Which is more accurate? GPS or Wheel magnet

[HJ]

Here is some comparisons between my Garmin Edge 305 and VDO C4 of some recent rides, all in hilly terrain, and difference is hardly nothing to worry about, the last two rides, I have pumped the tyres up to 125lbs max instead of the usual 120lbs and that may have made the difference of the VDO having a slightly higher mileage.:-

Miles

Garmin 32.18
VDO 32.10

Garmin 48.33
VDO 48.19

Garmin 80.15
VDO 79.84

Garmin 31.77
VDO 31.84

Garmin 52.63
VDO 52.91

As for speeds, they both read the same, but I have spent a lot of time setting up the VDO to be as accurate as possible, not just measuring one revolution but as many as I can up the footpath, usually about 5, and working in millimetres rather than tenths of an inch, and marking my tyre with a tiny bit of white paint so it leaves a tiny mark on the footpath to measure instead of the usual valve position, and then remeasuring for a double check. Using the white square markers that are painted on the roads are useful to check your computer if you come across any, as they do not seem so common now, providing that you have tested them previously in your car to determine their distance, and providing your car odometer is correct. Usually they are a mile apart but some can be an odd measurement.

I would put the variation down to the GPS, but it is still an impressive level of accuracy...

 

[GrasB]

Garmin 32.18
VDO 32.10

Garmin 48.33
VDO 48.19

Garmin 80.15
VDO 79.84

Garmin 31.77
VDO 31.84

Garmin 52.63
VDO 52.91

Impressive stuff. On group rides usually we don't get my Garmin Edge 500 & 2 other 305s to agree with each other to this degree of accuracy!

 

[zacklaws]

Just another comparison for a short hilly ride I did yesterday:-

Garmin 19.67, time 1:09:05.69, avg 17.1mph, max speed 34.4mph
VDO 19.70, time 1:09:48 avg 16.93mph, max speed 34.16mph

Never took much notice to all this till this thread started, but yesterday watched both computers carefully. After only about 3 miles there was a discrepancy of 0.03 miles and that never changed during the ride, both computers reset to 0 at start. Speedwise, there is no lag between them and both read the same, unfortunatly the VDO only registers to the nearest 0.5 of a mile but it always tallies with the Garmin. Wheelsize for 700x23, Spesh all condition elite I have set for 2100mm with a tyre pressure of 125lbs. Going by the ride time, the VDO's auto pause must kick in later than the Garmin.

I presume on club ride discrepancies, is there a possibilty of, riders taking different lines round corners, odd little wrong turns, little shortcuts, not zeroing at the start and including a bit of mileage from home and car parks etc whereas both my computers are next to each, variations as such would not effect them.

Also if you take the easy way of sizing your wheel there can be big discrepancies caused for example, for a 700x23 tyre, my Garmin manual states 2096mm, and the VDO manual 2133mm, yet I have measured mine to 2100.

Perhaps if there is discrepancies, then it probably can only be caused by incorrect setup for wheel sizing and never being consistent with tyre pressures as well, perhaps after a club ride, using the calculation mentioned earlier in the thread, that all riders synchronise their computers to one other riders in the group by making corrections to the wheel size just for an experiment.

 

[BenM]

[q]big discrepancies[/q]
Depends on your definition of big of course...

2100mm wheel = 476.2 revolutions/km
2096mm wheel = 477.1 revolutions/km
2133mm wheel = 468.8 revolutions/km

If my maths is right over 10km between the first two sizes there is a 'error' of ~18m or 0.18% and between the second two ~206m or 2%... obviously these are big handfuls numbers.

2% error on a distance measured over Km or miles is acceptable to most people - after all, for general use, why when you measure the run in Km or miles do you need accuracy of meters/yards?

This last question is the crux of the matter isn't it? the difference between want and need. Any calculation you perform on the data from a run will be only accurate within a tolerance and you have to take view as to what level of accuracy you need (not want - the two are different creatures!) for your final use of the data. IMHO 2% is pretty much good enough in most situations and if you work out the actual error bar for the end calcualtion I would be surprised if ANY consumer level equipment can achieve 2% or less overall accuracy.

B.

 

[psmiffy]

Assuming the wheel magnet is calibrated correctly then it is the most "accurate" - I calibrate the wheel magnet by putting a best guess in and then over then a few days correcting it using the GPS mileage (ferries and tunnels notwithstanding) difference over a days cycling is then consistantly less than 1% and most of the time less than 0.5%

Have to remember to turn the GPS off at the end of day as when stationary it will sometimes wander particulaly in the evening when the main saterlites are low to the horizon and DGPS signals can be poor

Speed wise always pretty much spot on (possibly a slight lag due to it having to calculate in the past tense) - Vertical GPS is generally within 5m when topping out passes (normally within the range given between the map and the sign at the top) and better than my wrist watch altimeter (normally calibrated every day when climbing)

 

[GrasB]

I presume on club ride discrepancies, is there a possibilty of, riders taking different lines round corners, odd little wrong turns, little shortcuts, not zeroing at the start and including a bit of mileage from home and car parks etc whereas both my computers are next to each, variations as such would not effect them.

Talking of different lines bellow is a snapshot of what happened on a roundabout I use a lot.

Okay it's out of sync with the satellite image but it's those long stretches of straight travel which were most definitely were I am cornering which interest me. I notice that it does this a fair amount where it'll ignore fairly noticeable changes in direction round corners, I'm thinking of saying all my bikes have an power meter to see if enforced 1s sampling changes this.

 

[Alien8]

More than likely it's due to rubbish software and the inappropriateness of any model used to filter the fixes within the expected dynamics of the application.

On the subject of accumulated distance, the software may be sampling at a higher rate than it saves data to file, hence the accumulated dot-to-dot error may not be as great as expected.