RTK Experiment

Then there is this one. In a particularly obscured spot, but the thing stayed initialized, and each daily pair agreed between themselves:

[pre]
Adjusted Observations and Residuals
===================================

Adjusted GPS Vector Observations Sorted by Names (FeetInt)

From Component Adj Value Residual StdErr StdRes File:Line
To

(V63 RTK Normal 0000003e 2015-06-08T13:13:27)
PDXA Delta-N -47832.4385 -0.0157 0.0733 0.2 3:187
1011_150608 Delta-E -12072.9476 0.0072 0.0594 0.1
Delta-U 6.8968 0.0016 0.1752 0.0
Length 49332.5276

(V64 RTK Normal 0000003f 2015-06-08T13:14:11)
PDXA Delta-N -47832.4385 0.0164 0.0743 0.2 3:192
1011_150608 Delta-E -12072.9476 -0.0075 0.0598 0.1
Delta-U 6.8968 -0.0025 0.1831 0.0
Length 49332.5276

(V86 RTK Check 0000001e 2015-06-09T13:29:10)
PDXA Delta-N -47831.8259 -0.0286 0.0599 0.5 4:71
1011_150609 Delta-E -12072.7928 0.0086 0.0531 0.2
Delta-U 6.4857 0.0083 0.1526 0.1
Length 49331.8958

(V85 RTK Normal 0000001d 2015-06-09T13:27:43)
PDXA Delta-N -47831.8259 0.0285 0.0599 0.5 4:66
1011_150609 Delta-E -12072.7928 -0.0086 0.0533 0.2
Delta-U 6.4857 -0.0083 0.1523 0.1
Length 49331.8958

[/pre]

But the 2 days disagreed by 0.6'

[pre]

Adjusted Coordinates (FeetInt)

Station N E Elev Description
1011_150608 151542.4635 352133.3601 117.8923 15
1011_150609 151543.0754 352133.5136 117.4798 60

[/pre]

Cool stuff. It'd be handy to have a "truth" value on that last point you posted. Were the 60's discerning something that the 15's weren't?

One thing that is interesting about the first data set, is that the StdRes are heavily weighted to <1. Seemingly the StdErr estimates are leaning pessimistic.

Thanks for sharing.

> This was intended to be a real world torture test for RTK positioning. This mark, for example, yeilds fairly small residuals considering. The greatest residual is in the 0.10' range, but most are well under 0.04'.

Although if you consider the problem from the standpoint of positioning accuracy, the standard error of horizontal position of the means of each day's observations is about +/-2cm or 0.06 ft., isn't it?

Are there any urban surveys where that would be tolerable?

What I find interesting is that the standard errors of the various vectors used in the adjustment aren't completely unrealistic, predicting a standard error of horizontal position about 0.08 ft. for each vector. Did you apply a scalar to the variances or were those the processor estimates?

> But the 2 days disagreed by 0.6'
>

Yes, it's great how the repeat positions generated very similar answers and it took returning on a different day (or possibly much later on the same day would have worked) to turn up the fact that the solution was in error by a huge amount. I take it from your comments that there was no clue or indication at all that something was seduced other than your judgment that the setting was a difficult one.

> ..Did you apply a scalar to the variances or were those the processor estimates?
A factor of 2 is applied to the vertical component.

I would not personally consider these points suitable for setting boundary monuments without augmentation with Total Station data in this neighborhood (~$500k for a 1920's bungalow), but (except for the 0.6' point) it would pass Oregon's minimum standards. It would sail past Oklahoma's with pennants flying.

> It would sail past Oklahoma's with pennants flying.

Yes, I'll bet that's true. :>

> Cool stuff. It'd be handy to have a "truth" value on that last point you posted. Were the 60's discerning something that the 15's weren't?
In term of standard errors there isn't a lot of difference between the 15 second and 60 second points, on average, but there does seem to be fewer outliers among the 60 second points.

I haven't collected any 180 second or 300 second sessions as part of this experiment yet.

So, is the residual that you are calculating the difference between the first record (before the 180 spin) and the second record (after the 180 spin)?

> So, is the residual that you are calculating the difference between the first record (before the 180 spin) and the second record (after the 180 spin)?

The output listing would indicate that it's the residual from the adjustment of all of the vectors together, treating the 0 and 180 vectors as independent observations.

I am a proponent of 180 epoch observations

Single base RTK or RTN?

> So, is the residual that you are calculating the difference between the first record (before the 180 spin) and the second record (after the 180 spin)?

The residual is the difference between the measurement and the weighted average of all measurements on the point. The weighting is determined by the standard error data included with each vector.

> Single base RTK or RTN?
Using ORGN. My location is in Sellwood, the vectors are recorded as coming off the base at PDX airport.

