distance jumping crazy near GZ

[+bilikituri]

So, I normally do geocaching with an iPhone 5 and either Groundspeak's app or MotionX-GPS.

 

They work great, but when I am near GZ and I use the compass view to get bearing and distance, the reported data goes often quite crazy. Like, it may report 3m, then 7m. I may walk a bit and get another jump to 12m. Then if I walk around, being at the same spot I was before, the reported distance and/or bearing may be different from what the phone told me just 2 minutes before. I may get the distance changing even if I stay at the same place. It is quite disconcerting.

 

I tried yesterday a GSPmap 62s, and it did the same. At the very GZ it could report 20m, then 12m, then 5m, then 8m, while the iPhone was like 13m, then 5m, then 8m... (I have returned the 62s).

 

I know there is a radius of accuracy/error margin. In that sense I would expect the distance to be steady, perhaps wrong by that factor, but steady.

 

Is anyone familiar with the theory behind GPS to explain these distance jumps? Why wasn't the 62s better? The hardware may be more sensitive in covered areas but this jumping phenomena is something perhaps fundamental to how GPS works?

[+stijnhommes]

You're right to mention covered areas. With a little bit of tree cover, a GPS unit can get seriously confused about where exactly it is located causing the distance to jump around.

[+Walts Hunting]

Not how it works. Your GPS reads the info from the satellites, computes your position and then draws the screen. It then does that all over without any weighting for the previous calculation. So if you have about a 20' foot EPE showing it can jump that far (or more) in any direction. Once you are that close you should have your gps in your pocket while you look.

[+bilikituri]

You're right to mention covered areas. With a little bit of tree cover, a GPS unit can get seriously confused about where exactly it is located causing the distance to jump around.

 

But this happens to me also in clear places. Well, around you there's always something, a small tree, a building... something, I don't do geocaching in the dessert . The jumping happens regularly to me. I believed that the 62s would fix this, but I was disappointed it didn't in my test (disappointed modulus maybe this jumping is fundamental and cannot be fixed).

 

Question is, in addition to the accuracy radius, this distance jumping near GZ is also normal and fundamental to GPS? To the way triangulation is done or whatever?

 

Or is my test suspicious and the 62s distance should be rock solid? That is, within the accuracy radius, but steady, contradicting the test I did yesterday?

Edited October 18, 2013 by bilikituri

[+bilikituri]

Ah, let me add I am in Europe and therefore can't benefit from WAAS.

[+bilikituri]

Not how it works. Your GPS reads the info from the satellites, computes your position and then draws the screen. It then does that all over without any weighting for the previous calculation. So if you have about a 20' foot EPE showing it can jump that far (or more) in any direction. Once you are that close you should have your gps in your pocket while you look.

 

So are you saying that the way GPS works implies that if you get data D now in GZ, and 10 seconds passes, the new data received by the GPS may be totally different and that explains the jumps?

 

If that was true, the same kind of jumps should be observed along the way, but when in the route I don't observe this as crazy as I do at GZ.

Edited October 18, 2013 by bilikituri

[+dprovan]

I know nothing about iPhones, but my GPSr -- and I think most others -- averages what it calculates, so any jinks settle out over time as it takes more and more readings. In practice, I've found that the longer I'm in the same place, the less the coordinates move and the more confident I can be that they actually reflect where I am.

 

I still tend to put the GPSr in my pocket when I get to GZ, but only because if I'm looking at my GPSr, I'm not looking for the cache.

[+bilikituri]

I know nothing about iPhones, but my GPSr -- and I think most others -- averages what it calculates, so any jinks settle out over time as it takes more and more readings. In practice, I've found that the longer I'm in the same place, the less the coordinates move and the more confident I can be that they actually reflect where I am.

 

I still tend to put the GPSr in my pocket when I get to GZ, but only because if I'm looking at my GPSr, I'm not looking for the cache.

 

Interesting.

 

My question is theoretical so to speak, would like to understand why does the distance jump like that.

 

But in practical terms, do you mean that when the GPS reports you are close, say 20 meters, or 15 meters, you already consider that to be GZ and start looking around? I thought people tried to get 0 distance, and then assume it was in a circle within 5-10 meters.

[+Gitchee-Gummee]

My question is theoretical so to speak, would like to understand why does the distance jump like that.

 

But in practical terms, do you mean that when the GPS reports you are close, say 20 meters, or 15 meters, you already consider that to be GZ and start looking around? I thought people tried to get 0 distance, and then assume it was in a circle within 5-10 meters.

That pretty much hits the nail on the head.

 

You are using a "consumer-grade" device. It is incapable of providing you with an EXACT location to any true degree of accuracy or consistency. Such devices are available, but few of us would want to spring for the cost of same.

 

As others have reported, when it starts "jumping all over", put it away. It has done what it is gonna do. Besides, at that time you need to look for the cache or its hiding spot, not wasting your vision looking at the device.

 

As an aside -- remember that the cache was hidden with a like device. True "accuracy" does you little good anyway because of that!

