radio modules that can send sync pulse

Hi,

Does anyone know of a wireless radio module which is capable of sending sync pulses to a number of microcontroller boards in a big building? We need the boards to all start doing something at the same time, ideally to an accuracy of about 10 microseconds. The pulses will be repeated several times a second, but don't have to be regular to the same accuracy, just to all be received simultaneously. It would also be useful if the boards were capable of two way data comms as well - this will be needed for another part of the system, but could be provided another way - e.g. wifi.

At the moment I've been testing the xbee zigbee modules from digi, but the best timing I can get with those is to about 0.2ms by triggering on the start bit of a data byte, which is not really good enough.

Thanks,

andy baxter

Most GPS modules can do this. The motorola Encore modules generate a

1 nS pulse every second. Of course they're synched to an *external* source, not one under your control, but they're synched.

Just use udp packs on giga-ethernet, to sync, and allow talk-back. Why wireless? You need a radio receiver at each point, and accurate sync will be hard to implement that way, assuming spurious signals mixed in causing interference.

Your joking, right ?

The only thing that needs this type of accuracy are explosives !

And seconds are good enough for that!

Andy,

please explain this problem you are having and why its so important.

thank you

hamilton

We are developing some microcontroller boards, which are capable of sending short messages to each other (not using radio but another medium). It is important for the project that the power consumption of the boards is as low as possible, so we are trying to design a system where the transmitters and receivers both wake up at the same time to send a signal before going back into a power saving mode. I.e. we want a simple, low power, module which is capable of sending a wake-on- interrupt signal to the processor on each of the boards.

andy

Some of the boards at least need to be wireless, as they will be portable handheld devices. The device which sends the sync pulses can be a static device.

See my reply to hamilton for more details about the application.

In principle we could send UDP packets over wifi, but the problem with this in practice is how to deal with the problem we've already had using broadcast packets over zigbee. I.e. the base protocol is capable of what we want, but something to do with the way the digi modules are designed is causing a 0.2 ms jitter in the received sync pulses (between one board and the next). Most likely this is to do with the way the embedded processor on the module handles incoming traffic and sends it out over its serial link to the main board - e.g. there is a polling cycle which can be interrupted at various points, causing the jitter.

P.S. I forgot to say where the 10 us figure comes from - it's the approximate length of a single serial bit at 115200 baud. I.e. we would like the boards to wake up for just long enough to send a few bytes at this rate before going back to sleep again.

We are also working on another application with similar timing requirements, which ideally should be synced to an accuracy of around

25us.

Thanks - that might be worth looking into. :-)

You may also be affected by transmission retries or use of repeater nodes which happens invisibly in Zigbee.

Perhaps consider using narrowband low-power VHF or possibly UHF transmitter and receiver modules which have much longer range than Zigbee and can give you direct control over the modulation and demodulation.

John

CDMA cell phone systems typically shoot for 1us synchronization and the requirement is 10us -- they use GPS receivers to lock the network together.

In terms of making a timing determination irrespective of protocol delays, have you considered an NNTP stack? I'm unsure of the recalibration frequency that your systems would require, but the basic protocols are an available resource for this kind of problem.

If this is about serial receivers, a self-clocking serial protocol would be more suitable than a 'master clock' for determining the start of transmission.=20

What lunatic wrote this requirement? Aren't you even allowing for the time it takes your "remotes" to wake-up and initialize?

What kind of interconnects do you have? You could use a wire to send your "wake-up"; that would simplify things dramatically.

Good Luck! Rich

Is this the same as 433 MHz? One of my colleagues was suggesting a 433 MHz radio module.

If you know of any particular products that are good, low power (for the receiver at least), and fairly cheap, that would be a big help.

I'll talk to my colleagues about this - it was just a quick meeting we had where that was stated as a requirement, so maybe it would help to clarify what's actually necessary.

We are working on another system though which also needs an external sync signal, which should be accurate to about 25 us to get the best out of the rest of the system. The 25us is coming from the basic physics of the problem in this case.

We can't use wires because both systems need to work on handheld devices.

The 10us accuracy is pretty loose requirement. Accomplishing that is not a problem. Just about any voiceband analog in - analog out radios would do; with the custom modem algorithms that have to run on your MCUs. The clocks of all devices should be PLLed to the master clock periodically transmitted by the central station, and then you only have to send the command to start doing something at the right time.

For the timing purposes, avoid ZigBee, WiFi and other "smart" solutions, as it is difficult to get the actual Tx/Rx timing from their physical layer.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

BTW, there is a whole lot of applications like synchronized sensor networks or distributed multi-channel speaker/mike arrangements, where the timing should be accurate to a fraction of microsecond or so. The OP's 10us requirement is easy.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

e

er.

It's been a long time since I worked on a 802.15.4 chip, but somewhere in the phy, there is information on the time of rx packets, in the preamble detection correlator most likely down to some fraction of a 2Mbit chip.

if you can get hold of it is a different matter, though I'd think theres is a way since it is probably used to align timing so the devices can wake up at the right time

I seem to remember that theres was an idea of using the rx timing to calculate distances kinda like gps

-Lasse

Please remember that any PLL based radio receiver (and/or transmitter) will require some time to achieve lock, thus the receiver must be powered up well in advance before the expected message.

A DDS based system will generate the desired frequency faster, but these consume typically a lot of power and you still have to wait for the quite long startup time for any high quality crystal oscillator feeding the DDS.

Going down to sub-microsecond timing accuracy will require at least 1 MHz bandwidth, which will consume a significant amount of power.

Of course you can use much smaller bandwidths and integrate over time and use much less power, but you have to measure the timing over a longer period of time, so the total energy requirement does not drop drastically.