Jump to content

Getting The Right Coordinates For Placing A Cache.


PastorDIC

Recommended Posts

I recently got my Etrex Legend and am looking forward to placing a cache.

 

However, I am concerned about getting accurate coordinates for it.

 

I know accuracy can vary based on where the sattelites are and obstructions, etc. but I have stood in front of my front door at the exact same place several times, and each time, even in the same spot, I get slightly different coordinates. If this happens a few feet from my doorstep, how will I ever get accurate cordinates when I am out setting a cache?

 

Thanks for the help.

Russ

Link to comment

I know accuracy can vary based on where the sattelites are and obstructions, etc.  but I have stood in front of my front door at the exact same place several times, and each time, even in the same spot, I get slightly different coordinates.  If this happens a few feet from my doorstep, how will I ever get accurate cordinates when I am out setting a cache?

 

Is normal to get slightly different coordinates on the same place. If your state accuracy there is, say, 6 m, you have to consider perfectly normal for your coordinates to vary inside more or less 10-12 m. Sometimes the difference will be greater, there's a thread on this on this same forum ("How close is close").

 

A method I'm trying with my legend (but I'm still not completely sure if it gives any REAL advantage) is to "emulate" the waypoint averaging of the Magellans. The legend has a configurable tracklogging which allows you to drop trackpoints at fixed time intervals (or fixed distance, or in automatic mode). If you set the time interval to 1 sec, you have de facto in the tracklog any position the GPS is calculating, and your tracklog will have the maximum possible resolution. This is useless normally, but if you:

1) put the GPS on the point you want to mark

2) clear the tracklog

3) leave it there for, say, 5 minutes

4) shut off the tracklogging before moving the GPS

you will end up with 300 measurements of the same waypoint, saved as active log in the GPS (DO NOT convert it in a saved track or almost all trackpoints will be dropped!). Any program capable to download it and calculate the average wp from a track (I use GPS manager under linux, but I suppose that also other programs will be able to do it), will provide you a (supposedly) high accuracy waypoint.

 

I know is a bit tricky, but is in fact an efficient version of the usual trick of moving around, return to the point and take another wp, doing it 10 times and then averaging at home.

 

But in fact, if you look to the other thread I told you above, you will see that no one expects to return to a waypoint with a precision better than 10 m: it is the physical limit of the GPS devices (unless you have a survey grade one...).

Link to comment

I had the same concerns and so I conducted this experiement in by large back yard:

 

1) I marked a waypoint on my unit and put a marker there.

 

2) Starting from various distances and directions I did a "go to" that position (A)and marked the positions that the unit computed. Then I marked those as waypoints.

 

3) When the lawn was littered with the kid' toys (I was using them as the markers) and after the nieghbors were thouroughly baffled I found that the markers were indeed an average of 30 feet or so from the target and averaging the way points did come close to the original position A.

Link to comment

I'm a stickler for spot on coords. I can spend an hour standing over a hiding spot trying to mark it. I've gone back on different weather days to take readings ( I'm a big loser! ;) )

 

Really, the thing I try to look for when I take coords for a cache-plant is good geometry. We all have that satellite screen on our GPS but few of us stare at the locations of the birds in relation to one another. To call a set of coords good I want to see an even distribution, hopefully in all 4 corners of the screen, as far out as possible, with as many birds in view as I can get. A big clump all in the center just won't do. I have an Etrex with a patch antenna so I hold that thing at different angles and point it thriough trees at open sky, etc, until I get a good 'spread'.

 

Then I'll watch the last two numbers of the North and West (I'm in Canada eh?) for several minutes and if it seems to stick on the same set I'll mark the spot and then do a goto from all four compasse points after walking 20 meters away each time.

 

Someone posted about setting the GPS on the ground for 5 minutes while you prepare the cache box. That seems to work too. The GPS seems to smarten up when you leave it alone for awhile.

 

And having said that, in really crappy places averaging can work, or simply taking a reading in the clear and then projecting it 20 meters to where the cache is from a few different points can give you something good to work with as well.