Edited October 18, 2013 by Gitchee-Gummee

[Fangamon]

When my E20 does this I put it on the ground and let it sort itself out. Slower I know but you get there quicker!

[+taeke]

When my E20 does this I put it on the ground and let it sort itself out. Slower I know but you get there quicker!

 

TBH, getting there quicker came for me only with experience of going out and cache. As a couple of people have said before me, don't rely on the GPSr when you arrive at GZ, but use your eyes. Think like a potential cache owner and look where you would hide a cache or tag. Most of the time your guess will be right and you will have found the cache without being driven nuts by a fluctuating GPSr reading

[+taeke]

deleted

 

double post due to forum server troubles

Edited October 19, 2013 by taeke

[+dprovan]

But in practical terms, do you mean that when the GPS reports you are close, say 20 meters, or 15 meters, you already consider that to be GZ and start looking around? I thought people tried to get 0 distance, and then assume it was in a circle within 5-10 meters.

Well, no, because even with the rough accuracy as I walk up, my PN-60 reliably takes me to within 20 feet, more like 6 meters. And I do have to let it average for a couple minutes before it really knows where it is. But after those few minutes, it's close enough -- I'd say about 2 meters -- that it's more important to look for the cache than try to use it to get closer to GZ, particularly because 2 meter accuracy is about as accurate as the coordinates themselves, considering the precision of coordinates expressed in thousands of minutes.

 

But do keep in mind that coordinates are sometimes unintentionally off, so you should always be weighing possibilities beyond where your phone is telling you to look, no matter how accurate it is.

[+bilikituri]

Yeah, we know that accuracy is not possible due to how GPS works. That's fine. We also know coordinates may be off.

 

But one thing is to get zero or near-zero distance, put the device in a pocket, and take a radius of about 5-10 meters for guessing, and a totally different thing is to have the slightest idea of whether you are close to GZ or GZ is actually 20 meters apart because the distance reading gets so inconsistent and tells you 6m now and 14m seconds later. In an area off-trail with rocks a radius of 20m is a Universe!

[+bilikituri]

But in practical terms, do you mean that when the GPS reports you are close, say 20 meters, or 15 meters, you already consider that to be GZ and start looking around? I thought people tried to get 0 distance, and then assume it was in a circle within 5-10 meters.

Well, no, because even with the rough accuracy as I walk up, my PN-60 reliably takes me to within 20 feet, more like 6 meters. And I do have to let it average for a couple minutes before it really knows where it is.

 

I'll try that. When it gets crazy, I'll leave the iPhone level somewhere and will wait a couple of minutes to see if it converges to something. Then a posteriori evaluate if the distance was off within accuracy error, or way off.

[+BAMBOOZLE]

You're right to mention covered areas. With a little bit of tree cover, a GPS unit can get seriously confused about where exactly it is located causing the distance to jump around.

 

But this happens to me also in clear places. Well, around you there's always something, a small tree, a building... something, I don't do geocaching in the dessert . The jumping happens regularly to me. I believed that the 62s would fix this, but I was disappointed it didn't in my test (disappointed modulus maybe this jumping is fundamental and cannot be fixed).

 

Question is, in addition to the accuracy radius, this distance jumping near GZ is also normal and fundamental to GPS? To the way triangulation is done or whatever?

 

Or is my test suspicious and the 62s distance should be rock solid? That is, within the accuracy radius, but steady, contradicting the test I did yesterday?

 

The 62S should not jump even under canopy and with a clear sky will be scary rock solid....we use 2 of them and have been around many more. Firmware can be an issue and on the 62S I have found ver 3.90 to be the best ( the latest is not always the best )......I've used almost every unit out there and prefer the 62S for geocaching.

 

Once in a great while I will visit a location where GPS operation is very unstable...metal buildings, overhead power wires, and sometimes no apparent issue ( maybe GPS jammer in the area )......again, this is rare. 99% of the time the 62 is VERY accurate and smooth at GZ......I recommend getting another, put on ver 3.90, put in fresh batteries, calibrate the compass, and have fun.

 

Re iPhone caching, I have a 4 and a 5 and while I use the GC app and its great features I'm not a fan of getting to GZ with an iPhone.

[+bilikituri]

But this happens to me also in clear places. Well, around you there's always something, a small tree, a building... something, I don't do geocaching in the dessert . The jumping happens regularly to me. I believed that the 62s would fix this, but I was disappointed it didn't in my test (disappointed modulus maybe this jumping is fundamental and cannot be fixed).

 

Question is, in addition to the accuracy radius, this distance jumping near GZ is also normal and fundamental to GPS? To the way triangulation is done or whatever?

 

Or is my test suspicious and the 62s distance should be rock solid? That is, within the accuracy radius, but steady, contradicting the test I did yesterday?

 

The 62S should not jump even under canopy and with a clear sky will be scary rock solid....we use 2 of them and have been around many more. Firmware can be an issue and on the 62S I have found ver 3.90 to be the best ( the latest is not always the best )......I've used almost every unit out there and prefer the 62S for geocaching.

