Is GPS more acurate that the Talivan?
Discussion
I got an NIP for 42 in 30 last week, at the weekend I downloaded my tracks from my GPS receiver to my PC, on the stretch of road they claim I was doing 42 my average speed was 32mph and my max speed was less than 42mph. Is this log admissable? Is GPS more acurate than what they use? Does it mean their laser needs calibrating?
I'm no expert (which is a good way of saying I'm guessing based on my limited knowledge of GPS), but I'd be amazed if any in car GPS speed readout was more accurate than an laser/radar gun.
Basically, GPS can be very accurate indeed (position within mm in fact), but only if the following things are true:
1) Lots of satellites are visible (the more the better).
2) The receiver is left in one place for several hours.
Without these two conditions being satisfied, GPS gets less and less accurate. As you're driving along, your position is only going to be known roughly - within 5m say, if you've got good satellite visibility. As a result, any speed readout is going to be approximate as well, IMHO.
A calibrated speed gun should be loads more accurate. Having said this, it might be worth checking to make sure the speed gun had been calibrated!
I should say my GPS knowledge is based on having messed about with fixed point GPS as part of a final year project at uni, so I could be talking through my hat!
Dan
Basically, GPS can be very accurate indeed (position within mm in fact), but only if the following things are true:
1) Lots of satellites are visible (the more the better).
2) The receiver is left in one place for several hours.
Without these two conditions being satisfied, GPS gets less and less accurate. As you're driving along, your position is only going to be known roughly - within 5m say, if you've got good satellite visibility. As a result, any speed readout is going to be approximate as well, IMHO.
A calibrated speed gun should be loads more accurate. Having said this, it might be worth checking to make sure the speed gun had been calibrated!
I should say my GPS knowledge is based on having messed about with fixed point GPS as part of a final year project at uni, so I could be talking through my hat!
Dan
Not quite true Dan.
The "error" from GPS is actually very constant. Speed just requires two measurements knowing two things, the time and the distance traveled.
Now - if the error in the distance is x at time t1, and y at time t2, you get
d = ABS((p1+x) - (p2+y))
If x tends towards y you get
d = ABS((p1+y) - (p2+y))
or
d = ABS(p1 + y - p2 -y)
or
d = ABS(p1-p2)
i.e. the errors become insignificant.
As to the time, GPS inbuilt decodes UTC from very very accurate clocks. As long as the thing that is analysing the data uses the embedded UTC time signature there is nothing that can touch it.
To that end, GPS speed is highly accurate over short distances assuming two things
1) The satellites being measured from the two measuring points are the same (not sure if the GPS logs will show this - if you are analysing the raw NMEA data you can verify this
2) The speed measurement uses UTC time.
Over longer distances you get problems with atmospheric effects etc.
Most GPS systems will quote an accuracy of GPS speed of no greater than 0.05 m/s, or 180 meters per hour, about 0.12km/h - first thing to do is to see what the manufactuer's quoted figure is for your device.
As to if it is admissable in court - certainly is, but whether it's taken into account is a different matter!
J
The "error" from GPS is actually very constant. Speed just requires two measurements knowing two things, the time and the distance traveled.
Now - if the error in the distance is x at time t1, and y at time t2, you get
d = ABS((p1+x) - (p2+y))
If x tends towards y you get
d = ABS((p1+y) - (p2+y))
or
d = ABS(p1 + y - p2 -y)
or
d = ABS(p1-p2)
i.e. the errors become insignificant.
As to the time, GPS inbuilt decodes UTC from very very accurate clocks. As long as the thing that is analysing the data uses the embedded UTC time signature there is nothing that can touch it.
To that end, GPS speed is highly accurate over short distances assuming two things
1) The satellites being measured from the two measuring points are the same (not sure if the GPS logs will show this - if you are analysing the raw NMEA data you can verify this
2) The speed measurement uses UTC time.
Over longer distances you get problems with atmospheric effects etc.
Most GPS systems will quote an accuracy of GPS speed of no greater than 0.05 m/s, or 180 meters per hour, about 0.12km/h - first thing to do is to see what the manufactuer's quoted figure is for your device.
As to if it is admissable in court - certainly is, but whether it's taken into account is a different matter!
J
joust said: Not quite true Dan.
The "error" from GPS is actually very constant. Speed just requires two measurements knowing two things, the time and the distance traveled.
[snip]
To that end, GPS speed is highly accurate over short distances assuming two things
1) The satellites being measured from the two measuring points are the same (not sure if the GPS logs will show this - if you are analysing the raw NMEA data you can verify this
2) The speed measurement uses UTC time.
Yeah, I was thinking this as I wrote it to be honest - you don't really care how accurate the actual location is, so long as the thing's consistently out as speed is distance over time, rather than having much to do with location.
