[time-nuts] Atomic Clocks: It is important that they keep good time, Part 1

[ed breya]

That's an interesting story and history about the clocks. I happen to 
have the original type digital one, used in a 5065A. I don't have an 
actual 5065A, but many years ago acquired a carcass of one - it was 
missing the Rb physics and a few other things, but was a nice box for 
re-use.

I eventually used it to repackage and improve a Z3801A GPSDO acquired 
later. It turned out to be a good size and layout for this purpose. I 
managed to squeeze in the rearranged Z3801A guts, an extra layer of 
mu-metal shielding and thermal insulation on the OCXO, and a couple of 
12V, 12 A-hr SLA batteries for backup. I used mostly the original power 
supply, and put additional supplies and control circuitry inside the 
original digital synthesizer box for shielding. The whole front panel 
and clock were intact, and I wanted to re-use all that stuff, except 
with different functions and labels - I especially like the door in the 
center with controls and monitoring functions behind it.

The clock appeared to be a very complicated mess at first, and a helluva 
thing to figure out without schematics. I eventually found the 
schematics, and the data sheet for the PMOS clock IC, then everything 
started to make sense.

It used a buck converter to make the low voltage for the LED display, 
which I eliminated since I was making new, cleaner low voltage anyway. 
The board also had some simple 1 PPS interface circuitry etc, that I 
moved elsewhere, so two boards were left for the clock function - one 
for the counting and control, and one for the LEDs. This made the whole 
assembly much shallower, and I found that a tin can (6-1/2 oz tuna can, 
as I recall) made an excellent shield/cover for it. I punched a bunch of 
ventilation holes in it, mechanically mounted it behind the clock 
assembly, and ran all the power and signals through feed-through 
capacitors.

I also changed the internal LED supply to a variable shunt regulator, to 
eliminate pattern-dependent load current. To the outside world, it looks 
like a constant-current load, depending on the selected LED brightness, 
regardless of the readout value. Finally, I fixed the "6" and "9" tails 
on the LED readout. The appearance of the top and bottom segments on 
these numbers depend on the particular IC used, and this one didn't have 
them. This always bugs me, and I try to make it "proper" according to 
me, if possible. It turned out to be very easy to add the tails.

Ed