[time-nuts] More GPSDO measurements

Matthias Welwarsky time-nuts at welwarsky.de
Wed Sep 23 22:12:57 UTC 2020


Hello time-nuts,

I'm back with some more data of my GPSDO. The OCXO meanwhile has settled in a 
bit and its drift has slowed down considerably, time to make some more 
measurements. Find attached the ADEV graph of a 24h run using the TAPR TICC 
connected to the 1PPS output of the GPSDO. I was able to use a higher time 
constant now for the PI loop (500 seconds). 

For comparison, I've added again the Trimble UCCM-P trace I've shown before. 
But with what I know now about how long OCXOs take before they settle again 
after being off for an extended time, I'm not sure I'm doing it justice. The 
Trimble OCXO that's inside is likely much better than the poor, mishandled 
MV85A I'm using in my own GPSDO, but I didn't give it more than a few days to 
settle. I'll really have to do a parallel measurement run at a point to 
subject them both to identical conditions.

Regards,
Matthias


On Donnerstag, 3. September 2020 09:46:23 CEST Matthias Welwarsky wrote:
> Hello, fellow time-nuts.
> 
> finishing the next PCB revision of my GPSDO cape project, after assembling
> some components that JLCPCB SMT service doesn't stock I did a test run to
> compare it to the previous revision. The measurement setup is identical to
> my previous test, 53131A with MV89A OCXO, frequency mode with 2s gate time.
> 
> There are only a few changes, I routed the OCXO power rail  through an LC
> filter to decouple it a bit from the digital noise generated by the
> Beaglebone. Another change was to add a 1PPS output, using an additional
> timer of the AM3358 in PWM mode, clocked by the 10MHz OCXO output. This
> will give a nice, stable 1PPS signal.
> 
> I also changed the analog output stage to be just a unity gain buffer for
> the DAC instead of a 6dB gain stage. This is probably the most notable
> change as it doubles the DAC resolution. It also shows in the results, see
> the attached picture. Green curve is the new rev2 board, blue curve is the
> best run of the previous revision I could do, magenta is the Rb GPSDO.
> 
> Unfortunately I did a bad pick with the OCXO. I bought three MV85 OCXOs from
> Ebay, two of them obvious fakes, the third one appeared to be genuine and
> it seems to be quite OK with regards to short term stability, but it drifts
> heavily and doesn't seem to find equilibrium. This means the GPSDO needs to
> kick it back on track quite frequently, preventing using of larger time
> constants (now 300s). But still, the overall stability is better than the
> previous revision, which I find quite remarkable. I'm pretty sure that I
> could now beat the noise floor of the measurement setup with a working
> OCXO. I've ordered a brand new OCXO from Digikey, I have another PCB
> revision to launch which will hopefully be the final one and the new OCXO
> will then go on the final PCB. The final PCB revision will feature a
> slightly more elaborate analog output stage which will allow down-scaling
> and biasing the DAC output to further extend the DAC resolution.
> 
> Regards,
> Matthias

-------------- next part --------------
A non-text attachment was scrubbed...
Name: BBB-GPSDO V2 1PPS.png
Type: image/png
Size: 65640 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20200924/5df88d21/attachment.png>


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