appears to be clock drift between the 
  playback and capture halves of the ice1712.  I also added some debug 
messages to show what sample rate the alsa driver was setting the 
hardware to and it agrees with the rate set in my jackd invocation 
(jackd -R -d alsa -n 2 -p 1024 - r 48000 -d hw:0,0)
Also adding diagnostics to jackd to show the time spent in poll waiting 
for the capture and playback interrupts shows that they start in sync 
and gradually drift until an 'xrun' occurs when they are '-p' samples 
apart.  According to the ice1712 data sheet the 'professional' 
multitrack part of this chip (the part my soundcard is using) runs its 
inputs and outputs from the same master clock so I'm puzzled as to why 
they are drifting.  Does anybody have any more information relating to 
the ews88mt?  - I think the drivers are working fine but the problems 
are being caused by the capture and playback not being quite in sync 
which renders it almost useless for professional recording.  Anyone have 
any suggestions?

Re: ice1712 and jack

[Takashi Iwai]

At Mon, 29 Dec 2008 23:22:54 +0000,
mike start wrote:
> 
> I've been having difficulty getting an ews88MT (ice1712 chipset) to work 
> in full duplex using the jack audio server (jackd).  Capture only mode 
> works fine, even down to 64 samples latency, Playback only mode works 
> fine down to 64 samples latency.  Duplex mode causes 'xruns' at even 
> 2048 samples latency.  Looking at the interrupt handler, if I add some 
> diagnostics to tell me the difference between the capture and playback 
> DMA counters timing out I see what appears to be clock drift between the 
>   playback and capture halves of the ice1712.  I also added some debug 
> messages to show what sample rate the alsa driver was setting the 
> hardware to and it agrees with the rate set in my jackd invocation 
> (jackd -R -d alsa -n 2 -p 1024 - r 48000 -d hw:0,0)
> Also adding diagnostics to jackd to show the time spent in poll waiting 
> for the capture and playback interrupts shows that they start in sync 
> and gradually drift until an 'xrun' occurs when they are '-p' samples 
> apart.  According to the ice1712 data sheet the 'professional' 
> multitrack part of this chip (the part my soundcard is using) runs its 
> inputs and outputs from the same master clock so I'm puzzled as to why 
> they are drifting.

I'm puzzled, too.  Not only clock but also sample rates are shared
(must be identical) both for playback and capture on ICE1712...

>  Does anybody have any more information relating to 
> the ews88mt?  - I think the drivers are working fine but the problems 
> are being caused by the capture and playback not being quite in sync 
> which renders it almost useless for professional recording.  Anyone have 
> any suggestions?

How did you check the DMA counter value?  From the raw register read
or on jackd?


Takashi