Distance/Speed Accuracy: Map vs. GPS vs. E-Bike Display: Serious Discussion Needed

[Stefan Mikes]

This has been already discussed in many threads in different EBR Fora without any real conclusion. Shall we try again? Let us discuss it once and for good.

Ride Moving Time
There is one ride parameter that is handled quite well (although not perfectly) by e-bike displays: The net riding time. E-bikes featuring ride time measurement give the net (or moving) riding time very accurately: The time is being metered from the moment the rear wheel starts rotating to the wheel stop. GPS sports trackers feature some kind of "auto-pause" or "auto-stop". That is, ride time measurement is being paused when the measured ride speed falls below some predefined value (and is restarted after the bike speed exceeds that value). The accuracy of GPS based apps such as Strava is rather spotty related to time measurement (it can differ from e-bike display value by many minutes on a long ride).

Map Accuracy (Distance)

No doubt, a digital map including elevation profile, would give the most conservative distance value between points A and B over defined route. As @mschwett has correctly stated, a bike doesn't ideally follow the route as it doesn't move in straight line really (on contrary, marathon champions are being guided to run over the shortest route possible). In any case, the map is the reference to which we all need to adhere for comparison.

GPS Devices (Distance/Speed)

GPS navigation works perfectly, even for bikes. Ride recording based on GPS is a different matter. @Mr. Coffee has quoted a good article:

Why GPS makes distances bigger than they are

That article is too complicated for me to really comprehend it. Suffice to say, the Austrian scientists have mathematically proven that GPS will overestimate the value of the distance ridden. For this reason, Strava offers Distance and Elevation Gain Correction Tools. If you want to be honest (or mistrust the GPS measurement), you can make the distance and elevation figures corrected by map, which gives the most conservative estimate of your effort.

E-Bike Measurement (Distance/Elevation Gain/Speed)

We are entering tricky terrain here. Theoretically, measurement by e-bike should be the most accurate. A magnet for the speed sensor is located either at a rear wheel spoke (less accurate) or at the brake disk (more accurate). A predefined value of wheel circumference is stored in the e-bike's computer/controller. Given measured rotational speed of the wheel, the distance ridden and actual speed should be the most accurate. Are these value correct?

No, they are not.

Have you noticed the actual speed of the car given by its speedometer is often overestimated, compared to the GPS readout? It is legally fine because the reported speed makes you actually drive slower, thus safer. The odometer/daily trip meter readouts are -- on contrary -- quite fine. There is a single e-bike situation that resembles the car: The European S-Pedelec, or, the EU 45 km/h e-bike. The EU S-Pedelec (like, Vado 6.0, Allant+ 9.9S and similar) has to be certified for conformity, as it is perceived by law as a moped, or a motorised vehicle. For this reason, the speedometer of the S-Pedelec cannot show actual speed lower than the actual, and the odometer should be quite accurate.

Take the Specialized Turbo Vado 6.0, EU version. Its Certificate of Conformity reads: "Solely certified for Specialized Electrak 2.0 Armadillo tyres". It is because Specialized had to calibrate the Vado 6.0 computer against the tyres used. Whatever the actual wheel circumference is, the "Wheel circumference" value of the system is a fixed value of 2250 mm for that e-bike. Outcome: The most accurate agreement for the actual speed and the distance ridden.

For all other regions and e-bike types, manufacturers actually give a s%$t for speedometer or distance meter accuracy. Let me give you three real life examples:

• Giant Trance E+ 2 Pro. "Wheel circumference" figure of the system is 2178 mm. Actual stock wheel circumference is 2250 mm. A distance of 120.95 km (by GPS) ridden; map-based distance of 120.32 km. The Trance display readout: 125.5 km! Another situation: a downhill ride: Trance display gave maximum speed of 62.4 km/h but Strava was closer to 59 km/h.
• Lovelec Diadem (a hub-drive classic e-bike): Very accurate readouts for distance ridden on 42-622 tyres but overestimated distance and speed on stock 40-622 ones.
• S-Pedelec Vado: very accurate on certified Electrak 2.0 tyres. However, the same 120.32 km map distance was reported as 122 km when I used 47-622 (smaller) tyres on that Vado.

Several EBR members (@Marcela, @GuruUno, others) have been patiently trying to adjust the wheel circumference of their e-bikes (working together with their LBSes), not to avail any real conclusions.

