[WriteLog] testing audio compression systems

[Eric Scace K3NA]

Hello --

   I recorded the WW CW contest using CCITT A-law and was not surprised to find the results were less than perfect.

   I am curious to find out if anyone has done any systematic tests of various compression choices available to us for our specific
application.  Below I've jotted down some thoughts about the problem... but I haven't tested anything and don't really have the
apparatus available to construct a proper test.

   Maybe someone on the reflector has the skills and laboratory access to construct some tests for evaluation.  I'd be interested to
hear more on this subject.

-- Eric K3NA

===============

   Random thoughts about audio recording requirements (recognizing that my technical knowledge is limited):

1.  Dynamic range:

   1.1  We expect our receivers to have much better than 100 dB dynamic range.  For an audio record to replicate what the receiver
presents during the contest, both the sound board and the compression technology need to support a similar dynamic range.

   1.2  Most of the time in the contest we don't need the full dynamic range; i.e., we are listening to a pretty loud signal.  But
occasionally we are pulling out a weak, just-above-the-noise signal from between some loud adjacent channel contesters.  This later
situation appears to be the most demanding; i.e., can the compressed recording recover the weak signal when very loud signals are
all around?

   1.3  I tentatively conclude that the compression algorithm should provide at least 100 dB of dynamic range.  That seems to point
toward CD-quality if one wants the audio record to faithfully recover what was heard by the operator.

   1.4  A 16-bit PCM technique provides 64,000 linear quantizing levels.  That is about 35 dB of range.  I have no idea what dynamic
range can be handled by MPEG, etc.

2.  Audio bandwidth:

   2.1  While we talk of a nominal "2.4 kHz" receiver bandwidth on SSB, in fact we are hearing signals far above this bandwidth.
That is because the filters are not perfectly cutting off at 2.4 kHz.  Therefore, it would be an error to assume that we could limit
the sampling rate to something like 5,000 samples/s (observing the Nyquist limits).  An 8 kiloample/second rate has a theoretical
upper limit of 4 kHz.  To avoid aliasing problems on playback (that could generate new QRM on top of the desired signals), the audio
being recording would have to pass through a 4 kHz low pass filter first.  Is it fair to assume that such a low pass filter is built
into the encoders?

3.  Disk space:

   3.1  WriteLog can be configured to send the recorded audio to a different computer over the Ethernet, if there is not adequate
space on the logging computer's disk.

   3.2  A 1m12s "CD quality" file on my computer is 1.570 MB in size.  That translates into 21.8 kB/s.... or 1.308 MB/mn... or
78.480 MB/hr... or 3.767 GB/48-hour contest for stereo.

   3.3  It's not difficult to have several GB of spare capacity on a disk to record a contest.  After the contest, the files can be
archived on CD (several, if CD-quality recording was used) or DVD.  DVD-writing drives are starting to become affordable.

   3.4  Disk space would be larger for multi-op stations.  W3LPL could use as much as 12 channels or 23 GB (if both listening posts
on 160m and 10m were left on 'record' during the entire contest period), for instance.

   3.5  How much space exists on a DVD?  Can a DVD-write drive keep up with the data rate during the contest?  If so, one could
record straight to the DVD drive.

   3.6  With inexpensive CDs available today, maybe it's just not worth attempting to compress below "CD quality" audio.

======== end =======