[ntp:questions] Garmin GPS18x LVC problem since firmware 3.50

Rob nomail at example.com
Wed Dec 29 09:12:23 UTC 2010


David J Taylor <david-taylor at blueyonder.co.uk.invalid> wrote:
> Following my previous message, I can confirm that the NMEA output from the 
> GPS18x LVC with the firmware 3.50 upgrade is coming out something like a 
> second after the PPS signal, and worse, it's very variable so sometimes 
> just before the PPS leading edge, and sometimes after.  No wonder NTP is 
> confused!
>
> I have contacted Garmin UK via the Web/e-mail form, but as it's almost a 
> two-week holiday here, if someone has a better contact with Garmin US it 
> might be worth alerting them as well.  I have requested that the firmware 
> be fixed so that the NMEA output be well under one second delayed, and 
> that the delay be more reproducible (although I haven't made that point 
> strongly enough, so it should be emphasised).

Technically, the timestamps in the NMEA messages are the "time of fix",
not the "current time".  So a receiver can rightfully keep a calculated
fix in memory and send it at some later time, with the timestamp of the
original fix included in the message.
This makes NMEA unusable for time synchronization.

I have seen the behaviour you describe from other receivers, from other
manufacturers, as well.  For example, I have seen it happen that the
offset between PPS and NMEA output varies every time the receiver is
restarted, and slowly wanders while the receiver is running.  Apparently
there is a software timer that sends the messages every second or other
interval, but is not locked to the PPS signal.

I think by expecting the NMEA to be locked to PPS, you are in fact
relying on an implementation detail that often is not covered in the
receiver spec, and that can be changed at will (accidentally or for
some reason).

It is better to switch to the binary protocol, which is manufacturer
specific.  There usually are messages in the binary protocol to communicate
the current time, as opposed to the time of fix, and there are specs
that define the relationship between the time and the moment the message
is sent.




More information about the questions mailing list