Hi,
I wonder if LimeSDR community would be interested in a wireless clock distribution system I have developed.
The prototype I have now demonstrates phase error jitter RMS below 25ns and P99% below 70ns. So far I have only tested over relatively short distances (5m), but it does works without line of sight, with multipath, severe signal attenuation etc. So I have reasons to think it will work outdoors over larger distances (tens of meters at least). There is also a lot of room for improvement in firmware, I think I can at least halve the metrics I mentioned.
It is a work in progress and I have a choice which feature to prioritize - increase working distance, reduce jitter, support specific clock frequencies or interfaces etc. If anybody sees a use case for it, I would greatly appreciate the some feedback 
I’d certainly be interested in hearing more about this!
Right now the test rig consist of a master device and two slaves, all based ESP32 S3, talk WiFi FTM between each other. The master and the slaves output square pulses in sync.
Measuring phase error between the slaves, I see a consistent 25ns jitter RMS. P99 depends on how badly you abuse the devices – changes in antenna orientation cause wireless delay variations, physical shocks induce rapid oscillator frequency changes which are poorly compensated (a defect of the current firmware, can be improved). But as long as you are reasonable, jitters is still below 25ns and P99 stays below 70ns. As I said, there is a lot of room for improvement, my ambition is to push jitter down to 10ns, should be doable (with C6 chips instead of S3 and better math / signal processing).
The current prototype uses whatever cheap XO the manufacturer have stuffed into ESP32 modules, so absolute accuracy and holdover are marginal. The output frequency is 2 kHz, there are options to increase it (but not into MHz range) and add more channels. Implementing 1PPS output is easy.
I am considering my options for next steps. An obvious choice is to mimic GPSDO modules and produce 10MHz and 1PPS outputs. This means adding TCXO / OCXO module and some FPGA. I am ready to go this route, because I have a couple of Ettus N210 SDR (one even fitted with GPSDO) and could use this platform to experiment easily (already have custom FPGA firmware for it which spits a stable 2 kHz itself).
But LimaSDR already has a local oscillator and FPGA and reference clock input, so chances are I can integrate my thing with LimeSDR in more cost effective way without sacrificing performance.
Any chance you share your use case and potential requirements? Why do you need to sync several LimeSDR? Are you already using GPSDO with 10MHz and 1PPS ref or something? How far these devices have to be separated physically? What is an acceptable jitter for you?