 

Once in a great while I will visit a location where GPS operation is very unstable...metal buildings, overhead power wires, and sometimes no apparent issue ( maybe GPS jammer in the area )......again, this is rare. 99% of the time the 62 is VERY accurate and smooth at GZ......I recommend getting another, put on ver 3.90, put in fresh batteries, calibrate the compass, and have fun.

 

Re iPhone caching, I have a 4 and a 5 and while I use the GC app and its great features I'm not a fan of getting to GZ with an iPhone.

 

Aha, that's what I expected. The accuracy error is fundamental to the way the location is computed, but I expected a good device to be more stable near GZ than an inferior device. I know accuracy is impossible, I just expect the estimated distance to behave more consistently. You clearly have a lot of experience, so my test had to be flawed or I had bad luck.

 

I would like to understand why that jumping happens. By now nobody has explained it, I purchased a book about GPS to research this.

 

On the practical side, I don't want to spend 300 euros to get to GZ, though I certainly have thought what you suggested. Navigate to the cache with Groundspeak's app, and then narrow to GZ with maybe an eTrex 10, since I only need bearing and distance. Assuming the eTrex 10 is going to do a better job than the iPhone 5 there (which is only a guess, since I don't know the technology behind them).

Edited October 19, 2013 by bilikituri

[+BAMBOOZLE]

But this happens to me also in clear places. Well, around you there's always something, a small tree, a building... something, I don't do geocaching in the dessert . The jumping happens regularly to me. I believed that the 62s would fix this, but I was disappointed it didn't in my test (disappointed modulus maybe this jumping is fundamental and cannot be fixed).

 

Question is, in addition to the accuracy radius, this distance jumping near GZ is also normal and fundamental to GPS? To the way triangulation is done or whatever?

 

Or is my test suspicious and the 62s distance should be rock solid? That is, within the accuracy radius, but steady, contradicting the test I did yesterday?

 

The 62S should not jump even under canopy and with a clear sky will be scary rock solid....we use 2 of them and have been around many more. Firmware can be an issue and on the 62S I have found ver 3.90 to be the best ( the latest is not always the best )......I've used almost every unit out there and prefer the 62S for geocaching.

 

Once in a great while I will visit a location where GPS operation is very unstable...metal buildings, overhead power wires, and sometimes no apparent issue ( maybe GPS jammer in the area )......again, this is rare. 99% of the time the 62 is VERY accurate and smooth at GZ......I recommend getting another, put on ver 3.90, put in fresh batteries, calibrate the compass, and have fun.

 

Re iPhone caching, I have a 4 and a 5 and while I use the GC app and its great features I'm not a fan of getting to GZ with an iPhone.

 

Aha, that's what I expected. The accuracy error is fundamental to the way the location is computed, but I expected a good device to be more stable near GZ than an inferior device. I know accuracy is impossible, I just expect the estimated distance to behave more consistently. You clearly have a lot of experience, so my test had to be flawed or I had bad luck.

 

I would like to understand why that jumping happens. By now nobody has explained it, I purchased a book about GPS to research this.

 

On the practical side, I don't want to spend 300 euros to get to GZ, though I certainly have thought what you suggested. Navigate to the cache with Groundspeak's app, and then narrow to GZ with maybe an eTrex 10, since I only need bearing and distance. Assuming the eTrex 10 is going to do a better job than the iPhone 5 there (which is only a guess, since I don't know the technology behind them).

 

If you're going to get an Etrex at least get an Etrex 20....they have mapping and good reviews. The Etrex 30 gives you a 3-axis electronic pointer.....it will always point at the cache you're seeking even if you're moving slow or standing still.

[Fangamon]

Etrex 20 is better as there are issues with the 30.

[+Team CowboyPapa]

......

They work great, but when I am near GZ and I use the compass view to get bearing and distance, the reported data goes often quite crazy. Like, it may report 3m, then 7m. I may walk a bit and get another jump to 12m. Then if I walk around, being at the same spot I was before, the reported distance and/or bearing may be different from what the phone told me just 2 minutes before. I may get the distance changing even if I stay at the same place. It is quite disconcerting.

.......

 

Is anyone familiar with the theory behind GPS to explain these distance jumps? Why wasn't the 62s better? The hardware may be more sensitive in covered areas but this jumping phenomena is something perhaps fundamental to how GPS works?

Well, I don't claim to all that familiar with the phenomenon, but here is a start.

[+bilikituri]

Well, I don't claim to all that familiar with the phenomenon, but here is a start.

 

Are you sure that's the explanation? I have a degree in math, can read the page if I grab a cup of coffee , but in what sense that models approaching GZ?

 

When the GPS devices I have tested report 1357 meters of distance to the target, they do not start jumping like crazy all the time, 1350, 1364, ... Sometimes there might be a little fluctuation, but generally speaking they report more steadily as far as I have tried. But that remarkably changes in my experience near GZ.

