Differential GPS accuracy?

[Pommie]

I've googled this lots and I'm seeing answers from 5M to 10mm accuracy quoted and thought I'd check if anyone here had any experience in this field.

What I want to know is, if I have 2 GPS units less than 100 meters apart and I compute the difference between their position, what accuracy can I expect. They will obviously both experience the same errors and so whilst their absolute position may have errors of many meters the differential position should be fairly accurate.

So, if anyone has two units they can connect to a PC, place them around half a meter apart and give me the two position strings from them, I would be very grateful. If you can also give me a bearing between the two that would be even better (actually marvelous). If anyone is able to do this then please let them sit there for 10 mins before recording the data.

Thanks,

Mike.

 

[picbits]

If you'd have asked this a couple of weeks ago I could have helped out

I had a couple of GPS units in the workshop for a project I was working on in exactly the configuration you were after

 

[j m]

I have expreance of using differential GPS, and have used system with quoted accuracy of 10mm. to get down to this level involves having L1/L2 frequency GPS units and using RTK to transmit between a base staion and a rover.
The base station is set up on a surveyed point(which normal involves logging the raw GPS for a hour or two and post-processing it to give a actuate position).
The base station then uses the known point to calculate the timing errors, and transmits the measured errors to the base station which uses the information to calculate to position for the rover.

so to get to your question, im not relay sure what sort of accuracy you could expect as it would greatly depend on how you intend to set them up. from reading you post you will be using a NEMA string and not comparing the raw GPS, it will be nowhere near 10mm. If one GPS unit was on a stationary know position and transimted the error to the roving GPS unit, it would be around 1-2m accurate (at a guess). but as you are relying on the calculated positions, it will be hard to know where the errors are happening (at the base or the rover).


I will be able to set up two GPS units and log both the positions, and have one on a surveyed point and give the the bearing between the two. but unluckily for you, but good for me, i wont be in work for two weeks, so will do it when i go back.
jm

 

[Diver300]

I may be able to help. I'll try to get you a log of two GPS receivers.

I think that how uninterrupted the view of the satellites is the most important thing for getting accuracy from standard GPS receivers.

 

[Pommie]

J M,

I was hoping that the errors would cancel out when the units were within a few meters of each other. I know the actual position of the units will be inaccurate but I was hoping their relative position would be fairly constant.

Diver,
That would be useful.

To add to my initial post, I read many years ago about an attempt to use two GPS receivers in the wings of a light plane in order to make an artificial horizon. This was when SA was still turned on and it was reported as being surprisingly accurate. As GPS is notoriously inaccurate on height measurement I thought this method may be accurate when used for horizontal measurement.

Mike.

 

[Diver300]

Selective Availability used to mess up the signals for a whole region of the earth. There is no way that the signal can be affected for one wing tip and not the other. Also, it would then be easy to average over a few metres or a few minutes and compensate that way.

Differential GPS is normally where a fixed GPS receiver transmits the difference from its real to its GPS position and the DGPS receivers use that difference to compensate their GPS positions. There isn't much point now that SA has been turned off.

Height is never as good with GPS as horizontal position. To get a good height reading you need a good view of the low satellites, the ones near the horizon. It probably depends on where you are and where the satellites happen to be at the time, but on an aircraft is a good start. I'm not sure if any test I do on land will be representative, as there are always buildings and trees and stuff so I can't get a perfect view of the sky.

 

[Pommie]

I don't want to measure height, that was just an example. I'm hoping that two receivers that are within 100 meters of each other will give me similar (erroneous) data so that I can calculate the horizontal distance between them to a resolution of a couple of centimeters.

Mike.

 

[blueroomelectronics]

Yes they should be the same error.

 

[Mikebits]

Pommie; I would be interested in your findings on this, as I have struggled over this same quandry for some geology projects I have worked on.

 

[j m]

Yes the error should be close to the same for the two receivers.
“I was hoping that the errors would cancel out when the units were within a few meters of each other. I know the actual position of the units will be inaccurate but I was hoping their relative position would be fairly constant.”
Yes this is how gps gyros work they are some times used on boats to give a heading. With both AE’s in good positions I would expect it to be very close.

As Diver, says
“Height is never as good with GPS as horizontal position. To get a good height reading you need a good view of the low satellites, the ones near the horizon. It probably depends on where you are and where the satellites happen to be at the time, but on an aircraft is a good start. I'm not sure if any test I do on land will be representative, as there are always buildings and trees and stuff so I can't get a perfect view of the sky.“


My figure of 1-2m was a knee jerk quote, which is properly a bit high. The main reason I gave a high figure was I imagined(as this is something I have been thinking about) the rover gps antenna world be mounted on a robot close to the ground, navigating over hills and with some limited view of the sky from building/trees. And because of the low antenna position having limited view, with satellites come in and out of view, Multipath from buildings and terrain, this is where errors would begin to creep in.
jm

 

[j m]

Another source of error, would be unlike commercial differential systems you are not comparing the errors from the signals from the satellites, but the computed positions. Antenna position to get down to a couple of centimetres would be critical. If one of the receivers could see one or two more satellites, this I think would change the computed position, making it harder to compare the two positions against the known point.
jm

 

[Pommie]

Your knowledge here is useful and i would like to ask a couple more questions,

If i get a 50 channel receivers rather than 12 channel will this give me greater accuracy or just quicker acquisition time?