Elevation gain measurement?! Forget it. Only the map can give you the true elevation gain (that's why Strava offers Elevation Correction). Barometric altimeter used in some e-bikes (typically, one of the smartphone) is good for nothing. Same with the inclination sensor found in some e-bikes.

• BLEvo app for Specialized e-bikes uses barometric altimeter and also tries to calculate the inclination. It gave the figure of 950+ m elevation gain in pretty flat terrain. (Strava has corrected that down to 148 m).

Radar

The only way to truly verify your e-bike speed (so, distance) measurement is to ride by a public radar with a display. You could be shocked to find out you're riding slower than you had thought...

I leave the subject open for a serious and honest discussion.

 

[RunForTheHills]

Radar has to be calibrated and may not be accurate. You could ride your bike over a known pre-measured course. You can also measure the exact circumference of your wheel by marking the side of the wheel nearest the ground with a pen and rolling it on the ground until the mark is near the ground again, then measure the distance along the ground between the starting point and end point. Garmin sells a wireless speed/distance sensor for their Edge GPS's that will also work with some of their GPS watches. It mounts to the hub of your wheel. This can be calibrated with the measured wheel circumference.

I have used the Garmin speed/distance sensors and Garmin GPS' over many miles on my bikes. I have also used their GPS watches and the very accurate Stryd footpod (calibrated on a track) for running. Aside from some really wonky tracks when running through tall buildings or under heavy tree cover, I find the GPS is accurate enough for most purposes. I am working from home these days, but I used to commute four miles to the office each way and the measured distance with GPS only would vary by only 1-2 tenths of a mile. Using the speed/distance sensor, it would hardly vary at all.

 

[retiredNH]

The article by the Austrian scientists is interesting, especially their experimental "data". It seems they chose circumstances that would maximize any error. The first experiment used a walking course that was a box 10M on a side. Since low cost GPS, like they claimed to use, have measurement errors in the same range, this experiment doesn't show much. The second, tracking car movement over 6 days, could have been better, but they did it in an urban environment, and compared against the car's CANBUS readings. Urban environments can be troublesome for GPS, and CANBUS has its own errors.

Bottom line? They didn't prove much.

More importantly, who cares? Since when do bike riders need a high level of accuracy? Terrain adjustments? Depends on how you define distance.

For those that might need high accuracy, there are high accuracy GPS systems available that can measure down to inches/cm, like Trimbles. But who cares?

 

[6zfshdb]

The main issue with distance and calculated speed measured by consumer grade GPS equipment is this:
GPS measures distance between satellite acquisition points which are rarely in a straight line and NEVER follow the exact line of travel. This can be seen in the pic below.

Assume your line of travel is the yellow roadway center line. The small hollow squares represent satellite acquisition points. The light & dark blue GPS tracks represent the distances between these points measured by the GPS on two separate passes. It's easy to see why they do not match each other as well as the exact measurement made along the yellow line. The GPS distance is usually longer and rarely the same on each pass.

When the white shoulder line is the line of travel, the red & green lines represent the GPS tracks using expensive commercial grade GPS equipment. The accuracy is better but still subject to the same track distortion error.

 

[PatriciaK]

And then, there are those of us - don't look at me - who will compare their bike data to their GPS, and choose the (usually infinitesimally) larger report every time .

Sorry, Stefan - that's not serious discussion, but it's the first thing that came to MY mind... Don't ask why .

 

[Deleted member 18083]

From today's ride:

• Ride with GPS logging interval set to 1 second.
• Note the different zoom/scales.
• 2 GPS tracks : out (lower distance) & return (higher number).
• Rough remote area rail trail : no need to keep to the 'correct' side of the track.
• Yellow X : difficult creek crossings – partly/fully explaining wandering track.

I doubt that there is any way that Ride with GPS 'knows' that I was riding on a track, in the way that my car GPS app (running on the same iPhone) seems to know what road or even lane I'm following.

Pythagoras' theorem is beyond my grasp, but someone else might be able to explain the extra distance travelled down a steep slope to cross a creek and then sharply up again. Let's keep it simple by imagining that one keeps to a straight line without wandering all over the place as I did.

 

[Stefan Mikes]

I am more interested in bad calibration of displays against wheel size.

 

[PatriciaK]

Map My Ride showed me as riding across our harbor on the water (it's a miracle!) once... That's when I first had an inkling it might be off just a smidge .

 

[Stefan Mikes]

Pat:

I was just waiting until a group ride started, remaining stationary for several minutes. See what GPS was doing. My friends were afraid I had ridden into Vistula and drowned

How can we say anything on GPS accuracy if Strava reports non-zero bike speed while stationary?