Is that using the nearest base corrector or the MAX corrector?

> Is that using the nearest base corrector or the MAX corrector?
I believe its the MAX, but I'm not certain of that. Will confirm. In this instance I would probably get better results using the nearest base - which is GLONASS capable. Results could probably be improved using a Single (nearby) Base as well.

In the context of this thread, I've been thinking about the challenge that Javad has issued to the land surveying community. First off, I have no connection with Javad, and I'm not paid to say any of this. As I understand it Javad has offered a reward of $10,000 to any land surveyor that can prove that his equipment has resulted in a bad fix. Has anybody tried? What would such a "proof" consist of? I'd guess something like the analysis that Mark has offered in this thread. Kent and others have pointed out that RTK fixes can be bad, and I've experienced this myself using Hiper Pro receivers and know it to be true. The advise that many have offered to deal with this possibility of a bad solution is to check, double check, and triple check the observation, to "make sure" that it is right, and generally speaking I agree that this is infinitely better than relying upon one observation. However, I am convinced that our equipment and software should be doing more filtering for us in the first place. That we can get a "fixed" solution that is incorrect is, arguably, a result of poor design in hardware and/or software, and a design that can be improved upon. I believe we will be seeing improvements along these lines in the years to come. In my opinion, Javad is at the forefront of these improvements, and when I purchase a new receiver, I'm going with J!

Big Al, post: 322198, member: 837 wrote: ... As I understand it Javad has offered a reward of $10,000 to any land surveyor that can prove that his equipment has resulted in a bad fix. Has anybody tried? What would such a "proof" consist of? ....

Certain acquaintances of mine insist that a bad initialization could never happen with Leica, either, as the software is constantly computing new, independent initializations every few seconds.

BTW, laying claim to a $10,000 prize is another good reason to record vectors and not just coordinates, if you needed one.

Norman Oklahoma, post: 322211, member: 9981 wrote: Certain acquaintances of mine insist that a bad initialization could never happen with Leica, either, as the software is constantly computing new, independent initializations every few seconds.

I know one of those acquaintances & he's got me convinced. We've double tied hundreds and hundreds of monuments in the last 6 years with our Leica RTK units and I can't recall a one that was way-out wonky if the 3DCQ was dialed in. Sure, sometimes an additional observation is needed, but as far as the results of two strong observations, based on the 3DCQ, being out by crazy amounts like 2 feet or 2 tenths, we just don't see it. I attribute it to (1) good crews who know how to assess the factors like GDOP, SV constellation configuration, local obstructions, and the data on the controller {read "not button pushers, but smart, experienced, motivated surveyors"} (2) correctors from the Ore. Real Time Network and (3) our county Low Distortion Projection.

And if that paragraph doesn't dislodge Kent from his Fortress of Moping, nothing will.

Five RTK ties on a brass monument. 3 are 15 second occupations. 2 are 90 seconds. Guess which ones. The monument is about 20 feet from a 2 story wood building. A three story brick and steel building is about 100 feet south. There are trees in the area partly obscuring the sky. In short, not an ideal RTK site.

(V149 RTK Normal 00000012 2015-06-16T12:17:40)
PDXA Delta-N -47738.5767 0.0065 0.0644 0.1 2:11
1020 Delta-E -13909.5413 0.0252 0.0566 0.4
Delta-U -82.2849 -0.0157 0.0854 0.2
Length 49723.7752

(V148 RTK Normal 00000011 2015-06-16T12:16:17)
PDXA Delta-N -47738.5767 0.0018 0.0607 0.0 2:6
1020 Delta-E -13909.5413 0.0188 0.0551 0.3
Delta-U -82.2849 0.0261 0.0788 0.3
Length 49723.7752

(V150 RTK Normal 00000014 2015-06-16T12:18:59)
PDXA Delta-N -47738.5767 0.0202 0.0579 0.3 2:16
1020 Delta-E -13909.5413 0.0393 0.0526 0.7
Delta-U -82.2849 -0.0215 0.0720 0.3
Length 49723.7752

(V185 RTK Normal 00000047 2015-06-16T13:43:43)
PDXA Delta-N -47738.5767 0.0091 0.0565 0.2 2:191
1020 Delta-E -13909.5413 -0.0306 0.0511 0.6
Delta-U -82.2849 0.0259 0.0792 0.3
Length 49723.7752

(V186 RTK Normal 00000048 2015-06-16T13:45:47)
PDXA Delta-N -47738.5767 -0.0253 0.0592 0.4 2:196
1020 Delta-E -13909.5413 -0.0430 0.0526 0.8
Delta-U -82.2849 -0.0084 0.0858 0.1
Length 49723.7752