Innacuracies : Strava vs cycle computer

funkyg

I’ve been using Strava since I got my bike six months ago, and it’s been brilliant for tracking my improvements (or not!) over the routes I do.
I did have the iPhone mounted to my stem but it looked a bit clunky and I also wanted something with current speed/cadence showing so I got myself a wireless cycle computer (Boardman) and now tuck the phone in my jersey pocket.
When setting the computer up, I used the default settings for my 700 x 23mm wheel/tyre but found that the distance travelled on that differed slightly from what Strava told me.
So I rolled the bike to measure the exact circumference of the wheel and it tallied with the standard setting that had already been set.

I forget which way round but one shows a lesser distance and therefore a higher average speed than the other.

So which would you believe, Strava on the iPhone or a dedicated cycle computer?
Cheers.

gethinceri

Computer.

larkim

A well calibrated cycle computer measuring the distance travelled by a wheel which is accurately measured and takes into account all necessary factors (e.g. tyre pressures) should be more accurate than any GPS reading, as the GPS is measuring the distance from one approximate location +/- a couple of metres to another approximate location +/- a couple of metres.

But assuming a decent "view" of the sky from the mounting position, a GPS will always be materially correct, enough for the purposes of tracking how far you've cycled.

If you cycle down a roman road 5 miles long in a straight line, a GPS would normally measure the course long, as it will naturally zig-zag a little from side to side. If you were to cycle a perfect circle, I would probably expect the GPS to measure slightly short as it would cut the corners and assume you'd travelled in a straight line between the two readings it takes every second or so. On a normal route, GPS is normally more or less OK.

CiB

GPS misses bits off, like taking a 90 left under tree cover or going under one of those disused railway bridges where the road is parallel on one side of the old line then parallel the other and requires a sharp left to go under it then a sharp right out the other side, and as it joins each plotted route point with a straight line it'll show that you did a direct line in the few seconds between each reading when you view the route on the web site. It's not a lot but it accumulates esp on very cloudy days in rural locations.

cougie

Does it matter that much ? Choose one and then you will be consistent.

larkim

CiB, you're right, but then it has the opposite effect on a straight road too - wiggling from side to side adding distance. Cloud cover doesn't meaningfully affect GPS reception, but tree cover does, as does shadows caused by buildings.

There's a good read on the subject mentioned in this link - http://forums.groundspeak.com/GC/index. ... opic=66303 - but in particular I like this summary:-

GPS technology is so impervious to normal weather effects, it is now being used to fly our missiles, our fighter planes, guide our ground troops, and drive remote hummers equipped for mine detection and removal. Iraq is all the evidence you should need.

Initialised

larkim wrote:

CiB, you're right, but then it has the opposite effect on a straight road too - wiggling from side to side adding distance. Cloud cover doesn't meaningfully affect GPS reception, but tree cover does, as does shadows caused by buildings.

There's a good read on the subject mentioned in this link - http://forums.groundspeak.com/GC/index. ... opic=66303 - but in particular I like this summary:-

GPS technology is so impervious to normal weather effects, it is now being used to fly our missiles, our fighter planes, guide our ground troops, and drive remote hummers equipped for mine detection and removal. Iraq is all the evidence you should need.

Yes but consumer grade GPS is considerably less accurate than military hardware, and for now at least a dedicated GPU unit like a Garmin is more accurate than the GPS on a smartphone.

larkim

DIsagree (about dedicated GPSr vs GPS on a smartphone).

When you've used a smartphone to get you to within 2m of a hidden tupperware box under tree cover as I have enough times (www.geocaching.com) you'll discover there is no meaningful difference in the accuracy.

GPS tech "knows" how accurate it is - its a function of the number of signals received etc - so it's trivial to get a smartphone to display the level of accuracy that it is achieving (download one of many apps to do this), and having compared a variety of phones that I've had to dedicated GPSrs that I've owned, I can say with confidence that there is no difference, neither is "better".

I have heard reports that *some* smartphones in the past were poorly designed so that in marginal circumstances (e.g. under tree cover) they might not be able to isolate a weaker signal that a decent Garmin could. But it's definitely wrong to state that smartphones are worse - in most circumstances they are identical.

