Rubidium Standard 5680B unstable? Possible?

[bson]

What is the 1PPS behavior on these 5680 variants?  I was considering rigging something where a 10MHz OCXO output is counted locally, and this counter is read by a PIC which also receives the 1PPS.  Over a period of time it would control the OCXO drift by adjusting its temperature.  I was thinking for precision/repeatability to have the 1PPS latch the counter output, and the PIC reads the latched value.  Whenever it changes it has a new interval sample.   It would also need to interrupt the PIC so it can detect no-lock.  Obviously, this can be no better than the PPS input...  Is a GPSDO like a Tbolt going to work better?  (BTW, my interest is in a low-phase noise freq std, not precision timekeeping.)

[Gyro]

 IK0OTG  did pretty much exactly this (albeit from a GPS 1pps source), might be worth a look....

http://www.ik0otg.net/index.php?option=com_content&view=article&id=53&Itemid=59&lang=en

[TSL]

What is the 1PPS behavior on these 5680 variants? 

99% of the 5680's don't have PPS out, it wasn't in the original spec and only found its way into the very latest models. I have the original Tech Manual for the Series 2 units and there's no mention of PPS anywhere in it, not even as an optional extra. I've yet to find an update of the manual that covers those that do have PPS. I suspect that such a doc doesn't exist outside of the OEM dept where the PPS units came from.

I was considering rigging something where a 10MHz OCXO output is counted locally, and this counter is read by a PIC which also receives the 1PPS.  Over a period of time it would control the OCXO drift by adjusting its temperature.  I was thinking for precision/repeatability to have the 1PPS latch the counter output, and the PIC reads the latched value.  Whenever it changes it has a new interval sample.   It would also need to interrupt the PIC so it can detect no-lock.  Obviously, this can be no better than the PPS input...

As Gyro has pointed out, there is a few people who have already done this sucessfully.

  Is a GPSDO like a Tbolt going to work better?  (BTW, my interest is in a low-phase noise freq std, not precision timekeeping.)

Probably. The specs for the Thunderbolt as a low phase noise frequency ref is exceptional. It is the go-to GPSDO for those who build transverters and such fo the Microwave bands and I know of guys here in Oz that are using it as a basis for gear up to 78GHz.

Some have made their Tbolt even better by removing the OXCO and replacing it with the venerable HP 10811D

Notes of such mods and resultant tests can be found here...

http://www.ke5fx.com/tbolt.htm

regards

Tim

[paulie]

I was considering rigging something where a 10MHz OCXO output is counted locally, and this counter is read by a PIC which also receives the 1PPS.

You maybe got the idea from that great (but recently departed) genius, philosopher, and humanitarian dannyf:

https://www.eevblog.com/forum/projects/toyo-tco-919-%27high-stability%27-crystal-oscillators/

Not only 1/20th cost of alternate units but more capabilities.

[Bryan]

If you want 1pps from 10Mhz May want to have a look at PD9 and PD10. Cost you about $2.00 if you have your own programmer.

http://www.leapsecond.com/pic/picdiv.htm

[bson]

I was considering rigging something where a 10MHz OCXO output is counted locally, and this counter is read by a PIC which also receives the 1PPS.

You maybe got the idea from that great (but recently departed) genius, philosopher, and humanitarian dannyf:

https://www.eevblog.com/forum/projects/toyo-tco-919-%27high-stability%27-crystal-oscillators/

Not only 1/20th cost of alternate units but more capabilities.

I hadn't seen that!  I think he's making it a bit hard on himself though, by taking the XO output and dividing it down to a 1PPS signal, which is then compared to a reference 1PPS input.  Instead, feed the OCXO to a counter, which can be internal to a PIC, or external, and use the reference 1PPS signal to read the counter delta from the previous second.  If it's <10M the OCXO is running slow, if it's >10M it's running fast.  Adjust the PWM output based on e/Tc, where e is the count offset (visavi 10M) and Tc the desired convergence time.  Keep a running average to detect when e is evenly distributed around 10M within that window; at that point it's locked.  Sum e and e^2 to produce a windowed error bound and std dev, and display on the front panel.  Obviously with an 8- or 10-bit PWM output a trimmer needs to be set to bring it to ballpark to begin with, and the PWM control should only adjust within a few degrees, as it's going to lack precision otherwise.  Anyway, that's what I had in mind...

My suspicion is the phase noise (windowed average of |e| while in lock) is going to dominate at some point.  This method can only discipline it to the 1PPS signal, it can't fix phase noise...  But it at least allows measuring it, which is a good start...

[paulie]

I think TCXO phase noise is minimal which is one of their main advantages.  And I think his idea is excellent way to actually obtain a 10mhz reference. Not to mention virtually any other frequency is available independent of the oscillator or GPS but just as stable.

Regarding your other point that is close to what I've attempted using an AVR instead of PIC. My conclusion is the dannyf way might be better but I won't be sure until finishing both.

BTW note that getting clean 10mhz from 1pps is extremely difficult compared to getting 1pps from 10mhz. XOR gates are good for multiplying couple octaves and ringing tanks a little more but beyond that PLL can be huge jitter bugs.

[bson]

Well, I bought a Thunderbolt with both 10M and 1PPS outputs...  Let's see what can be done with this.  Maybe I'll also get an HP 10M OCXO and experiment.