Jump to content

Two Identical GPS Units -- Different Accuracies?


Keo1

Recommended Posts

ecanderson,

Where are you in Colorado?

I'm now in W CO (GJ) but was in SW CO Lake City, and in that area (SW) Correct NAD83/WGS84 coordinates for a point if incorrectly entered as NAD27 Datum would show the point to be a small amount more East but a Bunch more North. (Z13)

EX: WGS84 NAD27

301338E 301379

4234554N 4234765

 

Here in W CO ( barely into Z12) it's a little different......700709 vs700850E.....and 4323314 vs 4323109N ....(84 first then 27 second)

Edited by Grasscatcher
Link to comment
ecanderson,

Where are you in Colorado?

Front Range area, up north of Denver. Our N/S on NAD27 produces tiny errors (0.001 ~ 0.002). It clobbers us the other way. As an example of one of my caches over here - which are in the same general area as the caches I was mentioning in the earlier posts:

 

GC1B035

 

NAD27

N 40°12.198

W 105°04.030

 

WGS84

N 40° 12.197 (almost no diff)

W 105° 04.065 (whopping big diff!)

Link to comment
We're talking about someone with literally hundreds of hides, some 75~100 or so of those produced during the "problem period". What other "user error" do you envision that can get these results?
Don't know - in spite of the amount of information you've given, anything else I offer is just more guessing. Good puzzle though, especially the bit about how the problem seem to pop up after an otherwise long and successful history with the device.

 

Rather than guessing, the only way I could think to narrow down the possibilities would be to get the "funny" GPS in hand for testing. When he got a new GPS, I hope he didn't sell the old one to someone here on the Garage Sale :D But if he did, it'd be interesting to follow-up with the new owner.

 

AND... rolling back to earlier in the thread: We've drifted into a discussion now of several other folks using a variety of different GPS units to look for caches from coords posted by someone else, without anyone tagging along on the original placements or doing a side-by-side check of devices. That's not what we were getting at earlier -- repeated tests of one device against a fixed known location. The whole point of mentioning benchmarks (or mailboxes!) was a general agreement that user placed caches are not a reliable way to guage GPS accuracy.

Edited by lee_rimar
Link to comment

Some interesting comments. Some that made me think I need to recheck some things. So far I've gleaned this out of all that's posted:

 

1. Should be referenced to a known geodetic marker. -- Done that, the SE Range & Township surveyor mark and my land-grabbing neighbors which started this all, which lead to my head-scratching about accuracy and why I started this thread. My original post about the GPS Accuracy Anecdote being within 2 INCHES of a surveyor's official spot to my GPS's >1/4th mile projected waypoint.

 

2. Accuracy vs. Repeatability. Well, the permanent and well-used geodetic survey mark is about 1/4th mile from my house through some dense bramble woods in 2+ feet of snow, it's not been above 0º F. for over a week now (-30º to -20º F. the norm) ... I'll have to test it this spring on that marker, as opposed to my mailbox for reproducibility which is a more realistic task that I'm willing to take on this winter. Using that official benchmark back in the woods, then I can test for both.

 

3. Others are reporting reproducible discrepancies. I feel somewhat vindicated with that. Some can have wide scatter patterns, some can have a consistent offset. What is creating that is another issue. User or environment GIGO or GPS unit manufacturing tolerances? Much of the reproduced discrepancies allude to manufacturer's tolerances. Lemons & Gems (I know, everyone else uses "Lemons vs. Cherries". I never cared greatly for cherries so I don't equate them to be the better of the two.)

 

4. It's started an interesting discussion that covers some points that I don't think all of us were aware of before, so I don't feel so bad for starting it now. :D

 

5. It would be interesting if those who publish reviews on these devices would give GPS units a score for these things, reproducibility and accuracy. But then if the reviewer happened to get a "lemon" for testing ... well, is that any different with any other consumer market and online reviews of any item sold today? Companies have been known to "cherry pick" merchandise to send to reviewers.

 

6. If nothing else, it might keep the manufacturers on their toes. Tightening up tolerances in their factories. Making sure their antennas and circuitry components are tuned to an F.R.C.H. (anyone remember what that means in electronics design? :D If you do, don't say it, I hear this is a family-oriented forum) with every batch to ensure the utmost in reception clarity, putting the whip to their programmers to produce the very best algorithms possible, etc. All good points for the consumer. If more people tested the accuracy and reproducibility of their units it might encourage better firmware upgrades from the provider.