 

There has to be something related to both points being close that affects this somehow. One point is your estimated coordinate, and the other point is the target coordinate which is an exact point in the grid. The target coordinate may be off relative to the cache, but that doesn't matter for this discussion, as far as the GPS system goes you've entered a concrete coordinate as destination, so the only error variation for doing the math happens in your end (I conjecture): once you have your estimated position, the target coordinate is known and exact relative to that pair and the computation of the distance carries no more error propagation (I conjecture).

 

Regarding the eTrex 10. Thanks all for the advice! I indeed purchased an eTrex 30 some days back. It couldn't calibrate the compass, systematically failed in the 3rd step (it is a known error, there are threads about it). Then I went and replaced the unit with the 62s and did a cache with it.

 

I am not a hiker, only recreational hiking to go find some caches in the backcountry, so I wouldn't use the majority of features in those devices. Also, the interface is from the 90s. Do caching with an eTrex, and then do it with your iPhone, OpenCycleMaps and all the cache metadata available offline (including photos). Drag the map, zoom in and out,... all fast and snappy looking awesome in your retina display. Case with battery pack to have enough for an entire day and done.

 

Now, if it was true (something I do not know) that these devices are more stable near GZ than the iPhone 5, I'd consider an eTrex 10 just to have bearing and distance. I can throw the rest of features, stable bearing and distance is all I need to complement the phone. Bearing does not depend on your orientation, it is an imaginary line between your estimated position and the target point, so I don't even need (and never actually use) a compass. In particular the lack of maps and type of compass of the eTrex 10 don't matter to me (Edit: though the device needs to know my orientation to be able to draw the arrow. In that sense the electronic compass would make the bearing arrow more responsive, but I don't really care where is North).

 

I am going probably to test one, do some caches, systematically compare their behaviour, and report my findings in this thread.

Edited October 20, 2013 by bilikituri

[39_Steps]

In my simulated geocaching experience, which generally occurs only after Delorme has released new firmware for the PN-40 or PN-60 handheld gps devices, I get more understandable results by creating a new waypoint(geocache)in an open clearing on a clear day.

 

(1) For instance, I first place a marker randomly on open ground and then stand still for a while before creating a new waypoint(geocache) at the coordinate point that will become ground zero.

 

(2) Then I make certain that automatic recalculation is enabled in the (non-road vector) Direct (Hiking)routing mode.

 

(3) Then I walk away from ground zero at least 15 meters (~50 feet) before selecting the waypoint(geocache)and Direct Routing to it.

 

(4) From what I have read of Garmin device "sticky" issues, I would likely back off at least 100 meters to create a route to GZ and then plan on walking at least 30 meters on past GZ while observing the route pointer. (Not necessarily to be confused with any available compass needle).

 

(5) Repeat as necessary from different angles to ground zero, with random deliberately erroneous parallel lines of travel.

 

(6) Then I might wait few hours for the satellites to circle and the clouds to roll in before repeating the tests to GZ.

 

Edit: (7) Forgot to mention calibrating any embedded 2 or 3 axis electronic compass before starting. Thanks to Bamboozle.

 

Disclaimer: I do not have any of the modern handheld Garmin devices typically used by Groundspeak forum posters. Suggestions made above are made based on reading of Garmin "sticky" issues on these forums. I infer that the 60 foot stickiness (or whatever the radii distance may be with different Garmin firmware) is deliberately programmed in by Garmin to head off forum threads such as this one about "jumpiness". Last time I checked one of my Delorme units, the "stickiness" seemed to be limited to about a 20 foot (6 meter) radius past GZ in a straight line. I have absolutely no experience with devices designed for cell phone tower triangulation.

Edited October 20, 2013 by 39_Steps

[+BAMBOOZLE]

Well, I don't claim to all that familiar with the phenomenon, but here is a start.

 

Are you sure that's the explanation? I have a degree in math, can read the page if I grab a cup of coffee , but in what sense that models approaching GZ?

 

When the GPS devices I have tested report 1357 meters of distance to the target, they do not start jumping like crazy all the time, 1350, 1364, ... Sometimes there might be a little fluctuation, but generally speaking they report more steadily as far as I have tried. But that remarkably changes in my experience near GZ.

 

There has to be something related to both points being close that affects this somehow. One point is your estimated coordinate, and the other point is the target coordinate which is an exact point in the grid. The target coordinate may be off relative to the cache, but that doesn't matter for this discussion, as far as the GPS system goes you've entered a concrete coordinate as destination, so the only error variation for doing the math happens in your end (I conjecture): once you have your estimated position, the target coordinate is known and exact relative to that pair and the computation of the distance carries no more error propagation (I conjecture).

 

Regarding the eTrex 10. Thanks all for the advice! I indeed purchased an eTrex 30 some days back. It couldn't calibrate the compass, systematically failed in the 3rd step (it is a known error, there are threads about it). Then I went and replaced the unit with the 62s and did a cache with it.

 

I am not a hiker, only recreational hiking to go find some caches in the backcountry, so I wouldn't use the majority of features in those devices. Also, the interface is from the 90s. Do caching with an eTrex, and then do it with your iPhone, OpenCycleMaps and all the cache metadata available offline (including photos). Drag the map, zoom in and out,... all fast and snappy looking awesome in your retina display. Case with battery pack to have enough for an entire day and done.

