[time-nuts] Strange GPS behaviour

bg at lysator.liu.se bg at lysator.liu.se
Sat Dec 29 21:59:20 UTC 2012


Hi,

> On 12/28/12 9:14 PM, Michael Perrett wrote:
>> Bob,
>> That is simply not accurate - if the solution rate is 1/second, then all
>> parameters are solved in that time frame. There are 4 indpendent
>> variables
>> and minimal processing power is required to solve all four equations.
>> Although I am not very familiar with commercial receivers, that is what
>> happens in the Rockwell, Trimble and IEC military units. If the output
>> is
>> more than once per second it is *usually* an output of the Kalman
>> Filter,
>> not a "true" measurement.
>>
>
> how is the output of a Kalman filter not a "true measurement".. it's
> essentially the composite of many measurements, weighted by the
> uncertainty of those measurements, and, if properly setup, the
> uncertainty of the filter output is less than any of the component
> measurements.

Often these receivers can output either a filtered solution or an
"unfiltered" solution.

The kalman filter will do measurement updates once a second with
pseudorange and delta range measurements. In between - to get 10Hz or
whatever - they will use the kalman filter to predict the
position/velocity.

> Or by "true measurement" do you mean a "point solution" at an instant in
> time, which I don't think you would get from any receiver that uses any
> form of tracking loop, because what is the output of that tracking loop
> but some estimate of the observable, filtered through the loop filter.
> After all, you need to measure/estimate both code phase and rate.

Sure, the pr/dr measurements are filtered both by the tracking loop and by
the integration time. But that filtering is on another scale than using a
filter to predict outputs at a higher rate than what the point solution is
available at.

--

      Björn




More information about the time-nuts mailing list