25 Jul
2012
25 Jul
'12
11:57 p.m.
On Wed, July 25, 2012 6:25 pm, Florian Paul Schmidt wrote:
On 07/25/2012 04:31 PM, Neil C Smith wrote:
Some people are reporting 512 period buffers are stable with 4.1 but the
claimed latency of ~10ms appears to be related more to the loading of
samples than realtime streaming. I.e jack would have a hard time still at
low latency. So far no one has reported better than 40ms for realtime
streaming. That is still wip.
You can install a cyanogenmod and get direct access to the alsa layer if
you want. I haven't seen any test results that have explored that option
for realtime latency yet. I can't do it because cyanogen mod doesn't run
on my devices.
I'm planning to get a new nexus at some point. I like the 10 inch form
factor for tablets.
IMO Google should have fired the person running the audio dev team for
screwing up so badly for three years running.
--
Patrick Shirkey
Boost Hardware Ltd
On 25 July 2012 15:01, SxDx <sed(a)free.fr>
wrote:
Note though, that in this example program, while the audio output is
rendered in C, the final pass to the AudioDevice goes through java (the
java code calls the JNI code to render into a buffer). But since under
the android OS, there's just a linux system, it should be possible
(maybe with root access and some hackery) to access the ALSA layer
directly. And then it only depends on the audio hardware in the device.
On my device (un-rooted) I have at best around
40ms.
So for me, android is a no go for live stuff. ...
We are in 2012 dear google, linux can go much lower than that,
the devices out there are blazzingly fast, 40ms is insane.
Android 4.1 brings
~10ms latency *on the right hardware*
But well, that's life. This is "consumer
electronics" after
all. If angry bird works...
There are use cases for low latency audio on consumer
devices, and I
think there is a push for a better performing audio stack.