On Mon, May 16, 2011 at 11:41:53AM -0700, Eric Steinberg wrote:
> Hi all,
>
> I'm trying to build a recorder that makes a 24 hour long audio recording,
> with multiple channels. So far, I can start jackd:
>
> $sudo jackd -d firewire
>
> and start ecasound:
>
> $sudo ecasound -a:1,2,3,4,5,6,7,8 -f:s16_le,8,44100 -i jack,system -a:1 -o
> channelone.mp3 -a:2 -o channeltwo.mp3 -a:3 -o channelthree.mp3 -a:4 -o
> channelfour.mp3 -a:5 -o channelfive.mp3 -a:6 -o channelsix.mp3 -a:7 -o
> channelseven.mp3 -a:8 -o channeleight.mp3
> ...this starts recording without problems. However, when I stop recording
> by issuing ctrl-c to ecasound and then start recording again, the files
> created now have zero bytes. If I stop and restart the jack server, then
> recording works normally. I would expect ecasound to overwrite the files
> and record new files; why must I restart jackd for audio to be recorded?
> Thanks for any info.
> -Steiny
Hi Eric,
Julien mentioned the issue of the selection and channel count of the
output channels.
Here are two (untested) ways to deal with it.
First of all, with -f:s16_le,8,44100, it looks like
you are declaring the input signal to be 16-bit.
In fact, JACK's internal format is floating 32-bit, so you
probably want -f:f32_le,8,44100.
Also -f specifies the format for *all subsequent input and
output signals*, so you'll need a -f:s16_le,1,44100 before the
outputs.
Since each chain gets a copy of the 8-channel signal, you'll
have to move the channel you want to channel 1 using
the -chmove operator.
ecasound -f:f32_le,8,44100 \
-a:1,2,3,4,5,6,7,8 -i jack,system \
-f:s16_le,1,44100 \
-a:1 -o channelone.mp3 \
-a:2 -chmove:2,1 -o channeltwo.mp3 \
-a:3 -chmove:3,1 -o channelthree.mp3 \
-a:4 -chmove:4,1 -o channelfour.mp3 \
-a:5 -chmove:5,1 -o channelfive.mp3 \
-a:6 -chmove:6,1 -o channelsix.mp3 \
-a:7 -chmove:7,1 -o channelseven.mp3 \
-a:8 -chmove:8,1 -o channeleight.mp3
Just for interest. Here is another way to do it:
ecasound \
-a:1 -f:f32_le,1,44100 -i jack_multi,system:capture_1 \
-f:s16_le,1,44100 -o channelone.mp3 \
-a:2 -f:f32_le,1,44100 -i jack_multi,system:capture_2 \
-f:s16_le,1,44100 -o channeltwo.mp3 \
....
However, there appears to be another wrinkle. According
to 'man ecasoundrc', ecasound will output a *stereo*
signal to lame:
ext-cmd-mp3-output
Command for starting mp3 output. Ecasound will write
stereo, signed, 16bit, little-endian samples to standard
output. Sample rate is configurable. Before execution, %f is
replaced with path to the output mp3, %S with sample rate in
kHz, and %B with encoding bitrate in kbps. Double-quotes and
backslash-espacing can be used to include white-space to
individual parameters. By default Ecasound will try to
launch lame. As a special exception (added to Ecasound
2.7.0), if the output program is "lame" and options include
"--big-endian", Ecasound will detect this and output
big-endian samples
So you may want to tell LAME to record in mono:
.ecasoundrc:
ext-cmd-mp3-output=lame -m m - %f
Regarding zero-length file, I will defer to the other
experts about which of Ecasound's JACK transport modes
you will want to use. There aren't that many, so you
could try them all, using the -G option, i.e.
ecasound -G:jack,system,notransport # ignore transport seek
I also wonder if you might get better behavior by
terminating Ecasound differently, for example with
a TERM signal, or via Ecasound-interactive mode
commands.
Regards,
Joel
--
Joel Roth
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user@xxxxxxxxxxxxxxxxxxxx
http://lists.linuxaudio.org/listinfo/linux-audio-user
