How does a cycling computer know the gradient?

[Dogtrousers]

Aircraft tend to travel several hundreds of miles in one flight, which means they will be traveling trough a variety of weather systems, with differing atmospheric pressure.

That is much less of an issue on a bike - though changes in weather/atmospheric pressure probably do explain some of the variations in elevation gain on rides of a few hours.

It is also rather more important to have accurate altitude in an aeroplane - if your bike computer says you are at 1000 feet when you are actually at sea level, that shouldn't be too catastrophic, whereas with a plane it could be.

As I understand it (definitely not an expert, just saw a video on the subject once), commercial aircraft have two modes of operating.

At high altitudes they just use the uncorrected pressure to give them the altitude. This may be wrong by a bit, but as long as everyone is wrong in the same way then they won't bump into each other, and as they are very high up it doesn't matter if what they think 50,000 feet actually is spot on 50,000 feet or is hundreds feet up or down.

But at lower altitudes it becomes more important to have an accurate absolute altitude as you don't want to bump into the ground, as that would be embarrassing. That's why airports broadcast their current ground pressure, and when they descend from high altitude they switch to a calibrated mode.

Plus they also have radar altimeters.

 

[Bollo]

A decade or two ago I did a few projects related to navigation systems, GPS and Inertial Navigation Systems (INSs). This was just around the time the US turned off the Selective Availability on their GPS system. Before that, civilian GPS had an accuracy of around 50m horizontal, which isn't great.

Anyways, even today most big expensive things with navigation combine an INS with GPS using some form of filtering, such as a Kalman filter. Modern GPSs have an accuracy of a few metres. Depending on cost INS can be incredibly accurate over a short time period but, due to the double integration of acceleration to get position, even the very best drift away over time due to dead-reckoning errors. The combined solution takes the best of both - the INS provides short term accuracy while the GPS 'grounds' the location ensuring that the estimate doesn't move away from the actual position.

I imagine the same basic maths is in play in most GPS-based cycle computers, combining the interpolated map elevation, GPS elevation and changes in barometric pressure. The first two aren't subject to any form of drift but will have quite poor precision (I think elevation for GLONASS was better but still not great) whereas the pressure reading doesn't really tell you how high you are but will give reasonably precise estimates of elevation changes.

Using something like a Kalman filter would integrate these three measurements, plus its estimate of where it thinks you're going to be based on your state (its estimate of your position, speed, possibly acceleration) in a more or less optimal way. This is why successive positions on a plot of your course don't jump around by anywhere near as much as you'd expect from the GPS error alone.

 

[Dogtrousers]

I imagine the same basic maths is in play in most GPS-based cycle computers, combining the interpolated map elevation, GPS elevation and changes in barometric pressure.

You say "basic maths". I say "black magic"

 

[Dadam]

Obviously no idea which model Garmin watch you have.

I have an Instinct and it is set to automatically recalibrate altitude via GPS at the start of each hike or bike ride.

It can also be set manually with a known value or by using the Garmin basemap (DEM - digital elevation mapping) that is built into the watch.

Seems to be pretty accurate too.

Vivoactive 3, so quite old now.
I don't think it has any settings to change that, at least I can't find anything in its menu or Garmin Connect.

 

[the_mikey]

That's hilarious. My local one is the Severn Bridge, I wonder if it's the same.

The Severn Bridge appears to the GPS device as if I'm riding on the surface of the water whenever I've ridden across it.

 

[Chap sur le velo]

The Severn Bridge appears to the GPS device as if I'm riding on the surface of the water whenever I've ridden across it.

That's why I prefer to walk.

 

[Ajax Bay]

The evidence from the bridge experiences ^^* suggests that those bar-mounted GPS devices are not relying on its altimeter/pressure sensor. On the sort of mapping used to support GPS devices, every discreet location has an associated altitude above datum. GPS is sufficiently accurate to locate the device accurately and the device then uses the embedded mapping to extract the data and display/record/use for differentials eg climb/descent rate.
Edit: * @ColinJ stated that his image is a route not a track (the latter derived from GPS, the former from a map) and same for the Severn bridge so the 'evidence' I drew was flawed (see posts below).

 

[Dogtrousers]

The evidence from the bridge experiences ^^ suggests that those bar-mounted GPS devices are not relying on its altimeter/pressure sensor. On the sort of mapping used to support GPS devices, every discreet location has an associated altitude above datum. GPS is sufficiently accurate to locate the device accurately and the device then uses the embedded mapping to extract the data and display/record/use for differentials eg climb/descent rate.
https://www.calcmaps.com/map-elevation/

Here's my recorded track of crossing the Humber bridge. It was a few years back so I've forgotten what GPS I was using. I think it was a non-bike-specific one, probably a Garmin Oregon but maybe Garmin 60CSX. You can actually see the curve of the bridge structure.

