[LinuxPPS] [PATCH]

[Bernhard Schiffner]

Am Freitag, 1. Februar 2008 14:30 schrieb Rodolfo Giometti:
...
> >  The other driver I have needs some cleaning up and rethinking some
> > things. The problem with the GPS I have (Trimble LassenSQ) is that the
> > device generates an undisciplined PPS signal when there is no
> > satellite lock and the manufacturer does not specify a tolerance.
> > Obviously using such a signal is not good for an NTP reference clock
> > and in my situation NTP will have no basis for deciding which clock is
> > defective (system or GPS).  So my other driver is actually a fairly
> > specialized GPS driver but it also interfaces to LinuxPPS; the PPS is
> > only delivered to pps_core if there is a satellite lock and this way
> > NTP is immediately aware that the problem is with the GPS.  I guess
> > such a driver is better placed elsewhere since PPS is only a minor
> > function of the driver - any comments on that?
>
> Can you please explain better the problem? How can you decide when the
> PPS is good or not?
>
> I think that the PPS signal should be reported anyway to the system
> but NTPD should decide (according to the GPS data) when it should use
> it or not.
>
> Ciao,
>
> Rodolfo

Yes.

1.) Commonly used GPS-Receivers derivate the outgoing PPS-Signal from their 
local XCO.
2.) In early stages of receiver-operation this signal is already provided 
but not at all synced to the GPS (TAI) clock and time.
3.) There is no way to  confirm TAI-locking from observing PPS-signal only. 
Phase and frequency error range from zero (!) to big values.
4.) Comparing it with other sources of NTP-Information doesn't real help in 
this situation too. 

5.) The only way out of this desaster is to check, if the receiver reports 
to be "locked". (NMEA message, SV used >= 4 is good too.)

6.) Some receiver may be set into "clock-mode". This means geographic 
location is known and regarded fixed. This knowledge provides help in 
calculating GPS-internals more reliable.
Side effect: Only 1 SV in sight is enough generate GPS-TAI synced 
PPS-signal...

So having a PPS-line _and_ the NMEA-information is crucial.

Bernhard