GPS Distance errors
#26
Senior Member
Join Date: Sep 2007
Location: Far beyond the pale horizon.
Posts: 14,240
Mentioned: 31 Post(s)
Tagged: 0 Thread(s)
Quoted: 4221 Post(s)
Liked 1,319 Times
in
915 Posts
Unfortunately not true.
It is update rate. the GPS only records a position every so many seconds (or fractions of one). It computes distance traveled by taking the distance between the two waypoints and dividing by time to get speed. If the GPS records one waypoint at the entrance to the circle and one at the end, well it doesn't take a math genius to figure out it won't compute an accurate distance.
It is update rate. the GPS only records a position every so many seconds (or fractions of one). It computes distance traveled by taking the distance between the two waypoints and dividing by time to get speed. If the GPS records one waypoint at the entrance to the circle and one at the end, well it doesn't take a math genius to figure out it won't compute an accurate distance.
It would have to be a quite small turning circle or you'd have to be going very fast.
At 3 mph, one moves 4.4 ft per second. A 100 ft cul-de-sac would produce about 50 measurements (1/sec).
At 10 mph, one moves 14.7 ft per second. A 100 ft cul-de-sac would produce about 20 measurements (1/sec).
At 15 mph, one moves 22 ft per second. A 100 ft cul-de-sac would produce about 9 measurements (1/sec).
The average car length is 15 feet, A really-small (not realistic) turning circle would be 3 car lengths or 45 ft. They are more likely 2-3 (at least) times as long.
There should be more than enough points to get a fairly accurate measurement of distance.
The distance might be off but not by that much.
Anyway, we have no idea at all how far off it supposedly is!
The device measures at the same rate but, with the "smart recording", it doesn't save the new point if it's not significantly different from the prior point. That is, if the location or time isn't much different.
Last edited by njkayaker; 05-25-21 at 08:26 AM.
#27
Senior Member
Join Date: Sep 2007
Location: Far beyond the pale horizon.
Posts: 14,240
Mentioned: 31 Post(s)
Tagged: 0 Thread(s)
Quoted: 4221 Post(s)
Liked 1,319 Times
in
915 Posts
If the site or application you use to determine your distance only uses the plots in the log file, then remember that the log files may not have that coordinate that takes you to the furthest distance into the cul-de-sac. As well around sharp corners, it might miss actual corner where you turned.
If you have 1 second recording intervals for your log, then that can still be 20 or 30 feet missed during that one second. Many devices by default are doing more than 1 second interval logging. Some Garmin's by default do what they call "Smart Recording". I've seen enough differences in my logged track with that to know there is nothing smart about it.
If you have 1 second recording intervals for your log, then that can still be 20 or 30 feet missed during that one second. Many devices by default are doing more than 1 second interval logging. Some Garmin's by default do what they call "Smart Recording". I've seen enough differences in my logged track with that to know there is nothing smart about it.
And not all of that distance would be "missed".
https://support.garmin.com/en-US/?fa...mK0P6l20SgpW28
#28
Senior Member
Join Date: May 2014
Posts: 2,549
Bikes: yes
Mentioned: 18 Post(s)
Tagged: 0 Thread(s)
Quoted: 1281 Post(s)
Liked 643 Times
in
329 Posts
I'm kidding mostly. Also your point is valid, cycling-based GPS options are generally in the realm of "good enough."
Likes For ksryder:
#29
I'm good to go!
Join Date: Jul 2017
Location: Mississippi
Posts: 14,800
Bikes: Tarmac Disc Comp Di2 - 2020
Mentioned: 50 Post(s)
Tagged: 0 Thread(s)
Quoted: 6099 Post(s)
Liked 4,730 Times
in
3,260 Posts
How fast do you think the OP is taking these turning circles?
And not all of that distance would be "missed".
https://support.garmin.com/en-US/?fa...mK0P6l20SgpW28
And not all of that distance would be "missed".
https://support.garmin.com/en-US/?fa...mK0P6l20SgpW28
I didn't know I was supposed to give the reason for the most error.
#30
Senior Member
Join Date: Sep 2007
Location: Far beyond the pale horizon.
Posts: 14,240
Mentioned: 31 Post(s)
Tagged: 0 Thread(s)
Quoted: 4221 Post(s)
Liked 1,319 Times
in
915 Posts
If there's a large difference, the thing you talked about isn't likely to be the cause. If so, talking about it wouldn't be relevant to the OP. The "reason for the most error" would be (which is why people need to "give it").
