[time-nuts] Low cost GPS module for < 100ns timestamping error

Hal Murray hmurray at megapathdsl.net
Sat May 3 01:47:04 EDT 2014


tnuts at toneh.demon.co.uk said:
>> In general, low cost oscillators make pretty good thermometers.
> True, but it's short term stability that matters here - over 10s of  seconds
> the temperature shouldn't change much - especially if a bit of  insulation
> is used around the oscillator. 

Ballpark is 1 PPM per degree C.  That's 1 microsecond/second.  You are 
interested in low ns, so you have to at least think about temperature 
changes.  Will your box be outdoors in the sun?  What happens when a big 
black cloud comes over?

> I'm not sure I understand this - I need to record the time that an event
> occurs so I need an accurate time reference. What can I store to
> post-process? The time reference drift and jitter relative to the local
> oscillator before and after the event? 

I was assuming that you would capture the "time" of your events and also the 
"time" of several/many PPS events surrounding your events.  By "time", I mean 
the counter value from the counter/timer module.

I'm not familiar with the chip you are interested in, but the ARM SOC chips 
I've worked with have a switching matrix between the input pins and their 
collection of IO modules.  I would try to set things up so the PPS signal can 
be used by both counters.  (maybe using 2 pins)  The idea is to switch the 
event counter to the PPS signal to find the offset between the counters.

You also have to monitor the serial port so you can label the time of a PPS 
pulse and/or do sawtooth correction.  You also need status info.  How good is 
the fix?  ...

What are you going to do if/when the GPS fades out?  (or gets confused by 
multipath)

One of the reasons to collect some data is so you know what happens just 
before and/or just after your GPS unit decides it can't get a good fix.



-- 
These are my opinions.  I hate spam.






More information about the time-nuts mailing list