Continuing the Crystal substitute using si5351 series…
Above is an example pair of inexpensive modules, less than $10 for the pair (incl shipping). Both boards are powered from 5V, the left hand module is a ATTiny85 dev board, it has a small 3.3V regulator on board. The dev board uses a DIP chip, so it can easily be programmed in a device programmer and then inserted in the socket.
The crystal oscillator on the si5351 board has error of 138ppm, and one might reasonably expect it will also have temperature coefficient quite likely around 3ppm/°.
A lot of 5x 25MHz 3225 0.5ppm TCXO was purchased on Aliexpress for about $15 shipped.
To test the concept, the onboard crystal was removed and the TCXO installed dead bug style near the old crystal pads. The output was connected via 0.1µF to the XA pad, 3.3V from the bypass cap to the left of the si5351 was wired to the TCXO power pad with 0.1µF to ground. The caps standing on end are 1206, 0805 would be better or even 0603 as a solution for evaluation.
Above is the voltage at the XA pin of the si5351.
In this instance, the error was -7.8Hz @ 10MHz, and error of -0.78ppm… outside the 0.5ppm spec.
A new configuration was created to add 0.77ppm.
Above is an Allan Deviation chart for this module referenced to a Trimble Thunderbolt (module freq=10M). Short term stability is good. Allan Deviation is much lower than with the supplied crystal (Crystal substitute using si5351 – part 2).
Above is a pic of GPSDO referenced frequency measurement over almost 12h with temperature ranging from 11-20°, frequency variation is 0.03ppm.
Comparison of supplied crystal with TCXO modified board
Here are the plots comparing XTAL (blue) and TCXO (magenta).
Above, Allan deviation.
Above, frequency difference.
… more to follow.