 

Now, if it was true (something I do not know) that these devices are more stable near GZ than the iPhone 5, I'd consider an eTrex 10 just to have bearing and distance. I can throw the rest of features, stable bearing and distance is all I need to complement the phone. Bearing does not depend on your orientation, it is an imaginary line between your estimated position and the target point, so I don't even need (and never actually use) a compass. In particular the lack of maps and type of compass of the eTrex 10 don't matter to me (Edit: though the device needs to know my orientation to be able to draw the arrow. In that sense the electronic compass would make the bearing arrow more responsive, but I don't really care where is North).

 

I am going probably to test one, do some caches, systematically compare their behaviour, and report my findings in this thread.

 

A few things :

 

1. Mapping = For driving to a cache my iPhone 5 does not compare to my 62S....zooming in and out, panning the map , and more are one hand operations with the 62.

 

2. Compass = the 3 axis compass serves as a pointer to the cache even when you're moving too slow for GPS compass operation....when moving among rocks or thick woods and swamps this is an incredibly valuable function...if you've caches for a bit using a unit with a 3-axis " pointer " you would never want to be without one.....the iPhone is O.K. but the 62 is GREAT in this respect.

 

3. Durability and weather....GPS units get dropped and wet.

 

Get a GPS with mapping capability and 3 axis compass.

[+Team CowboyPapa]

No need to complicate the issue by walking around. Simply place the GPSr in one place with a good view of the sky and after configuring one of the screens to display:

1. Non-auto calibrated barometric elevation,

2. GPS derived elevation,

3. Longitude,

4. Latitude, and

5. Time.

 

Now, for the 1st variable, jot down the displayed value every 5 seconds for several minutes.

Repeat for the other 3 variables.

 

Take that to a professor of math or physics of your local university and ask if that data is not described by the definition of random walk.

 

Note that none of that deals with the observations of the data as one approaches a particular location.

[+bilikituri]

No need to complicate the issue by walking around. Simply place the GPSr in one place with a good view of the sky and after configuring one of the screens to display:

1. Non-auto calibrated barometric elevation,

2. GPS derived elevation,

3. Longitude,

4. Latitude, and

5. Time.

 

Now, for the 1st variable, jot down the displayed value every 5 seconds for several minutes.

Repeat for the other 3 variables.

 

Take that to a professor of math or physics of your local university and ask if that data is not described by the definition of random walk.

 

Note that none of that deals with the observations of the data as one approaches a particular location.

 

But for that to be relevant to the thread, the device should have some tracking buffer and algorithms to correlate measurements using moving averages, markov chains, ... stuff that computes a new data point taking the previous ones into account, rather than the crude one-shot measurement from the satellites that ignores any other previous data.

 

Do GPSs do that? If they do, could it be the case that a handheld device like the eTrex has a better algorithm than a phone? And a 62s a more sophisticated algorithm than the eTrex?

 

Unfortunately manufacturers seem to be quite opaque about everything beyond "supports GLONASS", or "has a quad-helix antenna". Difiicult to compare other than trying actual devices.

Edited October 20, 2013 by bilikituri

[+Team CowboyPapa]

Nothing crude about it, just a start within a process consistent with the Grand Method of Science.

 

To continue, go to a large open space without overhead obstructions such as a shopping center parking lot, and:

1. Repeat the steps suggested above while stationary in this parking lot,

2. Repeat after a level walk to another location, to determine any variation due to location change,

3. Repeat again at the end of a return walk to the original location to determine repeatability to original location,

 

I suggest you calculate standard deviations for the data from these steps determine repeatability of the random variations.

 

With the exception of the auto correction of the barometrically derived elevations, I doubt there are other correction algorithms. However, is it not the object of these tests to reveal those, if any? Even while stationary, note the random variations in the elevations calculated from the GPS sources.

[+bilikituri]

Nothing crude about it, just a start within a process consistent with the Grand Method of Science.

 

To continue, go to a large open space without overhead obstructions such as a shopping center parking lot, and:

1. Repeat the steps suggested above while stationary in this parking lot,

2. Repeat after a level walk to another location, to determine any variation due to location change,

3. Repeat again at the end of a return walk to the original location to determine repeatability to original location,

 

I suggest you calculate standard deviations for the data from these steps determine repeatability of the random variations.

 

With the exception of the auto correction of the barometrically derived elevations, I doubt there are other correction algorithms. However, is it not the object of these tests to reveal those, if any? Even while stationary, note the random variations in the elevations calculated from the GPS sources.

 

Sorry, I don't understand the relationship between these tests and my question.

 

My question is: I observe that while following a route the estimated distance to the destination is quite steady overall, but when you are really close to the target, the distance often starts to jump. Why does it get so inconsistent?

 

Then you point to an article about random walks.

 

In what sense the article or your tests answer my question?

