We tied some monuments set in 1966 on a Record of Survey. The ground is flat ancient sea floor (uplifted Mendocino coast), a lot of pine and Douglas fir canopy plus thick brush in places. We located the south quarter corner of Section 22, the 1/64th corner nominally 660' north, and the 1/64th corner nominally 660' west of that using mostly a conventional network of control. They fit the 1966 R/S within less than two tenths.
Then I tied the next 1/64th corner with RTK nominally 660' west (1320' west and 660' north of the quarter section corner). A wind event had blown down small trees and brush. When they cleared the gravel road of course they threw all the slash on top of the monument. When I staked it out the Controller said it was about a half a foot away. I'm standing on a 12" log that missed the monument by a foot or so but I have the pole on the ground. I look down through the brush and there it is about a half a foot away. So I get the loppers and clear away the brush much faster than I could've done with a machete.
The machine says it lost initialization then it says initialization has been gained (the Trimble lady said it with a lisp) then she said poo-er Pee-Dop then it said poo-er precisions then I got a shot and it measured 3.5' short. I tied a reference rebar set on a 1975 Parcel Map (which calls the four concrete monuments at its corners off by small amounts without showing any other found monuments-figure that out) which fit very close. It re-initialized before the rebar when I crawled out of the brush.
I went back later that afternoon and got a shot that I believe to be good. I dumped the initialization and got a third shot which checked 0.03 from the second one. It checks very close to the 1966 measurement.
My limited experience with Access RTK is it seems like that if shows 0.04' or less with good PDOP it seems to be truthful. But if it shows a poor precision of say 0.20' it could be 10 times worse than that.
> My limited experience with Access RTK is it seems like that if shows 0.04' or less with good PDOP it seems to be truthful. But if it shows a poor precision of say 0.20' it could be 10 times worse than that.
It depends on the GPS unit you are using.
The R8 is very optimistic in the accuracies it gives
The R10's numbers are more realistic
But if it shows a poor precision of say 0.20' it could be 10 times worse than that.
Sounds right, when it's acting up, it's probably a bad init, suspect everything.
Hi Dave,
My RTK never gives bad fixes. Your problem is that you should be using Brand-J;-) . Ok, now that we got that out of the way, I think you're right about watching those RMS values. Not necessarily because the RMS values represent reality all of the time, but because it may be revealing when the fix is on the ragged edge. An example would be surveyors relying on the RMS generated by float solutions. Hmmm. I've seen float solutions provide statistics that would indicate the solution is submeter, but in reality were three to five meters off. The RMS didn't match reality, but it did indicate that the solution was not fixed.
In my opinion, there are precisely two variables a surveyor has control of with RTK (or any GNSS for that matter): time on site and repetition. Environmental considerations (trees, buildings, etc.) are generally beyond our control; satellite constellation is beyond our control (although we do have some control over when we observe after studying the mission planning); atmospheric considerations are beyond our control (ionospheric disturbance from the sun, tropospheric disturbance due to weather, etc.). For the most part, we have time on site and repetition.
Bad initializations, experientially speaking, only last for a minute or two, then the solution falls apart as the constellation changes. So, time on site alone can help identify a bad initialization. I recall when we were working on the Columbia shuttle recovery a long time ago, the nasally GIS professor from the local university was screeching to everyone "3 clicks and go", meaning collect for three seconds and go. Precision wasn't terribly critical for mapping the debris field, but the phrase kind of stuck as sophomoric and we still laugh about the phrase today.
Repeating the fix can also help. It is possible to get the same wrong answer twice in a very short time span, this is particularly true with NLOS (non-line of sight signals, where you pick up the reflection, but no the direct signal). Getting the same wrong answer twice is highly unlikely if there is some time separation between fixes (even just two or three minutes).
I suppose it is worth mentioning that there is a third option: putting the receiver in the truck and getting the total station out. But, who wants to do that? Especially if you'd try Brand-J;-) .
Bad initializations, experientially speaking, only last for a minute or two, then the solution falls apart as the constellation changes. So, time on site alone can help identify a bad initialization.
My experience is the same as what you found.
Now.... I need a new J....!
N
Second the experience;-) with Brand J
I set 6 lot corners once on a bad init, I went back to check them and they were all about .7' off; that bad init held for 20 minutes. so don't think it will always be quick, usually it will, not always.
I also have had bad intializations hold for a while. This is usually when you walk by something, like a tree or building and it loses lock, regains it while you are in poor conditions, then you walk out in the open. This happened to me with our old system, an early Hiper. Our newer system, HiperII's seem much better about this problem. As already said, if you are watching the screen, you can usually see the problem occur but you aren't always looking at the data collector when walking.
Moe
This is the kind of thing that bothers me. In Arkansas, we rarely get between any corners, without a re-init. But, this is serious. Thanks for sharing. I'd like to see a complete list, of things that can and do go wrong with RTK.
That one is definitely "one for the book".
N
Moe
That happened in the early days, maybe 1996 or so, you didn't want to reinit cause it could take a while, I would set a string of points and go back through them with another init, usually the next day. Having 6 of them be off was an eye opener, from that day forward I did more onsite checks, as I went, pulling those 6 up was a real pain. 🙁
Now the error trapping is much better, I haven't ever seen one with the R10, but we still check it all, you need "proof" just like with the instrument, running traverses and setting corners with a single tie, you still need another one.
I have run into a few environments with repeat bad inits that matched. It is rare but not impossible. I just had such a case several weeks ago.
I realize you were being tongue in cheek but this has to be said: Why anyone would NOT put the GPS in the truck and use a tool that works in bad environments is beyond me. The maps we publish impact things people work their whole lives for. If folks don't want to do proper work it's time to find another career...
Definitely tongue in cheek. It's possible to spend more time waiting for a questionable fix in a bad environment than it would require to simply use a total station and get reliable results quickly. This is true regardless of brand. The physics of differential positioning make some sites absolutely unsuitable for any GNSS positioning. In reality, I like to stress the absolute importance of professional discretion. At some point, it's up to the professional to make the call regarding what tool to use and how to use it, and the professional is responsible for the consequences of those decisions.
One time I tried to get a fix in tree cover during the winter. It was pretty thick with some large trees. I was trying to get the shot, then I was going to traverse in and shoot it with a total station. If it checked I wasn't going to loop the traverse. I got three fixed shots, none of which were good. It was pretty much waste of time but you don't know unless you try.