GPS - Single difference pseudoranges questions

Author

Single difference pseudoranges questions

VVX

2008-02-24, 7:33 am

What sort of accuracy should I expect using pseudorange single
differences?

I thought it would be around 1-2 meters or so.

But, using my own software (based on well known algorithms of course)
I am getting weird results.
On a baseline of 10 meters the calculated position wanders all over
the place, sometimes spiking up to 50 m. I mean it is not even close.

The setup is simple. The two antennas are hanging out the windows on
the same (east) side of a 5 storey building. The view to the east is
pretty much unobstructed except for some minor mountains approximately
20-30 km away. The west hemisphere (should I say quartersphere? :)) is
obviously unavailable, though the unit somehow manages to track
pseudoranges and even carrier phases of SVs from that side sometimes.
The two units employed are u-blox LEA-4Ts.

What can be the problem here? Multipath maybe?

Thanks,
VVX

Happy Trails

2008-02-24, 10:33 am

On Sun, 24 Feb 2008 04:54:09 -0800 (PST), VVX <vvxvvx@gmail.com>
wrote:

>The setup is simple. The two antennas are hanging out the windows on
>the same (east) side of a 5 storey building.

It really doesn't make any sense whatsoever to be doing what you are
doing without at least the base antenna, and better both, having a
clear, unobstructed 360 degree view of the sky.

You are wasting your time otherwise. How tall is your building, and
do you have access to the roof?

VVX

2008-02-24, 10:33 am

On Feb 24, 8:06 pm, Happy Trails <nom...@myplace.com> wrote:
> On Sun, 24 Feb 2008 04:54:09 -0800 (PST), VVX <vvx...@gmail.com>
> wrote:
>
>
> It really doesn't make any sense whatsoever to be doing what you are
> doing without at least the base antenna, and better both, having a
> clear, unobstructed 360 degree view of the sky.

I consider one of the antennas as the base one. The base antenna
'exact' position is calculated by averaging observations over a 20
hour interval.
I thought the 360 unobstructed view would be critical for the so-
called block-shit differential navigation method (when the base
receiver is sending position corrections, not pseudorange). Presumably
in case of _pseudorange_ single differences the _same_ set of SVs
visible to both receivers is enough. Is that wrong?

> You are wasting your time otherwise. How tall is your building, and
> do you have access to the roof?

The building is about 15-20 m (45-60 ft) high, and yes, I have access
to the roof (and actually was thinking about that already :) )

John

2008-02-25, 7:33 am

With the LEA-4T, you need to adjust your raw measurements to account
for clock error and doppler shift. You'll have to play with the signs,
but it goes something like:
clockdrift = clockerror - previousclockerror;
CorrectedPhase = Phase - clockerror*L1Frequen
cy -
clockdrift*(L1Freque
ncy - doppler);
CorrectedPR = PR - clockerror*C - clockdrift*(L1Freque
ncy -
doppler)*L1Wavelengt
h;

You should start with a "zero baseline" experiment, where you run both
receivers from a single antenna. In the beginning, don't even try to
calculate positions. Just compare the raw measurements. Your
pseudoranges will vary by a meter or so, and your phases should be
within 8mm. More specifically,
(PR[sat] - AveragePR) for receiver 1 should be close to (PR[sat]
- AveragePR) for receiver 2.

You'll need a signal splitter. You can build your own, or you can
purchase one for about $75. I use the wi-sys ws2903.

I agree with the other posters about your antenna location. You need
clear view of the sky. Head to a good location, collect raw data, and
bring it back. You can tackle multipath and cycle slips once your
basic software is working.

VVX

2008-03-30, 12:33 pm

On Feb 25, 4:37 pm, John <goo...@coyotebush.net> wrote:
> With the LEA-4T, you need to adjust your raw measurements to account
> for clock error and doppler shift. You'll have to play with the signs,
> but it goes something like:
> clockdrift = clockerror - previousclockerror;
> CorrectedPhase = Phase - clockerror*L1Frequen
cy -
> clockdrift*(L1Freque
ncy - doppler);
> CorrectedPR = PR - clockerror*C - clockdrift*(L1Freque
ncy -
> doppler)*L1Wavelengt
h;
>
> You should start with a "zero baseline" experiment, where you run both
> receivers from a single antenna. In the beginning, don't even try to
> calculate positions. Just compare the raw measurements. Your
> pseudoranges will vary by a meter or so, and your phases should be
> within 8mm. More specifically,
> (PR[sat] - AveragePR) for receiver 1 should be close to (PR[sat]
> - AveragePR) for receiver 2.
>
> You'll need a signal splitter. You can build your own, or you can
> purchase one for about $75. I use the wi-sys ws2903.
>
> I agree with the other posters about your antenna location. You need
> clear view of the sky. Head to a good location, collect raw data, and
> bring it back. You can tackle multipath and cycle slips once your
> basic software is working.

Thanks for the points everyone.

Having done the experiment in an open sky environment I got proper
results.
Pseudorange single differences give the baseline accuracy around 1.5
meters RMS.
The zero baseline experiment shows a few mm level noise on carrier
phase double differences.
A two-epoch carrier phase + pseudorange float solution is at 20-30 cm
RMS level using timespans of 60 seconds (between those two epochs)

Do I have to use longer timespans to get the _fixed_ integer
solutions?
If I understand the theory correctly the floats have to be quite close
to integers, but they are not in my case - their fractional parts seem
to be evenly scattered over (0,1) interval.

SamSvL

2008-04-02, 4:33 am

VVX <vvxvvx@gmail.com> wrote in news:9b1151eb-276e-4338-94d8-
c47f52e0348f@s8g2000
prg.googlegroups.com:

> On Feb 25, 4:37 pm, John <goo...@coyotebush.net> wrote:
>
> Thanks for the points everyone.
>
> Having done the experiment in an open sky environment I got proper
> results.
> Pseudorange single differences give the baseline accuracy around 1.5
> meters RMS.
> The zero baseline experiment shows a few mm level noise on carrier
> phase double differences.
> A two-epoch carrier phase + pseudorange float solution is at 20-30 cm
> RMS level using timespans of 60 seconds (between those two epochs)
>
> Do I have to use longer timespans to get the _fixed_ integer
> solutions?
> If I understand the theory correctly the floats have to be quite close
> to integers, but they are not in my case - their fractional parts seem
> to be evenly scattered over (0,1) interval.
>
Depending on the distance between reference and rover (and depending on
the efficiency of your ambiguity resolving lagorithm) you may require
many tens of minutes before the ambiguities can be fixed.

See e.g. experiment 2.1 on http://home-2.worldonline.nl/~samsvl/TIM-
LP.htm.
The LEA4-T will not behave very dirfferent from the TIM-LP.

Sam

LinkBot