Hi all
A few weeks ago I bought a Racal-Dana 1991 frequency counter. It is a great instrument! The one I bought have already been modified with a third BNC for an internal 1.3GHz prescaler, mounted exactly where the 1992 has channel C. The output of the prescaler goes via a relay in parallel to the A channel BNC. The relay is switched by a little slide switch on the front panel. All the work have been made very well, like it had been done at the factory :-). The internal prescaler circuit is hand made, on a perforated board, but it is well done in a HF tin box.
1. Prescaler divide by 1000, and here begin the problem: the first chip is a SDA4212, a good chip, but it divides by 64 or 256! To obtain simply reading values, it have been done a circuit like this:
https://archive.org/details/ALowCost1.2GhzPre-scalerIn that prescaler, for two times, every 5 pulses 1 is lost (4/5), then it divides by: 64*1.25*1.25=100 (1.25 is 5/4). Another divider by 10 gives the final division by 1000.
The problem is that the Racal-Dana is very fast, doing a reciprocal counting, then the periodic lost pulse makes jitter in the reading. It was a relevant problem for a high resolution instrument, then I looked for a solution. At the principle, I thought making a new prescaler. I tried MC12080, a divider by 10/20/40/80, but, I don't know exactly why, I was obtaining a division by 4! I suspect it is a problem due to the input signal: that chip could need a well squared signal, if it has been done to be used in PLL circuits.
After a moment, i realized that the counter has a (R-X)/Z math function! "I could simply divide by 64, then multiply by 64 dividing by 0.015625!" I tried and it was ok! I only had to accept to insert the value each time before using the prescaler: .015625 shift store Z shift (R-X)/Z. It required a few seconds, but why do it manually if I can use a microcontroller?... :-) Then I studied the keyboard, then I bought 9x PC817 optocouplers and driven them by an atmega328P (Arduino microcontroller). Each time I switch on the prescaler, the microcontroller makes the sequence for me. Each time I switch off the prescaler, it sets Z=1 and deactivate the math function. It works!
2. OCXO was not very good, then I bought a used CETC CTI OSC5A2B02 10MHz square wave ocxo for $6 and I mounted it on a perforated board with a salvaged TL431 and a 20-turns trimmer. That's OK! Its frequency error is worse when it is cold (+22ppm against -0.40ppm), but it stabilize in 3 minutes instead of 40 minutes! During warming, it draws 0.47A @5V.
3. "It would be wonderful if I could know when the oven has reached the final temperature..." I used a little 10k NTC mounted near the upper side of the OCXO can, with some heat conductive paste and made a thermal insulating cover over all. The same microcontroller read the voltage from the NTC and compare it with the stored values: when the OCXO is cold, the standby LED bllinks at 0.25Hz, when OCXO is warm so a little precision have been reached it blinks at 1Hz, when temperature gives a good precision it blinks at 4Hz and when the maximum precision have been reached the LED turns off.
3 jumpers store the voltages from NTC when I cut them (or re-close and open) for the 3 conditions.
4. Chips inside were very hot! Some of them are custom made, then I had to do something to preserve them! I made some holes on the right side and put a little fan inside. I don't need the GPIB card, then I removed it, so less heat is produced and the connector holes on the back are useful as warm air outlet. I've reduced the fan speed to reduce noise. Sometimes the fan did not start, then I bypassed the series resistor with a 47uF capacitor giving a strong pulse at power on.
If someone is interested, in the next days I will publish schematics and software.
Gianluca
Home
Help
Search
Login
Register