I was thinking that as the disagreement between the speed gun and the GPS seems to be fairly large, the satellites seen probably weren't consistent - a 30 limit makes me think that buildings/trees may have been blocking line of sight to them, or that only a couple of sat's were visible at the time (needs at least 3 to be 'properly' accurate, if memory serves).
It could be worth a challenge, depending on how much info your GPS device can provide about the number of satellites visible at the time, etc.
Mind you, given that J's managed to bring up UTC time, it sounds as though he knows a load more about the subject than I've remembered - I'd go with what he says!
Dan
Dan was basically right.
If anyone is interested, my company did a breat deal of applied research in the area measuring the speed of ships in the 90s. Then the GPS signal was degraded by the US Dept of Defense deliberately. We achieved accuracies for ferries ( at 17-25 mile/h of +/- 0.15 mile/h over a 40 sec average using Kalman filtering of data.
Errors occur when the vehicle changes direction as it travels further between the 2 points at which position is measured. Thus it can under-read speed.
With a position error of +/- 1.5m and a car doing 30 mile/h, and position fixes every 2 sec,
1 part in 26.8m corresponds to an error over +/- 1.8 mile/h. Averaging 2 reads reduces this to under 1 mile/h. But, the car could be accelerating by about 20 mile/h in that 4 sec so the reading will be no better than a 4 sec average speed. In other words a single reading of 30 on the GPS, could be an instantaneous speed of between 20 and 40 – that must be borne in mind if offered in evidence of adherence to speed limits.
The LTI 20-20 works in a similar way. I do not know the fine details of the circuit settings. It sends out a 1/10 sec burst of 0.9nm infra-red light pulses. These strike the target they are aimed at which should be a flat vertical surface such as the number plate and are reflected back to the optic receiver. If the car travels towards the device the time of flight between each outgoing and reflected pulse reduces – the rate of reduction of the time of flight divided between the time between pulses presumably averaged out over the whole burst is the vehicle speed.
Errors are down to accuracy of timing, discrimination in the input that gives the precise start point of outgoing and reflected pulses will probably give an accuracy of 0.1 to 0.3 mile/h ( basically as the speed reading is taken from distance fixes over 1 /20 of the time interval of GPS.
Gross errors must occur when the laser is aimed at sloping surfaces such as bonnet or windscreen. These do not reflect as well as the flat vertical surface but will scatter enough IR light to get a returned signal. If the car is rising 10 or 20 cm or the beam is lowered even a hundredth of a degree in that 1/10 sec then errors of 5-15 mile/h or even more will result. The equipment will surely have some kind of signal processing to reject this possibility but then maybe it does not ( see www.pepipoo.com for explanation of how errors were demonstrated in a courtroom – the 4 mile/h towards the device that registered from the courtroom wall is consistent with aiming it at the wall at an angle and moving it closer to 90 degrees)
It is possible to use this information to challenge the accuracy of the reading in court and use the GPS device to strengthen this case. I have successfully avoided a ban in the past challenging the reading produced by a Vascar. But against you is
The Approval certification this equipment carries
The Salesman who imports the equipment from the states who is trotted out as an ’expert independent witness ‘ in such cases
The fact that few outside the engineering profession understand what is meant by accuracy, errors, precision etc And with a few exceptions that includes the bench.
If anyone is interested, my company did a breat deal of applied research in the area measuring the speed of ships in the 90s. Then the GPS signal was degraded by the US Dept of Defense deliberately. We achieved accuracies for ferries ( at 17-25 mile/h of +/- 0.15 mile/h over a 40 sec average using Kalman filtering of data.
Errors occur when the vehicle changes direction as it travels further between the 2 points at which position is measured. Thus it can under-read speed.
With a position error of +/- 1.5m and a car doing 30 mile/h, and position fixes every 2 sec,
1 part in 26.8m corresponds to an error over +/- 1.8 mile/h. Averaging 2 reads reduces this to under 1 mile/h. But, the car could be accelerating by about 20 mile/h in that 4 sec so the reading will be no better than a 4 sec average speed. In other words a single reading of 30 on the GPS, could be an instantaneous speed of between 20 and 40 – that must be borne in mind if offered in evidence of adherence to speed limits.
The LTI 20-20 works in a similar way. I do not know the fine details of the circuit settings. It sends out a 1/10 sec burst of 0.9nm infra-red light pulses. These strike the target they are aimed at which should be a flat vertical surface such as the number plate and are reflected back to the optic receiver. If the car travels towards the device the time of flight between each outgoing and reflected pulse reduces – the rate of reduction of the time of flight divided between the time between pulses presumably averaged out over the whole burst is the vehicle speed.
Errors are down to accuracy of timing, discrimination in the input that gives the precise start point of outgoing and reflected pulses will probably give an accuracy of 0.1 to 0.3 mile/h ( basically as the speed reading is taken from distance fixes over 1 /20 of the time interval of GPS.
