[time-nuts] OT - GPS and North

[Magnus Danielson]

bg at lysator.liu.se wrote:
>> Mark Spencer wrote:
>>> Would Time Difference of Arrival techniques combined with an array of
>>> four closely spaced antennas work with Gps signals as a means of
>>> determing the orentation of the antenna array vs the gps satellites ?
>>> (I'm thinking traditional TDOA techniques may not work with gps
>>> signals.)
>>>
>> There are commercial products available:
>> http://www.hemispheregps.com/Default.aspx?tabid=379
>> http://www.hemispheregps.com/Default.aspx?tabid=412
>>
>>
>> These devices claim less than 1 degree RMS heading accuracy.
>>
>> Here the antennas are integrated and a fixed 0.5 meter apart.  They do
>> (or did) make a board that could be used with separate antennas, but I
>> can not find it right now.
>>
>> I have no idea of how the math works for computing heading.
>>
>> Gary
>>
> 
> Any (almost) pair of GPS-receivers with phase measurement outputs can be
> used to make an attitude GPS receiver - sometimes called a GPS compass.
> You need decent antennas for good results.
> 
> This is a special case of a phase ambiguity problem with moving base
> receiver and extremely short baseline. If the antennas are mounted at a
> fixed relative position, knowing the baseline gives a simpler problem.

By having the additional antennas hooked to the same GPS core with 
multiple frontends, the central antenna can act like a traditional 
receiver. The carrier and code tracking-loops for additional antennas 
can then be aided by the central antenna, which will help to reduce 
ambiguities. After inital ambiguities have been resolved, maintaining 
tracking should not be too hard.

Using multiple receivers does not give the same effect as tight integration.

> I have read papers about using only one receiver and one antenna. The
> trick is then to use the antenna diagram and SNR from the currently
> tracked satellites to estimate an orientation. I have seen no commercial
> product trying this. Accuracy was not spectacular - a few degrees - if I
> remember correctly. Would need a very stable environment to work. A
> groundapplication (say car moving in urban environment) will influence the
> received SNRs to randomly to make a one antenna approach possible.

Such an approach is indeed possible but fragile. You could also include 
shift in phase center, in which case a "bad" antenna would be good. 
Knowing the orientation of the antenna could allow for post-processing 
to compensate for phase-shift. You would loose precission from position 
tracking but gain an estimate in heading.

Cheers,
Magnus