-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hi Clemens, Hi ALSA-devs,
I'm working with Damien Zammit on the Digidesign-003 driver which is based on your alsa-kprivate dice code.
We were able to work out most of the issues so far but are now stuck with timing problems. More specifically: the jitter of async message timing is quite large and the AMDTP ISO stream has regular drop-outs.
Comparing 1394 data dumps, OSX sends async messages at regular 125 (+-5us) intervals. On GNU/Linux we've seen a delta -40us, + 120us. For async messages this is not a big deal. But..
Just playing audio via aplay: there are short dropouts every couple of seconds but otherwise the devices plays fine indefinitely. Interestingly, when running cyclictest while playing sound: there are no dropouts but the stream drops after ~10-15 seconds.
Any hints why that may be or how to debug the issue further?
In order to rule out kernel related issues (3.0.0 is quite old), we've rebased the work (and alsa-kprivate's firewire-kernel-streaming branch) onto alsa-driver.git. The issue remains the same with 3.2.0-4-rt
Alas, neither of us has detailed knowledge nor specs for AMDTP, not to mention in-depth knowledge about Linux' firewire stack timings. Any help or leads are appreciated.
As we believe we are close to a fully functional driver for the 003, we are very keen to resolve the following two separate issues:
1) The timing problem we have observed with AMDTP over firewire. 2) A framework for the capture stream in ALSA firewire.
The rebased driver code is at git://gareus.org/alsa-driver.git http://gareus.org/gitweb/?p=alsa-driver.git
The broken-out digi003 driver is available from: https://github.com/zamaudio/snd-firewire-003 https://github.com/x42/003amdtp
1394 bus data dumps of both OSX as well as Linux can be found at http://www.zamaudio.com/?p=715
Cheers! robin