[ntp:hackers] idev vs cdev in /etc/ntp.audio config

Tim Shoppa shoppa at trailing-edge.com
Sun Oct 5 09:30:12 PDT 2003


Now that I've got the latency problem solved for refclock_wwv on Linux,
I'm looking at a few other hitches I had in getting it going.  In particular,
the comments in libntp/audio.c lead me to believe that the /etc/ntp.audio
config lines work as follows:

  idev /dev/input_device    # I assumed this to be /dev/audio
  cdev /dev/control_device  # I assumed this to be /dev/mixer

But I was exactly wrong; idev is supposed to specify the mixer device,
and cdev is supposed to specify the device that delivers the audio stream.

Is this a problem in how I read the comment, or a problem in how the
code actually implements things?

Also, the agc config line is used to find the number of the IOCTL setting
the input gain... and then audio_gain() hardcodes two IOCTL's (neither of
which is right for Linux) for setting the AGC.  I just re-hardcoded
SOUND_MIXER_WRITE_IGAIN for my usage, but it'd be nice to have this
really be an actual configurable parameter.

I realize that the population of people interested in refclock_wwv is
probably counted on one hand at this point... but the design of the
decoder, the maximum likelihood stuff inside, and the performance are
all so remarkably stellar that it's had me fascinated and obsessed for
the past couple of weeks.

Tim.



More information about the hackers mailing list