I want to do some Chung aero testing and having an accurate elevation change would be very helpful.
I have an iPhone GPS app and I notice that the elevation at my starting and ending points varies ± 5-7 ft and then settles down to only drift maybe 2-3 ft. I don’t exactly know the details of how GPS works and what factors affect the accuracy, but if I get something like 10 or 20 updates and average them, will this be an accurate measure of the true elevation? If there would be a bias in this measure, would the same bias apply at my second location 600m away?
I actually looked a LOT into this last summer when I spent a couple weeks trying to make an automated rolldown tester using an algorithm that RChung posted the year before.
To answer your questions from my memory:
-Elevation may stop drifting over 10 seconds or so, but over 30-60 second timescales, I found that it’d suddenly jump - this is probably because it’ll suddenly gain or lose satellites, and so the elevation fix will shift substantially.
-The places that get the most accurate GPS reception (open areas) are also the places you don’t want to do aero testing (because they’ll be exposed to wind)
That said, you can probably accurately determine the height of a hill via surveying. Since the accelerometer in a phone is pretty accurate and nonbiased over longer timescales and GPS location is accurate and nonbiased over longer timescales, it should be possible to accurately survey a hill, as long as you’ve got an object to point at at the top. Since all you need to know is the total drop of the hill, that should work.
I imagine there’s already surveying apps out there, but I figure the procedure would look like this:
-Mark the spot you’ll be sighting at via GPS
-Walk 100m, point the phone at the target (it’d probably show you the camera image at max zoom so that you can center on the target). Hold it there for 10 seconds while the accelerometer figures out the angle
-Repeat until you’re at the bottom of the hill
-Now you’ve got a complete and fairly accurate elevation profile of the hill. You won’t have the absolute elevation (since we’re assuming GPS elevation sucks), but you’ll have the drop from a fixed point, which should be just as good.
While I was digging around for my old elevation research data, I did find a hurting diagram I made last year. The labels aren’t quite right (you would know the horizontal distance from GPS, and a bike computer measurement wouldn’t be useful).
I had a bit of standing-around GPS elevation data last year, but I can’t find it anymore. It was probably on my old, dead, phone.
This isn’t quite what the OP is asking for, but here’s an example of the lack of repeatability with GPS elevation as I went down a hill a ton of times. Notice especially how some runs have phantom hills or flats in them.
Most GPS units are set up to be most accurate in position. If the solution to the position problem means that the altitude is a bit wrong then so be it. For most applications on land this is preferable.
Many units use barometric altimetry to augment the GPS to detect altitude change. This is good, but suffers from drift as air pressure changes over time. I’m looking into ways of correcting for batometric drift, but it’s early days for me.
Iirc GC will apply map corrections to GPS positions.
I want to do some Chung aero testing and having an accurate elevation change would be very helpful.
I have an iPhone GPS app and I notice that the elevation at my starting and ending points varies ± 5-7 ft and then settles down to only drift maybe 2-3 ft. I don’t exactly know the details of how GPS works and what factors affect the accuracy, but if I get something like 10 or 20 updates and average them, will this be an accurate measure of the true elevation? If there would be a bias in this measure, would the same bias apply at my second location 600m away?
If you have the time to take multiple measurements over a period of hours (when the satellite constellation is different each time) then you would eventually get a good approximation to the true elevation. It is probably easier and faster to consult your local highway department or utility company and get precise location information for your hill of interest.
GPS accuracy in the vertical direction is poor because the geometry of your line-of-sight to the satellites is generally much poorer in that direction than it is horizontally.
Using a phone as a surveying instrument would make pretty steep demands on its capabilities and you’d have to find a viewfinder app that included a reticle overlay (or make one) so that you could exactly center your object of interest in the viewfinder every time. As you get further away from the object, it becomes easier to have the phone just a bit off-angle. But, if your object is 500 feet away then a one degree angle error works out to almost 10 foot elevation error; and this error is independent for each sample (picture or angle measurement) you take. Also, your phone accelerometer’s measurement capability may start to become a factor.
About a year ago I downloaded an Android app onto my phone called “Geosurvey Transit” and played around with it for the same purpose…the results, especially with the small screen on my phone, were less than satisfactory. That app doesn’t appear to be in the Google Play store anymore though…
That said, now that I have an iPad, I think I might need to download THIS:
You really can’t. Altitude GPS error is much greater than lat/lon error. That’s why we have baro altimeters in GPS gadgets.
To measure elevation really precisely you need a calibrated barometric altimeter: contact ATIS at the nearest airfield and punch in the atmo pressure into the watch… or set the altitude at a known elevation and hope pressure doesn’t change too much during your ride. Neither is an option on something like the Garmin 910, and it should be, or else the baro altimeter could be way off.
Another way is to overlay your route onto an accurate topo map and then the computer tells you exactly how much elevation that was. We have accurate topo maps of the entire world, but to my knowledge this approach hasn’t made it into sports yet.
I imagine there’s already surveying apps out there, but I figure the procedure would look like this:
I’d worry about sighting accuracy with a phone.
Another way might be:
Fix phone on bike
Log gravity vector and GPS position at highest sampling rate possible.
Roll bike down the hill at constant velocity and heading. (as much as possible).
You can then use the pitch angle and distance traveled between each sample to calculate vertical distance traveled between each sample. Integrate over the whole run.
Both GPS and the phone IMU generally have built-in Kalman filters that should track pretty well to smooth, continuous motion.
It’d be better to use a wheel-sensor based odometer (with interpolation) instead of GPS, but that might be difficult to time synch with the data from a phone given that most wheel sensors are ANT+ instead of bluetooth. Although I think there are ANT+ dongles for phones?
Only testing would determine if this could actually be more accurate, but the benefit is that it could be almost entirely automated and would be very easy to perform. It’d just be pressing start and stop buttons on the phone.
