Yet another Garmin GPS question

[Billy]

Garmin 60CSx...trip odometer not agreeing with track distance. Examples from three specific hikes:

1) T.O shows 6.0 miles, track shows 5.2 miles when track is downloaded to computer. Out-and-back hike.
2) T.O shows 31.5 miles, track shows 27.5 miles when track is downloaded to computer. Lollipop loop...only the first and last 1.4 miles are repeated.
3) T.O shows 5.1 miles, track shows 3.6 on computer. This was a "nested" out-and-back hike with one section repeated many many times. It was the North Skyline Trail in the Blue Hills near Boston. The final set of rock steps before reaching the summit...I went up and down that section 14 times (4 times then to the summit; then 5 mores times then back to the summit; then 5 more times then back to the summit). When I look at the track / elevation profile on the computer (on the Garmin "basecamp" software), it only shows 2 of those 14 trips up toward the summit. The coverage was good....GPS had at least 5 or 6 birds at all times, with good signal strength.

So the question is: Is it possible to confuse the GPS by walking back and forth over the same ground repeatedly in a short amount of time? Another way of asking the same question: Is the data from a loop hike more reliable than from an out-and-back?

I'm willing to concede that this might be cockpit error, but at the moment I can't figure out how.

 

[Kevin Rooney]

Billy -

I've been using a 60CSX for years, and usually upload the track into MapSource. The mileage on the GPS versus MapSource rarely agree - close, but almost never the same. My hunch is that they use slightly different algorithms, but that's just an educated guess on my part. Doug Paul may be able to give you a definitive explanation.

Personally - since they're so close, and everything becomes an approximation if reduced far enough - the differences don't impact anything I need the info for.

 

[Peakbagr]

I'm with Kevin and agree that Doug Paul will know.

 

[Amicus]

Billy:

I too have a Garmin 60CSx but my experience has been the opposite of yours. The TO distance is almost always shorter than the track distance displayed when I download onto my Nat Geo TOPO map, usually not by much but occasionally significant. The latter have tended to be when I'm hiking in a steep notch, as in one traverse I recall that included the Twinway a few years ago, and I've assumed that occasional dropouts produced the lower TO readings.

Yes, Doug Paul must know!

 

[DougPaul]

The core GPS generates a PVT (position, velocity, time) output once per second. (Think of a GPS as a core GPS which generates the PVT and a computer which computes statistics from the PVT and drives the map display.)

The T.O. trip distance is based upon the 1/sec PVT--most likely the sum of the distance between the PVT points.

The track generally saves only a subset of the PT (no V) points--depends on the settings for recording the tracks. The track distance is presumably the sum of the distances between track points.

Thus the T.O. distance and the track distance are based upon different data sets and it is perfectly reasonable to expect the numbers to be slightly different. (In fact, one can prove that the track distance must be <= the T.O. distance. Proof via the triangle inequality.)

So the question is: Is it possible to confuse the GPS by walking back and forth over the same ground repeatedly in a short amount of time?

There are two issues:
1) The rate of track sampling vs how often your velocity (velocity=speed and direction) changes. For instance, lets say you are hiking a trail with 100 ft switchbacks. If your track samples your position every second and you are moving ~10 ft/sec, you will represent the trail fairly accurately (~every 10 ft), but you could miss bits of the ends of some of the switchbacks. If, on the other hand, you sample every 20 seconds, you could miss entire switchbacks.

So if you move at at steady velocity between samples the track will accurately represent your path, if you change your velocity between track points, then the track will lose some of your path info. (There are algorithms for selectively removing points from the 1/sec track to create smaller tracks with minimal loss of accuracy.)

2) Back and forth movement on the order of 1 second (such as holding the GPS in your hand and swinging it along with your stride). A GPS averages over several seconds (often using a Kalman filter) to produce more accurate PV estimates. If your velocity vector is unsteady on this time scale, it will reduce the accuracy. (Thus the GPS will be more accurate if you attach it to your shoulder than if you hold it in your hand or attach it to your arm.)

Another way of asking the same question: Is the data from a loop hike more reliable than from an out-and-back?

In general a hike will be much longer than the sampling interval and therefore loop vs out-and-back is unlikely to be a factor. The complexity of the trail vs the GPS (or track) PV sampling interval is much more important.

Note 1: The PVT from the GPS is only an estimate--it has errors from a variety of sources. The nominal long-term average position error is 95% probability of being within 10 meters of the correct position. These position errors will tend to increase the T.O. and track distances.

Note 2: The error will vary with time--thus the PV from a stationary GPS will wander a bit and a stationary GPS will show a distance traveled over some time period.


