Valentin MISSIER wrote: > I am now trying to use the analog (later I2S) sound output. I build the > 2.6.11.7 kernel with sound support and devfs, so I have /dev/sound/dsp and > all. > Sound does not work for me (I am trying to build aumix to see if that's a > mixer problem...). A simple "cat sound.au > /dev/sound/dsp" does not work for > me. > However, mpg123 seems very slow (even though I can't hear any sound from the > output), as I use kernel soft floating point (the regular kind, as with the > fast kernel soft float engine I cannot login). The 9302 eval board arrived with an windoZe ce app which could decode an mp3 file and render PCM to the onboard codec which I believe is operating in I2S mode. In my attempt to decode the audio section of the data manual it appears only I2S mode furnishes a clock from the 92xx to external circuitry. So this agrees with usage of the CS4271 codec. I was curious at what frequency the codec was running so I measured the MCLK input of the CS4271 which was 11.06835Mhz. This appears odd as the audio frame rate (LRCK) is divided by MCLK for one of: 64, 96, 128, 192, 256, 384, 512, 768, 1024.. of which only a subset apply depending upon master/slave mode and oversampling rate [x1,x2,x4]. My point here is MCLK running at 11.06835Mhz doesn't divide to a familiar frame rate, Eg: 32Khz, 44.1Khz, 48Khz, etc.. divisor LRCK ---------------------- MCLK/128 86.471Khz MCLK/192 57.647Khz MCLK/256 43.235Khz MCLK/384 28.823Khz Some of these look closer to baud rates than conventional audio sample rates. So it is possible Cirrus created the mp3 at an oddball frame rate, or just let it decode at the wrong frame rate, or this is somehow a bug in the wince driver. Regrettably the wince bits are no longer available for experimentation as they have been overwritten by redboot when that battle started. Anyone out there using I2S under linux successfully (either with the CS4271 or another codec) and able to generate a standard LRCK frame rate? -- uhmgawa@xxxxxxxxxxxxxx www.gnu.org