I'm guessing it's the imprecise nature of the GPS, but I’m getting different numbers on an out and back course I ride on a regular basis. There is consistently about a 25 to 35 foot difference between the total number of feet ascending and feet descending on a relatively flat 25 mile out and back ride. Is there any other logical explanation?

If you have, for example, a Garmin Edge 205, there is no barometer inside it. The altitude measurements come from the GPS data, which isn't very accurate in terms of altitude.

If you use Garmin Connect, you can correct that altitude data based on the data from the road maps.

Here's a comparison on one of my loop rides.

I think you're right NOS88. GPS positioning isn't all that precise, about 15 feet on average for civilian GPS. Obstructions and atmospheric conditions can further degrade the accuracy.

Even the barometric pressure 305 and 500 units show differences. I attribute it to a combination of smoothing and changes in temp and pressure.

I have a Garmin Edge 705 which uses barometric for altitude. I've noticed that I have to start it at least 20 minutes before riding for it to settle in.

Also riding into the wind, or going fast downhill, will give a different reading than with the wind to my back or going up hills. It may be due to my Garmin not being mounted completely flat on the handlebars.

Then there's differences when riding while a front is moving in. On longer rides this can make the altitude of the start and finish quite a bit different even though it's the same place.

Geometry. Consider that you'll generally have satellites are various points over 360 degrees of azimuth. In elevation, at best you'll have them over 180 degrees because the earth blocks any below the horizon. Over all possible visible elevations and azimuths, the probability of having them overhead is relatively small. This is the main contributor to having less altitude accuracy the lat/lon accuracy with GPS.

Also consider that at sea level, atmospheric pressure only varies by .002%/ft. A 25' change in altitude is only a 0.05% change, or 0.0075 psi out of 15 psi of nominal atmospheric pressure.

If you pointed the pressure port on your device directly into the wind at 30 mph, the ram air pressure would make the altitude indicate ~800' low.

The main problem with ascent or descent calculations is that some of the noise gets rectified and included in the ascent or descent.

GPS position and altitude and barometric altitude are all noisy. When computing the distance you travelled, the noise isn't much of a problem. The cross track errors make almost no contribution to the computed distance and the in track errors tend to cancel out leaving only the errors at each end or a few dozens of feet for distances of several miles or more.

On the other hand, to get the total ascent or descent, one must try not to include all the noise. For example, if you just set your GPS out in your backyard for a day and then add all the ups and all the downs, you might well find several thousand feet of ascent and almost same amount of descent (differing by a typical error in the altitude - several tens of feet - the difference between the ending and starting altitude) even though the GPS unit was stationary. So some scheme for differentiating noise from true altitude changes is necessary. I don't know the algorithm Garmin uses in their units. I do know the algorithm I use. It goes something like: a segment is classified as up if the increase in altitude is greater than some threshold or the previous segment was "up" and this segment did not decrease in altitude. Then I add all the ups to get the total ascent. Downs are similar. The problem with this algorithm is that decreases in altitude happen much faster than increases in altitude (because you are going downhill rather than uphill, doh!), so true decreases are much harder to mask with noise than true increases. Hence the calculation leads to a greater descent than ascent.

I have an idea for a better calculation but haven't gotten around to implementing and testing it yet.

- Ed

Al -

You rode the route! How about the difference at about mile 42. The GPS shows about 400 ft of descent followed by about 500 ft of ascent in about two miles. The corrected elevations show almost flat right there (about a 50 ft dip). So which is correct? Surely you would have noticed such a steep hill or the absence of such a steep hill!

My own experience is that GPS is usually better than maps. Elevations on maps are usually determined from the DEMs (digital elevation models) which involve interpolation on a fairly coarse grid and which can miss small features like deep cuts, or steep cliffs.

- Ed

This gets more and more complex. Putting aside algorithms, rectification of noise and the effect of dynamic pressure on a barometer, Most error from GPS devices is intentional inaccuracy built into the satellite network and as mentioned, geometry. Being that altitude is calculated, position error causes error in altitude. For those with a barometer, the barometer is a fixed setting and measure difference in barometric pressure. Your computer can record elevation change sitting in your garage as the outside barometric pressure changes.

Ok, so all of these posts over the years of X number of feet of climbing are just estimates or SWAG.

I've actually learned a good bit reading these posts. Thanks, folks.

Apparently they stopped blurring GPS a while ago, so it is no longer the deliberate inaccuracy. It's really just the geometric difficulty that causes problems. With a higher power GPS (e.g., on a plane), one can use more satellites and get a more accurate altitude.