By the way, I'm very sceptical about the idea that "the device then uses the embedded mapping to extract the data and display/record/use for differentials eg climb/descent rate."

Maybe some very very fancy ones do. But I suspect (and I could well be wrong, but it's a strong suspicion) that for elevation GPS units use either GPS elevation or barometric elevation (or a combination of the two) but do not interpolate elevations from spot heights on maps.

The maps are there to show on the the display, that's all (I think). The only external elevation data that GPS units use (I surmise) is that embedded in route files, which requires no interpolation. (I suspect)

 

[ColinJ]

The evidence from the bridge experiences ^^ suggests that those bar-mounted GPS devices are not relying on its altimeter/pressure sensor.

My example was referring to the digital OS-map-derived elevation plot. I'm sure that my GPS had a better idea of my elevation. Having said that, the deck of the bridge is only about 30 m above the water and I know that the GPS can be out by that much, so maybe not!

I'll have a look to see if I kept any tracklogs of those rides...

I have one from the earlier rides that approached the bridge from the south.

The GPS did pretty well. You can see where the mapping software ignored the bridge.

 

[Ajax Bay]

Using GPS derived elevation is a waste of time: the standard error deviation makes the data unusable. Jolly good for position: useless for elevation and derivatives.
How do you think route files (eg gpx) have embedded elevation if not taken from mapping data? And that's what's loaded on a mapping capable GPS device.
Pressure sensors can produce accurate altitude, (only) provided the user has 'zeroed'(calibrated) it at a known height. Typically this is the end of the road (from one's house, say) or from the junction the club ride goes from. But if the isobaric gradient (eg weather front) is significant during the ride, the amount of climb will be way off.
As an aside, I used one of the earliest wrist altimeters competitively bitd for navigation in fell/mountain marathon races. Pre-race I annotated the map with heights of cols or points on a ridge, so that I could re-calibrate regularly (to negate inaccuracy from atmospheric pressure change). Height (contouring) was an additional navigational tool but the display needed to be numerically accurate. Btw altimeters are proscribed in orienteering competitions (map and compass only).

 

[ColinJ]

My tracklog came from an early Garmin Etrex which does not have a barometer OR any mapping capability whatsoever. I think it made a reasonable stab at measuring the altitude profile of the ride. You can see little fluctuations above and below what the actual elevation probably was, but nothing too drastic.

PS My later Edge 500, which does have a barometer, and is set to the elevation at my front door when I set off, can gain or lose 50+ m in a 30 km loop!

 

[Alex321]

Using GPS derived elevation is a waste of time: the standard error deviation makes the data unusable. Jolly good for position: useless for elevation and derivatives.
How do you think route files (eg gpx) have embedded elevation if not taken from mapping data? And that's what's loaded on a mapping capable GPS device.
Pressure sensors can quite accurately produce accurate altitude, (only) provided the user has 'zeroed' it at a known height. Typically this is the end of the road (from one's house, say) or from the junction the club ride goes from.
As an aside, I used one of the earliest wrist altimeters competitively bitd for navigation in fell/mountain marathon races. Pre-race I annotated the map with heights of cols or points on a ridge, so that I could re-zero regularly (to negate inaccuracy from atmospheric pressure change). Height (contouring) was an additional navigational tool but the display needed to be numerically accurate. Btw altimeters are proscribed in orienteering competitions (map and compass only).

Most of us aren't too worried about absolute altitude when cycling, we just want to know about changes in altitude during the ride, and the pressure sensors can be quite good for that, so long as the pressure at a given altitude stays reasonably constant for the duration of the ride.

I suspect the ones where we see loops that show us gaining or losing significant elevation overall are probably usually caused by changes in weather (and hence datum-point pressure) while we are out.

 

[Solocle]

The Severn Bridge appears to the GPS device as if I'm riding on the surface of the water whenever I've ridden across it.

47m clearance, so 5m off

 

[SpokeyDokey]

Vivoactive 3, so quite old now.
I don't think it has any settings to change that, at least I can't find anything in its menu or Garmin Connect.

The settings are on the watch itself.

Xmas pressie upgrade?

 

[Ming the Merciless]

The Severn Bridge appears to the GPS device as if I'm riding on the surface of the water whenever I've ridden across it.

That’s because the elevation came from the mapping service you used to plot the route. The mapping derived values are only roughly every 50m and thus won’t take into account the bridge. If you use the actual gps derived elevation it won’t put you in the water.