Note 3: all of the above assumes the GPS has a good skyview. If the skyview is poor all results will be less accurate. Trees, ground, rocks, and your body all block the satellite signals--orient the GPS vertically and carry it in the top of your pack, high on your shoulder, or on your head for best results. If I am mapping a trail (ie want the highest accuracy), I use an external antenna in the top of my hat. A pro would use a survey grade GPS with choke ring antenna (a ~1 ft diameter disk) on a pole to raise it above his head.

Doug

 

[DougPaul]

Doug Paul may be able to give you a definitive explanation.

I'm with Kevin and agree that Doug Paul will know.

Yes, Doug Paul must know!

I see that I have managed to pull the wool over you guys' eyes...

Doug

 

[DougPaul]

A few additional comments:
1) Consumer GPSes generally start with a 1/sec PVT. Tracks generally save a subset of the 1/sec PVT sequence. (There are other methods of generating a track from the PVT sequence.)

2) Manufacturers of GPSes and GPS software make their own decisions about how to compute distances and they don't reveal their algorithms to the public. So one can get different distance numbers from different sources (GPSes and software packages) for the same track.

Eg 1: A track consists of one or more segments. In computing the length of a multi-segment track, one manufacturer might treat the segments as if they were one (ie add in the between segment distances) and another might only include the within segment distances (ie drop the between segment distances).

Eg 2: Different manufacturers might also might use different degrees and methods of smoothing (intended to reduce the effects of the PVT errors). Smoothing would, in general, tend to reduce the computed distance. (I have experimented with smoothing of the track altitudes--it can make a very significant change in the computed altitude gain.)

Doug

 

[Kevin Rooney]

Yup, I knew we could lure Doug out ...

Thanks for taking the time to detail the technical bits.

 

[DougPaul]

Yup, I knew we could lure Doug out ...

I would have answered even without the flattery...

Thanks for taking the time to detail the technical bits.

Sure. GPSes and the associated algorithms can be complicated beasts... I certainly don't understand all the details myself.

Doug

 

[Billy]

Doug, if you're going to reply, would it kill you to put some time and thought and science into your answer?

Seriously, excellent info. Thanks. Just to add a few more details. For the Blue Hills hike I had the GPS at the top of the outside pocket of a paper-thin backpack....so it was between my shoulder blades...probably not too much crazy swinging there. I ran as much of this hike as I could...maybe that creates some of that swinging you mentioned. I had my distance interval set to track every 0.1 mile. Now I'm thinking that 0.1 mile might not be ideal for running up and down a big set of rock stairs over and over and over...that's essentially doing switchbacks that don't go anywhere, so there were no doubt lots of times when the turnaround points of these "switchbacks" didn't coincide with a 0.1 mile interval...so lots of these endpoints got cut off. I wonder if, for shorter hikes like this, especially when I'm just screwing around running up and down the same sections, using a smaller distance interval for tracking might yield a more accurate result.

Analogy: I draw a long irregular chalk line across my lawn...lots of curves and switchbacks. I want to measure it. I can choose:
1- a yard-stick.....lots of corners would get cut off
2- a hypothetical millimeter-stick.....no corners would get cut off, but the inherent error in each measurement would add up.
3- a happy-medium-stick...practical number of measurements, "good-enough" accuracy.........this is what I want from my GPS tracking log.....27.5 miles vs. 31.5 miles isn't good enough. I'll keep trying, i'll get there. Thanks again.

 

[billski]

I'll concur with Doug and Bill,

I've noticed that my gps-logged time is lower than the mapped time most often when I'm on a trail where I lose satellite reception.

I really should check the sampling period; never realized that was tweak-able.

I also agree with Billy on switchbacks, which is exacerbated when I lose signal. I usually have the gps on my shoulder, so switching back and forth changes the gps position relative to satellite. Although since switchbacks are slower travel on the way up, I would expect better accuracy on the way up. Hmm.

 

[JoeCedar]

Although we can't change how the GPS does its calculations, we can (within limits) change how it collects data. On the Tracks page, hit Setup, and you can choose between Distance, Time, or Auto in the "Record Method" box.

The minimum distance which can be entered in 0.01 (miles or km; about 53 ft or 33 m); the minimum time is 1 second; Garmin decides what "auto" does. The default is auto.

