prj71

Who T F cares about the Nyquist rate fercripesakes. Just go ride your bikes and have fun.

njkayaker

The files are getting points every second.

The interval doesn't have to be fixed either. The time just can't be too far apart.

Speed is a separate issue.

You are getting "all sorts of fluctuations" because the measurement interval is short, the speed is relatively low, and because there's a fair amount of variability in the location.

Making a significant change to any one of these would reduce the fluctuation in speed.

Iride01 

njkayaker You might have told me I didn't read the OP well. I do agree with you now that what I was talking about is crap with relation to the issue. <grin>

Pop N Wood

Is that what I was doing.

Interesting.

njkayaker

???

You said you were seeing fluctuations in speed.

This isn't hard to understand.

prj71

You guys are still arguing about this? Really?!

mschwett 

It would actually be really smart if the interval was variable based on speed, attitude, etc.

Assuming a one second interval, it's hard to get more than about a 1% variation from the "straight lines between periodic samples" vs a real bike tire path, say a 10mph u-turn in a 60 foot wide street.

njkayaker

And then there's you, still whinging about it.

njkayaker

This illustrates it better. I'm assuming your "samples" are every 1 second.

I pointed the same thing a couple of days ago.

https://www.bikeforums.net/22075066-post26.html

Real measurements would not be quite as neat. The real locations hop around quite a bit.

I don't think this would help much. The variation is going to be constant. You'd just get more noisy points with increasing the sample rate at a lower speed.

It would be better to have smaller location variation.

mschwett 

right, illustrated at one second with perfect accuracy. on modern GPS devices with a clear line of sight to the sky i actually don't see much of the wandering that used to characterize GPS paths. the accuracy drifts for sure relative to the actual location but the points are quite good relative to each other.

if you were going very fast and turning, more samples would help. looking at actual traces, i see quite a few chamfered 45 degree corners which cut across an area which is not road, even through the air sometimes on a winding mountain road. at 100 feet per second, lots can happen between samples. but typically the smoothness of the path increases as speed goes up, which is why my example is from a low speed u-turn.

it would take a VERY windy and fast path to get even close to an overall 1% variation though; a very steep ascent and decent (10%?) would give you another half a percent at best.

njkayaker

I slowly walked a 150ft rectangle with an iPhone and and an Edge 1030+ mostly in the open. The traces looked nothing like the actual path.

Sure. But, in the real world, cyclists aren't really doing both of these things at the same time.

Also, supporting a higher-sampling rate might make the devices more expensive or consume more power.

mschwett 

i definitely get much cleaner results than that with my iPhone; i can't really find any in the last 20 or 30 rides that are grossly off except for going through tunnels. perhaps the iPhones augmented GPS is doing something, perhaps it's just luck. i've chalked up major differences (more than .5%) in rides to tire inflation changing the accuracy of the wheel sensor, and the GPS always being just a tiny bit short of the real distance thanks to the straight-line-interpolation.



i've

njkayaker

I think I generally get better results. I think part of the issue was because I was going fairly slowly. I was mostly seeing what the sampling rate was.

iPhones might be a bit weird.

https://regex.info/blog/2015-12-03/2651
https://regex.info/blog/2016-01-19/2666

============================

The purpose of the "assisted" part of the "assisted GPS" is to get a location when there is no GPS signal (like indoors) or to "warm up" the GPS.

It also provides a location quickly (much faster than GPS on startup). For many things, the phone cares that it's quick not very accurate.

It takes a while for GPS to stabilize and the more accurate the initial estimate for location, the faster it can stabilize. Once it's stabilized, GPS is very fast and much more accurate than other methods.

prj71

prj71

prj71

JackieYan

I use MMR and Strava on the same Android phone simultaneously. I only have minor errors between the two.
It must be the way the software interprets the GPS data, as they both read the same GPS from my phone.
I also have a Cateye. On a typical 35km ride I find there's just under a 1% difference between the three readings.

wkc

Dear Ron
There’s a lot of talk about technicalities of GPS accuracy which I am finding a little superfluous.

I use MapMyRun on iPhone SE and a VDO bike computer, and I was finding quite large discrepancies. My first thought was - is the circumference in my bike computer correct? My second thought was - is the tyre pressure consistent on every ride? I figured that regardless of the accuracy of MapMyRun and/or phone-based GPS, I’m likely to get discrepancies from those two variables.

I believe calibration as well as those two variables were the cause of the differences. So I set about trying to calibrate my bike computer with the phone GPS app measurements.

Firstly - I inflated my tyres to my usual pressures. And I measured my actual wheel circumference, using a pencil, a tape measure and rolling my bike with me on it on a flat surface. The physical measurement was not quite the same as the “translation table” stated for my wheel specification.

Secondly - I used more care about inflating to same/similar pressures before each ride.

Then, I started to record the distance given for each ride by the bike computer and the phone app in a spreadsheet. And tracked a dozen or rides in this way. The differences were anywhere 2.6% to 5.9% between the two, but showed(proved) that my bike computer was consistently over-stating the distances by those percentages.

I then reduced my wheel circumference setting on my bike computer by 4%; 4% was the “average” difference between the two measurements over those dozen rides.