[+Team CowboyPapa]

I'll return to my earlier suggestion. Go to a local college and ask a professor there if you choose not to collect and interpret your own data.

 

My apologies, I forgot to post my disclaimer. Here is a link describing me as totally unknowledgeable:

http://forums.Groundspeak.com/GC/index.php?showtopic=189245&view=findpost&p=3438512

 

Actually, this was concerning similar subject matter: obtaining and analyzing elevation data statistically from a GPSr.

[+bilikituri]

I'll return to my earlier suggestion. Go to a local college and ask a professor there if you choose not to collect and interpret your own data.

 

Still do not understand.

 

So, if I collect data points by hand and run a statistical analysis of some sort, you claim I'll obtain a random walk.

 

That's fine, but, so what? What is the point? If the devices ignore previous data points, the measurements are independent. Why is it relevant to this discussion that the data points collected by hand by you yourself have this or that property?

[+bilikituri]

OK, I have done three known caches carrying the iPhone 5 and an eTrex 10. These caches are in Barcelona in wide streets. Buildings are not very tall, and there's one near the stadium of FC Barcelona. The streets had trees in the sidewalks, you know, but nothing like the backcountry of course. While going to the caches I walked often in the very center with clear sky and good margins in both sides.

 

I walked all the way monitoring both devices. These are my findings:

 

• While navigating to the cache, the estimated distance rarely matched. They were often off of one another by 10 meters or even 20 meters. Sometimes less, but almost never equal.
  
• None of them seemed to be inconsistent when they were away from GZ. The distance more or less was decreasing in a reasonable way. I stopped sometimes and did some random steps in different directions. Both devices remained more or less stable (we are talking 300 meters away from GZ, for example).
  
• To my surprise, the iPhone reported 10m accuracy often. The eTrex had full signal strength always.
  
• At GZ, the iPhone was erratic as I have explained in this thread, and the eTrex stable. The iPhone could report 6m, then 14m, then 12m, only doing small steps. And the eTrex consistently reported 3 meters, 2 meters, 1 meter. The eTrex did NOT jump at all, and the distance was actually much better than 5 meters.
  

 

So that may suggest the erratic behaviour I was asking for it is not due to the way GPS works in itself, a theoretical limitation. If that was the case, the eTrex could not be stable. There has to be something else, either the iPhone receiver is really inferior, or the eTrex has smarter software or hardware. But definitely the eTrex was stable.

 

Next time we go to the backcountry I'll run a similar test.

Edited October 21, 2013 by bilikituri

[+TriciaG]

I'm following this thread with interest. My Google phone jumps all over at GZ also, and wonder if there's a good explanation. Thanks for starting it, and I wish you success in finding a cause!

[Fangamon]

I'll admit that I have averaged 3 waypoints near a GZ and then triangulated on a map to zero the cache, good old analog maps..

[+Team CowboyPapa]

Nothing crude about it, just a start within a process consistent with the Grand Method of Science.

 

To continue, go to a large open space without overhead obstructions such as a shopping center parking lot, and:

1. Repeat the steps suggested above while stationary in this parking lot,

2. Repeat after a level walk to another location, to determine any variation due to location change,

3. Repeat again at the end of a return walk to the original location to determine repeatability to original location,

 

I suggest you calculate standard deviations for the data from these steps determine repeatability of the random variations.

 

With the exception of the auto correction of the barometrically derived elevations, I doubt there are other correction algorithms. However, is it not the object of these tests to reveal those, if any? Even while stationary, note the random variations in the elevations calculated from the GPS sources.

 

Sorry, I don't understand the relationship between these tests and my question.

 

My question is: I observe that while following a route the estimated distance to the destination is quite steady overall, but when you are really close to the target, the distance often starts to jump. Why does it get so inconsistent?

 

Then you point to an article about random walks.

 

In what sense the article or your tests answer my question?

Several points here:

1. Not having the same equipment, I am not able to test to verify and analyze your observations.

2. I agree, the manufacturer will provide no information to you regarding the observed performance.

3. Consequently, what other avenue could you pursue than gathering data yourself and analyzing it?

4. Random walk: Its existence is certain, the question is how much. Now watching the non-auto calibrated barometrically derived elevation, I see -47,-45 -43, -43, -44, -46 feet at 5 second intervals. (I have not calibrated and we are having higher than standard ambient pressure today.) If you are not sure of such characterization of instrumentation performance as random walk, I suggest that you take this post to a college for review and interpretation by a physics professor.

[39_Steps]

Sorry guys, but if you had not noticed the Nobel Prize for Random Walk studies has already been awarded.

[+bilikituri]

4. Random walk: Its existence is certain, the question is how much. Now watching the non-auto calibrated barometrically derived elevation, I see -47,-45 -43, -43, -44, -46 feet at 5 second intervals. (I have not calibrated and we are having higher than standard ambient pressure today.) If you are not sure of such characterization of instrumentation performance as random walk, I suggest that you take this post to a college for review and interpretation by a physics professor.

I do not question your claim.