 

I have WAAS on my GPS but if I turn it on I better be prepared to stay there awhile. I find WAAS needs some time to settle down and I only really trust it if it's very close to the first set of cords I had without it turned on.

 

Woohoo! :)

Link to comment

Averaging for 5 minutes now that SA is discontinued is really not worth the time spent doing it. Averaging can in fact make the position solution worse as there is basically a 50/50 chance of actually getting improved coordinates. Which 50% one might be in, well that's the unknown.

 

GPS accuracy is a statistic and all one can expect is to get the best possible result relevant to the circumstances at the time. At a particular location accuracy will vary throughout the day so there will be times during the day when one should be determining a position (best "possible" soultion) and times that range from not so good to a complete waste of time. It's called planning.

 

Cheers, Kerry.

Link to comment

An interesting factor is if you don't average about half the visitors will find your coordinates 'spot-on' as they're variance matches yours! (The other half'll be way away.)

 

Averaging over multiple days at different times of day I've found really useful.

 

I usually mark the spot when I find where I want to hide a cache.

 

When I come back with a container, it's usually a good distance away, and I mark a 2nd waypoint.

 

After hiding the container and prepping the spot (usually 15 mins. or so of the GPS sitting) I'll mark a 3rd and average them all together.

 

After the initial onslaught of visitors (next weekend) I'll revisit at a different time of day and take another waypoint to average in with the initial three. (Note, if you average the new one with the 'averaged set' you posted on the cache page, you'll be giving the new one more weight.)

 

Since things vary so much day to day, this really helps. Oh yeah, I'll not even bother marking the point if all the sats are directly overhead or in a line (epe/accuracy will be in the hundreds of feet then).

 

I think this is the extreme end of the spectrum but I've never had a complaint. On local cacher consistently had bad coords. (sometimes 70' off) until enlightened and since has had wonderful coords.

 

The one thing you want to be sure NOT to do is turn on your GPS, wait for lock and instantly mark the spot.

 

hth,

 

Randy

 

PS: An easier technique I've seen good success with is to wait for good lock, mark the point, leave, then 'Go To' it. Not as accurate but they don't get complaints either. This is particularly good in an area with lots of signal bounce as you'll see if it's impossible or not.

Link to comment

I placed my first cache on Saturday. I went to where I wanted to hide the cache and took a waypoint. I then walked away, came back, took another way point, moved the GPS, put it back and took another waypoint. .

 

Two days later I averaged those points and used that to guide me back to the spot. I was at least 30 feet away. I then took 3 more waypoints and averaged those while I took out ones that I knew were too far away. The new average was spot on. I used it to track my way back.

 

One person has since found my cache and I have a question in to him to let me know how accurate my coordinates were. BTW, I am using an Etrex Legend.

 

I have noticed that while we are looking for caches, I am usually withing 25-ft of the cache. That's why I find it helpful to have a second person out looking with me.

Link to comment
Averaging for 5 minutes now that SA is discontinued is really not worth the time spent doing it. Averaging can in fact make the position solution worse as there is basically a 50/50 chance of actually getting improved coordinates. Which 50% one might be in, well that's the unknown.

 

GPS accuracy is a statistic and all one can expect is to get the best possible result relevant to the circumstances at the time.

Well, Kerry, in fact 5 minutes of averaging is an overkill, but the trick works also for smaller times. And you are supposed to have 3-5 minutes left during which doing other things. Is surely far less time consuming than going around twenty times to produce twenty readings of the waypoint.

 

What I don't get is why you say that averaging has a 50% chance to produce worse coordinates.

 

If we assume the successive readings to be uncorrelated (which is false, but I will go back on this below) ther number would peak on a certain value (the "right" one) with some kind of error distribution around it (unlikely to be a gaussian) of which stated EPE is supposed to embrace a certain extent (in my experiments, about 66% of the measures are coherent within 1 EPE, but this figure may vary). This is exactly the case in which averaging will actually improve the result, in the sense that the result of the averaging will fall where the maximum number of results lie.