(We still have no idea what the "real" distance is.)
#31
I'm good to go!
Join Date: Jul 2017
Location: Mississippi
Posts: 14,800
Bikes: Tarmac Disc Comp Di2 - 2020
Mentioned: 50 Post(s)
Tagged: 0 Thread(s)
Quoted: 6099 Post(s)
Liked 4,730 Times
in
3,260 Posts
You are exaggerating the amount of distance lost. There shouldn't be much difference (so, the number should be fairly close to what is expected).
If there's a large difference, the thing you talked about isn't likely to be the cause. If so, talking about it wouldn't be relevant to the OP. The "reason for the most error" would be (which is why people need to "give it").
(We still have no idea what the "real" distance is.)
If there's a large difference, the thing you talked about isn't likely to be the cause. If so, talking about it wouldn't be relevant to the OP. The "reason for the most error" would be (which is why people need to "give it").
(We still have no idea what the "real" distance is.)
#32
Senior Member
Join Date: Jun 2015
Location: North Central Wisconsin
Posts: 4,598
Mentioned: 14 Post(s)
Tagged: 0 Thread(s)
Quoted: 2964 Post(s)
Liked 1,163 Times
in
760 Posts
GPS makes small curves into straight lines. Thus the mileage calculated by a GPS will always be less than actual mileage.
If you need to satisfy your mileage OCD then you need a wheel sensor that syncs to your GPS.
Weird that people get so engrossed on how accurate it is.
If you need to satisfy your mileage OCD then you need a wheel sensor that syncs to your GPS.
Weird that people get so engrossed on how accurate it is.
#34
Senior Member
Join Date: Sep 2020
Location: SE Wyoming
Posts: 607
Bikes: 1995 Specialized Rockhopper,1989 Specialized Rock Combo, 2013 Specialized Tarmac Elite
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 213 Post(s)
Liked 588 Times
in
278 Posts
Don't have that problem. I have a Garmin 'Edge 25 and an Apple watch. Over a 20-30 mile ride, the two units are often within .2-.5 mile for distance. The spread seems wider if I stop multiple times. Sometimes lay out a route on Google Earth Pro. Garmin and Apple are usually fairly close to the mileage reported by Google Earth. I'm not concerned over relatively small discrepancies.
Likes For Inusuit:
#35
Senior Member
Join Date: Sep 2007
Location: Far beyond the pale horizon.
Posts: 14,240
Mentioned: 31 Post(s)
Tagged: 0 Thread(s)
Quoted: 4221 Post(s)
Liked 1,319 Times
in
915 Posts
Don't have that problem. I have a Garmin 'Edge 25 and an Apple watch. Over a 20-30 mile ride, the two units are often within .2-.5 mile for distance. The spread seems wider if I stop multiple times. Sometimes lay out a route on Google Earth Pro. Garmin and Apple are usually fairly close to the mileage reported by Google Earth. I'm not concerned over relatively small discrepancies.
The OP was talking about over one mile. We have no idea what the difference the OP saw.
(Cul-de-sacs and turning-circles aren't really a particular problem,)
#36
I'm good to go!
Join Date: Jul 2017
Location: Mississippi
Posts: 14,800
Bikes: Tarmac Disc Comp Di2 - 2020
Mentioned: 50 Post(s)
Tagged: 0 Thread(s)
Quoted: 6099 Post(s)
Liked 4,730 Times
in
3,260 Posts
Wouldn't take but a 100 or so not big deals to add up 1000 feet of error. And that's using only 10 feet of track missed on turns or doubling back on a track. And not all of those not big deals have to be the one that I mentioned.
#37
Senior Member
Join Date: Sep 2007
Location: Far beyond the pale horizon.
Posts: 14,240
Mentioned: 31 Post(s)
Tagged: 0 Thread(s)
Quoted: 4221 Post(s)
Liked 1,319 Times
in
915 Posts
Again, cul-de-sacs don't have any particular problem being measured reasonably accurately with GPS.
Over long distances, it could add up to a "1000 feet of error" but that error would be very small over the total distance (you might not even notice it).
Outside of reception issues** and a really twisty path, the differences between GPS and wheel counter rotation are close.
