GPS as a low-speed Speedometer

[Andy StopfordParticipant @andystopford50521]

Posted by Michael Gilligan on 09/11/2021 17:28:08:

But this time

Max Speed was recorded as 19.9 kph

and again whilst I was standing on the doorstep, the Speed continued to display 2.something kph until it recognised the loss of reception.

Despite John’s assertion to the contrary … I think the incredible max walking speed of 19.9 kph must illustrate that my GPS does rely upon location data for its calculation of speed.

Unless anyone has a better explanation: I can only assume that ‘warp speed’ was engaged during one of the “lost satellite reception” glitches.

In my experience, using the gadget described above, sometimes the GPS receiver will record a sudden jump, which may be 100 metres or more off the real track, for a single fix or two. If speed is calculated from distance between fixes, this instantaneous translation might be responsible for your warp speed.

One time, when I was at a storage unit in Tonbridge, Kent, my GPS device recorded a single fix some 220 miles south of Accra, Ghana, a round trip of about 6930 miles accomplished in the 20 seconds between two less surprising locations. I'm not up to calculating the peak speed that this might imply, given acceleration from and to rest, but an average of 1.25 million mph seems about right. Sadly, I had left the GPS gadget in my car whilst I was in the storage unit, so I missed the trip.

[John HaineParticipant @johnhaine32865]

What is the target spec for a loco speedometer? I would guess it isn't very different from a bike computer except distance and route travelled probably aren't so important. Driver wants to see the speed at a quick glance, needs a reasonably big high contrast display. Resolution of 1 mph probably adequate, best if the number isn't oscillating wildly.

My CateEye Stealth10 dates from 2014. It reads to 0.1mph. Pedalling at a steady rate on the flat as far as I can see the LSD hardly changes. Responds fairly quickly when one accelerates or brakes. When stopped outside (good signal) it shows a steady 0mph. I have previously checked the distance travelled display against the OS map app on my phone, it generally slightly over-estimates but only by a small percentage. I did notice that if one brings it inside it can give spurious speed readings but only until either you move outside again or it loses satellites. So perhaps you wouldn't get a reliable reading in the "shed" but then you probably shouldn't be looking!

Based on this I think the cycle GPS would make a perfectly adequate loco speed display. More recent models probably have better GPS chips in them and could work better.

Regarding GPS on the phone. I've previously checked the speed as indicated by the cyclometer against that displayed by the "GPS Status" app and found they agree. It's also interesting that the OS maps app, if you use it to track your route using the "on road cycling" option shows a path that accurately follows roads – but it may be deliberately corrected to do that. Next time I go out I'll try the "off road" option on the road and see if it's any more erratic. Certainly when recording walking routes it doesn't show signs of the sort of erratic route plotted above.

[Michael GilliganParticipant @michaelgilligan61133]

Posted by John Haine on 09/11/2021 22:09:44:

What is the target spec for a loco speedometer? […]

.

I don’t know, John … and frankly, it doesn’t really interest me [which is why I spun this thread off from the other one]

I simply want to understand how [un]reliable my GPS speed indications are, at low speeds, and why.

Your Doppler-based description is very impressive; but I am not yet convinced that my old Garmin GPSMap 76 uses that technique. … If it does: How do the spurious unrealistically-high speeds get recorded ?

It may be a generational thing, in that my device pre-dates the wide availability of the Doppler system that you referenced … or it may be that I am missing some important point.

I am very happy with the location performance of the device [*], but I have always been a little wary of the speed display.

MichaelG.

.

[*] If I walk the same route on two days, but on the other side of the road; it will typically show a parallel track … and it does a very decent job of O.S. Grid References.

Edited By Michael Gilligan on 09/11/2021 23:02:33

[peak4Participant @peak4]

Michael, I wandered out for a stroll this afternoon, with the intention of comparing my old Garmin 3 GPS to my modern mobile.
Sadly I fails as the Garmin won't turn on for some reason.
However, the mobile was running two different apps at the same time, an Olympus one for geo tagging photos and Anquet OMS for a tracklog on OS 1:25k OS maps.

