I have several loggers (CR3000) and an experimental problem with the dual requirements of time synchronization and also time interval consistency at high frequency (20Hz). So, rather than using a solution that could intermittently skip a time interval (such as a GPS master/slave setup) I’m considering allowing the logger clocks to drift and measuring the drift relative to an external signal. The external signal would be an intermittant and ~ instantaneous measurement that is logged by all of the loggers simultaneously. The loggers are separated by no more than 100 m so direct wiring to each logger is a possibility. The experiment would only last a few days so I don't anticipate the loggers clocks will drift much over that time frame. I’m interested to hear if anyone has tried this idea or has insights into how best to implement it or suggest a better solution. Thanks.
Regarding the GPS synchronization;
If using the PPS signal from the GPS, and the GPS does not lose signal lock, the CR3000 has internal time "nudging" which is intended to eliminate the intermittent skipped intervals you refer to. The internal time is adjusted on a small scale (~20uS) resolution to align the top of the datalogger second to the GPS PPS signal. After initial startup and "lock", if all CR3000s are seeing the same GPS serial stream and PPS signal, they should not have intermittent skips.
If you have seen other behavior than descirbed, please contact us and give us the details.
I would recommend the master GPS driving the times of the CR3000s as the preferred solution here. This method should synchronize all CR3000s to within ~10uS. Using the proposed method of passing a common "sync" measurement to all loggers and aligning the measurements in post processing would give ~50mS (the scan interval) resolution and requires post processing to shift time series.