[time-nuts] Cheap GPSDO's

Nick Sayer nsayer at kfu.com
Thu Mar 3 14:13:31 EST 2016


> On Mar 2, 2016, at 6:00 PM, Keith Loiselle <keith.loiselle at gmail.com> wrote:
> 
> Below are some comments from Said:
> 
>>> The TCXO variant is powered by an LDO rather than a switcher, FWIW.
> 
> The published schematics only showed the DC-to-DC switcher version, sounds
> like the TCXO version should have better PN and spur performance with an
> LDO.

You might be behind. Go to https://hackaday.io/project/6872-gps-disciplined-tcxo and check in the “files” section. There are PDF schematics of the 1.5.3 (DOT050V) and 1.6.2 (OH300) hardware versions which are presently shipping.

I had plans for the next released variant to run at 20 MHz, which requires the digital section to be powered at 5 volts (with a separate dedicated 3.3v LDO for the GPS module), with the controller offering variable output frequencies by dividing the 20 MHz clock. That variant is going to be superseded by the one with the hardware phase detection. It will still run the digital section at 5 volts, but with a 10 MHz oscillator and no output frequency selecting. One big gain from powering the digital section at 5 volts is that the 3.3 volt supply now only powers the oscillator and the analog section - further isolating it from variances caused by the digital section. It also gives greater output amplitude, FWIW. The output frequency selection infrastructure had to go because I had to switch to a different controller (the ATTiny841), and it doesn’t have enough pins (and the feature isn’t sufficiently valuable to fuss over it. The original justification for the 20 MHz variant was the higher clock frequency afforded greater phase detection granularity, but that’s mooted by the hardware phase detector).

> 
>>> In my bench experiments with the OH300, its stability did not differ when
> powered by the switcher as opposed to a massively heat-sunk LDO.
> 
> Stability is likely not affected by the design issues I had mentioned (even
> though using the LDO or switcher as a voltage reference for the DAC will
> give poor tempco and thus influence stability over temperature especially
> when using DOCXOs as opposed to using a good, dedicated voltage reference).
> What the issues that I had flagged will affect most are phase noise and
> spur performance. If the switcher happens to operate at 2.50025MHz for
> example, you will see large and wide spurs at 1KHz, 2KHz, 3KHz, etc offsets
> in the output spectra of the 10MHz output as the fourth harmonic of the
> switcher frequency beats with the crystal output.
> 
>>> Still, most of the items of which you have expressed concern would, in
> principle, have an impact on low tau
>>> ADEV measurements, and my results there are in line with the expectations
> set forth by Connor Winfield.
> 
> For low-taus of less than 1s, yes. The items listed mostly affect phase
> noise and spurs. I am not familiar with the expectations that CW sets, but
> what we likely should use as a benchmark instead are some low-cost
> secondary market alternatives such as Z3801A’s, Thunderbolts, etc.

Those are easily at least an order of magnitude more stable than the OH300, which is itself an order of magnitude more stable than the DOT050V.

The OH300 has a 1s tau ADEV of around 1e-11. The DOT050V datasheet claims 1e-9, but in my own testing I’ve seen closer to ~2e-10. I e-mailed Connor Winfield about this and they acknowledged that the average performance is far better than the claim on the datasheet. For the OH300, I’m seeing performance on a par with their claim, with the caveat that my TIA’s own measurement abilities peter out at around that level (it can’t distinguish an OH300 from a Thunderbolt, for instance).

I do not pretend at all to make a GPSDO that’s as good as those alternatives. What I offer is a less expensive GPSDO (comparing prices new rather than used), and that’s open hardware and has open source firmware. I believe the rest of the hardware performs in a way commensurate with the performance of the oscillator(s). If the oscillators were better, then I agree that the rest of the hardware probably would need to be improved to keep pace. I believe the market for oscillators at that level is already well served by Jackson Labs and Trimble and a host of others with whom I don’t believe I can reasonably compete.


> Those
> units have good to superb phase noise performance. Have you taken Phase
> Noise measurement of the output?

I have not. My TIA (53220A) lacks the ability to do so (unless there’s a capability in there I haven’t found yet - which wouldn’t surprise me in the least).