If satellites on the horizon give better vertical resolution, will overhead satellites give better horizontal resolution? (I'm assuming my receivers will see more overhead satellites.)

Is there any way to tell receivers to ignore satellites? My application will be outdoor and so the receivers should be able to register many satellites, if I can tell them to only use the ones they can both see then my accuracy should increase. Shouldn't it?

Thanks,

Mike.

 

[j m]

Im glad if I can help. My knowledge is mostly gained from having an interest in what I do and wanting to find out about it, rather than just pushing buttons.

My understanding is that gps units with lots of channels is just help to get a quicker lock, not the position. I would preferably leave a gps unit on for 30min before relying on the position especially if it has not bean used for some time.

“If satellites on the horizon give better vertical resolution, will overhead satellites give better horizontal resolution? (I'm assuming my receivers will see more overhead satellites.)”
Yes I will give you a better position.

Quote from the gps page on wikipedia,
”Inconsistencies of atmospheric conditions affect the speed of the GPS signals as they pass through the Earth's atmosphere, especially the ionosphere. Correcting these errors is a significant challenge to improving GPS position accuracy. These effects are smallest when the satellite is directly overhead and become greater for satellites nearer the horizon since the path through the atmosphere is longer”

With the kind of units I normally use, you can enter a Elevation mask, around 8-10deg, is from memory what they are normally set to.

“Is there any way to tell receivers to ignore satellites? My application will be outdoor and so the receivers should be able to register many satellites, if I can tell them to only use the ones they can both see then my accuracy should increase. Shouldn't it?”

From memory without any manuals, im not sure if there is a way to ignore satellites on gps units i use. The only time I can think of when I have been able to disable sv, is when the raw gps is outputted from the gps card on a serial port to a computer and the position calculation is done on the pc(this is with Proprietary software, and the one I know of is very expensive)
As for the errors in accuracy from comparing the to gps positions if they see a different number of sv, I don’t know how much this will affect it, but am very interested to find out.

I take it your application needs to be fairly low cost? Because one of the easiest improvements would be to use a L1/L2 frequency gps receiver(and future proofing with L5) but these are lot more than just a standard L1 gps receiver, thousands of USD$ instead of hundreds.

 

[j m]

two GPS NEMA strings logged for 10min. AE spaced 50cm apart. forgot to get the bearing will measure this tomorrow.
jm

 

[Pommie]

Hi j_m,

Thanks for taking the time to do that, unfortunately, the data shows that no satellites had been acquired and so the data is not valid. When I requested that someone do this I had forgotten that most GPS receivers won't work indoors. Again, thanks for trying.

Mike.

 

[j m]

hi mike,
i didn't have much time yesterday, i had noticed that one of the files has no SV number. This seems to be a bug on some of our new units, it will display the SV number but only when it has a better quality solution eg. differential GPS. This also adds the station 0004 at the end of the string this changes to 0100 when it has a kinematic fix.
the reason this happens is the units i use have a GPS card that is interfaced to a embedded computer, and then the embedded computer outputs the serial data i logged. this is done because its the embedded computer also is interfaced to ether a combined GPS/demodulator or a external demodulator and computes the differential solution from the combined data.
both units had the same satellites when ever i look at them when they where logging.

i also spent some time to day playing about with some software, i had it displaying both GPS positions and the range and bearing calculated in real time.
the range and bearing between the two antennas was not very constant at all.
the antennas where in the same place as before 50cm apart but as you can see from the attached screen grab its displaying the range at 3.19m and from just looking at this over the day it went from 20cm to 4meters.
so i don't think this would provide anything near the accuracy you where looking for.

i enjoyed playing about with this anyway, but hope you find a solution to your problem.
james

 

[arrie]

I've been in the vehicle tracking field for quite a few years, but about five years back. On good days, we could get 2m accuracy.
Problem with GPS signal is that so much influence it, the accuracy is also affected by it bouncing of other stuff, such as buildings. Clouds and fog, foliage, buildings, metals, etc, all influence the accuracy.
SA is past, differential can make a huge difference, but proper base stations does not exist freely for a good reference to rule out all error.
Then there is the issue of map accuracy, a complete different topic.

It's quite nifty though, and the units are so small now, it's amazing.

 

[Pommie]

In case this comes up in the future, I finally connected two units and read the position strings when they were next to each other and should produce the same position. Whilst they sometimes agreed with each other, the error was up to 12 least significant digits. I calculate this as approximately 2.2M (each LS digit is 184mm at the equator). So, the differential accuracy and repeatability is not as good as I'd hoped.

Mike.