 

7. So far, I'm going to sit warm in my sub-zero home for now, and think I got the Herbie Love Bug or Christine of all GPS units ever made. One of the lower-cost units now outperforming even the $200,000 ones used by the government. :) (I'm allowed some psychoses due to cabin-fever and living alone in the woods without seeing another human face for months at a time. The voices said so.)

 

(p.s. 8. Yes, I could find more challenging caches to find (my disinterest in geocaching for game's sake), but I've never been much of a gamer (I'll occasionally play computer monopoly just to tromp the computer-opponent bunny who pops out of the hat token, but that's about it). I've always invented my own rules in life. I'm never much compelled to follow other's rules or ideas. Jus' who I am is all. Which ... has lead to some interesting discoveries in life, some that nobody else has ever made. "Doesn't play well with others" has its good points. "Discovery consists of seeing what everyone else has seen and thinking what nobody else has thought." by I. Forget.)

 

edit: some minor typos

Edited by Keo1
Link to comment

....

[*]ecanderson: "...consistent offset..."

[*]Renegade Knight: "...always 12' left for no particular reason..."

[*]GrassCatcher "...Consistently wrong is just as probable as consistently right..."

....

When I've done that kind of testing, a small (good) pattern has always been centered on the right place; and a wide (poor) pattern doesn't show enough consistency to look for an offset....

 

Because of how a GPS works what you observe, is what I would normally expect to see. What ecanderson and I said was essentially the same type of error. Where I would NORMALY expect this to show up is when a datum error is involved. That would not be a GPS problem, but it's how you might figure out your using two datums and a fairly common nOOb issue when they are 300 feet from the cache consistantly and can't figure out why. However there are other (and likely very rare) conditions that could cause this. For example a corrupt figure in an internal table or number getting corrupted that the GPS uses for it's calcuations.

 

A simpler example source of this kind of error would be a steel tape like surveyors use to measure distance. It expands and contracts. This is fairly linear with temp. By taking the temperature when you do a survey you can adjust your measurmetns to be more accurate using a simple equeation. Most of these kind of errors in the GPS system are no doubt already factored into the equations leaving you with random error like you have observed. It's only going to be the oddball and goofy things like a corrupt table where you may ever see it.

Link to comment

RenegadeKnight, I think we're talking about a couple of overlapping (but different) subjects.

 

Point granted: A benchmark is better than an arbitrary fixed point in that you know what the number should read when you get there. That way you can spot "n00b issues" like wrong datum, or filter location & situation issues (if you're in a canyon, or perhaps next to a powerplant throwing off a lot of electrical interference).

 

But...

 

In the question's original context -- Keo1's "lemons and gems" scenario, and comparing supposedly identical units? In that realm, comparing hardware for precision and repeatability, there is no difference between using an established benchmark or any other known fixed location.

 

Ecandersons' cacher posting coords that were "consistently offset" is an anecdote without enough controlled testing to conclude it was a hardware problem. And "...other (and likely very rare) conditions ... a corrupt figure in an internal table or number getting corrupted that the GPS uses for it's calcuations..." are just what you've said -- very rare, if they exist at all. For such defects to exist but still allow repeatable, precise, and WRONG answers would be even more rare -- far more likely that a defective unit would wander, lose lock, give imprecise results or fail to get a fix at all. A GPS receiver is not a steel tape measure.

 

If you ever lay hands on a GPS receiver that does gives precise, repeatable, yet incorrect results or you find one showing a "consistent offset" -- and you're sure it's a hardware issue rather than user errors or situation/location specific causes -- let me know. I'd like to see it and discuss the matter further.

Edited by lee_rimar
Link to comment

If you ever lay hands on a GPS receiver that does gives precise, repeatable, yet incorrect results or you find one showing a "consistent offset" -- and you're sure it's a hardware issue rather than user errors or situation/location specific causes -- let me know. I'd like to see it and discuss the matter further.

I'm going to be contacting them this weekend, and will see if we can sequester that unit. If they don't still have it, I hope it's in a landfill somewhere and not in some poor noob's hands.

 

My additional hope is that it turns out to be something simple: a bias that exists only prior to allowing the unit to settle down. I've never tried taking "snap readings" with either of my units - always do long averages when placing caches, and won't even take a reading if the EPE happens to suck on a given trip - so I have no idea at all as to what might be "typical" for Garmin's units when used/abused in that fashion.

