Where Is My Cache Really

[+geckoee]

Yet another way point averaging thread.

 

I have been asking myself this question for a while, and seen other people ask, and seen answers as well. There are some good links out there that people have referenced, but from a geo caching perspective I wanted a clearer answer.

 

What do I have to do to record the “correct coordinates” for my Geo Cache I placed?

Definitions as used in this question:

correct coordinates – known accurate to 0.001 minutes as observed by a Garmin ETrex Yellow. This model was chosen as a rough approximation of the standard 12 channel consumer GPSr.

 

Two common answers:

1. Average waypoints over a time interval (usually people suggest 5 to 10 minutes)

2. Take multiple waypoints on different days at different times.

1.5. Do both. Average over a time interval at different times throughout the day.

 

The question remains; how long must I average for? and how many way points?

 

Background understanding:

Satellite geometries at any specific location repeats on a 24 hour cycle. This makes different times during the day important when way point averaging.

GPS locations reported by a consumer receiver “walk about” and don’t jump sporadically. This has to do with atmospheric effects, satellite geometry, the mechanisms employed in the GPSr software to create a position, as well as other factors. By zooming in on data being collected by a programs (such as ExpertGPS, GPS Trackmaker, or any other program that displays NMEA or similar data, ) this can be observed.

The DD MM.mmmm NMEA format was chosen over UTM because of the higher resolution. On the display of our GPSr we are used to seeing 3 digits of precision after the decimal point, but NMEA actually sends out 4, or a resolution of about 7 inches. This is about five times higher than UTM.

 

Set Up and Analysis:

GPS was placed on the dash of a passenger vehicle. A car was chosen for it ability to be placed out in the open and have the content of the vehicle securely locked. This particular vehicle was chosen for its large slanting wind shield, for a better satellite view, and lack of polarization or tinting, so signals are not disturbed too much. Data collected using ExpertGPS and ported over to MSExcel for data analysis.

Results:
                 Worst case   average precision (50%)
No average        0.00654      0.00131
5 min average     0.00470      0.00106
10 min average    0.00300      0.00095
20 min average    0.00188      0.00074
30 min average    0.00154      0.00068

5 point average   0.00208      0.00059
10 point average  0.00151      0.00057
5 10 min averages 0.00219      0.00107

 

Explanation of Results:

All data is in Minutes. At my location the majority of the error was in latitude.

 

These definitions get a little rambly. 5 min average is a rolling average over 5 minutes. The 10 to 30 min averages are rolling averages over 10 to 30 minutes. The 5 point average is 5 waypoints taken about 2 minutes apart (worst case and average precision are derived from similar rolling average methods) The 5 10 min average is five different 10 minute averages, all average together.

 

Conclusion:

 

Method 1: 0.001 minute resolution requires over a 30 minute average, but most of the time a 5 minute average will suffice.

 

Method 2: Ten waypoints averaged through the day yields nearly 0.001 accuracy.

 

Method 1.5 Five averages of 10 minutes will get you close most of the time.

 

The data used for these calculations was very limited. Only 4500 data points, 2 seconds apart were collected. All data was collected during a 2 ½ hour ideal geometric configuration as determined by Trimble Software.

 

I would be curious to see what data other people have collected in other locations… or know if someone actually read this far! If you want to see the raw data, drop me a line, and I will email it off to you.

 

Edit to clear up the table a bit

Edited February 8, 2004 by geckoee

[+bob393]

Wow, bravo!

I did read it all the way through.

I use an eTrex Vista and it dosn't support waypoint averging, I wish it did. I never went as far as you did on your experiment. I usualy just approach ground zero from several directions and compare the results. Sadly it dosn't always work.

Thanks for the data and the ideas.

[+Tiwica]

All the time it took to do the experiment and the time it took you to write it all up would have been better spent caching. Too much free time???? Averaging does not make effective use of time. It's just not worth it. The best thing is to just let the GPSr settle into a spot for five minutes or so and go with that. Just MyOp. Where's the fun if I can just walk up to it. I wouldn't be able to anyway, unless I was using YOUR GPSr. There all different........even two of the same kind. Average away if it pleases you.

[+4agers]

Well done if you ask me. I do have one question though. How can you really test accuracy if you don't have a single point that is known to be accurate? I don't think we can get that with a consumer grade etrex. (Anybody have military targeting system we can borrow) What is really being tested is the variation.

 

I just picked up GPS Tuner for my PocketPC to see how it does with averaging. I will be placing a cache in the next week or so and will record some results and see how it comes out.

 

If nothing else, this is a good exercise in understanding.

[+geckoee]

<snip> ... What is really being tested is the variation.

<snip> If nothing else, this is a good exercise in understanding.

You're right about the variation being under test, not really the accuracy.

 

The average of the data was used as the "real" location. From topo maps, aerial photos, and bearings to nearby targets it is pretty close, but that can only get you within about a meter. Considering the variation was up to 20 feet I guess it is relatively close to the real position.

