Synching 2 PICs?

[arrow]

Hi

I have one PIC that has a Xstal=6MHz, and is outputing a pulse of 25ms every 2seconds (or any other duration for that matter).

I would like to build a 2nd PIC (with any Xstal) that will lock on to this pulse and output its own 25ms pulse every 2seconds.
I can connect the two PICS for say 10seconds to achieve this synching, before I disconnect them.

The aim is to have these two PICs produce the same pulse in time.

Can someone please point me in the right direction of how to achieve this synching?

Thank you in advance.
a.

[Ken Johnson]

If I understand your question correctly, I don't think this is possible, in the real world. Once the 2 pics are disconnected, the pulses will slowly drift. No matter how close the tolerances of the parts, they are different. Sorry.

Ken

[SherpaDoug]

There are three ways to go, none of which is perfect.

1) Buy the most stable crystal oscillator modules you can afford, and you will probably still want to tweak the timing with either a software count or hardware.

2) Use some external timing source such as GPS, WWV, or maybe the sync signal from a local TV broadcast station. It is possible to use the carrier of an AM radio station as well.

3) Have one PIC send some kind of signal to the other such as RF, IR, or a pulse over a wire or fiber.
_________________
The search for better is endless. Instead simply find very good and get the job done.

[arrow]

Hi

Thank you for your replies.

I understand that thigns will drift.

However, what I would like to know is the typical drift? My specs require that after 4 hours I have the two PICS different by no more than 0.5ms.

Is this possible? If I use a 40MHz clock on PIC #2 will that help?

What is a realistic drift that I should expect? (both PICS will be subjected to very similar temperatures).

Regards
a.

[rnielsen]

There are several variables that will affect the drift; temperature, accuracy of the crystals, internal circuitry of the parts, to mention a few. The parts may be sycronized, initially, but will drift apart after being separated and this drift will continue to get larger until the signals start to approach sychronization again.

It's kind of like a clock that has stopped. It shows the correct time, twice a day. The pulses, of the parts, will slowly drift further apart until they start to come together again, like the clock. Once they pass each other, the time between the pulses, will get larger again.

Clear as mud?

Ronald

[arrow]

Hi Rnielsen

Thank you for your explanation- yes I think I understand what you write.

However, do you think that the drift can be as large as 0.1ms over 4hours if my 2nd PIC has a 40MHz clock? i.e. the time is 10ns for each instruction.

Perhaps I should look at *decreasing* the number of clock cycles in those 4 hours?

Regards
a.

[mskala]

+/-50 parts/million is an error spec for decent oscillators you can buy. The error is 10000 times worse than what you want (worst case, one pic +50 other -50). I think you need to stay synced with some external method in order to get to 0.1ms over 4 hours.

[arrow]

Hi

Surely I can start off by matching the oscilations with an Oscope of the two Xstals to be used with the 2 PICs?

Regards
a.

[SherpaDoug]

For $15 from Digi-Key you can get an oscillator (10MHz) good to 2.5 ppm. In 4 hours that is 36mS each. Worst case would be 72mS but probably they would drift together so the difference would be much smaller. Still 0.1mS is an awful lot to ask for.

From Fox Oscillators (http://foxonline.com/pdfs/fts301ah.pdf) for a lot more money, 1 watt of power and a 2 minute warm-up you can get an ovenized oscillator that can be as good as +/-20ppb which is still +/-.288mS. But if these drift together they may work. The devil will be in the details. Maybe you should talk to an application engineer at an oscillator company.
_________________
The search for better is endless. Instead simply find very good and get the job done.

[ckielstra]

A GPS receiver is an extremely accurate time reference with no drift at all, you would never have to sync both units. In low quantities these modules can be bought for less than US$35.
Only problem is that you can't use them indoor (or add a long antenna cable to the outside world).

Before the GPS systems became popular here in Europe we used a radio transmitted timing signal, a German long wave radio station connected to an atomic clock. Cheap radio receivers are able to pick up this signal even inside most buildings. Search the internet for DCF77.

[rnielsen]

Are these two PIC's physically separated on different PCB's? If they are on the same board it might be easier to have an oscillator drive both PIC's. This way you would be sure they are running at the same frequency. If this is not possible, then you will definately need some way to get the same reference pulse to keep them in sync. Like I said, the different variables will affect each component differently.

Ronald

[andrewg]

The mains power could be a potential external signal to sync to. It's probably not too accurate in an absolute sense, but should be spot-on in relative terms.

I remember going on a power station tour many years ago, and being told that they go to some effort to get as close as possible to the correct number of cycles/day (i.e. 50/60 * 60 * 60 * 24) since there are many clocks that operate off the mains frequency for time keeping.
_________________
Andrew

[arrow]

Hi All

Thank you for your replies. Unfortunately I do not have the board space to use any external signal.

But let me understand this:
if the 10MHz Xstal reports a 50ppm accuracy, does that mean:
(a) it oscillates at 10e6+/-500 times per second which is fixed from second to second, or
(b) each second it oscillates a variable number of cycles per second

Thank you
a.

[RLScott]

[SherpaDoug]

There is Accuracy and Stability. You really need to look at the data sheet of the crystal. The crystal will have some nominal frequency which is close to 10MHz but slightly different. That is the accuracy spec. The frequency will drift with temperature, power level, aging, and countless other minor factors. That is stability. Those two are usually combined in the simple 50ppm spec.

There is also jitter, which is the cycle by cycle difference in frequency. It does not sound like reasonable amounts of jitter will be a problem for you.

Get a data sheet that separates accuracy from stability, or talk to an application engineer at the crystal company.
_________________
The search for better is endless. Instead simply find very good and get the job done.