Unfortunately I failed on this comparison for speed as the OS app no longer saves it on the tracklog. (I'm sure the previous version used to)
The distances and height gains were both believable and comparable, give or take a few metres on the vertical and a couple of hundred in distance.
The Olympus app showed a maximum speed of 2.7MPH, which is probably OK
I Did however have a scroll through previous Olympus tracklogs, 6.2mph, 3.3 MPH, 168.2MPH, 2.9 MPH, 1407.5MPH, 50.1MPH. (The various average speeds of each trip were believable)
It's several years since I flew anywhere, so it looks like some of the max speeds are a tad optimistic.

I don't know how often the Oly track program takes fixes, but I suspect it must be quite frequent as it's able to geotag photos taken a few seconds apart with reasonable accuracy.

The OS app tells me it only takes fixes at 60 second intervals, so would be hopeless for showing variations in low speeds.

Next one to try is my Garmin satnav, which I can set to walking mode.

Not sure any of this helps your cause though, other than my OS GPS app only locates every 60 seconds, so no good for speed variations at walking pace.

Bill

 

Edited By peak4 on 10/11/2021 01:34:22

[Michael GilliganParticipant @michaelgilligan61133]

Thanks for the interesting input, Bill

This ‘investigation’ of no enormous consequence [I’m really just on a quest for knowledge for its own sake] but I will be most grateful for any more examples.

It’s a very different issue, but this all reminds me a little of the ‘blind faith’ that people seem to have regarding angle measurements from things like the Wixey. … The Devil is in the detail.

MichaelG.

[Ady1Participant @ady1]

If you want a really accurate system then you need to stick your own waypoints down round the track

Even close up stuff like laser/IR measuring of a few inches isn't that great and is a compromise, and that scenario has no dynamic movement to deal with, yet a reliable 1/10th of a mm is still difficult to achieve, would be perfect for a lathe or a mill with no rails needed

Perhaps an atomic clock would help in certain cases

Edited By Ady1 on 10/11/2021 12:03:02

[Joseph Noci 1Participant @josephnoci1]

I will try add my cent's worth…

Any GPS worth its name will measure speed using Doppler computations, based on signals received from multiple GPS sats.

Speed derived differentially from position fix changes is very inaccurate and is worse at low speed – typical position accuracies with 6 to 9 sats in view, PDOP less than 1 ( Position Dilution of Precision) will typically provide position accuracies of 3 to 5 meters. 4 Sats with PDOP 2 to 3 will give 10 to 25meters. Your average commercial handheld GPS unit has a fix rate of 1second. Walking at 3.6km/hour ( a 'convenient number..) , around 1meter/s , with a PDOP of 3, means the GPS position can jump a few meters from 1 walking step to another. So a long-period Kalman filter is often implemented to smooth out the jumps in some sort of predictive fashion. But typical speed accuracies in this mode will easily vary by 20 to 30% from true, at speeds less than 5km/hr. At higher speeds the error reduces since the distance travelled in 1 second is a larger proportion of the GPS 'jumps' so the speed comes closer to truth…

Doppler speed determination is much faster, and caters for low speed ( 0.05m/s ) with ease.

Doppler speed measurement is however not the Holy Grail – it suffers severely from multipath effects in urban and built up areas. The reflection of signal and reception of such results in false doppler motion and gives way to some large speed jumps – which are again smoothed by predictive Kalman Filters ( Other filter type are also employed, huge advances in predictive filter types in this field).

You can get a feel for the accuracy of the speed measurement by looking at the SDOP parameter ( Speed Dilution of Precision) –

If serious analysis is contemplated, there are 4 main parameter that need to be understood and then monitored in the GPS messages – these will give a good feel as to how accurate the item of interest may be –

These are PDOP, SDOP, HDOP ( Horizontal Dilution of Precision – – related to the position accuracy) and VDOP, Vertical Dilution of Precision.

There are abundant texts explaining the above elements.

The bottom line is – for 'good' ( 2 sigma) low speed measurement, you need clear sky view, 6 to 8 sats, NO building or structure taller than yourself within 100meters – PDOP and SDOP of 1 to 2 MAX.

