RTK Part Zzzzzz
> > I think that can be fairly characterized as essentially cheerleading. This has been a recurring topic, i.e. how to estimate the uncertainties in positions derived via RTK and essentially all you've had to offer is "why don't you buy an RTK system?". Either that or "it works great in England".
> Your are selectively mischaracterizing what I and others have posted.
Actually, that's a fair summary of at least 80% of your posts on this topic over the past year. It's not as if this is a new topic or a trivial one, but pretty much all you've had to say was to suggest that the answer would be revealed simply from using RTK equipment, which was and is laughable as a proposition.
You've cited a study made in the United Kingdom of the accuracies of network RTK as delivered by a network there. "It works great in England" isn't very wide of the mark as a summary of that.
RTK Part Z
>The same problem that you don't seem to want to think about now will continue, i.e. how to estimate the uncertainties of positions derived via RTK methods.
INDEPENDENT, REDUNDANT OCCUPATIONS!!!
insert into favorite least squares program and voila
Not sure why this is so difficult to understand for a GPS master...
RTK Part Zzzzzz
[sarcasm]Just look at the bright side. Kent wasn't around when man invented the wheel so we get to drive cars![/sarcasm]
RTK Part Z
> >The same problem that you don't seem to want to think about now will continue, i.e. how to estimate the uncertainties of positions derived via RTK methods.
>
> INDEPENDENT, REDUNDANT OCCUPATIONS!!!
> insert into favorite least squares program and voila
>
> Not sure why this is so difficult to understand for a GPS master...
The missing piece, of course, consists of the weights of the observations and what factor to apply to processor-estimated variances. This is standard for other GPS work and should not be anything novel for RTK GPS.
If you are working efficiently, you'll have enough redundancy to identify blunders, but won't be trying to independently derive the standard errors of conventional observations and variance factors for GPS vectors on each survey. That is anything but efficient.
The whole point of testing is to identify what the reasonable limits of the variance factors are.
RTK Part Z
> > >The same problem that you don't seem to want to think about now will continue, i.e. how to estimate the uncertainties of positions derived via RTK methods.
> >
> > INDEPENDENT, REDUNDANT OCCUPATIONS!!!
> > insert into favorite least squares program and voila
> >
> > Not sure why this is so difficult to understand for a GPS master...
>
> The missing piece, of course, consists of the weights of the observations and what factor to apply to processor-estimated variances. This is standard for other GPS work and should not be anything novel for RTK GPS.
>
> If you are working efficiently, you'll have enough redundancy to identify blunders, but won't be trying to independently derive the standard errors of conventional observations and variance factors for GPS vectors on each survey. That is anything but efficient.
>
> The whole point of testing is to identify what the reasonable limits of the variance factors are.
Referring to one of many of Gavin's post, the a-priori error estimates generated by modern RTK DC's are MORE then adequate for a least squares adjustment.
It's like anything for a surveyor, you won't trust until you use it many times with empirical tests.
I supervise over 10 survey crews, and have not seen a false integer ambiguity appear on ANY measurement in the last 7 years. That's 15 times infinity of the amount you have tested I presume.
The adjusted results are often EXACTLY the same as post processed results, as we often log data while running RTK because I had many of the same fears you do.
I wish you the best in attempting to stay competitive.
RTK Part Zzzzzz
> Refusal to test is the perfect out.
I have no earthly idea what you're going on about. Shawn offered up some data that was anything but a test. Testing processor-estimated weights of RTK vectors shouldn't be a challenge at all. I think I've already described a good way to do it, but since some of us have short memories, here it is again:
A surveyor would want to test his or her RTK system in conditions as close to actual use as possible, meaning: for single base RTK over baseline lengths at both the outer limit of normal use and on short baselines under 100m.
The surveyor would want to conduct the test on an array of stable marks that were in multipath and sub-optimal locations similar to those the real world presents as well as in perfectly ideal locations. The idea is to characterize performance over the best and the worst conditions under which it is likely to be used.
