[time-nuts] Adapting my GPSDO to the FE-5680A

Bob Camp kb8tq at n1k.org
Thu Apr 21 00:48:49 UTC 2016


Hi

Most of these telco Rb’s start out around 2x10^-11 at 1 second tau. Their ADEV then drops
as square root of tau. At 100 seconds they hit 2x10^-12 and at 10,000 seconds they (might) 
hit 2x10^-13. A *good* OCXO will do 2x10^-12 at 100 seconds. An excellent part will hold
sub 1x10^-12 out to 1,000 seconds.

Very roughly speaking the Rb will beat out the OCXO somewhere past 100 seconds and somewhere 
before  1,000 seconds. 

Bob

> On Apr 20, 2016, at 6:31 PM, Nick Sayer via time-nuts <time-nuts at febo.com> wrote:
> 
> Thanks. I’ve taken your suggestion for the sine-to-square converter.
> 
> I believe there are two separate commands for tuning the 5680 - one is “temporary” and one writes through to the EEPROM. I’ll be using the latter, of course.
> 
> http://www.ka7oei.com/10_MHz_Rubidium_FE-5680A.html
> 
> I do agree that the short term stability of the 5680A isn’t as good as an OCXO, but at tau ~2s or so, the tables are turned. I’m getting a good measure of my undisciplined 5680A as we speak to get a good control, but it’s difficult, as I’m testing it against a Thunderbolt, and I think I’m seeing its “hump” (http://www.leapsecond.com/pages/gpsdo/) from 10s to 300s coming through. In any event, I’m getting awfully close to the limits of my 53220a. I may go down the road of trying to make a mash-up as you suggest, but I’m going to start by seeing if I can give myself a choice between whether I want short term stability (OCXO) or medium term (Rubidium) for my reference.
> 
> 
>> On Apr 20, 2016, at 1:57 PM, Attila Kinali <attila at kinali.ch> wrote:
>> 
>> On Wed, 20 Apr 2016 07:17:58 -0700
>> Nick Sayer via time-nuts <time-nuts at febo.com> wrote:
>> 
>>> I spent some time yesterday mashing together my FE-5680A "breakout" board 
>>> with my GPSDO to make a GPS discipline board for it. Before I send the board 
>>> off to OSHPark, I'd like to open the design to criticism (and I mean that in 
>>> its neutral sense) here first.
>> 
>> Looking at your schematics, I would replace the input squarer (IC4)
>> by something different than a schmitt-trigger with an input bias
>> voltage. For one, schmitt-triggers are more noisy than normal buffers
>> for an other, the bias voltage will result in a slightly skewed duty
>> cycle. If you want to use a gate, then the canoncical way would be to
>> use an inverter with an input capacitor like you did, but let it self-bias
>> itself by using a 100k-1M resistor from its output to its input.
>> Important: don't use a buffer, as this will only work with an inverter.
>> But I'd rather use a different squaring circuit, if you want to use
>> the input directly for the output. There are many discussions on
>> squaring circuits in the archives. Probably the most simple, yet sofisticated
>> is the one you can find in the TADD-2.
>> 
>> 
>> But:
>> As you can see on http://www.ke5fx.com/rb.htm the phase noise of the
>> FE 5680's is horrible at best, hence I wouldn't use it as source for
>> anything directly. Additionally, the tuning word you write through
>> the RS-232 is stored in an EEPROM inside the FE-5860 (unless i mix
>> it up with another Rb). Writing this tuning word often will wear out
>> the EEPROM pretty quickly. Hence you should not do this too often.
>> 
>> What I would do instead is, use your current GPSDO design, with OCXO
>> and all, but add something with which you can measure the phase/frequency
>> of an external 10MHz reference. One way would be to use a digital DMTD[1,2].
>> Another would be to sample the reference using an ADC and build DMTD in
>> the digital domain. For this you wouldn't need a high sampling rate, a
>> couple of kHz should be enough, as long as the analog bandwidth of the
>> ADC is high enough (>10MHz, better >20MHz). What you need is some PLL
>> though, as you need to create a frequency that is not an integer divisible
>> of 10MHz, as the ADC clock is used to downmix the reference frequency.
>> 
>> Eg:
>> If you can generate a 10001Hz ADC sampling clock from the OCXO,
>> you will get a 1kHz beat frequency. You can "lock" to this using a
>> digital PLL combined with an NCO (numerically controlled oscillator).
>> Then use the steering word for the NCO as an input for the control
>> loop of the OCXO, toghether with the corrections calculated from the
>> GPS PPS.
>> 
>> The advantage of this is, that you get the low phase noise and good
>> short term stability of the OCXO, but can use the Rb to get the nice
>> mid-term stability (somewhere from 1 to 10s up) while getting the
>> accuracy of a GPSDO, whithout ever the need of writing to the tuning
>> word of the Rb. That keeps your Rb more stable (the internal conditions
>> of the Rb do not change) and allows you to compensate for quite large
>> frequency offsets for Rb refernces that are working outside the spec,
>> but are otherwise fine.
>> 
>> One thing that you have to take care of is spurs, though. Because
>> the ADC does some heavy down-mixing, or rather sub-sampling, this
>> approach is quite sensitive to spurs. In order to not introduce some
>> weird oscillations in the control loop due to spurs in the reference
>> signal, you should use some narrow 10MHz filter at the input (at most
>> half the sampling frequency wide). One way to achieve that is using a
>> ceramic resonator which are available at 10MHz.
>> 
>> 			Attila Kinali
>> 
>> PS: I'm pretty sure I am not the first one with this idea. But I have never
>> seen anyone else mention it, much less implement it. Does anyone know why?
>> 
>> 
>> [1] "Digital Dual Mixer Time Difference for Sub-Nanosecond Time
>> Synchronization in Ethernet", by Moreira, Alvares, Serrano, Darwezeh and
>> Wlostowski, 2010
>> 
>> [2] "Digital femtosecond time difference circuit for CERN's timing system",
>> by Moreira, Darwazeh, 2011
>> http://www.ee.ucl.ac.uk/lcs/previous/LCS2011/LCS1136.pdf
>> 
>> -- 
>> Reading can seriously damage your ignorance.
>> 		-- unknown
>> _______________________________________________
>> time-nuts mailing list -- time-nuts at febo.com
>> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>> and follow the instructions there.
> 
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.




More information about the Time-nuts_lists.febo.com mailing list