Digital doping might make you a Tour de Virtual cycling champion A few weeks ago, Huawei asked if I'd like to take its new P8 flagship smartmobe for a spin. Which is why I found myself with two Android smartphones in my pocket just before I hopped on my bicycle to ride home from work. Before I braved the traffic, I wondered if different phones running the same cycle-tracking app might …
A friend has a Garmin on his bike and, whilst the ride profile graphs are interesting, it is always putting 'corrections' in. We rode over 70 miles one day, and the height at some points was around 200' above the OS map. The correction applied was a 200' cliff/shear in the profile, with a couple of 50' drops elsewhere en route. It was also about 1.5 miles shorter over the whole trip than my on-wheel sensor. It could be that neither is right, of course, though I set mine by riding on it to measure 5 wheel revs of a mark on the rim to counter the tyre squash. Another friend rode using Strava on an iPhone and he got somewhere in between, with no cliffs. The route had about 10 miles under rain-soaked trees and we reckoned that these interfered with the signal enough to give the Garmin problems.
I've also noticed an in-car Garmin nuvi vary the current speed by 4-5mph passing bridges and other obstructions when compared with 0-2mph on a more expensive Becker unit at the same point. YMMV but in my experience with admittedly their lower-cost units, Garmin satnavs have worse reception - it may be different with their £200+ in-car units. I'd like to think all satnav devices are as good as each other, but there are differences in design that make them sensitive to positioning and orientation so the same unit can give varying results depending what's around it.
All journalists should go to science, or at least stats, school. Not only are the discrepancies here effectively negligible (you really didn't think GPS was millimetre perfect, did you? I was actually amazed about how close they were!) but as you've only done each run once you have no idea at this stage whether the variability you are observing is due to the runs being different (different times of day, different 'wiggle' from the precise route, or even just random error) or whether it is significantly different between the devices.
@JHW re. "...due to the runs being different (different times of day, different 'wiggle' from the precise route..."
You missed this?: "Each device scored one of the three pockets in the back of my club jersey. I couldn't turn them on and off simultaneously, but started and stopped each within seconds of the others."
All at once. So, the runs were not different as it was one run, not different times of day but at the same time, not significantly different wiggles.
"You missed this?: "Each device scored one of the three pockets in the back of my club jersey. I couldn't turn them on and off simultaneously, but started and stopped each within seconds of the others." --- JeffyPoooh
You're right, I did miss that. But it's still only one run per device, so whilst my off-the-top-of-my-head suggestions for variance were wrong, I think my complaint about the test method still stands. Several runs with the same three devices are needed to determine what the intrinsic variability of each device is before meaningful comparisons can be made between them. Otherwise you end up with the "Which? effect" where, because one particular washing machine, vacuum cleaner etc. outlasted another, they mistakenly think they can form judgements about longevity of each model.
Precision is another factor. At .1 of a kilometre displayed you might only be 50m difference between measured distance. Eg 50.05 km vs 50.10 km measured displayed as 50.0 and 50.1 though equally it could be 49.96 vs 50.15 km giving. 200 m diff.
Definitely a case of more data required to eliminate rogue results and start estimating actual size of issue.
Why is cycling still allowed to continue in 2015?
Manufacturing an average bicycle results in the emission of approximately 530 pounds of greenhouse gases. If you consider that the average Londoner needs to buy two or three bicycles a year, it actually works out more environmentally friendly to drive a bin lorry everywhere.
"it actually works out more environmentally friendly to drive a bin lorry everywhere."
But a bike has a more favorable manufacturer's fuel economy figure. And EVERYONE knows that's the single only important thing the idiot plebs look for when buying a vehicle.
Unless you count the number of airbags as being important. Pussies.
This post has been deleted by its author
"John, apart from you missing the fact of using all of them simultaneously,"
I was aware of this, and it's still irrelevant.
You could take the same trip multiple times WITH THE SAME DEVICE, and still get that kind of variation.
You might take a different path (say, stay on one side of the path rather than the other), you don't know when the actual GPS fix will come in, meaning you may or may not "cut" corners. The only stable and repeatable distance measurement is the bike odometer. Otherwise, there's a thousand reasons you get discrepancies.
And yes, you as well as others have stated the accuracy is close, almost too close to be right. There is only so much one can do to allow corrections for GPS, but at the end of the day, comparing distances travelled and trying to make sense of that is meaningless when using GPS for this purpose.
I assume the author has never seen GPS accuracy figures...