Gross errors must occur when the laser is aimed at sloping surfaces such as bonnet or windscreen. These do not reflect as well as the flat vertical surface but will scatter enough IR light to get a returned signal. If the car is rising 10 or 20 cm or the beam is lowered even a hundredth of a degree in that 1/10 sec then errors of 5-15 mile/h or even more will result. The equipment will surely have some kind of signal processing to reject this possibility but then maybe it does not ( see www.pepipoo.com for explanation of how errors were demonstrated in a courtroom – the 4 mile/h towards the device that registered from the courtroom wall is consistent with aiming it at the wall at an angle and moving it closer to 90 degrees)
It is possible to use this information to challenge the accuracy of the reading in court and use the GPS device to strengthen this case. I have successfully avoided a ban in the past challenging the reading produced by a Vascar. But against you is
The Approval certification this equipment carries
The Salesman who imports the equipment from the states who is trotted out as an ’expert independent witness ‘ in such cases
The fact that few outside the engineering profession understand what is meant by accuracy, errors, precision etc And with a few exceptions that includes the bench.
Richard C said: Dan was basically right.
The LTI 20-20 works in a similar way. I do not know the fine details of the circuit settings. It sends out a 1/10 sec burst of 0.9nm infra-red light pulses. These strike the target they are aimed at which should be a flat vertical surface such as the number plate and are reflected back to the optic receiver. If the car travels towards the device the time of flight between each outgoing and reflected pulse reduces – the rate of reduction of the time of flight divided between the time between pulses presumably averaged out over the whole burst is the vehicle speed.
Errors are down to accuracy of timing, discrimination in the input that gives the precise start point of outgoing and reflected pulses will probably give an accuracy of 0.1 to 0.3 mile/h ( basically as the speed reading is taken from distance fixes over 1 /20 of the time interval of GPS.
Gross errors must occur when the laser is aimed at sloping surfaces such as bonnet or windscreen. These do not reflect as well as the flat vertical surface but will scatter enough IR light to get a returned signal. If the car is rising 10 or 20 cm or the beam is lowered even a hundredth of a degree in that 1/10 sec then errors of 5-15 mile/h or even more will result. The equipment will surely have some kind of signal processing to reject this possibility but then maybe it does not ( see www.pepipoo.com for explanation of how errors were demonstrated in a courtroom – the 4 mile/h towards the device that registered from the courtroom wall is consistent with aiming it at the wall at an angle and moving it closer to 90 degrees)
It is possible to use this information to challenge the accuracy of the reading in court and use the GPS device to strengthen this case. I have successfully avoided a ban in the past challenging the reading produced by a Vascar. But against you is
The Approval certification this equipment carries
The Salesman who imports the equipment from the states who is trotted out as an ’expert independent witness ‘ in such cases
The fact that few outside the engineering profession understand what is meant by accuracy, errors, precision etc And with a few exceptions that includes the bench.
Richard C : Thanks for that. Could it be said that these LTI's could have problems with reading the speed of say, a TVR Tuscan:
Lets say
operating device from elevated section, on a bend, pointing down some 350 metres away. Tries to target the number plate of a Tuscan, of which ground clearance is only 120mm. Could it be possible that the speed reading obtained would be very innacurate?
>> Edited by lucozade on Monday 10th February 17:30
Richard C said: Dan was basically right.
If anyone is interested, my company did a breat deal of applied research in the area measuring the speed of ships in the 90s. Then the GPS signal was degraded by the US Dept of Defense deliberately
I thought that Selective Availability was a more or less a slowly changing random error, so although it would give quite large errors for absolute positioning, measuring speed would still be quite accurate.
In any case wasn't SA removed quite some time ago?
Richard,
Remember that selective errors have been switched off now (probably until the next war!), which improves the error rate by 90%. Also, all current chips I could find do Kalman filtering as standard. There have also been some serious advances in other sources of errors - UNav's Zoom Correlators that significantly removes multipath (always a problem in built up areas) is just one of those.
You are quite right about the "curve/bend" effect thought, and acceleration / deacceleration.
However, over a sample of 4 or 5 that is steady you can be sure the error rates I mentioned are pretty much correct.
J
Remember that selective errors have been switched off now (probably until the next war!), which improves the error rate by 90%. Also, all current chips I could find do Kalman filtering as standard. There have also been some serious advances in other sources of errors - UNav's Zoom Correlators that significantly removes multipath (always a problem in built up areas) is just one of those.
You are quite right about the "curve/bend" effect thought, and acceleration / deacceleration.
However, over a sample of 4 or 5 that is steady you can be sure the error rates I mentioned are pretty much correct.
J
Gassing Station | Speed, Plod & the Law | Top of Page | What's New | My Stuff