I started a new spreadsheet and have been tracking about another ten or so rides. So far, with the amended bike circumference setting in my bike computer, the differences between distances recorded by my phone and bike computer is down to 0.74%!!! Yes, a variance of less than 1%.

By my calculation, if I further reduce the wheel circumference setting on my bike computer by 0.74%, I think the two measurements for all future rides would be absolutely negligible.

The only times that the distance measurements between the two devices vary by “a lot”*** are -
1. on routes where I’ve had to stop many times for lights, etc.
2. where I have not checked my tyre pressure after a few days.
3. really short rides - below 5km (to get milk, bread etc ).

*** I consider “large” variances that are markedly different in each spreadsheet, whether they were plus or minus.

I hope the above makes sense and is useful for you in somehow calibrating your phone or navigator GPS for tight turning circles.

For me, I was aiming to get accurate distances and accurate speeds on my bike computer; as I believe that my bike computer after calibration will be way more accurate than any phone GPS.

I am hopeful that if I tweaked my bike computer circumference by the second variance, my next lot of measurements will contain absolutely minuscule differences. (I will try to remember to post a further note here at that stage.)

Please PM me if you would like a link to my Google Sheet or I can send you a copy of the sheet.

(edited for spelling mistakes.)

wkc

PRJ71: I “liked” your GPS route posting... but because I’m a curious cat with time on my hands, I looked on a map and saw that there are no diagonal streets near to Lord Vadar’s head/helmet that you could’ve ridden. The area near Oxford, McKenzie, Linden and Chester Sts do not have any diagonal streets.

spelger

why do you believe that the bike computer is over stating the distances? i am curious what your proof is. if the bike computer has the correct information to work with then if anything GPS should be understating distance traveled.

wkc

Well...I didn’t assume anything to start with. I just wrote down the distance travelled as shown on my bike computer and on MMR for each trip. That first list showed the BC recording bigger mileage for most trips than MMR. From that, I concluded that the wheel circumference in my BC was too big and thus reduced it for my second list.

The first list could have gone either way. I had no assumptions at the very start. I was curious as to what each recorded and whether or not there was a “consistent” error. And after a dozen rides, the BC was consistently showing higher/bigger distances than MMR.

My BC wheel circumference was set to 2150 at the start.

Observing the consistent overstatements by BC, I reduced my BC wheel circumference to 2110. If I were to use the actual difference of 4%, I should’ve reset my BC circumference to 2065 (after rounding). But as I wanted to “creep up” on the accuracy (in case something was amiss), I did not reduce BC circumference as aggressively.
2150 x (1-0.04) = 2064

Having reduced my BC to 2110, I am still getting BC distances greater than shown on MMR. But on this current/second list, the difference is averaging only 0.74%. This suggests I should reduce my BC circumference down to 2094 (or more likely 2095 or 2100) and start a new/third list.
2110 x (1-0.0074) = 2094

I have to admit that my main assumption is that a “properly calibrated” physical device on a physical rotating wheel on a physical road surface, will always be more accurate than a “thin air” device pointing to some satellite or three floating thousands of km/mi above the physical space being measured. Besides, it is generally accepted that phone GPS is no where near the accuracy of military grade GPS. (Please refer to above posts for discussion about accuracy or otherwise of GPS devices; which I do not believe helps in calibrating my BC. )

(edited for typos)

njkayaker

For a curvy path, the GPS (measuring once per second) will measure a distance shorter than a wheel sensor (measuring about 3 times per second) using the correct circumference.

All you have done with this effort is make the wheel sensor less accurate.

The phone GPS is less accurate but you modified the wheel sensor calibration so it matches the less accurate distance. That doesn't make much sense.

wkc

Spelger, you correctly say “if the bike computer has the correct information”. IF!!!
I have always been a little suspicious about using “standard” conversion tables for translating wheel “size” to wheel circumference. (eg 700 x 23C = 25-622 = 2096)

My general knowledge tells me that standard wheel size will only equal distance travelled for any given tyre pressure or even tyre wear. A flatter, less inflated 700x23C will have a smaller circumference compared to a well/highly inflated 700x23C. Tyre wear and even manufacturing tolerances (exact amount of rubber all the way around a tyre) would affect circumference in a similar, albeit to a smaller extent.

Therefore, using 2096 in a BC assumes that your actual wheel on your bike rolls 2096mm for every revolution. Thus, using standard conversion table circumferences is assuming your BC is using the “correct” information. To me, that is not necessarily a good assumption and necessitated some sort of recording and comparison.

wkc

NJK, you could be right with your two comments. But thinking about it, at least my eventual outcome is some sort of consistency between my GPS using MMR and my BC, no???

wkc

When you say “curvy path”, yes, I agree if there were lots of relatively “small” curves. But as I was measuring rides between 2km and 30+km with straights and curves, I would think that the impact of “curves” over those distances would be somewhat negligible when compared to the circumference of my tyre.

I think you are absolutely correct if I was riding “short” and “curvy” distances every time. But that is not the case.

I think I will start my third list next week and see what the outcome is. Maybe I could compare my third list measurements/distances with measurements on Google Maps or some other map to check if gentle* curves make any difference over “larger” distances.
* gentle road curves vs concentric or tight circles and curves.