Re: [alsa-devel] [Jackit-devel] position_fix issue snd-hda-intel (was jack issue)
Donnerstag, 7. Februar 2008 Stephane THOMAS:
I found position_fix=3 works for the chip I have (ALC888).
I confirm that with this parameter passed to the module, jack works fine. But using ALSA directly produce (? samplerate mismatch).
Just to sum it up:
For, at least, our two snd-hda-intel devices (ALC888 and 82801I (ICH9 Family)) both position_fix values (POSBUF and FIFO, 0 and 3 respectively) produce ok sound, only that
position_fix=0 makes alsa fine and jack crackle/sizzle position_fix=3 makes alsa crackle/sizzle and jack fine
"alsa" meaning direct use of alsa and "jack" meaning (at least in my case) jack with the alsa backend.
Here I have alsa-1.0.14-31.2 and jack-0.109.2-0.pm.1 (0.pm.1 stems from the packman repository, I guess)
Wolfgang
Hi,
I'm sorry I just kind of cc-dumped this in here without any why's and how's. It's the sum-up of a communication on jack-devel that left us irritated because our devices would perform so strangely.
Could someone confirm this or hint me towards anything?
Thanks, Wolfgang
Donnerstag, 7. Februar 2008 Wolfgang Woehl:
For, at least, our two snd-hda-intel devices (ALC888 and 82801I (ICH9 Family)) both position_fix values (POSBUF and FIFO, 0 and 3 respectively) produce ok sound, only that
position_fix=0 makes alsa fine and jack crackle/sizzle position_fix=3 makes alsa crackle/sizzle and jack fine
"alsa" meaning direct use of alsa and "jack" meaning (at least in my case) jack with the alsa backend.
Here I have alsa-1.0.14-31.2 and jack-0.109.2-0.pm.1 (0.pm.1 stems from the packman repository, I guess)
Wolfgang
At Mon, 11 Feb 2008 22:22:27 +0100, Wolfgang Woehl wrote:
Hi,
I'm sorry I just kind of cc-dumped this in here without any why's and how's. It's the sum-up of a communication on jack-devel that left us irritated because our devices would perform so strangely.
Could someone confirm this or hint me towards anything?
Usually position_fix=0 (is almost equivalent with position_fix=2) should work. But this seems not working on some loose hardwares.
JACK does nothing special but do a very basic full-duplex operation. The difference is that you use short period size and likely two periods. So, the position calculation isn't accurate on your hardware. Maybe the positions of playback and capture streams differe. An often used workaround for this kind of problems is to use three periods instead.
(BTW, this problem really depends on the hardware. That's why there are such a strange module option. So, don't argue this as if it's a general problem.)
Takashi
Thanks, Wolfgang
Donnerstag, 7. Februar 2008 Wolfgang Woehl:
For, at least, our two snd-hda-intel devices (ALC888 and 82801I (ICH9 Family)) both position_fix values (POSBUF and FIFO, 0 and 3 respectively) produce ok sound, only that
position_fix=0 makes alsa fine and jack crackle/sizzle position_fix=3 makes alsa crackle/sizzle and jack fine
"alsa" meaning direct use of alsa and "jack" meaning (at least in my case) jack with the alsa backend.
Here I have alsa-1.0.14-31.2 and jack-0.109.2-0.pm.1 (0.pm.1 stems from the packman repository, I guess)
Wolfgang
Alsa-devel mailing list Alsa-devel@alsa-project.org http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
Dienstag, 12. Februar 2008 Takashi Iwai:
JACK does nothing special but do a very basic full-duplex operation. The difference is that you use short period size and likely two periods. So, the position calculation isn't accurate on your hardware. Maybe the positions of playback and capture streams differe. An often used workaround for this kind of problems is to use three periods instead.
Takashi, thanks for your explanation. The position_fix=0/jackd problem occurs with bigger period sizes too, -p[16..1024] -n2. But: With position_fix=0 and -p[16..512] _-n3_ jackd is fine.
Not that I'd understand why but I'm happy there is a workaround. If someone knows the how and why of this workaround I'd much appreciate to learn it.
(BTW, this problem really depends on the hardware. That's why there are such a strange module option. So, don't argue this as if it's a general problem.)
Like I said, the jack-devel thread was about 2 specific devices and I promise we won't make a big fuss about it :)
Wolfgang
participants (2)
-
Takashi Iwai -
Wolfgang Woehl