** This appears to be a real issue for mountain bikers since they are often under heavy tree cover and are doing more "twisty" turns. But you aren't talking about either of these issues.
Last edited by njkayaker; 05-25-21 at 10:31 AM.
#38
I'm good to go!
Join Date: Jul 2017
Location: Mississippi
Posts: 14,800
Bikes: Tarmac Disc Comp Di2 - 2020
Mentioned: 50 Post(s)
Tagged: 0 Thread(s)
Quoted: 6099 Post(s)
Liked 4,730 Times
in
3,260 Posts
Well instead of me and you filling up the thread with our own infatuations of ourselves. Why don't we agree to hush and let the OP fill us in on some of the other stuff. <grin>
Maybe you can go help that guy in the bicycle mechanics forum find all the un-solved threads.
Read this with a grin. I know you give good advice. You just like to argue unimportant things on technicalities to a greater extent.
Maybe you can go help that guy in the bicycle mechanics forum find all the un-solved threads.
Read this with a grin. I know you give good advice. You just like to argue unimportant things on technicalities to a greater extent.
#39
Tragically Ignorant
Join Date: Jun 2018
Location: New England
Posts: 15,613
Bikes: Serotta Atlanta; 1994 Specialized Allez Pro; Giant OCR A1; SOMA Double Cross Disc; 2022 Allez Elite mit der SRAM
Mentioned: 62 Post(s)
Tagged: 0 Thread(s)
Quoted: 8186 Post(s)
Liked 9,094 Times
in
5,053 Posts
Sounds like the sampling rate is almost synchronized with the amount of time it takes you to ride around the circle, so there's two readings that are almost in the same place, making it appear you are going very slow.
#40
climber has-been
Join Date: Dec 2004
Location: Palo Alto, CA
Posts: 7,002
Bikes: Scott Addict R1, Felt Z1
Mentioned: 10 Post(s)
Tagged: 0 Thread(s)
Quoted: 3335 Post(s)
Liked 3,435 Times
in
1,737 Posts
But if the sampling rate is too low, the accuracy of the "distance traveled" and "speed" calculations will be affected by this "high frequency" deviation into and out of the cul-de-sac.
Let's say you want an accurate measure of "distance traveled" and "speed", based solely on GPS measurements.
If, say, a route deviation into a cul-de-sac lasts 10 seconds, the GPS sampling rate must be 5 seconds or shorter. This 2x sampling rate requirement is based on digital signal processing theory called the Nyquist frequency.
Reduce the GPS sampling rate below the Nyquist frequency, and the distance measurement of routes with or without the cul-de-sac deviation will be the same (they will be aliases of each other).
__________________
Ride, Rest, Repeat. ROUVY: terrymorse
Ride, Rest, Repeat. ROUVY: terrymorse
#41
SE Wis
Join Date: Apr 2005
Location: Milwaukee, WI
Posts: 10,516
Bikes: '68 Raleigh Sprite, '02 Raleigh C500, '84 Raleigh Gran Prix, '91 Trek 400, 2013 Novara Randonee, 1990 Trek 970
Mentioned: 40 Post(s)
Tagged: 0 Thread(s)
Quoted: 2731 Post(s)
Liked 3,356 Times
in
2,034 Posts
Just an FYI. I'm not sure what the BF uses to be a "Senior Member" but if you look at some of the join dates it ain't much. So take "Senior member" with a huge grain of salt.
Likes For dedhed:
#42
Senior Member
Join Date: Sep 2007
Location: Far beyond the pale horizon.
Posts: 14,240
Mentioned: 31 Post(s)
Tagged: 0 Thread(s)
Quoted: 4221 Post(s)
Liked 1,319 Times
in
915 Posts
But is the sampling rate too low?
And, if the sampling rate is too low, would it result in a large difference in distance?
For anything I've used, an iPhone or a Garmin, the sampling rate doesn't appear to be too low.
It looks like the sampling rate for an iPhone XS Max is every second. The track walking a 25 ft rectangle didn't look very good.
Last edited by njkayaker; 05-25-21 at 12:54 PM.
#43
Senior Member
Join Date: Sep 2007
Location: Far beyond the pale horizon.
Posts: 14,240
Mentioned: 31 Post(s)
Tagged: 0 Thread(s)
Quoted: 4221 Post(s)
Liked 1,319 Times
in
915 Posts
====================
Here's a better description of the "smart recording". It shows that some detail is lost (but it still isn't clear it would matter in this case).
https://support.garmin.com/en-US/?fa...T27Zr4hxZDvjv5
It looks like the sampling rate for an iPhone XS Max is every second. The track walking a 25 ft rectangle didn't look very good.