I would establish the array of marks close enough together than they could be connected with conventional observations with a total station to get their relative positions with uncertainties on the order of +/-2mm or less. Some of them would be in the open, some at the edges of tree canopy, etc.
You'd establish two base points, one at the inner range limit of about 100m and the other at the outer range limit of whatever. I assume that most surveyors won't be doing RTK work much more than a mile from the base for single base RTK.
You'd survey a network of static GPS vectors to connect the two base points to each other and to at least three points in the test array and adjust the works, deriving exceptionally accurate NAD83 positions for all.
Then you'd be able to separately add RTK vectors from the base points to the points on the test array and examine the realism or lack thereof of the processor estimates of variance attached to the vectors.
It should be a fairly clean exercise.
RTK Part Z
> The missing piece, of course, consists of the weights of the observations and what factor to apply to processor-estimated variances. This is standard for other GPS work and should not be anything novel for RTK GPS.
RTK vector data includes standard errors just as static does (there are some receiver/dc combinations for which this is not true but those are exceptions). StarNet formatted RTK data looks exactly like static vector data. It is true that the errors are often - but not always - somewhat greater than with static, but only slightly.
I agree with you, Kent, that all too many people are using their RTK as a magic black box coordinate generator to perpetrate all kinds of mayhem. But you should hate the playa, not the game. It is quite possible to do excellent work using RTK.
Checking into someone else measurements to a high degree of precision is a near perfect check of that measurement but it doesn't say much about other measured vectors.
Gee Kent, I always thought a Saturday
Night Special was a cheap weapon of inferior quality. I have always assumed that most GPS users purchased well-made and therefore, expensive equipment, but perhaps decided to save money by eschewing the educational part of the purchase. In the second scenario, the not-so-good results are probably the result of poorly informed users.
More than once you have enlightened me by explaining to me better ways to ensure higher quality results for my GPS usage.
Jack
Gee Kent, I always thought a Saturday
> Night Special was a cheap weapon of inferior quality.
No, a Saturday Night Special was something you bought on the spur of the moment to be dangerous. Same deal with RTK in the hands of the folks in a hurry.
RTK Part Z
> It is quite possible to do excellent work using RTK.
I don't doubt for a minute that it is possible to do as good work with RTK as can be done with PPK GPS, but the fact that so few RTK users seem to have any sort of an idea about the uncertainties in their work and seem to be so poorly conversant with the general topic of estimation of uncertainties in general does not reassure me at all.
> Checking into someone else measurements to a high degree of precision is a near perfect check of that measurement but it doesn't say much about other measured vectors.
Yes, exactly. What a *professional* is concerned with is an entire measuring process and the limits inherent in the process, not a sort of Easter egg hunt for happy-looking results.
RTK Part Z
> Referring to one of many of Gavin's post, the a-priori error estimates generated by modern RTK DC's are MORE then adequate for a least squares adjustment.
Well, that's exactly what I meant when I described RTK as the Saturday Night Special of surveying techniques currently available.
RTK Part 2
> I assume that most surveyors won't be doing RTK work much more than a mile from the base for single base RTK.
I just finished profiling 8 miles of levee crown every 100 feet or so, and did it all from a single base setup in the middle of the project, so my vectors run from near zero to about 4 miles. I had the receiver set to record only when 2 or more RTK engines fixed, plus some nominal accuracy stats that I don't fully trust yet.
I checked into points that are part of an overdetermined static network, with check vectors ranging from 1 mile to a shade over 4 miles. The checks were done over the course of 2 weeks, with one point checked 9 times (1-mile vector), one checked 6 times (1-mile vector), one checked 5 times (2-mile vector), one checked twice (4-mile vector), and two checked once (1- and 2-mile vectors). The 6-check point was known to have subsided due to massive groundwater pumping and was on the rebound following the end of the pumping season, so I didn't pay much attention to the absolute comparison at that one, just to the spread. I also didn't care much about horizontal as this was a vertical project, but horizontals were coming in at well below 0.1 foot.
Not counting the subsided point, the worst vertical checks were about 0.15 foot, with most around 0.10 foot. The spreads were all 0.1' or less.