There is a further complication however – GPS Sat ephemeris is updated constantly – there is an approximate 3 hour cycle where major updates are required. Inn addition, when 2 or more sat signals fall below the limit, Sats are swapped out per the prediction tables stored within the GPS – during this time, Doppler measurements show glitches in the results, which have to be smoothed out in the predictive filters – this does glitch the speed accuracy as well – couple this with momentary body shielding the GPS signal, some multi-path reflections, and Murphy, and accuracies are never achieved.

I have some considerable experience in this field, having spent many years developing autopilot and flight control systems for Military drones – 300kg/8 hour flight time class. GPS NEVER played a role in the auto Take-off function as the initial runway roll was to slow – sub 10km/hour. MANY GPS units were tested and evaluated – UBLOX 7, 8 and 10 perform very well, TRIMBLE was very good at 5 to 8 times the price..

Joe

 

Edited By Joseph Noci 1 on 10/11/2021 16:04:04

[Michael GilliganParticipant @michaelgilligan61133]

I went for a short, but reasonably brisk [for a man of my decrepitude] walk along the main road today … outward on one side, returned on t’other:

 

The Trip Computer recorded the following:

Odometer 6.01 km

Stopped time 00:01:11

Moving Average 5.6 kph

Moving Time 01:04:04

Overall Average 5.5 kph

Total Time 01:05:16

Which, again, all seems reasonable, and consistent within itself.

.

But this time a [highly improbable] Max Speed of 11.0 kph was recorded

There was only one audible “lost satellite reception” warning, and that occured considerably later than the spurious Max Speed was recorded.

For the moment, I can only assume that the issue relates to jumps in location accuracy.

MichaelG.

.

Edit: __ Just read Joe’s post, which crossed with mine

Clearly my device is not “worth its name” … but we have been happy together since about 2003

Edited By Michael Gilligan on 10/11/2021 16:25:01

[Michael GilliganParticipant @michaelgilligan61133]

A question for you please, Joe

The spec. sheet for my GPSMap76 is here: **LINK**

https://www8.garmin.com/specs/GPS76_GPSMAP76.pdf

… and I note that it claims a Velocity accuracy of “0.05 meter/sec steady state”

Does that mean that it does use Doppler after all ?

MichaelG.

[Joseph Noci 1Participant @josephnoci1]

Michael,

Yes , it uses doppler for sure.

GPSMap76's circa 2003 use the first of the SiRFStar GPS chips – SirfStar II ( now up to SirfStar VI), and was the first practical implementation of Doppler speed determination in the SirfStar series.

They have made many changes and 'improvements' in the later versions and the lacklustre low speed performance ( within the boundaries of the application) is excused by the Specification qualifier 'Steady State'….

That says it all – It implies that the reception conditions provide a PDOP and SDOP of 1, that there is no Multipath reception effect, at least 8 Sats are in view and at least 6 of those show a received signal strength of -100dBM minimum, AND, importantly , if measuring position, to achieve the 3 to 5meters within 2 Sigma, the GPS must be static.

Likewise, to achieve the 0.05m/s speed to 2 Sigma, the speed must be CONSTANT and not changing…

The rest is smoke and mirrors.

Walking is quite a severe test on the GPS – the speed is nowhere near constant, with rapid changes from step to step, etc. The Kalman filter tries to smooth this out, and present the resulting speed with a large lag, That lag varies greatly depending on the rate of speed change, and its amplitude – changing from 1m/s to 3m/s and back to 1m/s ( you start to cross a road slowly, see a car approach and 'run' a few steps to cross safely, then slow down again) – that spike in speed can take 5-10 seconds to filter through to the display, and a number of seconds to dissipate. The peak speed you achieved during your short run will also never be shown – the filter killed a lot of it..

Couple with all that the fact that the GPS is strung around your neck, shielded by the body, swinging about, poor antenna, low signal strengths…It is a wonder it works at all – and attests to the amazing technology lurking there.

But 'working' means different things to different users – for the serious hiker, such devices do everything needed with good performance.

To a Farmer using Autonomous Farming method, ie, A tractor plowing without a driver, assisted by GPS and a Autopilot, The GPS has to be very accurate and work at low speed – no use the plowed furrows overlapping, when they are spaced 2 meters apart. Such applications use an integrated approach , generally of two possible types-

One is a good quality GPS, with a very good Survey Grade antenna ( it is large and specially shaped to mitigate Multipath effects) – this GPS then integrated into a 'navigation' package – with 3 axis gyro's and accelerometers which provide very high short term accuracy, with the long term accuracy aided by the GPS.