 

The reason why I'm not sure that this actually works is that subsequent GPS readings are far from being uncorrelated. As is well known, the solution presented is actually the result of a complex algorythm including a kalman filtering. This means that the actually derived solution is compared with a prediction of it based on the previous readings, and the two things are averaged (with appropriate weighting) to produce the reading the GPS actually displays.

I suspect that the actual game is more complex than a pure kalman filtering, but in fact the result is that any solution is dipendent from the previous ones, and that acts like a sort of inertia of the GPS solution.

 

This is the simple reason why is useful to give some time to the GPS to "settle down" its solution when we want a good reading: if your "first" solution is somewhat off (think to the first one you have at the moment you turn the GPS on) the "leverage" the new readings are able to apply on the displayed solution are only partial and need a bit of time to move the result to the "right" point.

But this "leverage" depends on how much the new reading differs from the prediction. When the difference is small, when you "almost" converged on the point, the leverage may become too weak to "move" the solution, which settles down. And it will do it not necessarily on the same point it would have stopped another time, when converging from another point.

 

So (excuse me for the technical blah blah above) this is the reason why I'm actually not sure that averaging works. At any time, averaging will in fact average around the point where the solution settled, and average to zero only the remaining solution noise. If it is smaller than the difference that may exist within different settled solution, if the systematic part of the uncertainity may be greater than the "noise" part, averaging become totally useless.

 

I don't think so: EPE as you say is a statistical quantity and may not take into account numerical convergence errors. There's no way to account for it. So EPE has to work on satellite geometry and error propagation on the noise like error sources (timing iuncertainities and so on). If the noise component is minor, EPE would be an useless quantity.

 

Anyway I'm experimenting on it, which is always the better way... :)

 

P.S. Sorry for the lecture-like post... ;)

Link to comment
Well, Kerry, in fact 5 minutes of averaging is an overkill, but the trick works also for smaller times.

 

.... What I don't get is why you say that averaging has a 50% chance to produce worse coordinates ....

 

.... Anyway I'm experimenting on it, which is always the better way... :)

 

P.S. Sorry for the lecture-like post... ;)

Experimentation :D now that's the way. It really all comes down to what actually happens in a practical sense.

 

The 50/50 reasoning is simply that, based on actuals and really as GPS accuracy (95% confidence) by definition is based over a 24 hour period then depending on the time frame averaged 50/50 is about the figure especially considering that one doesn't know exactly where one is in the first place.

 

Based on the system spec most days (on average :) ) will display the following traits in a practical sense, that is if one really knew where one actually was in the first place, which of course is what these results are based on.

 

GPS Averaging - A practical example

 

and

 

GPS Averaging - Part II

 

and

 

GPS Averaging - Affect of obstructions

 

Really to trap ALL errors that might affect the position solution then that's a 30 day exercise and I'm sure nobody has the capacity or time to average 1 second data for 30 days.

 

The other thing that can decieve averaging principles is that mnay receivers might display/output a position every second it can be the same position simply output over several consectutive seconds (sometimes 'can be" up to 5-7 seconds).

 

Averaging principles over enough time will (generally) provide a GPS solution that one would (will) expect to be more reasonable than others. However most users simply don't have the time to adequately do this.

 

Cheers, Kerry.

Link to comment

Experimentation  :)  now that's the way. It really all comes down to what actually happens in a practical sense.

 

Well, I know... But I have my legend since a very short times, and with the "Yellow" such an experiment was impossible, since it was impossible to gather all the points in the tracklog (of course, if you have an external datalogger would have been possible). So until now I made only a couple of tries.

 

The pages are extremely interesting (sadly on the second one I can't see the figures...) , I will study them more carefully... many thanks for the links :)

 

Inside there I saw many interesting clues to thing I actually always asked myself.

 

Surely I was aware that averaging with a bad geometry would provide worse results than a single measure with an excellent geometry. But I would have not said that no improvement may be given with respect to a single measure on the SAME geometry. Quite interesting. I'm actually trying to figure out WHY exactly it happens...