Using each of the three record methods, I took a short walk down the driveway and back. The total time was about 4 minutes for each. Here are the results:

Time (1 second): 245 points; distance = 286 meters
Time (10 seconds): 25 points, 272 meters
Distance (0.01 km): 21 points; 241 meters
Auto: 23 points, 253 meters

It is not surprising that the distance calculations are different, given the short total time and differences in the number of points, especially with the 1 sec setting. This comparison does not indicate what might be expected on a multi-hour hike or which mode may be more accurate, but shows that the Time setting permits collection of a lot of data (whether it is useful or more accurate in a given situation is another question). The GPS did not reduce the number of points when I saved the tracks. At 3600 points per hour with the 1 sec setting, however, the unit would exceed Garmin's 3000 maximum points per internal track log in less than an hour.

I have been satisfied with the distance setting at 0.01 miles for most hikes. At about 50 ft +/- 20 ft for the recorded points (about 100 points per mile), it is a reasonable compromise and compares well with published distances in trail guides. Unlike Time or Auto, it does not keep collecting points when I am not moving (unless it thinks I have moved >50 ft).

 

[DougPaul]

Doug, if you're going to reply, would it kill you to put some time and thought and science into your answer?

Sorry...

Seriously, excellent info. Thanks. Just to add a few more details. For the Blue Hills hike I had the GPS at the top of the outside pocket of a paper-thin backpack....so it was between my shoulder blades...probably not too much crazy swinging there. I ran as much of this hike as I could...maybe that creates some of that swinging you mentioned.

Sounds like you are carrying the GPS in a reasonable manner for what you are doing. Dry fabric is transparent to the GPS signals. My packs have top pockets where I place my GPS. I try to orient it as close to vertically as possible, but it can shift.

I had my distance interval set to track every 0.1 mile. Now I'm thinking that 0.1 mile might not be ideal for running up and down a big set of rock stairs over and over and over...that's essentially doing switchbacks that don't go anywhere, so there were no doubt lots of times when the turnaround points of these "switchbacks" didn't coincide with a 0.1 mile interval...so lots of these endpoints got cut off. I wonder if, for shorter hikes like this, especially when I'm just screwing around running up and down the same sections, using a smaller distance interval for tracking might yield a more accurate result.

Sounds like the sampling might be a bit coarse... The turnaround points could have up to .05 mi errors with an average of .025 mi error.

The 60CSx can be set to sample by distance, time, or auto with 5 levels of detail (which I presume is a combination of distance and time). I simply use "auto, norm" (the default) for casual use, but switch to by time when I am trying to be rigorous. The tracklog is 10K points long and the most frequent sampling is 1/sec which would allow recording up to 2 hrs 47 min. (You can record a longer track to the memory card.)

When I am attempting to measure the accuracy of the GPS in use (for instance in the "Bakeoffs"), I set it to every 5 seconds when walking (~6 meters of travel) or every 2 seconds when biking (~14 meters of travel). These are out-and-backs on reasonably straight paths so even a much coarser sampling will represent the path adequately. (In the bakeoffs, I compared the outgoing track with the return track to estimate the cross-track repeatability of the GPS as a surrogate for the accuracy.

Analogy: I draw a long irregular chalk line across my lawn...lots of curves and switchbacks. I want to measure it. I can choose:
1- a yard-stick.....lots of corners would get cut off
2- a hypothetical millimeter-stick.....no corners would get cut off, but the inherent error in each measurement would add up.
3- a happy-medium-stick...practical number of measurements, "good-enough" accuracy.........this is what I want from my GPS tracking log.....27.5 miles vs. 31.5 miles isn't good enough. I'll keep trying, i'll get there. Thanks again.

Sounds like you have been reading about measuring the length of a fractal. Strictly speaking, your walking path is also a fractal and therefore it has no unique length either... Pragmatically speaking, a 1/sec sampling (~3 meter walking travel distance) is probably good enough for most (walking) purposes and overkill for many. For your problem, guesstimate the maximum allowable error at the end of a switchback and sample often enough to stay below it. (Eg: 1/sec -> ~3 meter sampling at a walking speed -> 1.5 meter max err at the end of a switchback.)

If you want to be unbiased, sample by time. Any sampling scheme which includes distance (sample by distance or auto) includes what you are trying to measure into the data gathering and risks introducing biases.

Doing a good job of making physical measurements can be hard... There may be all sorts of subtleties waiting to trip you up.

Doug

 

[DougPaul]

I've noticed that my gps-logged time is lower than the mapped time most often when I'm on a trail where I lose satellite reception.

When you regain lost reception, the GPS will usually start a new track segment of a new track. (Output the track in GPX format and you can view it as text to see the breaks.)

I really should check the sampling period; never realized that was tweak-able.

You are probably using the default "auto, norm" setting.