Where a dedicated GPSr is undoubtedly better is that it will generally be built to withstand outdoor conditions and because all it is doing is GPS, they usually have a better battery life.

http://www.gps.gov/systems/gps/performa ... difference gives a decent description of the military vs civilian differences, noting that some augmentation systems that civilian systems use (I presume they are thinking of A-GPS which most phones use) can give higher precision for consumer GPS than military!

johnmiosh

When my train stopped earlier today in unfamiliar treelined territory, I used the smartphone to see where I was. It gave me a GPS location accurate to 1.6 km. When we started moving again and emerged from the cutting, it gave me a cell based reading with an accuracy of 150m. I would be disappointed with either level of accuracy on the Garmin.

funkyg

cougie wrote:

Does it matter that much ? Choose one and then you will be consistent.

No it doesn't matter, it'd just be nice for them to be the same. Like I said, the wireless comp was only really for a live sample of what I'm doing at that time and I clear the average/max values every ride anyway and use Strava to log because it gets uploaded automatically.

Thanks for the info though guys.

larkim

johnmiosh wrote:

When my train stopped earlier today in unfamiliar treelined territory, I used the smartphone to see where I was. It gave me a GPS location accurate to 1.6 km. When we started moving again and emerged from the cutting, it gave me a cell based reading with an accuracy of 150m. I would be disappointed with either level of accuracy on the Garmin.

Inside a train with limited vision of the sky in any direction other than to either side, I'd be surprised if a garmin got a lock quickly, if at all. I've used my Garmin in similar circumstances and failed to get a lock at all, where as I was sat at work 6 foot from the only window in my Victorian buiding office today and it got a lock within 5 seconds with an accuracy of 46 feet. In similar circumstances I would never expect a garmin to get a lock.

kajjal

The accuracey of GPS depends on how often it updates as well as how accurately it calculates your position. My Garmin 500 updates my position once a second which is fine for road riding as well as mountain biking. The Garmin 200 updates a lot less often so it OK for road riding but noticeably inaccurate if mountain biking on twisting trails. A properly calibrated bike computer using a magnet and sensor is more accurate as it continually updates and won't lose its signal. Downside is you get less info out of it.

wongataa

The Strava app only logs points every 4 seconds or so. That will degrade the accuracy of any logs quite a bit, especially if the track is twisty. This sort of accuracy also is a cause of some of the dodgy segment times on Strava. Strava doesn't interpolate between recorded data points which helps give dodgy times for shorter segments recorded on phones.

A decent GPS device or bike computer updates every second. A bike computer with a properly set up wheel size will most likely give a different recorded distance than a GPS device but as long as the GPS data is good the difference should be pretty small and not worth worrying about. I would not trust the Strava app on a phone to provide decent GPS data though.

Bozman

I had my Garmin and phone running while I was out on a ride a couple of weeks ago and when I got in I downloaded the Garmin to Strava, all had the same distance and average speed but the ascent differed wildly, 3960' on the phone, 3649' on the Garmin and roughly 3150' on Strava.

cougie

Altitude is bobbins on strava.

My iPhone seems just as accurate for strava as my garmin. The garmin has to be mounted on the bike facing the sky to work but my iPhone works great in a pocket.

harry-s

I'd go for the computer every time.
I've never figured out why the Garmins don't have an option with their magnet/wheel sensor where GPS is the default, but switching to the wheel sensor when the GPS signal is poor. And I'd take a lot of the marketing stuff with a pinch of salt though when it comes to accuracy, - I wrote this a little while ago if anyone's interested:
http://www.kingstonwheelers.co.uk/kwccf ... hlight=gps
I think most of it still stands, but it's worth mentioning that the new Garmin has GLONASS capability.

wongataa

Harry-S wrote:

I'd go for the computer every time.
I've never figured out why the Garmins don't have an option with their magnet/wheel sensor where GPS is the default, but switching to the wheel sensor when the GPS signal is poor.

Because a properly set wheel size with speed sensor is always more accurate than GPS, and it would be hard to program the Garmin to determine when GPS accuracy was too low and switch. If the wheel sensor is more accurate why not use it all the time when available?

ben@31

Hopefully future Garmins will allow you to choose which sat nag system you want to use.