Walking a 157 ft rectangle in a parking lot at a slow speed. the iPhone gave 109 ft and 1030+ gave 102 ft.
The inaccuracy is due to location errors. Not sampling rate.
The rectangle might be a worse case than a turning circle. If the 3 cul-de-sacs were measured low by 50ft each, the total would be 150ft over 5280 feet (or 0.3%).
Last edited by njkayaker; 05-25-21 at 01:24 PM.
#44
climber has-been
Join Date: Dec 2004
Location: Palo Alto, CA
Posts: 7,002
Bikes: Scott Addict R1, Felt Z1
Mentioned: 10 Post(s)
Tagged: 0 Thread(s)
Quoted: 3335 Post(s)
Liked 3,435 Times
in
1,737 Posts
For anything I've used, an iPhone or a Garmin, the sampling rate doesn't appear to be too low.
__________________
Ride, Rest, Repeat. ROUVY: terrymorse
Ride, Rest, Repeat. ROUVY: terrymorse
#46
Senior Member
Join Date: Jun 2020
Location: Maryland
Posts: 1,379
Bikes: 1982 Bianchi Sport SX, Rayleigh Tamland 1, Rans V-Rex recumbent, Fuji MTB, 80's Cannondale MTB with BBSHD ebike motor
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 667 Post(s)
Liked 529 Times
in
355 Posts
Reading through the OPs situation again where everything looks good up till the end of the circle, my guess is the unit is trying to use some sort of long term averaging to compute speed and doubling back on oneself is confusing it.
At any rate have to agree with the second post, add a wheel sensor.
#47
Senior Member
Join Date: Jun 2020
Location: Maryland
Posts: 1,379
Bikes: 1982 Bianchi Sport SX, Rayleigh Tamland 1, Rans V-Rex recumbent, Fuji MTB, 80's Cannondale MTB with BBSHD ebike motor
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 667 Post(s)
Liked 529 Times
in
355 Posts
By the way, Nyquist rate has nothing to do with this problem. These are not periodic signals that alias.
#48
Senior Member
I have not had time to test this with road riding so take it with a grain of salt and I feel the difference there will be minimal, unless you live in Colorado. I use a Garmin Edge 500 & a Garmin Instinct watch. Adding the wheel sensor made a huge distance difference when riding mountain bikes with significant altitude gain & loss. On one ride w/o the sensor pared to my Edge it showed 8.4 miles with an average speed of 6.3 mph, the watch pared to the wheel sensor showed 9.99 miles with an average speed of 7.0 mph. This was with 1220 ft elevation gain and the same loss. A 17.3% difference is pretty big to me. The watch reads 2 GPS signals, the Edge 1, they never matched distances but were not real far apart. After that ride I bought 2 more sensors for my other bikes. Get one to match the unit you are using, set it up properly by actually measuring the circumference of your tires at the running pressure to enter them, you won't regret it.
__________________
Two wrongs don't make a right but three lefts do! Unknown
Two wrongs don't make a right but three lefts do! Unknown
#49
climber has-been
Join Date: Dec 2004
Location: Palo Alto, CA
Posts: 7,002
Bikes: Scott Addict R1, Felt Z1
Mentioned: 10 Post(s)
Tagged: 0 Thread(s)
Quoted: 3335 Post(s)
Liked 3,435 Times
in
1,737 Posts
I’m far from an expert on discrete signal processing, but isn’t the Nyquist rate (for periodic signals) a lower bounds for the sampling rate of non-periodic signals?
__________________
Ride, Rest, Repeat. ROUVY: terrymorse
Ride, Rest, Repeat. ROUVY: terrymorse
#50
Senior Member
Join Date: Jun 2020
Location: Maryland
Posts: 1,379
Bikes: 1982 Bianchi Sport SX, Rayleigh Tamland 1, Rans V-Rex recumbent, Fuji MTB, 80's Cannondale MTB with BBSHD ebike motor
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 667 Post(s)
Liked 529 Times
in
355 Posts
None of this explains what short cuts Garmin and others take in their GPS units. They have little processors that need to juggle multiple tasks. Sometimes things get dropped.