For even more accurate performance – to centimeters – that installation is aided by accurate position surveyed Ground Stations that assist by providing GPS correction data to the Host GPS system – this is identical to the method used by land surveyors to mark out land sites.

There are now consumer GPS models that incorporate integrated Gyros and Accelerometers and these do provide greatly improved low speed performance, but one must remember there is a fair lag on that data, and for consumer use this can be some 10's of seconds. Some very accurate consumer models even increase that lag to prevent the device being used as part of a navigation/autopilot system in a weaponised drone..

I would suggest if you want better performance from your '76, obtain the best external antenna you can find, something better than those 40mmx40mm ceramic patch 'puc' antenna, and fix it to your Bunnet – you may be amazed at the improvement..

Another long story from me..

Joe

[Michael GilliganParticipant @michaelgilligan61133]

Thanks for the comprehensive, and very helpful reply, Joe

… I was just about to post a supplementary “philosophical” question:

In the context of a hand-held GPS receiver, measuring Speed :

What, on Earth, is ‘steady state’ supposed to mean ?

.

You have already debunked that one

MichaelG.

[John HaineParticipant @johnhaine32865]

If I may add to Joseph's comments. My GPS cyclometer operates in a reasonably optimal configuration being mounted nearly horizontal on the handlebar facing the sky, and normally operates in an open space. It probably uses one of the little ceramic pucks Joseph describes but at least the designer has some certainty as to how to locate it in the product. And bikes move at a reasonably constant speed. Same characteristics probably apply to small locomotives.

[Michael GilliganParticipant @michaelgilligan61133]

Posted by John Haine on 11/11/2021 09:13:48:

If I may add to Joseph's comments. […]

.

You certainly may, John … Your experience might be of great interest to some who are following this thread.

For myself, however : I am much more interested in Navigation than in Speed measurement … and therefore Joe’s explanation was exactly what I needed … I will help me to ‘manage expectations’.

The GPSMap76 does a superb job of locating my position and my track; the anomalous Speed readings are just a minor irritation.

MichaelG.

[SillyOldDufferModerator @sillyoldduffer]

To put some numbers on the problem, I've set up an Adafruit Ultimate GPS Module on my dining table with a fixed antenna aimed out of a south facing window. It's view of the sky is limited to about 45° above horizon by a house at the end of my garden, and the antenna is inside the double-glazing. Not good compared with an outdoor installation.

The receiver and antenna aren't moving, therefore any speed and direction of travel reported indicate errors due to the satellites moving.

The GPS module outputs an NMEA data stream on a serial pin. I've connected the output to an Arduino Nano, which reads the records and forwards them to my main computer via a USB cable. The Nano does no processing, it just implements a USB serial connection so my workstation computer can read whatever the GPS module says.

I wrote a Python3 program to decode the NMEA records. Most of the work is done by a python module called pynmea2. It validates NMEA strings, and if they are OK, unpacks them so the data can be manipulated by a program. The program can examine GPS innards in a way most user applications don't reveal. In this case, I'm not interested in the correct time, latitude, longitude or altitude. I am interested in how many satellites are in view and how many of them are being used at the moment to calculate the fix. The quality of the fix depends on how many satellites are used (more is better, but only if their signal strength is above a certain level), and how spread across the sky the are. The fix obtained from 6 satellites grouped close together, is worse than the fix obtained from six satellites spread across the sky: this error is reported by the GPS as Horizontal, Vertical, and Position Dilution of Precision. (HDOP,VDOP&PDOP)

The Ultimate GPS outputs GPGSA, GPGSV, GPGGA, GPRMC and GMVTG records. GSA contain fix quality, hdop, vdop, pdop and lists the satellites in use. GSV lists all the satellites in view and their signal to noise ratios. GGA reports time, position, altitude and fix type. RMC reports time, date, position, course and speed.