GPS isn't precise, it's pretty good, but it will suffer from "jitters", you can be sitting still and your location will still be wandering about a few meters in all 3 dimensions according the the GPS data. It's down to the software to try to smooth this out... If you are sitting still, and it's possible to take multiple readings, then averaging is fine and over a few seconds the exact location will become pretty precise... But you're not stationary, you're pedalling along... So if you're riding along a flat road, and the GPS jitter is showing you're going up 1m in one measurement, and down 1m in the next (in reality you'll be luck if it's as small as 1m!), what should the software do? Probably ignore it as jitter... But where should it draw the line of jitter vs you going over humps on a off-road BMX course?
No surprise that different programmers come up with different arbitrary levels, either based on figures off the top of their heads, or maybe some data they collected cycling to work in their own specific environment.
Combine these natural GPS drifts and programming assumptions over a long ride, and figures will be incredibly different. It's amazing they are as close as they were.
The Garmin edge 705 has a barometric altimeter which differ from GPS derived altitudes. (Not sure if those smartphones have a barometric sensors.)
The altitude "cliffs" mentioned above can be sensor blockages or atmospheric changes. If you ride for a while, have an hour pub lunch, then continue riding the conditions could have changed (atmospheric pressure) and it will show as a sudden cliff. Most websites allow "fixing" of elevation using their own topographical maps.
GPS recording is always going to be a "best-guess" effort, the only way to get a precision recording of performance is on a properly timed & measured course. GPS data is also wide open to modification to create fake times & distances. (Search digital EPO)
GPS altitude is a waste of space.
I use my cheap Samsung phone as a tracker for my kayaking training: this is flatwater racing so the *only* change in altitude comes at the occasional lock where I have to get out and run up/down some steps.
I regularly get altitude changes of a couple of hundred metres in a 10km session - it counts every little wobble in the altitude reading and adds them all up. What it disconcerting is that the final altitude is usually different to the start, although they're the same place.
Also, on the rare occasions when I have used the device to trakc a run up a proper hill and back down, the altitude profile is actually quite good.
Large variance in heights though and you might hit 30MPH on a hill, I would be on the brakes at that point unless I could see the road a long way ahead, but 56MPH is in bricking it territory. If you were not really bothered about height then you can get a wired SunDing for £2.99 which will give you Distance Travelled, Average Speed, Top Speed and Journey Time. It also tells you the real time. I suppose the biggest problem with it would be the lack of pose factor and the bloody battery is a replaceable AG13 that lasts more than two years.
"you might hit 30MPH on a hill, I would be on the brakes at that point unless I could see the road a long way ahead"
Bikes are very light, they stop very quickly - even with limited contact with the road. 50MPH feels like flying, it's quite the most exquisite sensation. My last 50MPH experience was curtailed because the BMW convertible in front of me couldn't/wouldn't descend as quickly as my bike and I had to wait for him to zoom off when the road straightened out before I go let got of the brakes and release all that lovely stored energy.
I remember doing 50mph on my pedal bike as a teenager, it was easy...
Only became terrifying once I let go of the sissy bar on my brothers motorbike and had to rely on my own little brake blocks and not just stopping myself via my arm braced against his bike.
I'll admit there was probably about a second of "weeeeeee" enjoyment, just before reality and outright terror set in.
In my experience, endomondo records less distance than Strava.
I now have a Cateye Bluetooth display & cadence/speed sensor, which is different again.
endomondo have stated they will not do altitude based challenges, as altitude data is so inaccurate.
I believe Strava get around altimeter data on road rides by mapping ride to data from OS maps or WHY (as Craig 2 hints).
I use a cheap ZTE Blade Apex 2 with Cateye Cycling app & their Bluetooth display/sensors. Battery life has lasted a 7 hour sportive with spare for another hour or so.
If you always have a smart phone with you, the Cateye is a good idea.
Bluetooth 4.0+ seems to be replacing ANT+ for sensors, so that's the way to go IMO.
Have fun!
"There is SO much wrong with the article."
Yup, accurate measurement can be difficult stuff. Well beyond the lycra-clad numpties of negligible traffic sense plaguing the roads hereabouts these days. (No I didn't miss the statement that the writer doesn't actually wear lycra into the office.)
