GPS multipath
Clarifying my previous question..
There's no doubt that multipath exists, and how to test is fairly
straightforward, whether with multiple antennas, cables, or waving
cookie sheets around..
What I was really asking is if anyone had observed this in the output of
their GPS receiver.
That is, say you were watching the 1pps output and comparing its to your
ensemble of active hydrogen masers.. As you place and remove the cookie
sheet, do you see any (fraction of a nanosecond) change in 1pps?
(unlikely, since I assume the 1pps has a fairly long time constant).
Or more interesting, if you happened to have a GPS receiver that puts
out raw observables of carrier or code phase, would you see a bump? Or
if you were experimenting with your KF implementation, where you were
comparing filter output (i.e. estimate of where it "should be") and
tracking loop output (i.e. "where it is") would you see any discontinuity.
Ultimately, the way to find out is just to get a GPS sampler, record
some raw bits, and then run the correlator and look for the second peak
from the reflection.
There's been a lot of discussion over the years about good and bad
locations for the antenna, and how multipath is a big issue with getting
very good timing performance. I was wondering if someone had a
practical anecdote of better or worse performance that could be
attributed to something on the order of a square meter. (position
inaccuracies in urban canyons are a good example of multipath from
hundreds of square meters)
Don't forget that the polarity of the reflection will reverse. RCP > LCP.
and a >20 dB loss will occur.
Tom
----- Original Message -----
From: "Jim Lux" jimlux@earthlink.net
To: "Discussion of precise time and frequency measurement"
time-nuts@febo.com
Sent: Saturday, August 09, 2014 11:31 AM
Subject: [time-nuts] GPS multipath
Clarifying my previous question..
There's no doubt that multipath exists, and how to test is fairly
straightforward, whether with multiple antennas, cables, or waving cookie
sheets around..
What I was really asking is if anyone had observed this in the output of
their GPS receiver.
That is, say you were watching the 1pps output and comparing its to your
ensemble of active hydrogen masers.. As you place and remove the cookie
sheet, do you see any (fraction of a nanosecond) change in 1pps?
(unlikely, since I assume the 1pps has a fairly long time constant).
Or more interesting, if you happened to have a GPS receiver that puts out
raw observables of carrier or code phase, would you see a bump? Or if you
were experimenting with your KF implementation, where you were comparing
filter output (i.e. estimate of where it "should be") and tracking loop
output (i.e. "where it is") would you see any discontinuity.
Ultimately, the way to find out is just to get a GPS sampler, record some
raw bits, and then run the correlator and look for the second peak from
the reflection.
There's been a lot of discussion over the years about good and bad
locations for the antenna, and how multipath is a big issue with getting
very good timing performance. I was wondering if someone had a practical
anecdote of better or worse performance that could be attributed to
something on the order of a square meter. (position inaccuracies in urban
canyons are a good example of multipath from hundreds of square meters)
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.
Jim,
On 08/09/2014 05:31 PM, Jim Lux wrote:
Clarifying my previous question..
There's no doubt that multipath exists, and how to test is fairly
straightforward, whether with multiple antennas, cables, or waving
cookie sheets around..
What I was really asking is if anyone had observed this in the output of
their GPS receiver.
Yes. Just look at the deviations when making position averaging.
After position lock, monitor the residue errors of the T-RAIM.
That is, say you were watching the 1pps output and comparing its to your
ensemble of active hydrogen masers.. As you place and remove the cookie
sheet, do you see any (fraction of a nanosecond) change in 1pps?
(unlikely, since I assume the 1pps has a fairly long time constant).
Or more interesting, if you happened to have a GPS receiver that puts
out raw observables of carrier or code phase, would you see a bump? Or
if you were experimenting with your KF implementation, where you were
comparing filter output (i.e. estimate of where it "should be") and
tracking loop output (i.e. "where it is") would you see any discontinuity.
Watch the T-RAIM output if you can't get the carrier or code phase. The
carrier phase will me much less sensitive in absolute time to multipath.
I've looked at the residue errors to some degree. The worst out-lines
will be dropped out of the average building, but yes, it will creep
through if you look at it. Also the residues from a T-bolt will
illustrate this.
Ultimately, the way to find out is just to get a GPS sampler, record
some raw bits, and then run the correlator and look for the second peak
from the reflection.
Quite possible, but... outside of a chip period, the Gold-code will
relatively effectively de-correlate multi-path, so it is only within the
chip period that multi-path will affect things.
There's been a lot of discussion over the years about good and bad
locations for the antenna, and how multipath is a big issue with getting
very good timing performance. I was wondering if someone had a
practical anecdote of better or worse performance that could be
attributed to something on the order of a square meter. (position
inaccuracies in urban canyons are a good example of multipath from
hundreds of square meters)
Well, in urban canyons the problem isn't as much multi-path, but lack of
visibility of birds that gives good geometry. You will typically see
that the position moves widely along some corridors, which is indication
of geometry problems, but multi-path can give similar issues, but it is
about moving the apparent phase of the chip.
So, real life positioning or timing errors is really a mixture of the
two effects. With good logging on good receivers, you can separate the
effects.
Cheers,
Magnus
Cheers,
Magnus
On 8/9/14, 9:33 AM, Magnus Danielson wrote:
Jim,
On 08/09/2014 05:31 PM, Jim Lux wrote:
Clarifying my previous question..
There's no doubt that multipath exists, and how to test is fairly
straightforward, whether with multiple antennas, cables, or waving
cookie sheets around..
Ultimately, the way to find out is just to get a GPS sampler, record
some raw bits, and then run the correlator and look for the second peak
from the reflection.
Quite possible, but... outside of a chip period, the Gold-code will
relatively effectively de-correlate multi-path, so it is only within the
chip period that multi-path will affect things.
If you have a correlator process that gives you all the cross
correlation bins, it's actually easier to see multipath that is greater
than one chip away (e.g. 1 microsecond .. about 300 m, for C/A code)
Jim,
On 08/09/2014 07:02 PM, Jim Lux wrote:
On 8/9/14, 9:33 AM, Magnus Danielson wrote:
Jim,
On 08/09/2014 05:31 PM, Jim Lux wrote:
Clarifying my previous question..
There's no doubt that multipath exists, and how to test is fairly
straightforward, whether with multiple antennas, cables, or waving
cookie sheets around..
Ultimately, the way to find out is just to get a GPS sampler, record
some raw bits, and then run the correlator and look for the second peak
from the reflection.
Quite possible, but... outside of a chip period, the Gold-code will
relatively effectively de-correlate multi-path, so it is only within the
chip period that multi-path will affect things.
If you have a correlator process that gives you all the cross
correlation bins, it's actually easier to see multipath that is greater
than one chip away (e.g. 1 microsecond .. about 300 m, for C/A code)
Indeed. An efficient way to do that is to use the FFT correlation trick.
If you only want narrow range, then you can have an additional
correlator bin.
Another approach is to use the full sample-rate sample-stream of I&Q
values after carrier frequency removal, and then to autocorrelation on
that, and then subtract the expected autocorrelation, possibly with some
shift. That would give you a fairly high resolution picture.
Feel tempted to try something like that.
Cheers,
Magnus