fooferdoggie 

I find on my rides where I have a lot of stops before I get going the gps gets pretty off sometimes 1/2 a mile in maybe a mile. then it evens out some by the time the ride is over.

Richard Cranium

Is there any point to what accuracy is "acceptable" as opposed to what is "questionable?"

In practice your chances of riding identical "tracks" for 15, 30 or even a 100 miles is impossible. I remember playing with tape measures and counting wheel revolutions 50 years ago when checking Lucas cyclometers for repeatable results. (do you have a wheel mounted on a hand held fork?)

Since your so interested is measuring things - can you tell me how many miles to travel to make a 700c bicycle wheel rotate a millions times?

For what its worth - I heard a rumor that GPS signals and calculations are in a constant state of estimation and updated and corrected to accuracy as each additional data point signals are received. (this is app dependent)

Trsnrtr

I did an experiment this week for a totally different reason than checking GPS/Garmin accuracy and repeatability. I was actually checking something totally different but I rode a route earlier in the week and saved the route on my Garmin 1030 and then followed the route a couple days later. My distance was 0.12 mile longer on the 23.5 or 23.6 mile route the first time that I rode it. I realize that there may be a zillion variables in play but my point is that I can live with a 0.48 mile difference in 100 miles.

It's been over 10 years since I used a wheel circumference measuring device so my post isn't really germane to the discussion at hand but I found it interesting.

unterhausen

I am trying to think of what the errors in a gps would be. I suppose since the devices record at intervals, there is a bias towards being short. If this matters, you can go to higher resolution sampling. Otherwise, they are probably fairly well distributed. I went on a 200km ride with someone where one of us went a longer distance, but the other climbed more. Edge 800 vs. 810. The difference can seem pretty significant over a distance like that.

Steve B.

There's a limit as to how often the devices will take a position snapshot. If you ride a course with a lot of turns, the distance shows up short.

But as Richard stated "Is there any point to what accuracy is "acceptable" as opposed to what is "questionable?".

I'm perfectly happy with the record I get from my Garmin.

Tourist in MSN

You are correct in the way a GPS does it, it calculates distances between a bunch of specific location points that it has measured, and takes the sum of those distances. You can think of it as a sum of distances between bread crumbs.

GPS units calculate the location of the unit based on the time it takes for a radio signal from at least four different satellites to be sent to the receiver. Speed of radio waves will vary with things like atmospheric conditions, etc. Plus, that is assuming that the GPS unit has downloaded the latest information from the satellites on what their exact orbits are. I am not an electronic engineer, I can't explain much more than that. But I know that on my GPS units I have turned off WAAS correction, by turning that off I have increased error by quite a bit in exchange for increased battery life. I am not going to try to explain WAAS correction when googling it would give you a much better description than I could, if you are interested.

The net effect is that I expect that my GPS will have an accuracy of plus or minus 50 feet in horizontal direction, up to double that in vertical. And it is rare when the error exceeds that. But if I ride around a loop half a dozen times and then plot up a map of my route, I will see distinct plotted tracks that are in slightly different places on the map even if I was careful to always stay on the same exact track on the road. There is some error. And each error is cumulative when you are summing up distances.

Yesterday I did an exercise ride on a trail that I had not ridden on in several years. I was curious, so I looked at the GPS map and profile when I got home. And I was surprised to see a low spot on the elevation profile that I did not recall biking over, so I looked closer at it. And for about a minute my GPS had an elevation that was as low as 684 feet above MSL. But on the USGS topographic map, my GPS was between the 860 and 880 contour lines. Thus, for about a minute and 15 seconds, my GPS had an error in elevation that peaked at about two hundred feet. Not sure why, but I have seen similar errors before, I suspect it was heavy tree cover and poor satellite reception that caused the GPS unit to temporarily lose signal from some of the satellites.