 

But you have not yet explained how such property would explain why the distance seems to be stable away from GZ, and unstable near GZ.

Edited October 22, 2013 by bilikituri

[+fizzymagic]

My question is: I observe that while following a route the estimated distance to the destination is quite steady overall, but when you are really close to the target, the distance often starts to jump. Why does it get so inconsistent?

 

I've been watching this thread for quite a while and nobody has stated the obvious, so I guess I better jump in.

 

The effect you are seeing is a simple consequence of geometry. When you are far from a cache, random variations make less difference in the distance to the cache than the same variations do when you are close to GZ.

 

Think about it: when you are far from the cache, variations in position to either side of the direction toward the cache make no difference in the distance to the cache. Only variations straight toward or away from the cache can give you large jumps. But when you are at GZ, variations in ANY direction give you large changes in the calculated distance.

[+bilikituri]

I've been watching this thread for quite a while and nobody has stated the obvious, so I guess I better jump in.

 

The effect you are seeing is a simple consequence of geometry. When you are far from a cache, random variations make less difference in the distance to the cache than the same variations do when you are close to GZ.

 

Think about it: when you are far from the cache, variations in position to either side of the direction toward the cache make no difference in the distance to the cache. Only variations straight toward or away from the cache can give you large jumps. But when you are at GZ, variations in ANY direction give you large changes in the calculated distance.

 

Hmmm, not sure I follow. Jumping 4, 6, 8, meters when you do small steps in random directions should happen no matter whether you are at GZ or at 500 km of distance, if the erratic behavior happened at any given spot.

 

Only your position is estimated, once the GPS assumes some coordinates, the distance to the target is a computation.

Edited October 22, 2013 by bilikituri

[+bilikituri]

Well, I have to admit this explanation would match my tests if:

 

1) At GZ the iPhone had 10m accuracy.

 

2) When I looked at the distance decreasing, accuracy was better (5m).

 

I'll try to do tests taking 1) and 2) into account.

Edited October 22, 2013 by bilikituri

[+bilikituri]

Well, I have to admit this explanation would match my tests if:

 

1) At GZ the iPhone had 10m accuracy.

 

2) When I looked at the distance decreasing, accuracy was better (5m).

 

I'll try to do tests taking 1) and 2) into account.

 

Actually, 2) may not be relevant. If the correct coordinate may be anywhere within the circle, being away from GZ there's much more area where being off affects less the estimation.... Yeah, your reasoning makes a lot of sense after thinking more about it.

[+bilikituri]

After sleeping on the problem, I believe a geometrical explanation would be the most reasonable one I have to say, since it would apply to all devices, only distance and accuracy would matter (assuming the software is equally smart, something I don't really know, and suspect is not the case if I had to bet).

 

However, I have run some calculations that, if I am not mistaken, do not fully explain what I observe.

 

Let's say we are at a point in the trail, far from GZ. Say 900 meters away. Let's say our position has a radius of margin of 10 meters. Which is the probability that our location jumps in a step of 4 meters or more? (Assuming within the error circle the distribution is uniform, that is, the GPS can locate you at any point evenly.)

 

The points 4 or more meters closer to GZ than our location are those that overlap our circle of radius 10, with the circle centered at GZ with radius 900 - 4 = 896. Visualize an overlap with the shape of a lens.

 

The points 4 or more meters away from GZ than our location are the complementary of the overlap with the circle centered at GZ with radius 900 + 4 = 904. Visualize an overlap with the shape of a moon.

 

According to my calculations, if you are at GZ you should see jumps of 4 with probability 0.84. But if you are 100, 200, 300... meters away, you should still see them with probability 0.50.

 

This is the Ruby program I've written to compute this: https://gist.github.com/fxn/7114945, and the output is below the token __END__.

 

While the probabilities are certainly different, the ones away feel still to high. I don't see jumps of 4 meters half of the time.

Edited October 23, 2013 by bilikituri

[+Team CowboyPapa]

My question is: I observe that while following a route the estimated distance to the destination is quite steady overall, but when you are really close to the target, the distance often starts to jump. Why does it get so inconsistent?

 

I've been watching this thread for quite a while and nobody has stated the obvious, so I guess I better jump in.

 

The effect you are seeing is a simple consequence of geometry. When you are far from a cache, random variations make less difference in the distance to the cache than the same variations do when you are close to GZ.

 

Think about it: when you are far from the cache, variations in position to either side of the direction toward the cache make no difference in the distance to the cache. Only variations straight toward or away from the cache can give you large jumps. But when you are at GZ, variations in ANY direction give you large changes in the calculated distance.

Exactly.

[+fishgeek]

My question is: I observe that while following a route the estimated distance to the destination is quite steady overall, but when you are really close to the target, the distance often starts to jump. Why does it get so inconsistent?

 

I've been watching this thread for quite a while and nobody has stated the obvious, so I guess I better jump in.

 

The effect you are seeing is a simple consequence of geometry. When you are far from a cache, random variations make less difference in the distance to the cache than the same variations do when you are close to GZ.

