[time-nuts] Z3816 Oscillator Saga
nebula at telus.net
Fri May 2 16:20:46 EDT 2014
The Z3816 has a few design short comings. In the circuit board pic the MTI
oscillator circuit board grounding is improved. Subsequently I opened and
modified the oven internally and added an extra oven feedthrough pin. The
issue is the oven heater current modulates the ground pin voltage and that
effects the EFC value (poor layout). I finally connected the oven power
pins to a separate floating PSU with the new internal oven EFC signal
ground pin the only common connection point, while hunting for the source
of a short transient EFC event.
I would caution you that the circuit board traces and hole plating are very
thin, and both are easily damaged by external forces. I used multiple irons
for un soldering problem pins.
The two tantalum capacitors with the black marker dots are not original.
These two capacitors had the wrong voltage rating and became noisy. These
might also be the source of your problem. It can cause unexpected jumps in
the reported EFC value and are not an oscillator problem, just a dumb
assembly mistake. FYI at the turn of the century cell phone sales had a
surge and tantalum capacitors became impossible to get at 10x the price.
I replaced the Z3816 oscillator temporarily with no problems but observed
at power up the smart clock algorithm used a much higher gain in the EFC
loop for about an hour then drops to a more normal value.
I also adjusted the oven temperature turning point pot, however, I suspect
the MTI internal oven circuitry self calibrates the crystal turning point.
I never figured out how to stop the oven from self calibration and had
intended to fix the oven circuit before I determined it was a feature. The
pot adjustment seemed to have more effect at power up. As I recall, during
the relatively short calibration event, the frequency changes by 0.008Hz
with a maximum gps phase error of about 400ns. I had wondered if the two
algorithms were interacting badly but I never verified this. Even with this
quirk, the 260 is a great oscillator.
More information about the time-nuts