The 2023 GPSDO Geomagnetic Impact Test John Ackermann N8UR jra@febo.com Revised: 03 December 2023 UPDATE -- I was able to capture data during the CME events of 1-2 December 2023. See the METADATA.txt file for details. It's a good bet that most of the radio experiments planned for the upcoming April 2024 eclipse will use GPS Disciplined Oscillators ("GPSDO") as the reference for measurements of things like Doppler shift as the eclipse changes ionospheric conditions. The thought came to me -- what if the very geomagnetic changes we are monitoring have an impact on the output of those GPS-based frequency references. It's been shown that GPS propagation delays are impacted by changes in Total Electron Count (and probably other effects). It wouldn't be good if the thing we're measuring also causes our reference to change. Now, my gut feeling is that this is not an issue. The purpose of a GPSDO in one sense is to act as a low pass filter (a phase locked loop with a narrow bandwidth/long time constant, perhaps of several hundred seconds). That filtering should smooth out any sudden change in GPS timing. The slow frequency steering applied to the crystal oscillator might completely ignore a sudden blip in the GPS signal. If that theory holds, GPSDOs with long time constants should be quite resistant to geomagnetic effects, but the shorter the time constant, the less likely that is to be the case. In particular, one of the most popular frequency references is the Leo Bodnar Mini GPSDO which performs extremely well, but has a relatively wide loop filter. Might it be susceptible? It seemed worthwhile to try to prove whether this is a valid concern. To do so, I connected three GPSDO of varying design and quality, and one plain GPS receiver, to a common antenna and monitored their output frequency (actually, their phase) relative to a very stable local frequency reference over a period of about 42 days. The 10 MHz output of each device was connected to a TinyPFA phase and frequency analyzer* with a passive hydrogen maser at its frequency reference. Ten phase measurements per second were logged to disk files in two runs with only a few seconds gap between them. During the measurement period there were several geomagnetic disturbances, but unfortunately not a really major one. The four sources measured were: (a) Bodnar Mini GPSDO. This is a relatively low cost and very popular unit. It uses a different architecture than more traditional GPSDOs and has a fairly wide loop bandwidth (about 9 seconds). Unlike the other sources, the Bodnar has a frequency offset of about -1.55e-11 at 10 MHz. I created two TimeLab files for the Bodnar, one with raw phase and the other with frequency offset removed. The original phase files are uncorrected. (b) BG7TBL GPSDO. This is one of the many variants of BG7TBL units available on eBay (this one belongs to Dave McGaw). It has a fairly decent 10 MHz OCXO but the control loop is very crude and the short term performance is not nearly as good as it should be. (c) u-blox NEO-M8T timepulse at 10 MHz. This is the raw TP output from the receiver. There are a couple of large glitches in the data files. I do not know what caused them; as far as I can tell there was no disturbance to the receiver and none of the other units were impacted at the same time. The glitches do not appear to correlate with major solar events, darn it. (d) Trimble Thunderbolt GPSDO. This is the best unit of the lot, with a good OCXO. The control loop was configured with a time constant of 250 seconds. All the receivers were in the same room and fed from the same antenna. Other than the Thunderbolt, all were sitting on the same table within a foot of each other. Along with the frequency data, I had a Mosaic-T dual frequency GNSS receiver that was connected to the same antenna and using the same maser frequency reference logging RINEX data at 30 second intervals throughout the experiment. By post-processing, this data should yield the TEC at the test location, along with the maser offset and drift. The RINEX files are included in the data set. The results of the phase measuremenets are in text files totalling about 900 MB. The file "METADATA.txt" describes the file format and provides some statistics about number of records, etc. I've also imported the data into the TimeLab (https://www.miles.io/timelab/beta.htm) time and frequency analysis program and saved it in TimeLab format. The challenge is to slice and dice this data to see if there are any traces of frequency changes that correlate to solar activity. * I was able to make four measurements at once thanks to the TinyPFA firmware (https://www.tinydevices.org/wiki/pmwiki.php?n=TinyPFA.Homepage) that Erik Kaashoek wrote for the low cost NanoVNA H4 analyzer. This setup can make remarkably high resolution frequency and phase measurements for a device that costs around $100.