[ntp:questions] Re: WLAN synchronization @ 10-100us

Hal Murray hmurray at suespammers.org
Wed Dec 10 08:41:24 UTC 2003


>All the nodes must agree to the same time, not a global time.

I think that will make the problem much easier.


>Actually we're designing a new product so we have control over the
>crystals and all nodes will have same same crystal.

If you order, say, a 10.000 MHz oscillator package, you will
find that it has some tolerance in the specs.  That's typically
20, 50, or 100 ppm, depending upon how much you pay.  That covers
temperature and power supply voltage (see the specs), initial
manufacturing accuracy, and probably aging over some time period.

Using the same part number will not get you parts that run at the
exact same frequency.  If you buy them all in one batch, I'd guess
they would form a tighter cluster.  Or there might be several clusters.

The vendor might be able to give you some data.  Unless you are buying
a lot of them, they might tell you to go read the data sheet.


>We will periodically sync the nodes. Maybe once every second or so.

I think you can do much better than "sync" if you use NTP like
algorithms.  The first thing to do is correct for the crystal
inaccuracy, usually called "drift" in NTP discussions.  The next
step is to filter the measurements rather than simply smashing
the clock to the answer from the latest sample.


-- 
The suespammers.org mail server is located in California.  So are all my
other mailboxes.  Please do not send unsolicited bulk e-mail or unsolicited
commercial e-mail to my suespammers.org address or any of my other addresses.
These are my opinions, not necessarily my employer's.  I hate spam.




More information about the questions mailing list