For barometric pressure, yup, that's pretty much it.

As far as Ed's (groth) point, we're kind a screwed either way. I agree that the elevation grid that is used is too coarse (particularly if you are trying to measure grade), but with my Edge 305 (which uses barometric pressure) I find 10% differences in the same ride on different days.

The corrected one is correct. The road only goes down a little and then up again.

There is tree cover there, and perhaps the unit lost contact with the mother ship for a second. I ride that route a lot, and that glitch only appears on that one ride.

Interesting. last GPS course I had (3 months ago) still indicated a 75 ft programmed inaccuracy. In aviation, approach minima for GPS approaches is still higher than ground based navigation unless Wide Area Augmentation is used in which case, that error is removed by ground transmitters.

The last time I carted my 705 from NM at 7,600' to NJ at about 200', it told me that my house in NJ was below sea level. It didn't change after a week and several rides. Like they say, "For entertainment value only".

So far as I know there is no deliberate tweaking of the GPS timing to produce an inaccuracy anymore. Current inaccuracies are inherent limitations of the system. One of the last things Bill Clinton did before leaving office was sign an executive order eliminating Selective Availability (the deliberate errors). Of course there is a caveat that in times of national emergency, the DOD reserves the right to re-introduce it. But by now, that would probably so damage the economy that they wouldn't dare. Even farmers use GPS to guide their tractors! (Can't fight a war if your soldiers are hungry!)

I think you might be confusing WAAS (Wide Area Augmentation System) with differential GPS. Assuming you are out in the open so you can see lots of satellites without reflections from buildings or power lines and there is no deliberate jamming (try riding next to a military base), the biggest error in GPS comes from the ionosphere. The free electrons in the ionosphere produce an index of refraction for the radio waves that GPS uses. In other words the waves travel through the ionosphere slower than the speed of light. Since the ionosphere is variable (depends on time of day, solar activity, latitude, etc.) this is an unpredictable delay in GPS timing. The WAAS uses a network of ground stations (I believe about a dozen over the US) to measure the time delays due to the ionosphere. These are then uplinked to geostationary satellites which broadcast the ionospheric delays (on a coarse grid). WAAS enabled receivers can use these broadcasts to make a correction for the ionosphere. The correction can only be perfect at the locations of the ground stations that measured the delays at the time the delays were measured.

Differential GPS uses receivers at a fixed location (an airport say) to measure the errors from each satellite as seen at that location. These are then broadcast locally and a differential GPS receiver can correct its GPS signals to produce much higher accuracy positions and elevations near the airport. Landing planes totally under the control of differential GPS was demonstrated some years ago.

The air force is currently launching the next generation of GPS satellites which have two civilian frequencies. (The current generation has only one frequency for civilians.) With two frequencies, the GPS receiver can measure the ionosphere delay to each satellite without the need for WAAS and accuracies of several feet (rather than tens of feet) should be possible.

OK, lecture mode OFF!

- Ed

A friend of mine has a business burying drainage tile in fields. He tells me that the in-cab computer in his tractor, using RTK (Real Time Kinematic) with GPS, can maintain the depth of the tile within 1 centimeter as he (or the computer, actually) drives across the field.

Yes - that's really impressive. A scheme to use the carrier phase to get relative positions to centimeters!

In Canada, DGPS is only used by the Navy and Coast Guard. It has no aviation application due to inaccuracies so I am not mistaking it for WAAS. It does however take GPS lateral position error from 90 feet down to 30 feet which is significant if you are looking for people or small vessels in a search area.

WAAS, right now, is the most precise multi dimensional navigation system but I still don't see a day where is will be accepted for sole source precision approaches in aircraft.

CPGPS (RTK)is impressive for survey work or farm but has too many limitations to be considered for airborne applications. Right now, it isn't even being considered by regulatory bodies. Sometimes however, developments work from the ground up. Pun intended.

Take the largest GPS recorded elevation gain, multiply by 1.5 and report that as your climb, works with average speed as well. Try it, you'll like how everyone in impressed...

Well that is obvious.. What we are trying to figure out, is why we don't report the descent the same way.

Think about why you even bother - therein lies your answer.

Could GPS electronics exhibit some sort of hysteresis? I'm not an expert, but would hope one would chime in here...

The engineering answer here is long and boring and probably would result in a lot of flashing of credentials.
So why start - just except the fact that it isn't perfect and leave it at that.

I have found that if a weather front is coming through your barometric altitude numbers will be skewed one way or the other.