About a decade ago I needed to have some land surveyed. Hired a land surveyor for that. They set up a device to provide differential correction at a benchmark near the town hall. Then at a section corner on the plot of land the surveyors said that there was a 0.07 foot discrepancy between their instrument and the county data for the location of that section corner. My point here is that with the right tools and the right conditions, GPS can be amazingly accurate, in this case the error was less than an inch. But the surveyors certainly were not using the low budget stuff that you or I use when we go out for a bike ride.

I have been using GPS units for two decades, I am a retired Geological Engineer and I worked with maps and location data for every working day before I retired. That probably explains why I almost always have a GPS on my bike, when you have worked with maps every day of your professional career, you develop a desire to know location data. I don't need to know exactly how GPS units work, just need to know how much to trust or not trust them. And with the low budget stuff we use on bikes, you should expect some error.

My GPS unit would be much more accurate if I turned on WAAS correction and turned on Glonass, but for a bike ride I do not need extreme accuracy and would rather extend my battery life.

Sure beats a magnetic compass and paper map.

JohnJ80

Best explanation in one place I've seen:

https://www.bicycling.com/bikes-gear...s-lies-to-you/

Steve B.

Also has inaccuracies. This comment “GPS relies on interpolation almost always means that your recorded distance is longer than your actual distance.”.

With a GPS device measuring using data points (breadcrumbs), the device is cutting corners usually. I typically see the trace on the inside of curves, not the outside, which would add distance to a ride. Anybody that’s ridden a mt. bike on twisty single track and doesn’t use a wheel sensor knows this.

As well, a quick look didn’t indicate to me that the Garmin Edge series devices use WAAS. Might have missed it but it’s not listed in the spec’s of the 530/830 & 1030

njkayaker

This is the full quote. It's staling about what a particular paper said. It's still odd. Talking about "interpolation error" (done later) is odd too. It seems it's "sampling error". I suspect the journal article was talking about something else.

Steve B.

I would also speculate that elevation is the most common error, not distance. But really, anything short of a straight course with no turns is going to have some error. A meaningless error in my opinion, I'm not doing world record attempts so a bit off is just fine.

Seattle Forrest

I wrapped a speed (only) sensor around my rear hub last night. My Garmin calculated the wheel size after I rode about a mile. In auto mode, does it always recalculate the size at the beginning of every ride? If not is there a way to force it? Different air pressures will change that slightly.

Ironically I don't even care about speed or distance, I just want to be able to use the Robert Chung method for testing aerodynamics. But if I'm going to the trouble of using another sensor, I feel like I should use it properly.

njkayaker

https://forums.garmin.com/sports-fit...e-speed-sensor

Seattle Forrest

Great to hear! Thanks @njkayaker!

biker128pedal

I have a Garmin Edge 1030. I found the auto calibration for the wheel sensor off. I use a 25m tape measure on the smooth flat garage floor to measure the distance for one rotation. This is at the tire pressure I run and my weight on the bike. I figure I get within 2 mm. Good enough for me. The standard tables are way way off.

JohnJ80

When you say that the tables are way off, did you use the marked size of the tire or did you measure it? The two are often way off depending on rim width mostly, but other reasons too. I’m wondering how accurate it is if you measure the diameter of the tire when inflated on a 700C rim and then use that number in the table.

njkayaker

You want a normal load on the wheel to include the tire being squished at the bottom.

The number you want is actually the height of the middle of the axle from the ground.

​​​​

Iride01 

How accurately can you measure that? Remember any error will be multiplied by Pi. As well, what is the actual diameter when a rider is riding the bike? It is not so easy to account for the shortened diameter from weight deformation and measure diameter.

To me roll out with the rider in place should be more accurate than anything. For me it's been easy to put a drop of something like paint from a paint pen or white-out (if anyone uses typewriters anymore) on the tire and then ride the bike on a clean surface and measure the distance between two or more spots.