 

Of course you have also the problem relate to the precision of the GPS output. UTM for example has a precison of 1 m (at least on the etrexes) so the discreteness of the data is not negligible with respect of you errors (~10 m) ...

Would like to know how accurate is the internal representation...

 

Well, I think that the question is settled now!

Thanks for the evidences!! :)

Link to comment

Hee hee hee. It's scary how the forums are always one step ahead of me. My wife is working on a Puzzle Cache using the concepts outlined above (particularly Team Tar's comments). My wife's work entails Research Meteorology and has to do with assimilating satellite data into numerical weather models (i.e. stuff only supercomputers can handle). The topic of "error" always comes up in her work.

 

The Puzzle Cache were conceiving has to do with the inherent error with GPS technology, but using a meteorological method of averaging of several biased waypoints with varying EPE's to find the cache site. She assures me that the math is not to bad on this one :)

 

I'll have to Markwell this thread on the cache page once it's finished. Great discussion. :)

Edited by Touchstone
Link to comment

When I first started placing caches, I spent a good bit of time averaging coordinates. I'd take a waypoint, walk a 20 feet away, return, mark another and repeat 15-20 times.

 

I've since found this to be a total waste of time. Now I make sure I have a good sat lock, take one reading and I'm done. My coordinates are no better or worse than when I averaged. In fact, the only cache out of the 50+ I've placed that I've received frequent complaints about was one that I averaged.

Link to comment

I do pretty much the same as briansnat, unless the signal I am getting is absolutely crap. Then I may wait to see if the signal improves before taking the coords. If the signal looks good, I take the coords and move on to the rest of my day. (Of course I have only been placing caches for 2 months and haven't really got enough data to tell if my caches are impossible to find.)

 

OzGuff

Link to comment

Depending on how good of a signal I am getting. If I have very good signal w/ WAAS I will just take a couple of way-points they tend to be the same anyways. But under tree cover I like to take a few paces north of the cache and take a couple of way-points then do the same to the south, west and east. Average them all together and I get a good location even in the worst of cover.

Link to comment
.... But I would have not said that no improvement may be given with respect to a single measure on the SAME geometry. Quite interesting. I'm actually trying to figure out WHY exactly it happens...

 

Of course you have also the problem relate to the precision of the GPS output. UTM for example has a precison of 1 m (at least on the etrexes) so the discreteness of the data is not negligible with respect of you errors (~10 m) ...

Would like to know how accurate is the internal representation...

Even with good geometry and even over a reasonable period of time (as far as averaging in a practical context is concerned) the position solution is not jumping all around in a random fashion but rather simply wandering/meandering unpredictably along at a generally steady rate.

 

So over the few minutes a person might be averaging for all that period of time the position solution has every possibility of moving further away from the absolute position until a point in time that it tuns around or heads off on a different meandering track.

 

These days (without SA) the position solution has a reasonably steady rate of change over time. Still unpredictable but not random (as such).

 

When Selective Availablility was at its prime a "stationary" receiver would effectively travel (on average) up to 24 kilometres in any 24 hour period but these days is can be as low as 100-200m per hour, which over a few minutes isn't a great distance (on average).

 

This lack of movement over a short period of time also negates any "possible" improvement in the position solution.

 

The following is another example based on what some referred to as "golf ball average". Basically 4 points fixed over a 5 minute period using dual freq post processed coordinates (being true and basically absolute), then the same 4 points located with a handheld over several minutes. Even though there is some slight out of shape movement based on the handheld the general error in position is fairly consistent.

 

1085_1.jpg

 

The 1m UTM precision stuff is one of those other subjects and if one has ever tried inputing a UTM data set in 1m increments some might have noticed that many units will echo different UTM values (due to internal precision issues) to those entered.

 

Basically input UTM/zone gets converted to Lat/Long then to Cartesian XYZ (for storing) then the whole process is reversed and in the process one doesn't get echoed back what one input. UTM is linked to lat/Long etc and really this 1m UTM isn't any different (accuracy wise) to the Lat/Long it reflects.

 

Cheers, Kerry.

Edited by Kerry.
Link to comment

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
×
×
  • Create New...