> With a good PN system it should be
> possible to fix these issues interactively.
> 
> Keith
> 
> 
> 
> Keith
> 
> On Wed, Mar 2, 2016 at 1:17 PM, Nick Sayer via time-nuts <time-nuts at febo.com
>> wrote:
> 
>> 
>>> On Feb 25, 2016, at 12:41 PM, Keith Loiselle <keith.loiselle at gmail.com>
>> wrote:
>>> 
>>> Below is a message Said asked me to forward to the group:
>>> 
>>> 
>>> Gents,
>>> 
>>> 
>>> 
>>> While I haven’t posted here for a while, I have been following Time Nuts.
>>> One recent post caught my attention, and here are my comments.
>>> 
>>> 
>>> 
>>> There are some serious basic design issues with Nick Sayers’ GPSDO hack
>>> mentioned by Attila in the previous post. Since Nick is asking up to $275
>>> for the kit, and there are various ways to get professional-grade
>> brand-new
>>> GPSDO’s for significantly less than that. Here are some comments:
>>> 
>>> 
>>> 
>>> 1.       The design is based on a 3.3V switching regulator. The output of
>>> this is used essentially without any significant filtering to power the
>>> DAC, the DAC opamp, the RF output drivers, and the OCXO/TCXO. All of
>> these
>>> items should be powered by a low-noise LDO and the DAC should be driven
>> by
>>> a high-stability voltage reference. The noise floor and spur levels will
>> be
>>> quite bad, and since the switcher is running at 2.5MHz, its fourth
>> harmonic
>>> and higher harmonics (at 10MHz, 20MHz, etc) will beat with the TCXO/OCXO.
>> 
>> The TCXO variant is powered by an LDO rather than a switcher, FWIW.
>> 
>> In my bench experiments with the OH300, its stability did not differ when
>> powered by the switcher as opposed to a massively heat-sunk LDO.
>> 
>> None of the Connor Winfield oscillators I’ve been using have a separate
>> reference voltage output, which would be the best driver for the DAC. In
>> lieu of that, I chose to reference the DAC directly from the oscillator
>> supply voltage, in keeping with Connor Winfield’s own recommendations.
>> 
>> 
>>> 
>>> 
>>> 
>>> Any switching noise on the 3.3V supply will also be directly transferred
>>> into the opamp positive input pin via R17 and will be amplified to find
>> its
>>> way to the EFC pin of the OCXO. There should have been a cap in parallel
>> to
>>> R16 to filter this noise out. Additional low-pass filtering at the output
>>> of the opamp would have also helped.
>> 
>> There are footprints in the most recent board for an RC filter on the
>> output of the OP amp, and for an additional cap across the feedback
>> resistor.
>> 
>> My experiments have shown no benefit to populating them, but they’re there
>> if you want to have a go.
>> 
>>> 
>>> 
>>> 
>>> 2.       The opamp positive input is driven by a 50K Ohms equivalent
>>> impedance. This high resistance will cause resistor thermal noise to pass
>>> right through the opamp, and will pass right into the EFC pin and
>> modulate
>>> the OCXO. Same issue with the negative input pin and the high-value
>>> resistors R14 and R15 connected to that input pin
>>> 
>>> 
>>> 
>>> 3.       The DAC only has ~200ppt resolution (16 bits). If using a DOCXO
>>> with stability in xE-012, this will cause issues as the LSB resolution is
>>> 10x to 100x worse than what the DOCXO could achieve in stability
>> 
>> The DAC has an ~200 ppt resolution with the DOT050V variant. With the
>> OH300, the resolution is closer to 20 ppt.
>> 
>> Given the stabilities of the two oscillators, I believe this is sufficient.
>> 
>>> 
>>> 
>>> 
>>> There are additional issues in this simple hw design. This design is not
>>> professional grade in my opinion, and at $275 not really priced as an
>>> amateur kit either.
>> 
>> Well, alrighty then.
>> 
>>> 
>>> 
>>> 
>>> For example at $220 in single piece quantities for a brand new LTE-Lite
>> kit
>>> you could also purchase a very good MTI or Morion DOCXO for about $25 on
>>> eBay that can be gluelessly disciplined by the LTE-Lite. That solution
>>> would still cost a bit less than Nicks kit overall and have DOCXO
>> stability.
>>> 
>>> 
>>> 
>>> There is one big advantage to Nicks design - access to the firmware for
>>> tweaking and expanding. That fw may be useful as a basis to design, or
>>> modify the hardware to follow the above design suggestions.
>> 
>> One thing I anticipate adding in the near future is a hardware phase
>> detector (and full credit here goes to Jim Harman). The goal of that is to
>> improve the responsiveness by a couple orders of magnitude, which will have
>> a lot of impact on the 10^1-10^3 ADEV, if all goes according to plan.
>> 
>> Still, most of the items of which you have expressed concern would, in
>> principle, have an impact on low tau ADEV measurements, and my results
>> there are in line with the expectations set forth by Connor Winfield.
>> 
>>> 
>>> 
>>> 
>>> Bye,
>>> Said
>>> 
>>> 
>>> Keith
>>> 
>>> On Thu, Feb 11, 2016 at 12:34 AM, Attila Kinali <attila at kinali.ch>
>> wrote:
>>> 
>>>> On Thu, 11 Feb 2016 00:13:20 -0700
>>>> Joseph Gray <jgray at zianet.com> wrote:
>>>> 
>>>>> I searched the list archives and found some discussion mid-to-late
>>>>> last year about several inexpensive GPSDO's made by bg7tbl. It seems
>>>>> that all of the better models (according to discussion on the EEVBlog
>>>>> forum) are gone.
>>>> 
>>>> Have a look at [1] especially at last two paragraphs.
>>>> 
>>>> If you want something cheap but ok GPSDO, get one of Nick Sayers [2].
>>>> He sells them on tindie [3]. The TCXO version goes for $175 and the
>>>> OCXO version for $275. If you already have a good OCXO you can
>>>> modify it to get the EFC voltage out. Maybe you'll have to adjust
>>>> the firmware for it, which you can find on github [4].
>>>> 
>>>> 
>>>>                       Attila Kinali
>>>> 
>>>> 
>>>> 
>>>> [1] http://www.ke5fx.com/gpscomp.htm
>>>> [2] https://hackaday.io/project/6872-gps-disciplined-tcxo
>>>> [3] https://www.tindie.com/products/nsayer/gps-disciplined-ocxo/
>>>> [4] https://github.com/nsayer/GPS-disciplined-OXCO
>>>> --
>>>> It is upon moral qualities that a society is ultimately founded. All
>>>> the prosperity and technological sophistication in the world is of no
>>>> use without that foundation.
>>>>                -- Miss Matheson, The Diamond Age, Neil Stephenson
>>>> _______________________________________________
>>>> 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.
>> 
>> _______________________________________________
>> 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 mailing list