On Tue, Feb 03, 2004 at 06:29:14PM +0200, Kai Vehmanen wrote: Ok, this works, and seems to copy in to out on all channels. Still trying to understand that. Could you give a practical example ? Say e.g. that I want to play t2.wav and t5.wav to outputs 2 and 5, and record from inputs 1 and 6 to t1.wav and t6.wav. This emulates an 8-track machine playing back two channels and recording two others. How would that be done ? Many thanks for your help. -- Fons

On Tue, 3 Feb 2004, Fons Adriaensen wrote: This kind of setup is probably easier to handle with JACK, but anyways, still possible without it: ecasound -a:1,6 -f:32,12,44100 -i alsa \ -a:1 -f:32,1,44100 -o t1.wav \ -a:6 -erc:6,1 -f:32,1,44100 -o t6.wav \ -a:2 -i t2.wav -ea:200 \ -a:5 -i t5.wav -erc:1,5 -ea:200 \ -a:2,5 -f:32,10,44100 -o alsa Input object with the most channels determines the channel count of the chain it is attached to. On the other end of the chains, if an output has less channels than the chain it is attached to, extra channels are discarded when writing to the output. With these simple rules you can route audio between channels and chains, and without bloating the syntax too much. The '-ea:200' at the end of chains 2 and 5 helps to keep the volume constant. This is ugly, but unfortunately something I cannot fix at this point (I'd like to change the chain-mixing semantics from add'n'divide to plain add'n'saturate like in most other systems, but unfortunately doing the change now would break far too many existing scripts/sessions :( ...). With JACK the above would be something like: ecasound -f:32,1,44100 \ -a:1 -i jack -o t1.wav \ -a:6 -i jack -o t6.wav \ -a:2 -i t2.wav -o jack -a:5 -i t5.wav -o jack ... so cleaner, but OTOH you have to manage the connections externally. -- http://www.eca.cx Audio software for Linux!

On Tue, 3 Feb 2004, Fons Adriaensen wrote: Yes, error in my example. Ecasound has to combine chains 2 and 5. Former has 1 channel of audio, the latter 5 channels. Now the problem is two-fold: first, Ecasound does not know that only ch5 of chain 5 has audio and others are in fact silent. Secondly, Ecasound mixes chains by doing channel-by-channel add+divide. So without the '-ea:200', volume of channels 1 and 5 in the stream going to the ALSA device would be halved. It's ugly, but ensures backwards compability. [...] Unfortunately the above examples are not quite correct as '-a:t2 -o alsa,hw:0 -a:t5 -o alsa,hw:0' is not equivalent to '-a:t2,t5 -o alsa,hw:0'. The former creates two devices (which will fail in this case as you can open an ALSA hw-device just once), while the latter creates one ALSA device and connects it to two chains). For this reason, I like to write Ecasound setups with one input/output per line. This helps to write valid setups. It wouldn't change the processing result in anyway, so it's ok. But it does add to CPU usage. -- http://www.eca.cx Audio software for Linux!