Link to comment

...a bias that exists only prior to allowing the unit settle down...

 

THAT would be fascinating and a useful bit of trivia of know about -- especially is it affected a whole family of devices instead of just one oddball. Kinda like older Magellan slingshot effect.

 

...no idea at all as to what might be "typical" for Garmin's units when used/abused in that fashion...

 

Glad you used the word "abused" there :P You might have me caught out on this one. If that one device or its whole family gives consistent, repeatable, but incorrect results (biased 25' north) in specific situations (until averaging kicks in? weak signal?) - is it a "hardware feature"? Or user error for not applying proper technique?

Edited by lee_rimar
Link to comment

If the weather EVER improves enough around here to cause me to get out into some open field testing (a fair weather cacher? me?) I'm going to try to settle some questions for myself once and for all. Have about 75% of a test plan completed that should be pretty comprehensive... right down to the shape of the constellation and position of WAAS/EGNOS satellites during each phase of the EPE test. Should be some good amusement value when it's completed. My only regret is that this site (for all of its income) seems to refuse to locally host any images, so to keep the graphs with the reports, it'll all have to be hosted somewhere else. You'd think Groundspeak could set a per-user budget in MB for images (perhaps enhance for those of us who pay the annual freight for this place). Other boards - even those who have no subscription mechanism at all - seem to be able to manage this on the weight of their ad revenue alone. Ah well ... that's another topic altogether.

 

Anyway

 

Will be focusing on the variables that drive EPE figures (understanding that the algorithms in different models and within that, different firmware levels are wildly different), repeatability, and accuracy.

 

One of the things that this particular thread will have added to the testing is a section on what I'm going to call 'snap' vs. proper settling vs. averaging readings -- what kind of results can be expected before the EPE settles down, and what kind of results even after they've settled? Even after getting a 'device minimum' EPE, what are the results of using the waypoint averaging functions on these units vs. taking a single reading after settling to that minimum EPE? What does it mean when you use one of the newer (e.g., Oregon or Dakota) units that shows a confidence level in percent for the displayed coordinates vs. a specific number of 1 second samples on a 60 or eTrex when doing the same kind of waypoint position averaging?

 

Here's a ringer ... What does it mean when a GPSr is indicating a distance to target that is below the granularity of any of the displayable coordinate systems of the unit? To what resolution is the hardware and firmware in these units really trying to operate?!? Using our typical decimal minutes, you don't resolve any tighter than a 4' x 6' (0.001 E/W vs N/S) area at my latitude. Using decimal degrees, you're looking at 2.5' x 3.5' resolution. The spec for the SiRFstarIII chip, if that's what you've got, is <2.5m (8.2 feet), 50% confidence, 24 hr static position (like that's gonna happen), assuming a -130 dBm set of signals or better.

 

In short, the objective is to come up with a "Best Practices for Placing" piece along with a reasonably accurate data set that shows what is likely to occur when you deviate from those practices. At the same time, I hope to come up with some hard numbers vs. people's expectations (including my own) of this hardware. FWIW, I have this sense that the repeatability of some units may be higher than we think, at least when the target is approached with comparable satellite configurations.

Link to comment
Here's a ringer ... What does it mean when a GPSr is indicating a distance to target that is below the granularity of any of the displayable coordinate systems of the unit?
That's a really good question, one I've specifically raised a few times myself. My PN-40 has on more thn one occassion displayed numbers like this:

 

4a170bd8-da0e-4535-9e85-a513b2d1c2df.jpg

 

For our French friends, 0.05 feet would be 15.24 millimetres. All this means is that internally the device is running a calculation past the real usable number of significant figures, and not bothering to round a result. But how do you know what the usable increment really is?

Edited by lee_rimar
Link to comment
My only regret is that this site (for all of its income) seems to refuse to locally host any images,
It doesn't? Where do users store their avatars and cache log/gallery items?

I was referring to the boards here, not the geocaching.com site where attachments are no problem. Here, the only "Insert" I can find is to add a URL to some outside source. Am I missing something?
Link to comment

You're not missing anything; it's geocaching.com that hosts the images, not the message forums. But any image you add to a cache log, or avatar/image in your public profile -- can be inserted into a message here. You just need to go round-about to get the location on the geocaching.com server, and use the IMG tag:

[img=http://img.geocaching.com/cache/log/4a170bd8-da0e-4535-9e85-a513b2d1c2df.jpg]

It would be nicer if you could just upload/attach images when composing though...