First, I thought GPS information included altitude, do these generally use GPS altitude or the one out of a database? I see three main potential sources of error here. First, in case of strong GPS signal, does an app use GPS location as-is or assume you're following a road or path? With an accuracy of a few feet, every time the biker went around around a parked car or someone along the bike path or anything, a system using GPS location would record a few extra feet compared to one assuming you're going dead straight along the road or path. Second, GPS filtering. I've seen those devices where the GPS location does just seem to jitter around a bit (well especially some phones). It can be totally stationary, claim a foot or two accuracy, but jump around 5 or 10 feet. The app's handling of this kind of thing could be important, I assume apps all handle this to avoid false movement. But if it overfilters, it could subtract some legitimate movement and reduce the measured distance. Third, weak GPS handling. Does the app try to use these 100 foot accuracy fixes and filter them to estimate location? Dead reckoning until the GPS gets better? Use a road and path database even if it doesn't for stronger GPS? Can it use the accelerometer?
Second... cheating by putting the GPS car? Really? How rugged are these GPSes? I'd just shoot if off the front with a slingshot, and pick it up Mad Max style when I catch up to it. I'd be getting 90MPH speeds in no time hahaha. For a quick sprint, launch the GPS with a trebuchet 8-).
"First, I thought GPS information included altitude"
They do, it's calculated along with the horzontal data, though, it requires four birds as a minimum rather than three just for location. Generally, you get many more than that all of which help with accuracy.
"do these generally use GPS altitude or the one out of a database?"
Always GPS. Elevation data requires a bit more number crunching, which is a challenge when they pick the smallest weakest processor to do the job. Also, good accurate elevation data is expensive, and only has limited coverage. As others have mentioned, using barometric pressure is a by far better way of doing it - if you need altitude to be close.
"With an accuracy of a few feet"
I'm nit picking here, but you only get 3 metre 95% accuracy (typically) with WAAS enabled systems. And since that only operates either on or around the US, most of the world is stuck with non-WAAS, which is about 15 meters 95%, and altitude would be around 23 metres 95%. I need to stress I'm just quoting widely available numbers, and leaving out some gotchas, but these numbers should be acceptable most of the time. From there, it gets really, really complicated really, really fast.
"GPS location does just seem to jitter around a bit"
That's normal, and some units have built-in code to prevent this, so it "looks" clean to the otherwise untrained eye, and a whole lot less annoying on maps... Some handle this gracefully, some not at all.
"Dead reckoning until the GPS gets better?"
I've only seen dead reconing being used in (some) car dash GPSs. But I don't have wide experience with that, so can't be sure.
"Can it use the accelerometer?"
On a bicycle? Don't think so. What would be the point? They're widley used elsewhere, but not on bikes.
"I'd be getting 90MPH speeds in no time hahaha."
Don't laugh, I did that once. One hell of a long and steep downhill, the whole time thinking, "If I blow a tyre now, they'll have to scoop me up with a wet&dry vacuum cleaner.". Scary.
This post has been deleted by its author
This man has spent a very long time looking at this problem, at least as far as climbing data is concerned:
The Strava app doesn't use sensor data, even if the phone has a barometric sensor. GPS altitude data is notoriously inaccurate, so Strava's altitude data is entirely dependent on how accurately they map roads to contours, My completely flat commute along the waterfront here in Wellington used to show a cumulative 200m of climbing, whereas my altimeter showed a total 10M of climbing. When riding with people using the Strava app or other GPS only apps, I often have only half the total ascent they have.
If you want proper digital doping, run the GPX/TCX through http://www.digitalepo.com/
Reception in "difficult places"
I use a Garmin Nüvi 200 car GPS on my bicycle ( with open street map ) and whenever I enter under a tree-leaf-roof the shown speed drops to up to half ( even when all other things are unchanged ) . Also if a significant part of a trip is under a leaf-roof the recorded trip distance turns our to short. . This does not matter as much as one would think, as I use the gadget as autocentered map essentially.
( with the route prepared at home on the PC shown in some bright color ) .
Of course cars are not usually under trees, unlike bicycles, so the device MAY be correct enough for its original purpose.
I mountain bike weekly, and the GPS tracking (saves the map route, yo) always shorts us on everything. It's only so good, and will never improve much.
First, GPS does not track paths, it records points in time. Inevitably, some turning and climbing is lost to straight lines between points. It would seem that more data points wold be more accurate, but then each point has a a certain inaccuracy. More points equals more combined error (noted above how GPS wanders when sitting still).
GPS tracking is marvelously accurate, but has limits that cannot be currently overcome. As such, it should not be used for competition. Just like home scales that say "not legal for commerce".
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
