10 Nov
2010
10 Nov
'10
8:43 a.m.
Question about the alsa_out tool in the JACK github.
How critical is the libresample portion of the code? What would happen if I removed it?
i.e. if I were sending, say, one stereo pair to one amp, and another stereo pair to a
different amp, and didn't care about the word clocks being in sync, do I really need
that libresample stuff sucking up valuable CPU? Or could I whack it? What would be the
difficulties created by doing that? Again, this'd be only for output channels and I
wouldn't need the word clocks to be in sync.
-ken
10 Nov
10 Nov
11:08 a.m.
On Tue, Nov 09, 2010 at 11:43:09PM -0800, Ken Restivo wrote:
Question about the alsa_out tool in the JACK github.
How critical is the libresample portion of the code? What would happen if I removed it?
i.e. if I were sending, say, one stereo pair to one amp, and another stereo pair to a
different amp, and didn't care about the word clocks being in sync, do I really need
that libresample stuff sucking up valuable CPU? Or could I whack it? What would be the
difficulties created by doing that? Again, this'd be only for output channels and I
wouldn't need the word clocks to be in sync.
you are confusing things here.
its not about wordclocks being in sync. if your wordclocks are in sync.
you dont need alsa_out.
but most likely your wordclocks are not in sync. this means your
soundcards run at different drifing samplerates, and one soundcard will
consume a different amount of samples per jack cycle. but jack is
feeding period samples each cycle.
first try using -q option set to 0 because that will use linear resampling.
reducing the value of the -s option (default 256) will also improve cpu usage.
but might yield worse performance of the algorithm. it depends a bit on
the soundcards you are using.
--
torben Hohn
4 p.m.
On Wednesday 10 November 2010 05:08:00 torbenh wrote:
On Tue, Nov 09, 2010 at 11:43:09PM -0800, Ken Restivo
wrote:
Didn't he mention having one physical soundcard with 4 outs that only
presented 2 outs to jack? If so, do any such single physical soundcards
actually have two clocks? The alsa-out trick would just get the other two
outs on the same card to show in jack. Or is my memory acting up this
morning?
all the best,
drew
Question about the alsa_out tool in the JACK
github.
How critical is the libresample portion of the code? What would happen if
I removed it?
i.e. if I were sending, say, one stereo pair to one amp, and another
stereo pair to a different amp, and didn't care about the word clocks
being in sync, do I really need that libresample stuff sucking up
valuable CPU? Or could I whack it? What would be the difficulties
created by doing that? Again, this'd be only for output channels and I
wouldn't need the word clocks to be in sync.
you are confusing things here.
its not about wordclocks being in sync. if your wordclocks are in sync.
you dont need alsa_out.
but most likely your wordclocks are not in sync. this means your
soundcards run at different drifing samplerates, and one soundcard will
consume a different amount of samples per jack cycle. but jack is
feeding period samples each cycle.
first try using -q option set to 0 because that will use linear resampling.
reducing the value of the -s option (default 256) will also improve cpu
usage. but might yield worse performance of the algorithm. it depends a bit
on the soundcards you are using. 
4:09 p.m.
On Wed, Nov 10, 2010 at 9:59 AM, drew Roberts <zotz(a)100jamz.com> wrote:
Didn't he mention having one physical soundcard
with 4 outs that only
presented 2 outs to jack? If so, do any such single physical soundcards
actually have two clocks? The alsa-out trick would just get the other two
outs on the same card to show in jack. Or is my memory acting up this
morning?
as usual, it depends.
some cards have an analog side and a digital side, and the clocks used
by the converters for the analog side are not related to the clocking
for the digital side. sometimes they are.
7:16 p.m.
On Wednesday 10 November 2010 08:43:09 Ken Restivo wrote:
Question about the alsa_out tool in the JACK github.
How critical is the libresample portion of the code? What would happen if I
removed it?
i.e. if I were sending, say, one stereo pair to one amp, and another stereo
pair to a different amp, and didn't care about the word clocks being in
sync, do I really need that libresample stuff sucking up valuable CPU? Or
could I whack it? What would be the difficulties created by doing that?
Again, this'd be only for output channels and I wouldn't need the word
clocks to be in sync.
The point of libsamplerate is exactly for word-clocks that are not in sync.
While both soundcards might say they run at 48 kHz, the master (which is
clocking jack) is running at 47990 Hz while the slave runs at 48010 Hz, that
is 20 Hz difference. Which means the slave needs the next block of audio 0.4 ms
before the master, otherwise it will produce silence for 0.4 ms. Such a small
silence every second, actually several smaller clicks of silence each second,
will be very audible as clicks in the slaves output.
The only way you can skip libsamplerate is when both really run from the same
word-clock but then the only reason to use alsa_[in|out] would be when the
first device is a firewire/network/macosx device and you want to use the sample-
synced alsa device too...
Have fun,
Arnold
4924
days inactive
4924
days old
linux-audio-user@lists.linuxaudio.org
4 comments
5 participants
participants (5)
-
Arnold Krille -
drew Roberts -
Ken Restivo -
Paul Davis -
torbenh