However with my Garmin 500, when I left it in auto calibrate, the figure it came up with was usually surprisingly accurate. It was only a few times that it was off, and I imagine it was due to it not getting a good GPS signal while the calibration was taking place.

As for the tables being off, well they can't make a table for every tire and tread and inflation plus load that is possible. But even if you use the table you'll get reasonable accuracy. Automobiles are not required to have anywhere near the accuracy in their odometers and speedometers that people here are complaining about. Mileage is mainly just a gee-whiz number for boasting. 5% and even 10% isn't going to make a big deal.

JohnJ80

i can measure that quite accurately. And I get all the math and geometry (I’m an engineer).

what I was getting at was selecting the proper table entry based on a measured tire size and wondering if that improves accuracy. I don’t intend on rolling out wheels for circumference or any of that. The few percent variance doesn’t bother me.

Trsnrtr

Rollouts are easy as Pi(e). Sit on bike, squirt a dab of your favorite condiment on the front wheel, I like catsup, ride a couple wheel revs and then measure between the dots on the driveway or wherever. Done.

Tourist in MSN

I just use one wheel revolution with a tape measure, if I had a quarter inch error that would still be less than a third of one percent error. I am careful to keep the wheel rolling in as straight a line as I can. It takes less time to do than it took to type these three sentances.

Iride01 

Never thought about using condiments. Bio-degradable and can be gone after rain or hosing. Since tires are more similar to hotdogs than hamburgers, shouldn't we use mustard? At least those of us from Chicago should. <grin>

I still have marks on my driveway from a rollout done 10 years ago.

Tourist in MSN

I mentioned above that I think my GPS with Glonass turned off and WAAS turned off to save battery life would give a location that is within 50 feet of horizontal location. I have several GPS units, the one I use on my bike is a Garmin 64.

Yesterday I put some fairly old but freshly charged up NiMH AA batteries in it, turned it on and let it sit in one spot until the batteries were empty. I left the screen on, but backlight turned off. Took 14 hours and 26 minutes to run down the batteries and shut off. During that time it moved around over time, the map screen shows that it stayed within a envelope of about 80 by 100 feet during that time, but it logged up a total distance of 1.8 miles of distance traveled.

My point is that you might think that a GPS that is not in motion will always show that it is staying in one spot, but it is constantly calculating location and as the satellites are moving and over time it acquires different satellite signals, there is measurement error, and that error accumulates over time.

And this also likely explains why two riders that are riding the same route together will probably come up with different route data, even if they both used the same brand and model GPS units.

In my case, I am sure if I turned on WAAS and Glonass that it would be much more accurate, but for my biking I am willing to trade accuracy for better battery life. I suspect most of you have WAAS turned on as a default and if your units can receive Glonass signals, that is probably turned on as a default too.

That is why I mentioned above that my wired bike computers that count wheel revolutions will be much more accurate to measure distance traveled.

I am sure that some of you have noticed that your GPS is much more accurate after you have used it for a half hour or more after it has been dormant for several weeks or months. For my "test" above, I used this GPS for a long bike ride a few days earlier, so the GPS had very recent satellite ephemeris and almanac data downloaded before I started the test, thus that should not have added any error.

Seattle Forrest

​​​​​​Garmin's newer GPSs do a better job of this type of error, at least the watches do. Software does a lot of filtering on the numbers the GPSr provides. They use the compass, your speed, etc, to eliminate a lot of side to side jitter.

Trsnrtr

I’ve used mustard, relish, mayo, basically anything that comes out of a squeeze bottle.

Steve B.

As I visualize what you are doing, you are doing a tire circumference measurement using a smear of condiment, whose spread along the tire and subsequently the ground, could add or subtract a few centimetres either way to the circumference, dependent on how much condiment you use and how much it smears. The point being to be able to sit on the bike while rolling along, so as to accurately depress the tire as if you were actually riding.

And this is accurate how ?. I probably should ask, how accurate do you expect this to be ?