[wdmaudiodev] Re: Modified kernel streaming example

  • From: uwe kirst <u.kirst@xxxxxx>
  • To: wdmaudiodev@xxxxxxxxxxxxx
  • Date: Mon, 15 Dec 2003 22:12:30 +0100

User for realtime audio wrote:

Hello Community,
I have modified the kernel streaming example (directKS.zip) to create a directKS host for audio processing
(input -> output) with low latency. I am using three different soundcards and have observed different
behaviors. One is however very strange:

I am using Win XP, 2 GHz PC, flushing 2 buffers with variable length for reading/writing in ONE thread,
using the EGO SYS U2A the result depends very much on the selected buffersize:
At 32 kHz samplerate, using buffersizes below 512 does not work (which is what I actually expected) but also
setting the size to more than 1024 also fails. In this case the read-process for buffer1 always immediately returns although
only in the very first case the buffer is really filled with audio data. After that the buffers contain the same data always. This
results in playing back the same buffer always.
The read-process terminates whenever coming to the position in the software, that is I get multiple reads while only one write thread
is indicated to being completed.

The same software works for SoundMax for any buffersize and for Terratec EWS 88 MT (even though I can not get any sound from
that card but at least write and read are completed one after another). Also talkthrough works fine with 512 <= buffersize <= 1024 for U2A.

the buffersize may depend on the dma type used in the driver: WAVECYCLIC or WAVEPCI. My soundblaster card is WAVECYCLIC and has no buffer limit, but another card of mine is WAVEPCI and has a limit of 32768 Byte.


WDMAUDIODEV addresses:
Post message: mailto:wdmaudiodev@xxxxxxxxxxxxx
Subscribe:    mailto:wdmaudiodev-request@xxxxxxxxxxxxx?subject=subscribe
Unsubscribe:  mailto:wdmaudiodev-request@xxxxxxxxxxxxx?subject=unsubscribe
Moderator:    mailto:wdmaudiodev-moderators@xxxxxxxxxxxxx


Other related posts: