This story goes back a few months to a couple of Racal counter/timers that I had repaired for resale. I decided that I needed a decent reference to calibrate them against and bought one of the FE5680A Rubidium frequency standards - like the one Dave reviewed. Job done, I thought.
Then I got to thinking - I bought it on ebay, from Hong Kong, but that amounts to buying it from a bloke halfway around the world that I know nothing about. He says he's got a lab full of equipment but I have no proof of that or that he knows how to use any of it - how the heck do I know that this thing actually puts out an accurate 10MHz signal? The chances were, of course, that it was fine but I had no way of checking.
So, I spotted the 1804M for about the same money as I'd paid for the FE5680A and much cheaper than the Thunderbolt references that are kicking around on ebay - sorted
Except it broke
and I couldn't find a 5MHz ocxo to fix it
I could find 10MHz ocxo's though, such as
this one which was the right size, right pinout and similar spec to the original judging by the data sheet for the Rapco (the data sheet for the 8663 ocxo is
here) - just the frequency was wrong.
So I decided to see if I could adapt the Rapco to a 10MHz oscillator. In any event it's a more useful output frequency and at least at some point the 1804M was available with 10MHz out as I have a few different revisions of the data sheet and some describe the unit as being 10MHz - not sure whether that was an option or a change to later units but as pretty much everything has standardised on a 10MHz reference clock it is much more useful than 5MHz.
I traced out the circuit to the block level rather than every individual component and was pleased to find it looked like this:
With a 10MHz ocxo fitted I could remove the 22R resistor and replace it with a 74HC74 to drop the 10MHz down to 5 - I'd get 10MHz out of the BNC - perhaps with a bit of tweaking to the LC filter used to shape it back to an approximate sine wave - and could feed 5MHz into the logic as before.
So I hunted around for a 74HC74 - lucked out on that one. I had just about every other family from a plain 7474, 74LS74 (lots of those) 74C74 (nearly used that) 74S74, 74F74 but not a "HC" - in the end I found a 74HC193 and used that. Slight overkill just for divide by 2 (it's a 4 bit presettable up/down binary counter) but it was on a junk board so free.
Made up a little bodge board using some tripad with the following on it
Here's a photo of the Rapco with the replacement oscillator and bodge board fitted (click on picture for large version).
General quality of construction is very nice - if a little dated with through hole components and an 8-bit Z80 CPU.
The CPU is actually a Hitachi HD64180, there's also a M48T08 2k non-volatile SRAM/timer, a Z85230 serial chip and the AD569 16-bit monotonic D/A used to generate the oscillator control voltage. There are also a lot of Philips PLC18V8Z PALs for the glue logic and a trio of 74HC590 8 bit counters which presumably divide down the 5MHz and provide a count of clock "ticks" for the firmware.
You can also see the separate GPS reciever - this is a Trimble Sv6 (or probably the SV8 as mine does occasionally give info on 8 satellites).
The cable retainer is loose simply because I hadn't gotten round to putting it back.
Here's a close up of the new ocxo and the bodge board with the 74HC193.
OK, so is it going to work? In the previous thread on the original ocxo Aurora mentioned that he'd been advised against changing the ocxo in the Rapco on the basis that the firmware could be quite fine-tuned to the oscillator characteristics. that is certainly possible but I figured if the control voltage range was about right and the slope was the same it was worth trying.
Initial impressions were good - 10MHz at the rear BNC at about 900mV p-p, a bit lower than it should be but it was enough to drive the external standard input on a Racal 1998 counter so "good enough" for the time being. Once it had a satellite fix it started the oscillator discipline routine. I confess I was a
bit dubious but after 24 hours I was rewarded with this.
Once the Rapco reports "fine" control it means that the error is no more than 2 in 10
10 - the 24hr average running with GPS control is supposed to be +/- 5 in 10
11 with a +/- 3 in 10
11 long term accuracy.
Actually that shot is a slight cheat as it was taken about 24 hours later - I didn't have chance to check the rubidium standard yesterday before it lost the fix and couldn't get it back long enough to achieve a "fine" control level for much of the day. So that photo was this morning when I had chance to check out the FE5680A.
As a little diversion here's the Racal 1998 displaying the 10MHz output - for a "repair or spares" purchase from ebay it was surprisingly accurate - it has the 04E "high precision" Racal ocxo reference oscillator. For some reason the displays on these counters are hard to photograph - it doesn't bloom like that in reality.
Finally the FE5680A using the Rapco as the reference clock source. The most significant digit has overflowed so the display should be 10.000000000MHz!
So I finally got to verify that the FE5680A was genuinely 10MHz plus or minus not very much at all - certainly good to 1-2 in 10
10 and probably a bit better. And I'm finally happy with the Rapco (except I need a better antenna but that's another problem).
It turns out that these units are fairly easy to modify for a 10MHz ocxo should the original fail. It might not be the best mod purely to get 10MHz output though as the oscillators aren't the easiest of things to desolder without a vacuum desoldering tool. I think adding a PLL between the Max 903 and the 74HC14 would be the best bet if the unit is otherwise working.
Edit: Added the bodge board circuit diagram for completeness.