Extracted and collected together, the relevant data, which is output once per second, looks something like this:

Quality of fix = 3 (good)

Altitude 110,1 metres ASL (about right)

Number of satellites detected, 13 (a bit low)

Number of satellites usable, 6 (a bit low)

Ground speed varies between 0.1m/s and 0.49m/s. At the same time, the course is reported to be between 79° and 311°, actually speed and course are wandering inside an error circle. As receiver and antenna are both fixed, the error must be due to the limitations of the system. Partly GPS itself, but mainly the poor view of the sky the antenna has.

I'm letting the computer log data for a few hours. I expect to see the number of satellites used change, and the quality of the fix vary with that. I think the DOP figures will alter too. I'll try and draw some meaningful graphs rather shower the forum with numbers.

In theory DOP improves when the satellites being used have wide angular displacement, and reduce when the satellites are in the same part of the sky. The poor position of my antenna probably means it can only see clusters of satellites. Better if the antenna were outside above roof level, but I'm too old for that sort of adventure!

Dave

[John HaineParticipant @johnhaine32865]

Another couple of "data points". This morning I did a test ride to a nearby village, tracking the route using the "on-road" option on the OS App on my phone and also recording the distance on the GPS cyclometer. Then I returned recording the same route exactly but using the "off road" option, on the assumption that the first will correct the positions since it "knows" they should be on the mapped road but the second wouldn't.  Most of the time the phone was just tucked into a jacket pocket.

Part of the outbound track:

I was cycling to the southeast. Note more of less in the middle of the grid square the mast on the SW side and the trig point which is a few metres to the NE. The "kink" in the track a bit further along is me stopping.

Now the return track:

The track is clearly a bit more wobbly but for the most part stays on the road. I did stop briefly just by the trig point and there it thinks there's a slight overrun and reverse which can't be seen at this scale.

Distance out recorded as 20.45km, back 20.4km. Both 1.04 hours. The cyclometer recorded the outbound trip as 12.74 miles, back as 12.70; 20.5 & 20.43 km.

I'm not sure what this indicates except that the OS map App is pretty good at tracking a cyclist on road and off road, though it clearly uses its knowledge of where the roads are. It's also pretty consistent and agrees well with a different GPS device. The phone is a Pixel 4a, uses a Qualcomm platform so therefore the SiRF GNSS, and is multi-standard GNSS. Also uses GPS assistance data.

Finally I took this screenshot while I stopped by the trig point.

I stopped on the righthand side of the road, The trig point is probably 5 metres into the field there, so the location accuracy is pretty impressive to me.

Edited By John Haine on 11/11/2021 15:02:21

[Neil WyattModerator @neilwyatt]

Posted by Michael Gilligan on 07/11/2021 17:36:13:

Posted by Michael Gilligan on 07/11/2021 14:47:58:

.

I would be interested to see the results of similar journeys by others.

MichaelG.

.

I am grateful for your comments thus far, gentlemen … but, if it's not too much to ask: Could someone please do me the simple kindness of [approximately] repeating my exercise, using a different GPS.

I have neither an Android 'phone with GPS, nor an iPhone, so I cannot do the comparison myself.

My car GPSs don't have the detailed 'Trip Computer' which the GPSMap76 provides.

I am more interested in the actual slow-speed performance of available/affordable devices than the hypothetical merits of cutting-edge developments.

This is simply a quest for knowledge and understanding.

Thanks

MichaelG

Simpler than that, I have an app I ocassionally use to record where I go during site visits 'GPS Logger'.

For the most recent walks the maximum speed and averages recorded are:

7 & 2.7 km/h over 303m

9 & 4.4 km/h over 272m

8 & 3 km/h over 791m

7 & 2.4 km/h 2.12km

All of which are feasible but probably a bit excessive.

Cycling the top speeds are more believable: 38 & 41 km/h given they included some fast downhills.

Sitting here at my desk typing this I have 13 to 17 satellites out of 17. It says 9m accuracy, an average speed of 0 km/h (but I did flicker to 3 km/h when I moved it by about 1m then back to zero).

This suggests that it does register short, rapid movements enough to create a peak.

It's giving my altitude as 111m, when it's actually nearer to 60m, which is definitely the weak point of GPS due to the geometry of visible satellites being fairly high up.

Hmm accuracy dropped to 49m, speed flashed up to 2km/h … and then it spent a while looking for satellites, then 13/13.

I suspect it would be vastly more accurate outdoors.

