
       12 September 1993      MGJ


1.  Borrow a high accuracy thermometer.  I happened to get a
    Fluke model 51 for a couple of hours.  You need one that
    is accurate to a half-degree Farenheit.

2.  Modify the source code in calc_farenheit.c so that it
    prints out the raw frequency (in Hz) that appears on
    the audio input.  There is a printf in the code which
    is commented-out; enable it (temporarily).

3.  With Scotch tape, mount the LM34 sensor and the thermometer
    so they are close together (0.5") and hanging over the edge
    of a desk (out in the air).  Wait 10 minutes for the temperature
    to stabilize.

4.  Using /usr/demo/SOUND/soundtool, record 10 seconds of your
    hardware's output at record volume 82.  (Or use /usr/demo/SOUND/record).
    Also write down the reading of the accurate thermometer.

5.  Run the program calc_farenheit on this file.  Write down the
    SPARCstation computed frequency that corresponds to the
    measured temperature.  For example:  339.57hz <--> 72.2 degrees F.

6.  Take a can of cola out of the refrigerator, wipe off any
    condensation, and tape the LM34 and the thermometer to the
    side of the can.  {We use a cola because it has a relatively
    high thermal mass}.  Note: a dab of white silicone Heat Sink
    Grease helps immensley, and also makes a hell of a mess.
    Repeat steps 4 and 5.  For example, 288.83hz <--> 54.0 degrees F.

7.  Get the biggest heaviest coffee mug you have, fill it with water,
    and microwave for 3 minutes.  Let stand 2 minutes, pour out
    the water, and let stand for 5 minutes more.  Now repeat the
    cola can maneuvers, using the warm coffee mug.  Put the sensors
    and the grease on the OUTSIDE of the mug!  For example,
    423.54hz <==> 95.4 degrees F.

8.  Take as many *simultaneous* readings on both the thermometer
    and the LM34 hardware gizmo as you can.  Get as many different
    temperatures as possible.

9.  Plot the thermometer measured temperature, versus the hardware's
    frequency in hertz.  The data should be a straight line.  You
    should be suspicious of any datapoints that fall a significant
    distance away from a straight line.

10. Edit the source code for calc_farenheit.c.  First, change the
    #define constant ENTRIES to be the number of trustworthy
    (hz, temperature) datapoints you have.  Next, edit in the
    values in the freqtab, temptab entries.  Do this in
    increasing order of temperature.  (The search code assumes
    it).

11. All done.  Recompile and you're off and running.  Congratulations.


       --- Mark Johnson
