[GPS_Standard] VE2ZAZ

[Dave Platt]

Bob Bownes wrote:
> So, 24 hours later, I still can't keep this thing locked if I crank
> the value of S over 16. Right now, it is set as follows:
> 
> S: 8
> F: 8

You might able to reduce the tendency to unlock, at least, by
reducing F to a smaller value.  This would allow the controller
to start making "coarse" adjustments more quickly.

As things stand, you've left very little margin between the F
value (the point at which the controller can start using coarse
adjustments) and the holdover threshold.  If the drift exceeds
one fine-adjustment per S period, it's likely that the system
will go into "holdover" a lot, and may unlock itself.

> L : 4
> H: 10
> W: 20
> N: 3
> 
> The coarse dac value swings from one end of the scale to the other.
> When I happen to catch it in the unlocked state, I may make a coarse
> DAC adjustment of 150-200 to get it lock again.
> 
> The DAC is changing the output of the oven, just not fast enough to catch up.

At this point, based on what you've said above (and also to me in
private email) I can think of several possible problems which might
cause such behavior:

(1) Bad oven - it's not maintaining the crystal at a constant
    temperature.  This might be due to a bad thermistor, bad
    heater, or bad control logic or heater driver.  The oven
    might be failing to track the ambient temperature (e.g.
    heater "stuck" at a constant power level) or the oven temp
    might be jumping around due to transient/intermittent
    problems.

    Diagnostic:  measure and record the oven current, both during
    steady-state operation, and as you place some sort of
    thermal stresses on the oven (e.g. insulate it for a while,
    then take off the insulation and fan it for a while).

(2) Other problem affecting the oscillator.  These sorts of
    oscillators usually have a manual trimcap adjustment for
    coarse adjustment - if that trimcap is drifting or jumping
    around, it would yank the oscillator off frequency, perhaps
    faster than the FLL control voltage could adjust for.  Or,
    possibly, you have a bad crystal in the oscillator which is
    prone to drifting/jumping for non-temperature-related reasons.

    Diagnostic:  try tapping gently on the oscillator a few times
    and see if you can trigger a sudden frequency jump.

(3) Bad PPS signal from the GPS.  Are you actually sure that the
    GPS has a good lock on the satellites?  Quite a few GPS
    engines will produce PPS signals even when they don't have
    satellite lock... but the timing of this "early PPS" can
    be very inconsistent, as it may be derived from the GPS's own
    onboard crystal oscillator rather than from the satellite
    signal.  It's possible that you're trying to "lock" your
    high-quality ovenized oscillator, to a time reference
    derived from a cheap utility-grade non-temperature-stable
    oscillator on the GPS board.

    Diagnostics:  check the NMEA sentences from the GPS to confirm
    that it's actually tracking an adequate number of satellites.
    Apply some thermal variance (e.g. blow cold or warm air)
    to the GPS module, and see if the error counts from the FLL
    controller suddenly start jumping around.  Check the GPS
    manual to see if it talks about the behavior of PPS when
    there's no good satellite signal lock... and, confirm that
    PPS from this type of GPS really is suitable for use for
    timing purposes (some aren't).

(4) Bad oscillator varicap.  If the capacitance variation in the
    oscillators voltage-controlled cap is much less than it
    should be, the FLL's control voltage would have little
    effect, and the controller wouldn't be able to track
    temperature changes rapidly enough.  [This would not,
    I think, account for the occasional sudden jumps you're
    seeing.]

    Diagnostic:  manually tune the FLL to its midrange value
    and use the manual oscillator trim to get as close to zero
    error count as you can.  Then, manually tune the FLL to
    minimum value, wait a minute or so, and see how high the error
    counts are.  Tune the FLL to maximum, and repeat.  See if the
    total tuning range is somewhere on the order of 1 Hz, much more,
    or much less.