[TowerTalk] Fwd: Telrex rotator

jimlux jimlux at earthlink.net
Mon Nov 19 10:13:51 EST 2018


On 11/19/18 6:31 AM, Hans Hammarquist via TowerTalk wrote:
> Hi Martin,
> 
> Decoding the signals from a selsych is a "little" complex as it require you to do A/D convert all three signals at the same time, then do the mathematical quirk to calculate the angle out of that. Not impossible. There are ICs that do it and I should be able to find leads for you about that. There might also be some software available for down-load on the net somewhere but I have no idea where. The decoding is frequency independent as the calculation only requires the amplitude of the signals
> 

Assuming you don't need microdegree accuracy, and 1 degree is good 
enough, you don't actually need "precisely synchronized" sampling, 
particularly if you're running them at 50 or 60 Hz.  An Arduino, for 
instance, can easily sample at 100kHz.  What you would do is sample them 
all in quick succession, then wait a millisecond or so and sample all 3 
again, etc.

The voltage on each winding is:
V1 = C* cos(theta)*cos(fline*2*pi*t)
V2 = C* cos(theta+2pi/3) * cos(fline*2*pi*t)
V3 = C* cos(theta+4pi/3) * cos(fline*2*pi*t)

C is a constant - related to the excitation voltage and resistances, 
etc.  It's always the same. What you really care about is the *ratios* 
between the voltages.

The important part is the first term: the cos(theta), cos(theta+120), 
cos(theta+240) (angles in degrees).  The tricky part is that an rms or 
peak voltage (trivial to implement) won't work, you need the negative 
values.

Your ADC is probably also measuring Voltage relative to ground, as if 
your synchro is wired star (Wye) - but you can make those pseudo delta 
by just subtracting (if you sample fast enough, the error is small)
V1 = ADC1-ADC2
V2 = ADC2-ADC3
V3 = ADC3-ADC1


Once you've calculated a signed amplitude for each winding, you can then 
just match it up against a table of values.  rather than try to do the 
trig to convert 3 amplitudes to angles, just have a table that has 360 
entries, and pick the one that's closest to what you measured.







More information about the TowerTalk mailing list