[time-nuts] Sawtooth Effect in GPS timing

Tom Van Baak tvb at leapsecond.com
Thu Aug 24 03:50:45 UTC 2006


Faisal,

The M12 manual, or at least a command reference,
should be online at: www.synergy-gps.com. You're
looking for one byte in the periodic @@Hn message
(these are binary, not ascii like NMEA). The code I
use looks like:

struct _Hn {
    byte header[4];
    byte pulse_on;
    byte pulse_gps;
    byte traim_solution_status;
    byte traim_status;
    byte svid_mask[4];
    byte sigma[2];
    byte sawtooth;
    struct {
        byte svid;
        byte frac[4];
    } local[12];
    byte checksum;
    byte cr;
    byte lf;
};

You're going to add or subtract this from your TIC
reading, depending on which start/stop channel you
used. You'll figure it out.

Do you plan to make the correction in software after
the fact, or make the correction in real-time using
hardware (e.g., programmable delay generator)?

May I ask what you using this for? When you're playing
at the nanosecond level there are a lot of details and
sawtooth correction is just one small, optional, part.

/tvb

----- Original Message -----
From: "Phaysal Khan" <faisalkhan1981 at hotmail.com>
To: <tvb at leapsecond.com>
Sent: Wednesday, August 23, 2006 19:49
Subject: Re: Sawtooth Effect in GPS timing

> Hi Tom,
> As it is know that when the Motorola receiver's navigation software makes
> its computation of the "best" integer to load into the counter to produce
> the next pulse, it also computes the residual error which it sends out
> (truncated to 1 nsec) as an 8-bit number embedded in one of the  Motorola
> binary messages.
> Could you tell me how could I access those msgs and make corrections to
> sawtooths. I have been trying to look for some info over the net, but was
> just unlucky to find anything :(
>
> Regards
> Faisal
>
> >From: "Tom Van Baak" <tvb at leapsecond.com>
> >Reply-To: "Tom Van Baak" <tvb at leapsecond.com>
> >To: "Phaysal Khan" <faisalkhan1981 at hotmail.com>
> >Subject: Re: Sawtooth Effect in GPS timing
> >Date: Sat, 19 Aug 2006 22:16:21 -0700
> >
> >Faisal,
> >
> >The "sawtooth effect" is expected and normal for
> >most GPS OEM timing receivers. It exists because
> >the hardware generated 1 PPS pulse is derived from
> >the on-board crystal oscillator.
> >
> >For example, a 20 MHz oscillator can generate
> >20 million pulses per second; each is 50 ns apart.
> >The receiver has a whole second to decide which
> >one of those 20 million clock pulses will become
> >the next 1PPS pulse.
> >
> >One second is plenty enough time for the CPU to
> >make all sorts of calculations based on past and
> >current GPS signal data and oscillator drift. But in
> >the end the firmware must choose which of those
> >clock pulses becomes the 1 PPS pulse. Clearly,
> >the chosen pulse may be up to 25 ns early or 25 ns
> >late of the "ideal" pulse.
> >
> >As the oscillator drifts and as the pseudo-range
> >solution drifts, so will the choice of pulse. The net
> >effect is that you get closer, or farther from the ideal
> >pulse, but never more than half of 50 ns away. Make
> >sense?
> >
> >The sawtooth isn't too much of a problem for timing
> >applications. Often a few minutes of averaging drops
> >the jitter down by a factor of ten. In addition, some
> >OEM boards give you the sawtooth error for each
> >pulse so one can accommodate this in software.
> >
> >/tvb
> >
> >----- Original Message -----
> >From: "Phaysal Khan" <faisalkhan1981 at hotmail.com>
> >To: <tvb at LeapSecond.com>
> >Sent: Saturday, August 19, 2006 17:25
> >Subject: Sawtooth Effect in GPS timing
> >
> > > Hi,
> > > I am currently working on problems with GPS timings. I visited your
page
> > > http://www.leapsecond.com/pages/m12/sawtooth.htm
> > > and found useful information there. I have been observing the M12,
> >NovAtel
> > > OEM4 rcvrs' timing pulses (1PPS) and getting a similar sawtooth
effect.
> >but
> > > I am unable to findout what is causing such thing. It would be quite
> >helpful
> > > for me if u guide me in this regard
> > >
> > > Regards
> > > Faisal
> > >
> >
>







More information about the Time-nuts_lists.febo.com mailing list