Edited by lee_rimar
Link to comment
ecanderson,

Where are you in Colorado?

Front Range area, up north of Denver. Our N/S on NAD27 produces tiny errors (0.001 ~ 0.002). It clobbers us the other way. As an example of one of my caches over here - which are in the same general area as the caches I was mentioning in the earlier posts:

 

GC1B035

 

NAD27

N 40°12.198

W 105°04.030

 

WGS84

N 40° 12.197 (almost no diff)

W 105° 04.065 (whopping big diff!)

 

Interesting that you compare NAD27 with WGS84!

Remember that these datums are a snap shot of time, i.e. the 27 in NAD27 is a mathimatical model taken for 1927 and I think you'll find it is superceded. WGS84 is 1984.

I also thought NAD27 was a transformation of WGS84, i.e a 2D mathimatical model transformed from a 3D model. If you input Northings & Eastings (NAD27) and the setting of your GPS displayed as Lats & Longs (WGS84) there may be a transformation difference, however I doubt it. The error should be within HHGPS accuracy.

Anyway, food for thought.

Link to comment

Interesting that you compare NAD27 with WGS84!

Only because that's the only alternate datum that I've seen accidentally used, and it serves to show that any accidental use of it would have been far more obvious in our area. There are indeed places where the difference vs. WGS84 would be 25'N/S and nil E/W, but NOT in our area.

 

Remember that these datums are a snap shot of time, i.e. the 27 in NAD27 is a mathimatical model taken for 1927 and I think you'll find it is superceded. WGS84 is 1984.
Yes. And?

 

I also thought NAD27 was a transformation of WGS84
I did not.

 

The reason for bringing up the delta was to point out that if the problem we had been discussing was operator error, in this particular geographic area, it wasn't due to a simple NAD27 datum selection error - the results of that mistake would have been quite different. The delta between the two varies all across the country. Here, the delta is almost nil N/S, and huge E/W. Ergo, whatever is causing one user here a 25'N shift cannot be attributed to that.

Link to comment

Apologies, I need to correct myself.

My quote "I also thought NAD27 was a transformation of WGS84" is not correct. NAD27 is a model to best fit for the North American Continent post 1927. When a world datum was required due to the role of GPS around the world, a new model NAD83, was adopted to fit as close to WGS84 (GPS Datum) as possible.

source:-North American Datum

 

Apologies again, I may have jumped onto this thread prematurely and may have missed a point or two.

GPS is a multi headed animal bound to give the operator a false sense of security, i.e. clock errors, multipath, datum variables, SV geometry, etc

Link to comment

Simple test for anyone who wants to find out just how consistent their GPSr is:

 

Ideally, first download the Trimble Planning Tool, and notice how the number of satellites and DOP will vary over time at your location.

 

Go to a known fixed location - preferably a benchmark with known accurate coordinates, but your mailbox will do. Wait a couple of minutes for your GPSr to "settle" (yes, it does matter!), and record the coordinates. Use a coordinate format which allows the highest inferred precision - UTM is good, because it will report your location to 1-metre precision in both E-W and N-S directions. Also, make a note of the EPE and number of visible satellites as reported by the unit, and make a note of the number of satellites and DOP as reported by Trimble Planning Tool.

 

Go back and repeat a few hours later the same day - try to choose a time when the Trimble Planning Tool reports a significantly different number of satellites and DOP to your first visit.

 

Repeat again later the same day.

 

Repeat the above series of tests the next day ... and the next .. and the next. (Note - if you check the Trimble Planning Tool, you will see there is general tendency for the number of satellites and DOP to be more or less repeated on a 24-hour cycle. For the most meaningful results, you should NOT repeat the exercise on a rigid 24-hour cycle. 9:00 a.m. on 11 January is going to be ROUGHLY the same as 9:00 a.m. on 12 January is going to be ROUGHLY the same as 9:00 a.m. on 13 January ... but 9:00 a.m. 11 January is probably going to be quite different to 3:00 p.m. 11 January.)

 

And now here is the clincher - to qualify as a meaningful scientific experiment, there has to be a prediction:

 

All of your position records taken at times of day when Trimble Planning Tool shows a low DOP are likely to be quite tightly clustered around a common location (the centroid of which SHOULD correspond to the true location) - say plus or minus 3 metres in any direction if you have a modern high-sensitivity GPSr, good sky visibility and limited multi-path at your location AND the reported EPE at these times will probably be quite low (perhaps 3- 4 metres?)