Tries in window… 23/28 satellites, accuracy 8m. Speed showing 0.

Neil

[Neil WyattModerator @neilwyatt]

My 'on site' experience and also that of overlaying tracks onto maps in a GIS is that the given precision is usually pessimistic.

Neil

[SillyOldDufferModerator @sillyoldduffer]

Posted by Neil Wyatt on 12/11/2021 13:31:16:

Posted by Michael Gilligan on 07/11/2021 17:36:13:

Posted by Michael Gilligan on 07/11/2021 14:47:58:

.

I would be interested to see the results of similar journeys by others.

MichaelG.

.

I am grateful for your comments thus far, gentlemen … but, if it's not too much to ask: Could someone please do me the simple kindness of [approximately] repeating my exercise, using a different GPS.

…

…

7 & 2.7 km/h over 303m

9 & 4.4 km/h over 272m

8 & 3 km/h over 791m

7 & 2.4 km/h 2.12km

…

Cycling the top speeds are more believable: 38 & 41 km/h given they included some fast downhills.

Sitting here at my desk typing this I have 13 to 17 satellites out of 17. It says 9m accuracy, an average speed of 0 km/h (but I did flicker to 3 km/h when I moved it by about 1m then back to zero).

This suggests that it does register short, rapid movements enough to create a peak.

It's giving my altitude as 111m, when it's actually nearer to 60m, which is definitely the weak point of GPS due to the geometry of visible satellites being fairly high up.

Hmm accuracy dropped to 49m, speed flashed up to 2km/h … and then it spent a while looking for satellites, then 13/13.

I suspect it would be vastly more accurate outdoors.

Tries in window… 23/28 satellites, accuracy 8m. Speed showing 0.

Neil

 

Consistent with Neil's comments, here are my results over a few hours for a fixed GPS, registering course and speed despite the receiver not physically moving at all. Why?

First of 3 graphs below, shows the number of satellites in view (orange line), varies between 13 and 7, while the blue line shows the number of satellites with signals strong enough to use. It varies between 9 and 3. The choppy blue line suggests my receiver frequently has to switch between satellites as they transit overhead, might cause trouble.

Second of 3 shows the Horizontal Dilution of Precision. Although HDOP remains fairly low, the graph shows my GPS loses the fix on 3 occasions, and suffers a major drop of quality on a fourth. The third graphs shows speed peaks during the HDOP crises, and HDOP crises correspond with the receiver only picking up 3 or 4 usable satellites.

All three graphs show that GPS loses the plot whenever satellite signals are compromised. Many reasons for that, such as reflections, solar activity, interference and obstacles, including 'weather'. In a typical consumer GPS, the receiver is unlikely to be of the best, but even the best receivers struggle with grotty antennas. The aerial in a portable GPS or phone is poor: prone to interference and reflections because they are too close to the ground, small and deaf, and maybe pointing in the wrong direction! In contrast, surveyor GPS perch a large sky-orientated antenna about 3 metres high on top of a pole.

Next graph is a histogram of speeds centred around average speed. The spread shows a mathematical filter applied over time would conclude the unit is barely moving on average. But it's hard to tell genuine slow movements and apparent movements due to the system apart. Another problem is the histogram scale shows there are fast outliers, the maximum being 7m/s. The outliers match the speed peaks noted when HDOP goes crazy due to too few usable satellites.

Next graph is harder to explain. The scatter shows the relationship between speed and course. I expected both to be random, showing the unit to be moving about inside a small circle and they're not. There are two clusters at 0/360° and 180° suggesting my unit's apparent movement has a North / South bias. Possibly to do with my antenna only seeing an east-west arc of sky because my house and neighbouring properties block the North and South views. Dunno!

Even though outdoors with a good antenna in the clear would reduce the errors, the results suggest GPS isn't a good way of measuring slow speeds at ground level. Not awful, just not reliably accurate.

Dave

 

Edited By SillyOldDuffer on 12/11/2021 14:51:31

[John HaineParticipant @johnhaine32865]

Though this thread has gone quiet, I just thought that I'd add that I asked a GPS guru of my acquaintance and he confirmed that today GPS (and I assume other GNSS system) receivers do track the rate of change of carrier phase to determine velocity.

[Author]

Posts