I also hit every 5th levee crown point (a paint spot) a second time on a different day. Those checks -- over 80 of them -- were all 0.15 foot or less (and mostly less than 0.1 foot), with the exception of two that came in at 0.20 foot and 0.25 foot. I reshot those a third time on yet another day so I could toss the outliers.
Although I used the g-file vectors to process the data in Star*Net, I didn't adjust anything, because 80% of the observations I was after had no redundancy. I just looked at the check shots as indicators of measurement quality, and was happy with the magnitude and consistency of those results. So while I don't have any rigorously determined error estimates, I reported to my client that the profile is accurate at the 0.15 foot level.
I'm still getting comfortable with RTK, but this was one job that would have been hard to accomplish efficiently without it.
RTK Part 2
> I just finished profiling 8 miles of levee crown every 100 feet or so, and did it all from a single base setup in the middle of the project, so my vectors run from near zero to about 4 miles.
That all sounds fairly plausible. I wouldn't think that +/-2cm s.e. was unlikely at all for the horizontal component (inferred from the vertical uncertainties you mention) of RTK positioning in the wide open skies scenario you describe.
Just out of curiosity, though, what was it about surveying the profile of some levees that couldn't be done via PPK GPS? Was it some spec that called for sampling at regular intervals?
RTK Part Zzzzzz
It's the fools not the tools.
I for one don't trust those new fangled electronic distance measuring things. I've seen way more surveys screwed up with those EDM things than I ever saw with a good old transit and chain. o.O
RTK Part Zzzzzz
> I for one don't trust those new fangled electronic distance measuring things.
Well, I don't think "trust" figures into land surveying. If you do, you may have missed your true calling in real estate sales. Just saying.
RTK Part Zzzzzz
> Well, I don't think "trust" figures into land surveying. If you do, you may have missed your true calling in real estate sales. Just saying.
You are exactly right Kent; just because:
> In 1914 you measured, with a solar transit and steel tape:
> NS Fwd Azimuth 147-49-39.4, Ground Distance 24333.07'
> and, before you, in 1897 another surveyor measured withb a solar transit and steel tape:
> NS Fwd Azimuth 147-49-39.2, Ground Distance 24333.06'
This only goes to prove that both of you did it wrong by precicly the same amount!
Best Regards
Dougie
We are just walking in the footsteps of the original GLO Surveyors. They said "It's good enough", and went to town, to get stuff.
We do the same. Good enough, and went to town!
(I make loose joke)
N
RTK Part 2
> Just out of curiosity, though, what was it about surveying the profile of some levees that couldn't be done via PPK GPS? Was it some spec that called for sampling at regular intervals?
No, the interval wasn't an issue -- I actually laid out station markers (fence post with PVC cover) every 2,000' using a Garmin eTrex, and paced the 100-footers in between. PPK would have worked -- though I have little experience with it -- but likely would have required more time and/or more site visits. Although the whole project appeared to be wide open, there were some areas that required much more time to get good results -- 30 seconds to as much as 3 minutes, versus 10 to 20 seconds in most areas. (Some of that was due to constellation, but in some spots it occurred even with a boatload of SVs in view, and was seen again on revisits.) Having the real-time indicators made it possible to determine when enough good data was in hand.
RTK Part Zzzzzz
RTK Part 2
> PPK would have worked -- though I have little experience with it -- but likely would have required more time and/or more site visits. Although the whole project appeared to be wide open, there were some areas that required much more time to get good results -- 30 seconds to as much as 3 minutes, versus 10 to 20 seconds in most areas.
If all you needed was 0.15 ft. uncertainty or better, that should have been a snap to survey with PPK, logging 15 seconds of data on each point. Naturally, you'd have had to pay attention to PDOP, but I'm sure you do that routinely. It would have been smart to run two or more bases and you probably would have wanted to do a static intialization just as a belt-and-suspenders solution, but if the site was as completely unobstructed as your photos indicated, I've never had a surprise, even with L1 PPK, as long as the PDOP was within tolerance.
