Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.naic.edu/~phil/hardware/windmeter/readme_windspeed.txt Дата изменения: Fri Oct 17 16:44:25 2008 Дата индексирования: Wed Jan 14 10:52:12 2009 Кодировка: Поисковые слова: http astrokuban.info astrokuban

This directory, C:\hagen\ecad\ECADWIN\windspeed, contains the schematic
diagrams of the control room windspeed monitor (in EasyCad ecw and fc7
formats and autocad dwg format. PDF file from R.M Young, Co. show catalog
pages and parts lists for the platform-mounted "airplane" wind sensor.

J. Hagen Aug. 30, 2005



The following was added by Dana Whitlow on 16 Oct 2008:

The windspeed sensor itself produces a pulse train whose frequency is
proportional to wind speed; 90 Hz corresponds to 19.7 mph. There are
3 pulses per revolution (in case that matters to you). This is one of
its two outputs.

The wind vane for direction is mechanically coupled to a special 10k
pot which permits wraparound with negligible (or no) gap at the
wraparound point. You drive the top of the pot with a DC reference
voltage and measure the voltage at the slider to get wind direction.
A direction of 0 degrees produces 0 output; a direction of 180 degrees
produces half the reference voltage; a direction of 360(-) degrees
yields an output equal to the applied reference voltage. This is the
other output from the sensor assembly.

The upstairs circuitry uses a 555 timer and a precision current source
and ramp integrating capacitor to produce a series of output pulses
whose width is proportional to the wind direction (modulo 360 degrees),
and whose PRR is the same as the output of the windspeed sensor.

Expressed more briefly, wind speed is encoded as pulse rate and wind
direction is encoded as pulse width. The output pulses from this
circuit are converted to optical form and sent down a fiber to the
IF/LO room.

Now at the downstairs end the fiber signal is converted back to TTL
pulses. The rate is measured by the "tachometer card" to obtain
wind speed, and the pulse width is measured by the "azimuth card" to
get wind direction.

The azimuth card takes in the input pulses through a 2-stage shift
register, whose first stage merely synchronizes the input pulse edges
with an internal 32 kHz clock. The second stage, along with U2, is
a trailing-edge detector. A two-stage 5-bit counter is enabled by
the synchronized pulse, then the detected trailing edge loads the
count at that time into a latch which drives 5 bits of a D/A converter.
The output of the converter is buffered and drives the analog outputs
of the system. Also the 3 most significant bits of the counter drive
a 3-bit (8-position) LED compass rose display of the wind direction.

I'm a little but unclear about one aspect of how this works- the
counter gets preset to 19 counts immediately after the speed-dependent
count is latched for the D/A, but the max count for a 5-bit counter
is only 31 counts. But one thing is clear- only 5 bits get transfered
to the D/A, so the wind direction resolution is 360 / 31, about 11.25
degrees.

I'm guessing that the presetting has something to do with avoiding
operating too close to zero pulse length. So perhaps the wraparound
is not at 360 degrees, but elsewhere.

The D/A's output also goes to a separate small microprocessor in the
downstairs chassis which measures the voltage and reports it to the
outside world via ethernet. Phil says that the microprocessor also
smoothes the data by running a 10-point rolling average.

The current complaint from Phil is that the output data currently
shows jumps of about 100 mV from the D/A output, and this is reflected
in his data via ethernet. He says that this problem started after
the microprocessor was replaced. One theory- the replacement processor
is being run at a much faster clock rate such that the rolling average's
impulse response is too short to filter effectively.

In any event, the ~100mV steps in the analog output seem consistent
with the 10 mV per degree nominal scale factor and the 5-bit resolution
over 360 degrees.

The business of the 19-count preset etc should be understood before this
tome can be considered complete. But, 5 bits is 5 bits, and I don't
think the 19-count preset mystery has anything to do with our present
problem.