GPS versus Speedo Accuracy

[OldButNotDead]

On some forum somewhere I once asked does GPS accurately measure speed in turns.  I asked the question because I noticed that indicated GPS would drop significantly in a turn.  That is if I went into a turn at an indicated 60mph, it might drop to 55 mid turn and recover to close to 60 after the bike straightened up.  On my GS it was hard to read the analog speedo but the GPS on the Nav 5 was easy to read.

Well I recently set my digital readout to show "speed", the speed that the analog needle was show or real dog on close.  So today I compared the digital readout and Nav 5. I went in and came out at 63.  It read 62 mid turn.   GPS went in at 63 plus or minus one mph.  It dropped to about 55 mid turn.  I say about because I did this several times and it was also plus or minus about 1 mph.

I have my theory but I want to see what you GPS gurus think.

[HawkGTRider]

My theory is that it depends...on how quickly the gps is connecting to the satellites and most likely several other variables. If the gps connected every 5 seconds, there's a significant probability that it could miss you making a turn and calculate your speed as if you'd gone straight across the corner. Since it would think you went straight and therefore a shorter distance, you must have gone slower. If it connected to the satellites every 1/10th second, the probability of that happening would go way down...unless you're in an area with significant trees covering the road (another variable).

That's my theory of how it could happen. It's probably a rather simplistic theory, but may still have some credence. I suspect if you had a military spec gps that sampled the satellites at an extremely fast rate, your variation "might" not be as significant.

[TN2Wheeler]

^ this. The gps speed is calculated based on the elapsed time between the locations it "obtained" from the satellites. So, it's gonna "square off" the corners to some degree.

That said it's still likely to be more accurate than a mechanical speedometer when traveling in a straight line.

[OldButNotDead]

To get a three dimensional fix four satellites are required, three for a 2D fix.  It's all geometry.  The data sent by the satellite is processed by your unit.  The unit gets the position of each satellite from the data stream of the that particular satellite.  The distance to the satellite is measured by the time the packet left the satellite to when it was received.  Where things start to mess with the basic solid geometry problem is that the unit may be choosing from four to as many ten satellites in view AND the signals can reflect off of objects making them longer than they should be.  Jim, what you said about the straight line is probably right and why your E911 dispatcher usually gets a more accurate position on cars than something standing still.  Actually I'm not as interested in "why", I'm pretty sure I know the "why".  I'm just wondering how many people notice this change in speed in the turns.

[DirtFlier]

I'm almost always looking through the curve and not down at my GPS.  :-)

[HawkGTRider]

I have a series of curves on one of my routes to and from home, and I'm guilty, at times, of seeing how quickly I can go through them. Like Tosh, I am focused on looking through the curve and only observe my speed with a quick glance as I exit the maneuver.

[OldButNotDead]

Tosh, Geoffrey,
I'm right with you on keeping eyes on the road.  My primary exit into the great beyond is a four or five mile stretch of road that has several "clear sight line" 20 mph turns.  That's where I do my experimenting.  I know the turns and can almost run them blind (blatant lie lol) .  Just wanted to see what speed I was actually taking them and that is where I noticed the difference between speedo and GPS, not to mention just by feel I knew I was not dropping speed in the turn.
Rod

[stevegrab]

I'm almost always looking through the curve and not down at my GPS.  :-)

This and I probably pay attention to my speed as much as anybody so that I know if and by how much I'm breaking the limits.
I have noticed on some mountain roads my GPS showing my bike dancing off road somewhere, because it is not accurately reflecting my location. So I presume similar things impact the speed readout, which is still far more accurate than any bike speedo I've had (newewst bike is over 10 years old).

[NinjaBob]

This anecdote may have no relevance to this topic but... I run with a GPS enabled trainer on my wrist, Garmin Forerunner 305.
It seems to be as accurate as my Nuvi and Zumo. But I noticed an elevation discrepancy of about 40 ft depending on whether I have my wrist at my side or have it raised to eye level. I figure the antenna is more accurate at a certain angle. So besides the reasons mentioned above for the speed differences maybe the lean angle of the bike in a turn affects the accuracy.

[JimRRides]

Corollary to GPS accuracy for speed

In the last few years I have observed officers and prosecutors referring to the GPS recorded speed on the video from the patrol vehicles as proof of the defendant's speed. Rarely does the defendant challenge the assumption that the GPS speed is an accurate measure. It is NOT an accurate measure of the suspect vehicle nor the patrol vehicle, although it may be referred to as an approximation. This source of speed measurement has never (to my knowledge) been vetted by the courts in a manner similar to speed radar. If you or an acquaintance face an accusation based on this measure,  challenge it. (Note that some video systems record the readout from the actual speed radar unit and thus is a different issue. Our traffic units had this capability but not our routine patrol units.)

[HawkGTRider]

That's pretty interesting...thanks for the words of wisdom.

[OldButNotDead]

Corollary to GPS accuracy for speed

In the last few years I have observed officers and prosecutors referring to the GPS recorded speed on the video from the patrol vehicles as proof of the defendant's speed. Rarely does the defendant challenge the assumption that the GPS speed is an accurate measure. It is NOT an accurate measure of the suspect vehicle nor the patrol vehicle, although it may be referred to as an approximation. This source of speed measurement has never (to my knowledge) been vetted by the courts in a manner similar to speed radar. If you or an acquaintance face an accusation based on this measure,  challenge it. (Note that some video systems record the readout from the actual speed radar unit and thus is a different issue. Our traffic units had this capability but not our routine patrol units.)

Never knew that was happening.  I do not believe GPS is a reliable source to write a ticket OR to defend your case either.  Also, NinjaBob, GPS has a lot of problems with vertical accuracy.  There is this thing called PDOP, Positional Dilation Of Precision which tells you how good your accuracy is from the geometry of the satellites.  It is always worse in the vertical.

[HawkGTRider]

I can believe that quite easily. When I was in Death Valley years ago, the official elevation was below sea level. At that point, my gps was off by quite a bit. Other times, such as going across some high mountain passes in Colorado, etc., it's been close to the posted elevation at times and not particularly close at others.

Good info...thanks for sharing OBND.

[BMWKeith]

I'm kind of late to this thread, but I'll post my observations and theory here now anyway.
I've also noticed the GPS speed being significantly lower coming out of turns than in straight lines, sometimes 10 or more mph slower.  Also, it is not very good for quickly establishing your speed as it changes, such as when accelerating away from a stop or getting on the expressway.
I believe that this a reflection of the accuracy of your position calculated by the GPS.  Every position that is calculated has what is called Circular Error of Probability (CEP).  This is essentially the circle that you will see displayed on your GPS when you zoom in really close on your position.  Your true position could be anywhere inside of that circle.  Every time it calculates, it has that error.  If it displayed your speed based on the distance per time of every position update it calculates, first, it could be difficult to read because the number would be changing so frequently.  Second it could potentially have variations of several mph from one update to the next because of where you actually are in that circle, whether ahead of or behind your true position.  Therefore, I suspect that the speed display is updated based on some number of points several samples apart, maybe 5 or 10, so that it minimizes the errors in the calculated speed due to the error in your actual position.  In a turn, the distance between points several samples apart is smaller than the distance traveled because it is calculating based on the cord length of the arc just traveled, instead of the sum of a series of points that almost form a smooth arc.  Hence, you see a lower speed on your GPS coming out of a tight hairpin turn than you see on your motorcycle speedometer.