I also agree with Billy on switchbacks, which is exacerbated when I lose signal. I usually have the gps on my shoulder, so switching back and forth changes the gps position relative to satellite. Although since switchbacks are slower travel on the way up, I would expect better accuracy on the way up. Hmm.

Not only that, but the mountain is partially blocking one side too. Thus if the mountain and the GPS are on the same side, the mountain is blocking one side of the sky and your body is blocking the other. If the mountain is on one side and the GPS is on the other they would both be blocking the same side of the sky.

I guess for best on-the-shoulder skyview while hiking switchbacks, one should move the GPS to the downhill side at every turn-around.

This is part of the reason that I put my GPS in my pack-top pocket. And, of course, I generally lean forward to counter balance the pack weight and turn my head down to keep an eye on the trail both of which reduce the blocking by my body. There is no problem if I am just recording a track, but it is a minor nuisance to get to the GPS if I want to use it for navigation. If I am using the GPS heavily for navigation I will carry it in a pouch on my shoulder strap or in a jacket breast pocket.

As stated earlier, I use an external antenna in my hat if I want the best accuracy. But then you have to be careful not to snag or get tangled in the wire...

Doug

 

[DougPaul]

Using each of the three record methods, I took a short walk down the driveway and back. The total time was about 4 minutes for each. Here are the results:

Time (1 second): 245 points; distance = 286 meters
Time (10 seconds): 25 points, 272 meters
Distance (0.01 km): 21 points; 241 meters
Auto: 23 points, 253 meters

It is not surprising that the distance calculations are different, given the short total time and differences in the number of points, especially with the 1 sec setting.

Nice little experiment.
If we assume your driveway is straight and ignore GPS position errors all of the distance errors would be underestimates because the endpoints of the track might not match the endpoints of your driveway.

This comparison does not indicate what might be expected on a multi-hour hike or which mode may be more accurate, but shows that the Time setting permits collection of a lot of data (whether it is useful or more accurate in a given situation is another question). The GPS did not reduce the number of points when I saved the tracks. At 3600 points per hour with the 1 sec setting, however, the unit would exceed Garmin's 3000 maximum points per internal track log in less than an hour.

You don't say what model GPS you were using, but the 60CSx stores 10K points and saved tracks are up to 500 points long. It defaults (settable) to displaying only the last 3K track points.

I have been satisfied with the distance setting at 0.01 miles for most hikes. At about 50 ft +/- 20 ft for the recorded points (about 100 points per mile), it is a reasonable compromise and compares well with published distances in trail guides. Unlike Time or Auto, it does not keep collecting points when I am not moving (unless it thinks I have moved >50 ft).

The best recording method and interval depends on what you are doing and staying within the tracklog size limit of 10K points for your trip. (However, if you save the track to the data card the size is limited only by the amount of space on the card.)

Auto also records fewer points when stationary than does pure time. Pure distance would tend to record lots of points (perhaps too many) when driving whereas auto appears to take your speed into account so that the default "auto, norm" setting does a decent job (IMO) both in the car and on foot. As noted earlier, I use pure time to minimize possible biases when trying to analyze the accuracy of the GPS itself.

Doug

 

[Raymond]

Billy, do you go by (and preserve) the Active Log rather than the saved track? You’ll see more points if you do, but I’m not sure how the distances compare.

Regarding the track being shorter than the odometer, DougPaul explained it to me a few years ago. (Or at least I understood his answer to mean that) the Track connects the various track points. So if you do a lot of back-and-forth, as while mowing your lawn, the track will come out a lot shorter than the odometer, because the track length depends on where the track points were recorded. If you go back-and-forth north and south, and if you are unlucky, the track points may all be recorded at the northern turnaround, so when they are connected, it would just be a straight west-east line, and all the travel to and from the south would be missed. I’ve exaggerated to make my point clear, I hope. Which is that some of the back-and-forth may be missed by the track.

But what I’ve been wondering lately is, does the odometer really take into account the distance traveled up the grade, or does it think I’m always on a flat surface? If someone were to climb, say, straight up El Capitan while carrying a GPS receiver, would the distance be accurately recorded, or would the unit think it was just staying in one place (because from overhead, it would not be seen to be moving)? All the track points would be atop each other, but would the odometer be accurate?

Whenever I’m out and think to check, the distances usually are shorter from one point to the next than what the on-site signs and books and maps all say they should be, so I’m wondering if the GPS measures short or if the measuring wheels measure long.

 

[DougPaul]

Billy, do you go by (and preserve) the Active Log rather than the saved track? You’ll see more points if you do, but I’m not sure how the distances compare.

The active log has a max of 10K points, a saved track 500 points. If you save a log longer than 500 points, it will be shrunk down to 500 points with some loss of detail.