 

Your position records which are taken when Trimble Planning Tool predicts a high DOP are likely to be scattered over a larger distance (although note that some will probably still be "spot on"!) AND the reported EPE will probably be higher.

 

When someone completes this exercise and can show that their unit consistently gives a fix within 2 metres of a single location at all times of day or night, I will be truly impressed! (Because it can't happen, according to theory which underpins the whole GPS system!)

 

(P.S. Of course, the easy way to collect data is to set your track log to record your position at regular time intervals - say every 10 minutes - then set your GPSr in a fixed location for 12 hours or more. You can then download the track log and see how much the recorded position moves around as the satellite positions change.)

Edited by julianh
Link to comment

I lost my vista hcx, and replaced it with another vista hcx. The first one was factory refurbished, but you wouldn't have known it, it even had the screen protector if I remember correctly. I had to do an update and reset but after that it worked fine. There was a backlight flickering problem but it never got any worse and I got to where I didn't notice it. It was considerabley dimmer than my venture but I figured that's why the batteries lived longer.

 

The new one is as bright as my venture, and still has a long battery life. Haven't noticed any flickering either.

 

My new one has a flourescent orange lanyard, and a reflective cord lanyard. Also has a caribeener on the garmin lanyard. I may put a floaty thing on it as well.

Link to comment

Simple test for anyone who wants to find out just how consistent their GPSr is:

 

Ideally, first download the Trimble Planning Tool, and notice how the number of satellites and DOP will vary over time at your location.

 

Go to a known fixed location - preferably a benchmark with known accurate coordinates, but your mailbox will do. Wait a couple of minutes for your GPSr to "settle" (yes, it does matter!), and record the coordinates. Use a coordinate format which allows the highest inferred precision - UTM is good, because it will report your location to 1-metre precision in both E-W and N-S directions. Also, make a note of the EPE and number of visible satellites as reported by the unit, and make a note of the number of satellites and DOP as reported by Trimble Planning Tool.

 

Go back and repeat a few hours later the same day - try to choose a time when the Trimble Planning Tool reports a significantly different number of satellites and DOP to your first visit.

 

Repeat again later the same day.

 

Repeat the above series of tests the next day ... and the next .. and the next. (Note - if you check the Trimble Planning Tool, you will see there is general tendency for the number of satellites and DOP to be more or less repeated on a 24-hour cycle. For the most meaningful results, you should NOT repeat the exercise on a rigid 24-hour cycle. 9:00 a.m. on 11 January is going to be ROUGHLY the same as 9:00 a.m. on 12 January is going to be ROUGHLY the same as 9:00 a.m. on 13 January ... but 9:00 a.m. 11 January is probably going to be quite different to 3:00 p.m. 11 January.)

 

And now here is the clincher - to qualify as a meaningful scientific experiment, there has to be a prediction:

 

All of your position records taken at times of day when Trimble Planning Tool shows a low DOP are likely to be quite tightly clustered around a common location (the centroid of which SHOULD correspond to the true location) - say plus or minus 3 metres in any direction if you have a modern high-sensitivity GPSr, good sky visibility and limited multi-path at your location AND the reported EPE at these times will probably be quite low (perhaps 3- 4 metres?)

 

Your position records which are taken when Trimble Planning Tool predicts a high DOP are likely to be scattered over a larger distance (although note that some will probably still be "spot on"!) AND the reported EPE will probably be higher.

 

When someone completes this exercise and can show that their unit consistently gives a fix within 2 metres of a single location at all times of day or night, I will be truly impressed! (Because it can't happen, according to theory which underpins the whole GPS system!)

 

(P.S. Of course, the easy way to collect data is to set your track log to record your position at regular time intervals - say every 10 minutes - then set your GPSr in a fixed location for 12 hours or more. You can then download the track log and see how much the recorded position moves around as the satellite positions change.)

 

dadgum, that's a lot of work! My brain hurts!

Link to comment

dadgum, that's a lot of work! My brain hurts!

Here's the short version! :)

 

Set your track log to record your position at regular time intervals - say every 10 minutes - then set your GPSr in a fixed location for 12 hours or more. You can then see how much the recorded position moves around as the satellite positions change.

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...