[Tizen General] Tizen Audio Stack

Patrick Shirkey pshirkey at boosthardware.com
Mon Sep 30 13:53:17 GMT 2013


Another update for those who are interested.

It turns out that there is a more recent version of jack_delay which has a
more advanced algorithm for measuring the loop signal and associated
latency. I have put the results up here:

http://boosthardware.com/pa-jack-latency-2.txt

They are encouraging as the variation is now between 16 - 76 ms. If we
include the approx 10 - 12 ms that the PA Stream Buffer is adding on this
machine then we can see latency going as low as 4 - 6 ms outside of the PA
Stream Buffer. However the variation that I am seeing with this machine is
still a concern for proper low latency audio (sub 5ms) and an average of
10ms for the PA Stream Buffer can go lower too. There is still some work
to be done to figure out the causes of the variation.

Comparing the results to Android is interesting because it shows that the
combination of PA and JACK provides potentially better results and
infinitely more flexibility than the current Android Audio Stack which is
capable of best case 20ms latency on the latest Nexus devices and OS
build. It took them almost 5 years and at least two full time resources to
get to this point where they now have a fragmented audio stack and a lot
of badwill generated in the Professional Open Source Audio Community.

That suggests good things will happen if the PA+JACK combo is accessible
on Tizen and some effort is spent to solve the issues that have been
flagged in this discussion and my tests including:

- Bluez API & Bluetooth device support
- Switching between PA and JACK automagically
- Solving the Latency variation issue

To make things easier, I have put together a basic document on the test
procedure here:

http://boosthardware.com/pa-jack-latency-test.txt

Artem, FYI, I have uploaded your script there for ease of use. It's very
handy :-)

We are now looking into the cause/s of the variation that I am seeing. It
would be helpful to have some additional test results so I have asked the
Linux Audio community to contribute. Running the test on a Sandy Bridge
and the new Atoms would also be enlightening.


--
Patrick Shirkey
Boost Hardware Ltd


More information about the General mailing list