Regarding the track being shorter than the odometer, DougPaul explained it to me a few years ago. (Or at least I understood his answer to mean that) the Track connects the various track points. So if you do a lot of back-and-forth, as while mowing your lawn, the track will come out a lot shorter than the odometer, because the track length depends on where the track points were recorded. If you go back-and-forth north and south, and if you are unlucky, the track points may all be recorded at the northern turnaround, so when they are connected, it would just be a straight west-east line, and all the travel to and from the south would be missed. I’ve exaggerated to make my point clear, I hope. Which is that some of the back-and-forth may be missed by the track.

The underlying difference is that the odometer is based upon the 1/sec position readings whereas the track length is based upon the [usually] coarser sampling of the track.

An interesting experiment might be to try recording a track at 1/sec and comparing the track and odometer distances.

But what I’ve been wondering lately is, does the odometer really take into account the distance traveled up the grade, or does it think I’m always on a flat surface? If someone were to climb, say, straight up El Capitan while carrying a GPS receiver, would the distance be accurately recorded, or would the unit think it was just staying in one place (because from overhead, it would not be seen to be moving)? All the track points would be atop each other, but would the odometer be accurate?

This question has been asked many times and the answer that I have seen always is that it is the 2D distance (ie elevation change is not taken into account). The difference in practical situations is less than most people assume: for instance, the difference is only ~1% at a grade of 1000ft/mile.

Whenever I’m out and think to check, the distances usually are shorter from one point to the next than what the on-site signs and books and maps all say they should be, so I’m wondering if the GPS measures short or if the measuring wheels measure long.

All methods of distance measurement have errors and biases. Since the errors may be different each time, you would have to walk it multiple times and look at the average and the variation. And the path hiked is a fractal so it doesn't even have a unique length. I simply don't worry about achieving a super degree of accuracy for hiking trails...

FWIW, for the Pemi lollipop ski trip the distances were: GPS track 27.6mi (Mapsource), GPS track 27.09mi (NG TOPO!), and 25.4 mi (WMG), however we weren't always on the trail and did some wandering and backtracking.

You could walk a measured mile on a road with a GPS and/or wheel and see what you get. Since it will be straight or nearly straight and have a smooth surface many of the problems go away. (Technically, you should probably walk on the centerline if it isn't straight. You could walk both sides and average if you wish to increase the chance of your surviving... )

Doug

 

[bikehikeskifish]

Saving a track not only shrinks it to 500 points, it (seems to) remove the timestamp/elevation info.

For your amusement, I was at my father-in-law's place on an island on a lake in NH and we were discussing plans for a new dock. The town rules say you can build so many feet of dock per 100 feet of frontage. A recent tax bill says he has 450 feet - which he thought rather high. I had brought my GPS for amusement purposes and so he and I walked the boundary, as close to the water line as possible, and amazingly came up with... 451 feet. Presumably the assessors walk the property line on the ice in winter, but we're not sure.

The frontage includes a narrow, unbuildable point jutting out which clearly accounts for 250' since pin-to-pin is close to 200'.

Tim

 

[DougPaul]

For your amusement, I was at my father-in-law's place on an island on a lake in NH and we were discussing plans for a new dock. The town rules say you can build so many feet of dock per 100 feet of frontage. A recent tax bill says he has 450 feet - which he thought rather high. I had brought my GPS for amusement purposes and so he and I walked the boundary, as close to the water line as possible, and amazingly came up with... 451 feet.

I suspect the small difference between the numbers involves a significant amount of luck. If you walk it additional times, you will most likely get different numbers.

Doug

 

[Raymond]

This question has been asked many times and the answer that I have seen always is that it is the 2D distance (ie elevation change is not taken into account).

Knew it!

I remembered what made me first wonder about this.

Last year, I was descending Square Ledge and stopped to reset the trip odometer when it reached exactly the 10.0-mile mark (I like the distance to be carried out to hundredths of a mile, not just tenths, even though I understand it’s not really that accurate) and also to change the batteries.

I descended a short distance, then noticed I had forgotten to turn the GPS back on. So I turned it on, climbed back up to the spot where I’d reset it, then went back down again. I noticed that the distance traveled (in either direction, up or down), according to the trip odometer, was exactly the same as the change in elevation, which meant I must have been climbing and descending exactly straight up and down. Well, it was steep, but it wasn’t 90° steep, obviously. So that was what made me first think that the trip odometer wasn’t measuring the actual distance covered so much as the apparent distance as drawn on a map.

Thanks for the confirmation of my suspicion.