 

Think about it: when you are far from the cache, variations in position to either side of the direction toward the cache make no difference in the distance to the cache. Only variations straight toward or away from the cache can give you large jumps. But when you are at GZ, variations in ANY direction give you large changes in the calculated distance.

 

Plus, when you are farther from GZ, you are usually moving and will not notice minor variations in the GPS's estimate of your position. At GZ, you are generally stationary and the variations are much more obvious.

[+bilikituri]

Plus, when you are farther from GZ, you are usually moving and will not notice minor variations in the GPS's estimate of your position. At GZ, you are generally stationary and the variations are much more obvious.

But in my tests I do stop and walk around GZ-style.

Edited October 23, 2013 by bilikituri

[+bilikituri]

Followup: I contacted Joe Mehaffey from www.gpsinformation.net by email, he was very kind to confirm that Garmin devices implement some kind of hysteresis. That is, they do some kind of weighted average or something like that that stabilizes the estimated position, though the exact algorithm is not disclosed.

 

That matches my experience, which I describe as "stability" as an end-user.

 

My hypothesis would be that the iPhone 5 (or at least Groundspeak's application, since I don't know if that should be done by the hardware, or the operating system, or the end-user application), does not do that averaging and thus appears to jump more.

 

I would also conjecture that this jumping does not mean the eTrex does not jump the same way behind the scenes (assuming same accuracy), it would only mean that it remains stable on the surface. It is also the case that in my tests the accuracy reported by the eTrex is often smaller (smaller estimated error/radius) than the one reported by the iPhone.

 

Unfortunately they have not reviewed the iPhone, so could not provide information related to that and other questions that have arisen in this thread.

Edited October 28, 2013 by bilikituri

[+bilikituri]

And probably my last post here.

 

I decided to do another comparison between de iPhone and the eTrex 30, but this time with total certainty about the correctness of the GZ coordinates. In order to do that, I chose a geodesic marker in Barcelona. This marker is at the top of a castle, which is located at the top of a hill near the sea: the Fortress of Montjuic. Sunny day, totally blue sky, no clouds. This is the marker data:

 

ftp://ftp.geodesia.ign.es/Red_Geodesica/Hoja0421/042112.pdf

 

I parked my scooter at about 90m (300ft). I let both devices to settle, and the iPhone reported 95m (accuracy 5m), while the eTrex reported 85m (accuracy 3m). I walked a few steps, the iPhone would report 82 all of a sudden, then jump to 88m. The eTrex transitioned steadly.

 

To reach to the geodesic marker at the top of the fortress you need to go inside a tunnel, and then go up some stairs. I was in the terrace for about 15 minutes. When I was exactly at GZ with a tolerance error way beyond what these units can offer, the eTrex reported accuracy 3m, and estimated distance of 2m. When walking around near GZ the iPhone told me anything from 1m to 12m, jumping and with any bearing you can think of. The eTrex did not jump, distance and bearing where really really consistent and accurate.

 

My conclusions of this thread:

 

* I don't believe the test that started the thread was significative.

 

* The eTrex 30 is consistent in being accurate and steady.

 

* In comparison, the iPhone 5 has proven to me at least to be much more erratic. Either because the hardware is inferior, the software less smart, or a combination of both. I can't really tell because there is little information out there about how these things actually work.

 

* I no longer believe the behavior far away from GZ is significantly more stable than at GZ, as disproved by my test today. The geometric argument explained above yields some light, and my simulations report some probabilities to test. But I do not know if the model in the simulation is too simplistic, in particular if the distribution is uniform in the disc, if you can actually trust the estimated accuracy, etc.

 

Those are conclusions not based on technical knowledge unfortunately, so take them with a grain of salt. They are based on the information I have been able to gather from trusted sources, and my own field tests.

 

Thanks all for your contributions!

Edited October 28, 2013 by bilikituri

[39_Steps]

My thoughts on testing a specific handheld gps device are that a known accurate benchmark in an open area is as good a place to create a new Waypoint as any.

 

Then ignore the "accurate" coordinates assigned to the BM by someone else and quickly test your own gps device with your own waypoint before the satellite constellation changes significantly.

[+bilikituri]

My thoughts on testing a specific handheld gps device are that a known accurate benchmark in an open area is as good a place to create a new Waypoint as any.

 

Then ignore the "accurate" coordinates assigned to the BM by someone else and quickly test your own gps device with your own waypoint before the satellite constellation changes significantly.

 

Your waypoint has an error itself that you need to add to the later estimation. The coordinates of a geodesic marker are virtually exact for our purposes, so only the estimation error of the device remains.

[+NotreDameFan11]

Handheld Gps units work best when moving at a pretty good speed. Lets say you are travelling in a straight line going north. Each time it determines your position it could be off 20 to 30 feet in any direction of where you are so if you travel 100 feet north before it determines your next position the general direction will be relatively north of your last position. Plus the gps is programmed so the pointer moves slower than with each reading it takes. Once you stop and it determines a position 20 ft. to your right and then next position 20 ft. to your left you will see the pointer starting to point to your left.