The Russian GLONASS is accurate upto 2.8m

Some modern receivers are able to use both GLONASS and GPS satellites together, providing greatly improved coverage in urban canyons and giving a very fast time to fix due to over 50 satellites being available. In indoor, urban canyon or mountainous areas, accuracy can be greatly improved over using GPS alone. For using both navigation systems simultaneously, precisions of GLONASS/GPS navigation definitions were 2.37—4.65 m

The European Galileo (when up and running) will be accurate <1m and upto 30cm. Yes centimetres !

dilatory

Always find the Strava iPhone app to be out of whack. I ride with a 500 and sometimes ride with a chap who uses his iPhone. He can come out even up to a few seconds ahead on a segment, when he was on my wheel all the way through it. I also know another chap who is fast as hell and makes it a bit of a hobby to take every KOM he can. He uses his 500 and a phone running at the same time because 9/10 he'll be faster on the phone and take the KOM.

harry-s

The Russian GLONASS is accurate upto 2.8m

Nope, it's not. And neither will Galileo be, - read my link above.

For using both navigation systems simultaneously, precisions of GLONASS/GPS navigation definitions were 2.37—4.65 m

Wrong again I'm afraid. Out of interest ben, where did you get those figures?

CiB

Can't read your link as it requires a log in.

larkim

That forum post of yours is only available to registered users on that site I think, and I don't want yet another forum registration - can you re-post the text here?

My Moto G phone has GPS and GLONASS, matching the Garmins.

wongataa

dilatory wrote:

Always find the Strava iPhone app to be out of whack. I ride with a 500 and sometimes ride with a chap who uses his iPhone. He can come out even up to a few seconds ahead on a segment, when he was on my wheel all the way through it. I also know another chap who is fast as hell and makes it a bit of a hobby to take every KOM he can. He uses his 500 and a phone running at the same time because 9/10 he'll be faster on the phone and take the KOM.

That's because the Strava app only logs the position very 4 seconds or so and the Strava segment matching system does not interpolate between data points. This means a segment recorded on a phone can be a good bit shorter and therefore faster than the same segment recorded with a device updating every second. Over long segments this is less of a problem but it is a big problem with shorter segments.

navrig2

johnmiosh wrote:

When my train stopped earlier today in unfamiliar treelined territory, I used the smartphone to see where I was. It gave me a GPS location accurate to 1.6 km. When we started moving again and emerged from the cutting, it gave me a cell based reading with an accuracy of 150m. I would be disappointed with either level of accuracy on the Garmin.

The sensitivity of a GPS inside a train varies between trains. Cross country trains seem to block all GPS signals and downgrade phone signals whilst East Coast allow GPS signals to be picked up relatively accurately.

Cross Country trains have a coating on the glass which, I presume, interferes with air borne signals.

larkim

wongataa wrote:

That's because the Strava app only logs the position very 4 seconds or so and the Strava segment matching system does not interpolate between data points. This means a segment recorded on a phone can be a good bit shorter and therefore faster than the same segment recorded with a device updating every second. Over long segments this is less of a problem but it is a big problem with shorter segments.

I don't think this is true - I've just exported a GPX file from my most recent ride on Strava - data points are there for every second of the ride, with occasional gaps (presumably due to lost signal).

I can't answer for iPhones as I've never had one, but on all of the Android devices I've used and all of the apps I've used to record progress I've always seen GPS traces recording once per second so long as the phone is set up to use the most resource intensive GPS settings.

Bozman

I saw an 800' difference on a 50 mile ride at the wknd, a Garmin 810 recorded 3400+' and a 500 recorded 2490' when the ride was downloaded to Strava. Slightly confusing for the two riders.

Edit

Sorry.......ascent.

larkim

Elevation? That's another kettle of fish. GPS generally is poorer at measuring altitude anyway.

And in any event, strava (usually) overrides GPS elevation data with the data about elevation it gets from the topographic maps it uses instead. They are not always that precise (there is a bizarre "lump" which shows around where I live which Strava thinks is a downhill, but most certainly isn't, but when you see the contour lines on the mapping you can see, sort of, why it thinks the way it does).

larkim

https://support.garmin.com/support/sear ... 0000000%7D

harry-s

Aah, sorry about that, didn't realise it awas on the 'members only' page. It's a bit long, but here you go...

GPS gadgets seem to be a bit of a hot topic these days, and no doubt letters to Santa will be being sent up a few Wheelers chimneys, with the authors promising to be good cyclists and wishing for a Garmin. So it seems like a good idea to put a few facts down about GPS, for purposes of either inducing excruciating boredom, mild interest, or maybe just to give fellow Wheelers a few more tools with which to cut through the marketing fog that the manufacturers pump out like dry ice on a prog rock comeback tour.

