[time-nuts] TBolt: UTC PPS

[bg at lysator.liu.se]

>
>> Referring to the GPS ICD, the size of A0 in subframe 4 is a signed
>> scaled 32
>> bit integer so if a receiver chooses to convert it to floating point it
>> needs a double precision format to contain it without loss of precision.
>> A1,
>> on the other hand, is a signed scaled 24 bit integer and so it just fits
>> in
>> a single precision floating point number. So Trimble did the right thing
>> here.
>
> Then why do they need 32 bits in the GPS packet format?
>
> A single precision floating point would work fine until the 24th bit gets
> turned on.  That's 16,000,000 ns or 16 ms.  So back in the days when they
> were designing the packet formats, somebody must have thought the
> satellite
> clocks might drift a long way from UTC.
>
> Maybe their control loop wouldn't be good enough to keep the clocks locked
> but the clocks would be stable enough for navigation.


But A0, A1 is not related to SV clocks. They are for translating t_GPS to
t_UTC. A_f0, A_f1 and friends are used to get t_GPS from t_SV.

It seems they wasted a few bits, since the USNO goal is having t_GPS
within 1us. Actual performance is as told in the some ns range.

--

   Björn