I found the missing piece of paper on which I wrote down all the data directions, FSK etc for the microwave station. This may be useful in deciphering operation, upside down data etc...

Considering the *higher* frequency microwave box (10450):
   Microwave frequency FSKs up
   105 MHz i-f FSKs up
   45 MHz i-f goes down
   data goes up
   pin 11 13055 goes down
   pin 10 13055 goes up
   tune meter goes up

Considering the *lower* frequency microwave box (10345):
   Microwave frequency FSKs down
   105 MHz i-f FSKs up
   45 MHz i-f goes down
   data goes up
   pin 11 13055 goes down
   pin 10 13055 goes up
   tune meter goes up

For the above logic "1" the output of the 2n2222 (inverting)data corrector goes low. This voltage when added at the AFC integrator summing node causes the reference, and thus the frequency of a locked carrier, to be offset from the discriminator center. Now instead of the AFC trying to drive the voltage on pin 11 (which is lower than center for a logical "1", see above) to discriminator center it will hold it low of center, the logical "1" side. Thus, very slow data will not be high pass filtered by the corrective action of the AFC and data rates all the way down to DC will get through the receiver. A long carrier (relative to AFC time constant) will therefore sit on the side of center consistant with what the data output is rather than being dragged to center and causing noise on the data line due to short term frequency changes.

When data toggles, the *average* correction to the AFC center is (if the resistors are correct!) zero and the AFC holds the average frequency of the data centered on the discriminator.

If the data line ever gets in the wrong state and the signal is locked on the wrong side, as soon as data starts toggling the action of the AFC to hold the average frequency to center will correct matters. This might happen the first time after power is turned on. However once the carrier which the rcvr incorrectly thinks is (say) "1" data starts toggling, that previous "assumed 1" level which is really "less 1" than the "assumed 0" will be dragged across discriminator center and will start the data line toggling in the correct manner. The AFC will now hold the discriminator centered on the data for all data rates from DC up as described above.

This offset circuit *does* hard code information about the expected deviation into the receiver since it says how far off of data center a logical state is to be. If the transmitted deviation is *less* than this value, it could be possible for the above corrective action (receiver wrong about the data state) to fail. As it is, I think we have about +- .2 Volts correction and you measured +- .3 Volts of data swing. This actually means that the definition of data center moves .1 volts between long carrier and toggling data. This should have only a slight effect on duty cycle of incoming data within an afc time constant of the start of data. The squarer in the 13055 should have enough gain to make it negligible. It may be observable on the tuning meter but probably not in a very direct manner since the tune meter uses the same reference as the afc. Perhaps this should be changed to make it more truly a discriminator output. Part of the "click" in the audio on data may be caused by this too, I don't know. Maybe that is a feature and not a bug (:>)

73

Glenn