First off, although I work with GPS for a living, it's scientific and/or precise GPS that I use on a daily basis, - what I'm trying to do here is give you a bit of background info, - I don't know a great deal about recreational GPS, so I can't tell you what button to press on your new Hokey Cokey 2000 so that it beeps when you pass a cafe or flashes when you need a change of shorts.

Anyway, for what it's worth....

GPS as you no doubt know, stands for Global Positioning System, - so you are allowed to snort when you hear the media stutteringly refer to it as the GPS system, - I know I do. The satellites that make up the GPS constellation are owned and operated by the United States Air Force, under the Department of Defence (and there's nothing to stop them turning the satellites off at any minute should the inclination take them!), and orbit the earth and Kingston at an altitude of about 12,600 miles. That's a long way, even by Audax standards. The Space Shuttle, by comparison, hauls its lardy backside up to the space station at the less dizzying height of around 170 miles, a distance achievable perhaps on a particularly mazy club run. The system was declared operational in 1993, - although the US military had been using it for about 10 years previously, mostly for Naval manoeuvres, I have no information about any military Sunday runs at that time.

Once the system was up and running, they were a tad taken aback by how good it was, they'd been expecting something in the range of +/- 400m but, holy mackerel, with the wind in the right direction it turned out to be good for about 100m. The top brass decided this would offer strategic advantages to foreigners in general, and probably the God-damn Rooskies in particular, so induced a random error to the GPS signals that only military hardware could by-pass, thus degrading the precision for Joe and Jane Public. Luckily for us, one of President Bills last acts, as soon as he got back from the dry cleaners (or perhaps, tragically delaying his visit there in the first place), was to ensure the error software, known as SA - Selective Availability (a bit like Alex Fergussons attitude to international games) was turned off. Anyone who was using any hand held units around then would have noticed a significant increase in accuracy. I did, and it was possible to get back home from the pub even when my vision was severely compromised, up until then I was just as likely to end up in Mrs Higginbothams down the road. Or even back at the pub.

To start with, the system barely ran to double figures, but additions and replacements mean there are now 24 satellites up there, plus a few spares, you can check out their vital statistics here: http://msl.jpl.nasa.gov/Programs/gps.html. It's not unusual for one or more to be 'down' for service work, not literally obviously, that would be a real faff, but usually remote software tweaking. Each satt has an orbit that takes them 12 hours to complete (11 hours 58 minutes for the pedants out there), the constellation however, ie the pattern the satellites make above our heads varies throughout the day, which sort of brings us round to the horny (or is it thorny? - I'm never sure which...) topic of accuracy. I'll try and ease into this one gently.

Before I go there though, it's worth mentioning that the Russians have their own GPS, known as GLONASS. Due, perhaps to Boris's Vodka bills, it fell into disrepair during the nineties from lack of funding. It is however, now back up and running (sadly, unlike Boris, who I think lies pickled in Moscow somewhere), and they are regularly adding to it. The GLONASS signals are available to the scientific/professional community, and you too if you want to pay for it, but none of the recreational units utilize it (as far as I know). This does mean though, that if Garmin and all the rest develop or buy software to decode the GLONASS signals, there will be another hike in the accuracy achievable. And it will be a sad farewell to Mrs Higginbotham.

The satts or SVs are basically accurate clocks, each have four atomic clocks on board, and they broadcast time signals on several frequencies, - recreational GPS, including your Garmin, uses only one of these frequencies. The signals are very weak, by terrestrial standards, and just about anything will block them, - twigs, light foliage or a nice pair of Oakleys, for example. Dense cloud cover is not a problem. By comparing the time signal broadcast from the SV, with the time in your GPS receiver (your bike/running unit) then the distance from the satellite can be calculated, - it's a straightforward calculation (but not that straightforward, as we'll see later!) using time and the speed of the signal from the SV, - the speed of light, your distance from the SV at that instant can be calculated. Obviously, the clocks have to be pretty good, - at the speed of light 20m is covered in fractions of a millionth of a second, quicker even than Steve Saunders, - Which is ok with the atomic clocks (they're good for about a billionth of a second per day) but the quartz clock in your bike unit is the lanterne rouge of timekeeping in comparison. Even good old Einstein has to be invoked in the calculations as there are relativistic effects on the signals journey from the satt to you. Using more than one satellite to work out distances to the receiver, your position in space can be calculated by intersection (similar to triangulation). It takes a bare minimum of 4 SVs to get a position. Further software in your unit works out where in space you are on the earths surface with regard to whatever local mapping system you're using, - Ordnance Survey Grid for us. Personally, after all that, I'm severely impressed that these little units get a fix at all, they're a great testament to the microchip and software authors.

