I need to wirelessly (not optically, due to the distances and 3D volumes involved) transmit a "sync" pulse train from some master transmitter (the "time keeper") to be received by basically any number (lets say 50, to put an upper bound on it) of devices within 100m.

The idea is that the devices that receive this all have a common sync they can share for using as a reference time. Measurements will be taken and reported as relative times to their best knowledge of the master clock's most recent rising edge.

The system will be deployed in a factory environment, but there will be Wifi and the occasional large electric motor/servo/magnetic interference in the area.

I have a requirement for approximately 100 nanoseconds of phase noise/jitter from the receiver to the master. Latency isn't an issue as long as it's deterministic and repeatable to a similar 100ns magnitude. What is the lowest latency and highest repeatability/deterministic-timing RF device, technique, or protocol available that will allow a master to transmit a uni-directional "sync pulse" at 60Hz with ~100ns TX->RX jitter?

I have looked at a few options, and done some research on the matter. I've read some questions on here, such as this one Low latency rf link answered by @Andy aka and I had a look at the Radiometrix RF modules, which seem like a simple and possibly effective solution to my problem.

The next thing I looked at was the ANT Protocol, which according to the information on this Digikey article 6 years ago the ANT protocol allows almost zero latency on the TX/RX pipeline, due to the receiver always listening. If the pipeline from master TX to the ANT device is streamlined enough, the latency and repeatability should be in the couple of microseconds region. Does this sound right? Has anyone used ANT protocol for this purpose before? I am considering trying out the ANT capabilities of the NRF52832

I also saw that Texas Instruments has something called the CC1121 - "High Performance Low Power RF Transceiver" which appears to have extremely fast wake-up time, and only needs 4 bits worth of preamble. This means I could effectively transmit 1 byte, with the last few bits just being a 0x01. However, in "4.13 Wake-up and Timing" on the specs sheet, it still has values in the hundreds of microseconds for going from IDLE to TX. I'd have to test for repeatable Idle->TX transition jitter, in basically any of the above devices/protocols.

As a last resort, I could try for a custom made RF "spike" generator, and design my own receivers to be matched/tuned to it. I'm sure this is illegal, and is not really a useful option (or is it? would I have to get it certified? I guess it would immediately fail EMC emissions standards!).

The RF transmitted "clock" could be PLL-locked (in software) to filter out jitter (which is why latency isn't much of an issue now), and I could calibrate out any constant phase delays due to TX->RX delay and speed of light) in order to create an "estimator" of the original master clock at the instant in time that it's sent. Are there any techniques I should know about for dealing with this problem?

  • 2
    \$\begingroup\$ You're missing some key specs, like is there an upper power limit to the receivers, and what accuracy/jitter you are actually looking for. Could you deploy a GPS receiver at each point and extract the PPS output? Could you deploy a mobile comms (2G or 3G) receiver at each point to lock on to the local basestation frame sync? \$\endgroup\$
    – Neil_UK
    Mar 7, 2017 at 6:29
  • \$\begingroup\$ "What they do is not relevant" seriously, how often did you encounter that in engineering. Since your specs are lacking, and you don't tell is the purpose, we can't help you. Voting down until you fix at least one of these. \$\endgroup\$ Mar 7, 2017 at 7:32
  • \$\begingroup\$ Do you actually need low latency, or is it sufficient for the devices to PLL onto the transmitted signal? Do you care about the 500ns of radio delay within your 150m sphere? \$\endgroup\$
    – pjc50
    Mar 7, 2017 at 10:26
  • \$\begingroup\$ @pjc50 Well that's related to my rambling further down, I can phase-lock to the pulse train so that the actual latency isn't relevant, and jitter could be partially rejected as high freq.. the speed of radio in 150m can be compensated/calibrated for later on, once the distances to the receivers has been determined.. \$\endgroup\$
    – KyranF
    Mar 7, 2017 at 16:37
  • \$\begingroup\$ In that case you could probably achieve acceptable results with those cheap 433MHz OOK-modulation modules by transmitting a recognition code followed by a trailing "pip". At either 60Hz or 1Hz. This is assuming you've ruled out GPS 1pps and the MSF or your country's equivalent: npl.co.uk/science-technology/time-frequency/… \$\endgroup\$
    – pjc50
    Mar 7, 2017 at 16:46


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.