 

[alphacarina]

Keep in mind that any measurement using the bike's wheel is going to give you the distance the wheel has traveled, rather than the distance you have actually traveled. None of us ride in a perfectly straight line - We wobble a bit as we pedal, left to right, so the tire actually goes farther than it would if we were able to stick to a perfectly straight line. I'm pretty sure even with a properly calibrated wheel, the distance it shows would be greater than if you drove the same route in a properly calibrated 4 wheeled vehicle. Personally, I would trust my GPS speed and distance measurement over anything generated by the bike itself

Don

 

[RunForTheHills]

I have over a thousand GPS activities logged in Strava and thousands more logged in other tools before I started using Strava. Occasionally I get a glitch because I didn't wait long enough to lock in enough satellites or there were tall buildings or heavy tree cover on the route. It might be one activity out of a hundred or less than that. It can be annoying when it happens, but not enough to discard a valuable tool. If I am measuring my activity for fun or exercise purposes, GPS is accurate enough for me. I do have the Garmin speed/distance sensor on a couple of my bikes. I got it to use with an indoor trainer, but I don't take it off the bike outdoors. Similarly, I have the Stryd footpod for running on treadmills. I sometimes use it outdoors and it is more accurate than the GPS, but there is not enough of a difference to make me want to use it on every outdoor run.

 

[Alaskan]

Is not the real issue, how do we get our ebike systems to accurately assess our true speed so that we can get assistance all the way up to the legal limit of the bike's capabilities rather than coming up short because the bike is reading faster than actual speed?

 

[mschwett]

Interesting discussion! First, there’s a conceptual question - are we concerned with the distance between our start and end, or how far we actually rode? @alphacarina points out that a measurement from the wheel rotation is measuring how far the wheel traveled. To me, that’s the distance I’m interested in since it corresponds most directly to effort. I’m not really interested in how far two places are apart by way of roads or trails, I’m interested in how far I (and the motor) moved, wiggles and backtracks and wrong turns included.

GPS accuracy is obviously a subject of much technical discussion, but looking at many route traces with the latest iPhones assisted GPS I see very few of the drift and wiggle type errors that are observed otherwise. In an urban route, it’s quite easy to know if the trace is correct due to the density of landmarks and limitations of route. Ride up to a red light, do a quick 20 meter diameter circle in the street while waiting, then cross on the green directly adjacent to the parking lane or in a bike lane and even the casual eye can see exactly where the trace is wrong or right.

My last ride shows 21.66 miles at the wheel from the bike app, and 21.3 miles via GPS. A close inspection of a known portion of the route against very detailed GIS drawings of the city (I’m an architect) leads me to be believe the actual distance is in the middle. There are clearly sections of the GPS trace which connect points with straight lines that are by the nature of geometry shorter than the actual curving paths, and more than a few wiggles or jogs which are completely omitted.

Does anyone know if distances reported by strava, rideWithGPS, etc, are intended to be horizontal measurements (as land is measured) or following the upward and downward slope of the land? Obviously the wheel based measurements are the latter!

 

[RunForTheHills]

Does anyone know if distances reported by strava, rideWithGPS, etc, are intended to be horizontal measurements (as land is measured) or following the upward and downward slope of the land? Obviously the wheel based measurements are the latter!

GPS does provide elevation data, but it is not very accurate. My Garmin Fenix 5 watch has a built in barometric altimeter and can also measure elevation gain/loss (also not super accurate). Some mapping software will compare the GPS track to elevation maps and calculate an elevation profile.

This is from Garmin:

GPS heights are based on an ellipsoid (a mathematical representation of the earth's shape), while USGS map elevations are based on a vertical datum tied to the geoid (or what is commonly called mean sea level). Basically, these are two different systems, although they have a relationship that has been modeled.

The main source of error has to do with the arrangement of the satellite configurations during fixed determinations. The earth blocks out satellites needed to get a good quality vertical measurement. Once the vertical datum is taken into account, the accuracy permitted by geometry considerations remains less than that of horizontal positions. It is not uncommon for satellite heights to be off from map elevations by +/- 400 ft. Use these values with caution when navigating.

 

[Stefan Mikes]

Is not the real issue, how do we get our ebike systems to accurately assess our true speed so that we can get assistance all the way up to the legal limit of the bike's capabilities rather than coming up short because the bike is reading faster than actual speed?