At this point, you might be wondering how do the satts know where they are, - it's all well and good us knowing where we are relative to the SVs, but what about them? Fortunately (thank you Sir Isaac!), the orbits are fairly predictable, this is why your unit probably finds them pretty quickly when you use it regularly, but takes a long time if you haven't used it for a while, -it can predict where they'll be from the last time it locked on and knows where to look, - but only for so long. To keep things sweet, there are a few base stations on the earth, that track the satellites, and once a day it gives them an update on their real position, rather than the position they think they're at, in a sort of "I know you think you're here mate, but you're not, you're there. Now toodle off and I'll see you next time round". The differences are only millimetres, but still important, - NASA logs the exact orbits, known as 'precise ephemeris' and they can be downloaded to use for any high accuracy work. Recreational units use the 'broadcast ephemeris', ie where the satt believes itself to be when it sends out it's time signal.

The quality of your fix is directly proportional to the amount of satellites your unit can 'see', and the strength of the geometry of the constellation over your head. If, for example, only 8 SVs are visible, but they are mostly in the same part of the sky, then the fix won't be as good as if they were evenly spaced around you. It's unusual, even if the horizon were perfect, at sea for instance, that more than 12 satts would be visible. Mostly you'll be looking up at between 6 to maybe 12 satts, probably averaging 8 through the day. Also, the signal received from an SV low in the sky is not as good as that from a position higher up, - the signal has more atmosphere to travel through and is degraded more. Factor in a compromised horizon, ie hills, tree lined roads (sound familiar?) buildings etc and it will be less. And don't forget the sweating, wheezing cyclist hunched swearing over the bars, where the unit is probably mounted, - depending on the size of the cyclist, that could be a considerable portion of the sky blotted out. And by and large, most of the SVs are in the southern sky, with only a few of them making an occasional foray up north, so if you are wheezing and swearing in a northerly direction, your generous bulk (or waif like girth) will be creating a satellite shadow.

It's probably fair to say that with the wind in the right direction todays units are accurate to around 25m (in plan, - forget about height with recreational GPS), at times it will be much worse. The factors affecting this are (approximately, as they'll vary through the day):

2.5 metres, Clock and Orbit errors
7.5 metres, Troposphere and ionospheric interference
1.0 metres, general receiver noise
1.5 metres, Multipath (signals bouncing off surfaces such as buildings before reaching the receiver)
7.5 metres, Range errors (quality of intersections, number of SVs)
1.5 metres, constellation geometry weaknesses

Giving us, ooh, about 21.5m.

Sadly, most of this is all controlled in a black box sort of way, and I don't think there's much a user can do to help things. But I think it helps to be aware of what's going on, and that maybe your log should be treated with a little caution if you were in an area of reflective buildings, tree lined roads etc, - or at a time of day when there are not many SVs around. I'm told that recreational units use a lot of (very educated) guesswork when the signal or position quality isn't good, ie rather than just stop recording they will extrapolate their last position, speed and direction. The unit also locks onto the mapping details, ie if the fix says you're travelling parallel to a road, but you appear to be off in the field by 50m or so, it will assume, quite reasonably, that you are travelling on the road and will display just that, as opposed to it's calculated, more rural fix. When I've used recreational GPS in the past for navigation, I've used a separate antenna/receiver combination, - the antenna is about the size of a matchbox and can be tucked into your hatband, with a bluetooth connection to the PDA based receiver. This arrangement would work well on a bike as it takes the cyclist out of the horizon, whether it's available with Garmin type units or not I don't know, - probably not, I'd imagine it's a lot easier to market a single unit than a pair of components.

Err, that's about it. I suppose I better get me coat...

Mikey23

My gob is well and truly smacked... Thank you for that detailed explanation... All things being equal im amazed that they work at all

bobmcstuff

Mikey23 wrote:

My gob is well and truly smacked... Thank you for that detailed explanation... All things being equal im amazed that they work at all

Indeed - one of the few applications where engineers have had to include both special and general relativity in their calculations