>  >You should only need *one* measurement in each of the two axis (N-S, E-W)
>  >to get the wind vector.  Of course it would be wise to take multiple
>  >samples and average them.  You could even have some indicator of the
>  >*spread* of the min. and max. to get an idea of whether the wind is
>  >"gusty" or "steady".
>
>  The speed of sound varies significantly over temperature (it also varies
>  with pressure/altitude, though maybe not enough over the range of pressures
>  encountered in a fixed location to worry about)  Either the temperature
>  reading needs to be factored into the computation, or soundings taken in
>  opposite directions and the *difference* attributed to wind speed.

This doesn't sound right.  It looks like you need to factor in speed of
sound differences/temperature even going both ways.

If the speed of sound in 40 degree air is 100 units/sec, and in 20 degree
air is 200 units a sec, then:

        40 degrees - 100 units/sec
      1 unit/sec wind  ----->     sender    100 units distance   receiver

  The transit time is roughly .99 seconds.  The return trip is roughly
1.01 seconds.  The difference then is .02 seconds.  So far so good.  Now

        20 degrees - 200 units/sec
      1 unit/sec wind  ----->     sender    100 units distance   receiver

  The transit time is roughly .495 seconds.  The return trip is roughly
.505 seconds.  The difference is .01 seconds.  Same wind speed, different
temperature.

  I realize my numbers are not quite precise (Since the transit time downwind
is less than a second you will not get a full extra unit in the time, etc
etc) and the example speeds are extreme, it is just to demonstrate the
problem.

If there is a flaw in my math feel free (as if the list wouldn't) to correct
me.

Roger