That is the major point of the discussion I have brought up.

Now, something very interesting. I was on a 70 km ride today. The least expensive of e-bikes I own:

• Reported exactly the same distance ridden as GPS (Strava) did.
• The speed measured by a public radar was exactly the same as shown at the e-bike's speedometer at the same time.

My Giant Trance E+ is notorious of reporting wrong speed and overestimating the distance ridden. Interestingly, the motor cut seems to happen at exactly 25 km/h (when the speedometer reads 26.7 km/h). How comes?

(My Vado reports correct distance and speed only on certified tyres).

 

[Catalyzt]

I'm going to stay on the sidelines for most of this one, because, being an out-the-front-door-after-work / between-clients rider, most of my rides are short-- 10-20 miles-- and at least for now, Google Maps, Strava, Speedometer+, and the Shimano E6100 cycle computer are providing very similar values for average speed, high speed, and distance for these short rides.

The cutoff in motor assistance here in the US is higher, so it's less of an issue, if I were in Europe, I'd want every possible .1 KPH before that cutout! THAT should be based on forward motion, not any number including wheel wobble or whatever. So I feel your pain-- wish you had our limit (though not our other problems!)

My own pet peeve, as a rookie rider, is range estimates on the cycle computer, which are, of course, far more complicated. I think these have limited value and seem to be designed to trigger a simian OCD rage outburst: "You have 10 miles of range remaining! No, sorry, that's 7. Just kidding, now it's 12... whoops, make that 5 miles... 1 mile going up the final hill to your house... oh, look at you, at your front door with 5 miles of range left!"

Just shoot me before I scream like an enraged macaque and start throwing...

Well. You get the idea.

 

[6zfshdb]

GPS accuracy is obviously a subject of much technical discussion, but looking at many route traces with the latest iPhones assisted GPS I see very few of the drift and wiggle type errors that are observed otherwise. In an urban route, it’s quite easy to know if the trace is correct due to the density of landmarks and limitations of route. Ride up to a red light, do a quick 20 meter diameter circle in the street while waiting, then cross on the green directly adjacent to the parking lane or in a bike lane and even the casual eye can see exactly where the trace is wrong or right.

Most consumer grade GPS equipment has built in "snap to" software that aligns the track with the internal map of the road you are riding on. As long as the internal GPS map displays the road you're on, this software will align the track accordingly. This however does not alter the measured distance and the track error still applies.

The "snap to" software will not work unless the GPS knows about the road or street you are riding on. Many bike trails & paths are not mapped and the zig zag track will still appear under these circumstances.

 

[Deleted member 18083]

Does anyone know if distances reported by strava, rideWithGPS, etc, are intended to be horizontal measurements (as land is measured) or following the upward and downward slope of the land? Obviously the wheel based measurements are the latter!

Someone else might be able to explain the extra distance travelled down a steep slope to cross a creek and then sharply up again.

Here's the issue (not that we customarily ride slopes as steep as these!):

 

[BlackHand]

Here's how Strava calculates distance:
https://support.strava.com/hc/en-us/articles/216919487-How-Distance-is-Calculated?mobile_site=true

Short answer is no, Strava does not account for vertical distance.

Clearly all the gps tracking methods have built in error correction. Example: if I look at the speed graph from today's ride in Ride With GPS I can see a couple times that it recorded speeds in excess of 50 mph, but it reports my top speed as 40.0 mph (on a steepish descent).

Fyi: Strava shows 37.8 (using data from my Fitbit) and my bike shows 38.0 mph.

 

[mschwett]

Here's how Strava calculates distance:
https://support.strava.com/hc/en-us/articles/216919487-How-Distance-is-Calculated?mobile_site=true

Short answer is no, Strava does not account for vertical distance.

Clearly all the gps tracking methods have built in error correction. Example: if I look at the speed graph from today's ride in Ride With GPS I can see a couple times that it recorded speeds in excess of 50 mph, but it reports my top speed as 40.0 mph (on a steepish descent).

Fyi: Strava shows 37.8 (using data from my Fitbit) and my bike shows 38.0 mph.

aha! that’s a very useful explanation from the folks at strava. it confirms my theory that the “straight line between GPS points” results in lower than actual distances, and also confirms other’s suspicion that the wheel circumference is typically not correctly accounted for, leading to inaccurate wheel measurements. throw in the wheel measurements being along the diagonal of a slopr in cross section, and the “small” differences are mostly explained.