 

After collecting the data, and analyzing it, I feel I have a better understanding of GPSr location data, and how to use it.

 

Thanks for the feedback.

[+briansnat]

I find averaging to be a complete waste of time. When I first started, I'd take about 20 readings and average them. Did that for the first 10, or so caches I placed. Then I got lazy and started taking one reading and going with it. 60+ hides later, I've found no difference in accuracy. I've had complaints about exactly two of my caches. One was one I averaged and one was for one I didn't (and the latter was in a ravine with bad reception to begin with).

 

If you have a good lock, go with it. If you get more than one complaint, then go back.

Edited February 8, 2004 by briansnat

[ArtMan]

How can you really test accuracy if you don't have a single point that is known to be accurate? 

To test your GPSr against a known location, I recommend you check out the benchmark section of this site. Wherever you are, chances are you are within a short distance of a location with latitude and longitude determined very accurately by professional surveyors, possibly using tens of thousands of dollars worth of survey-grade GPS equipment, with results measured in hundred-thousandths of a second, e.g. 58º 18' 25.12345"

 

But a cautionary note, some benchmarks (true bench marks, in surveyor parlance) are very accurate in terms of elevation but are only approximate in terms of latitude and longitude. Watch for this language - "The horizontal coordinates were scaled from a topographic map and have an estimated accuracy of +/- 6 seconds" - at the bottom of the page. In my experience such scaled locations are generally rather accurate, but you can't rely on it, so if you're testing out your GPS stick to triangulation stations or others that do not use scaled coordinates.

[+Renegade Knight]

Since my caches are never where I placed them when I check up on them, even if I started with perfect coordinates, the coordinates become nothing more than an estimate of location.

[Trogdor!]

Yes I read it. There is a big difference between precision and accuracy. Example; you shoot at a square target that is one foot on a side. You take four shots and completely miss the target by 2 feet, but have spacing between hits of less than 1/32 of an inch. You would be considered to be precise but not accurate. You take four more shots and each shot hits each corner (so all four corners have been hit), your average would tell you that you were accurate, but you wouldn’t be very precise.

 

Here is my deal, I don’t understand why one would be worried about trying to be precise when by the nature of the of the equipment they can’t. Your GPSR displays a reading of a calculation based on best available data at any given time. There is no right or wrong answer as the conditions are always changing. Even if you were able to somehow continuously average the data at a specific point for an extended period of time (which would yield what is hoped to be an accurate and precise location) those who are looking to find that location are still limited by what they are using to find it.

 

Too many times I’ve seen people log finds and then complain about poor coordinates. As long as your GPSR gets you close enough to find it, you’re good.

 

All of that being said, when I place a cache, I take 10 readings over a period of half an hour, I then throw out the highest and lowest values and average the rest. I think of this as being more of a ritual I do than as being a process in an attempt to be accurate or precise.

[+briansnat]

I don’t understand why one would be worried about trying to be precise when by the nature of the of the equipment they can’t. Your GPSR displays a reading of a calculation based on best available data at any given time. There is no right or wrong answer as the conditions are always changing.

 

This is absolutely right and the reason averaging doesn't always provide better accuracy. In fact at times it can provide worse if you're averaging bad data. Even if you average over several days, how do you know which day is better? Say the coords for day one were dead on correct, day two they were off by 20 feet and day 3 by 50. By taking an average, you're definitely going to be off. If you took 1 reading, you'd have a 1 in 3 chance of being dead on, and even in the worst case, you're only 50 feet off, which is well within anybody's searching range.

 

I don't average anymore and if you read the logs for my caches you'll see comments like "Brian's coords were dead on as usual". Instead of wasting your time and beating your head over getting the coords accurate down to a foot, use the time to place, or find another chache!

Edited February 9, 2004 by briansnat

[+geckoee]

Artman, I’ve only been able to find vertical stations near my area. There are some triangulation stations, but I can’t park a GPS on it for over two hours unless I sit there and watch it the whole time. I’m procrastinating, but I can’t get away with that much

 

Renegade Knight, good point.

 

Trogdor, I think you answered your own question. The idea is to get a more precise reading from something that is not that precise. Also, thanks for explaining accuracy vs precision I left that out of the post, but I guess it’s needed in case some one is un clear on the concept.

 

briansnat, The idea is that if you have a large enough continuous data set the average will be close to the real. Coordinates tend to walk about 20 feet around the real location in 45 mins. If you average all of those points, or a little smaller sub set you will get a result that is close to the real. Keep in mind this is a small sampling of data, and all of it was collect when the satellites were in an optimal geometric configuration.

 

I may not have go into enough depth about what a rolling average is, or the patterns of walking position reported by a GPS. Oh well, I’m not going to go into it. Also it might not be practical as it turns out for Geo-caching, but an interesting line of thought any ways.

 

Edit to clean things up a bit. It gets dirty on the first read through

Edited February 9, 2004 by geckoee