I remember as a kid going to the park with friends and playing on the merry-go-round. We'd spin it as fast as we could then run and try to jump on. I was reminded of this reading the thread https://surveyorconnect.com/community/threads/javad-integrates-magnetic-locator-into-the-rover-pole.327651/ The thread was a collaborative exercise in attention deficit disorder. It was difficult to find a place to jump in, so I thought it best to start a new thread.
I started my own business a few months ago. I live in an area known as the Piney Woods of East Texas. Pine trees are difficult on precision GNSS results. Because of this and other considerations there are times that GNSS is not the right tool for the job at hand. Having said that, I've been continually amazed at the results I've obtained from RTK. I'll remind readers at this point that I provide sales and support for Javad GNSS, but this thread is meant to discuss RTK in general and not intended to be brand specific.
I've been using RTK steadily for four years. Prior to that, I have dabbled with RTK since around eight years. I've been using post processed GPS (static and kinematic) for 16 years as well as sub-meter GPS for about 14 years. Early experience with post processed GPS and conventional (total station) surveying revealed that least squares adjustments provided the best method for blending the results. Least squares adjustments provided a rigorous process that could apply the statistical strengths and weaknesses of each measuring technology to arrive at the best estimation of the position of points surveyed in a network. Often times the task of using least squares adjustments was arduous as the preparation of the data required no small degree of effort, however the results always proved to be very good when tested with additional cross ties not used in the adjustment (for example surveying a 100 acre tract, adjusting the survey measurements with least squares and then comparing a diagonal GPS vector that was not initially included in the adjustment to find agreement within hundredths of a foot).
Collecting data in the field for post processing was always done with an eye toward insuring good results with extra time in the field. The idea of a return trip due to insufficient data was typically unacceptable logistically and economically, so we would collect more data than necessary. This was particularly true in marginal environments as there were no standards on observation time near obstructions or under canopy.
Starting in 2008 I had the opportunity to provide product reviews for American Surveyor Magazine. I tested several brands of RTK systems. At the time, there was a noticeable difference in precision between some of the brands. Two stood out as remarkable from the others: Altus APS-3 and Javad Triumph-1. Others I had tested at the time were Leica 1200, Spectra Precision Epoch 25 (which was based as I recall on the Trimble 4800 internals), and ProMark 3 and ProMark 500 from Ashtech. The Altus and Javad systems produced precisions comparable to my experience with static GPS results. This ultimately led me to convince my Dad to invest in an Altus APS-3 system when we made our decision to buy RTK. In 2014 I tested a Champion GNSS system which used a Trimble board. This receiver also produced results that were comparable to my experience with static GPS. When I say results were comparable, I mean that we would usually set pairs of points that we would then occupy with a total station. We would always check the backsight (horizontal and vertical) and the results would typically be less than 0.02 foot horizontal and less than 0.04 foot vertical, which was also true for static GPS pairs. The advantage of RTK over static was that I knew the coordinates immediately in the field and could perform COGO on the spot. This did indeed speed the process of collecting field data immensely.
Fast forward to the last few years. I've been consulting for Javad which has caused me to step up my understanding immensely. I've performed numerous tests in many environments with his equipment. I was prepared for some of this from my days providing technical reviews for American Surveyor, but the testing has only grown from those experiences. One of the tests I performed last year was an analysis of accuracy. I should note that in my observations, similar to the accuracy stated for OPUS, time on site plays a key role in expected accuracy. The longer the time on site, the better the precision of the results. Similar to OPUS, the improvement eventually comes to a point of diminishing returns, but I am persuaded that 4 minutes of RTK can produce an accuracy of:
5mm+0.7ppm at 1 sigma and 10mm+0.7ppm for the Vertical Precision at 1 sigma.
I based this on tests at three ranges:
at 13560' BL: HSD (estimated) 0.025', VSD (estimated) 0.041'
at 41930' BL: HSD (estimated) 0.045', VSD (estimated) 0.061'
at 50100' BL: HSD (estimated) 0.051', VSD (estimated) 0.067'
To determine the 2 sigma error estimate of the horizontal, we would multiply the 1 sigma value by 1.6, which would give 8mm+1.1ppm, close enough to call it 1cm+1ppm. For the vertical, we would multiply by 1.96 to get the 2 sigma estimate, which would give 20mm+1.4ppm, close enough to call it 2cm+1.5ppm.
NOTE that these error estimates are based on 240 epoch observations. "Outliers" have been seen at around 3x the standard deviation in these tests.
It was suggested that the error estimates could not be improved by repeat observations. This seems like a weird assertion. Most surveyors would agree that averaging multiple observations will improve the precision of the result.
The processor estimates for my current RTK system seem reasonably accurate, bordering on pessimistic. By observing a point twice and averaging (weighted by error estimates), the resulting point shows a marked improvement in precision. If I need to meet an ALTA/NSPS specification, I will typically observe the boundary points twice. On a typical ALTA/NSPS survey this isn't too onerous. If the site is very large, making repeat observations less practical, there is little consequence as the larger size allows for larger error estimates, negating the need to repeat at all.
In the hands of an intelligent user with imaginative problem solving skills, RTK can be transformative in performance and efficiency. The intelligence comes in to play because precision GNSS is not a total station. Understanding the geodesy which underpins GNSS positioning is critical to sustained success, as well as the requirements for receiver to the satellite communication and receiver to receiver communication. The imaginative aspect relates to problem solving and how best to leverage the tools available to provide a result that exceeds some minimum standard (whatever that standard might be).
This week, I had two field jobs, both of which were expedited by use of RTK. In the first, I was staking out three small parcels on an industrial site that Dad and I surveyed in 2009. I found three monuments from that 2009 survey to which I related my new survey. The 2009 survey was done with a mixture of static GNSS and total station observations. The inverse between the three points as measured by RTK:
2016 RTK
87å¡51'09" 652.471'
28å¡41'15" 103.440'
80å¡40'37" 711.060'
The 2009 survey:
87å¡51'44" 652.461'
28å¡42'33" 103.437'
80å¡41'16" 711.065'
A best fit transformation reveals there is a rotation of 38" between the two surveys. I then set a pair of points for total station occupation because some of the parcel boundary was along a metal building, unsuited for RTK. The pair of points, by RTK, has an inverse of:
AZ: 333å¡49'11" HD: 343.896' dU 2.304'
A truck parked between these two points and I had to resect between them. The inverse by total station:
AZ: n/a HD: 343.922' dU 2.298'
this gave a difference of 0.026' horizontally and a vertical difference of 0.006'.
nm, just now understood the answer.
The second job from the week was an unimproved lot survey in an older subdivision (1948). The dimensions on the plat were nearest half degree and half foot. There were several lines that had no direction and some that had no distances in the subdivision. I was able to tie in corners several lots away to prove my boundary location. My lot was approximately 100' by 250'. It would have not been feasible to have tied the extra boundary corners without RTK. Also, I did this entirely solo. There were some neighborhood corners I would have liked to have tied, but were not suitable for RTK. I didn't need them though, so I didn't fret about them. Because I tied the additional corners I was able to find one of the back corners much, much faster, an old 3/4 inch pipe. Without the RT part of the RTK I would not have had this information available to me in the field. The realtime aspect allows me to evaluate data immediately. I was also able to put a couple of line stakes in before I left. The total time spent with RTK on this brushy lot was 4 hours with about thirty minutes before and after spent preparing the site and wrapping up.
There was some question as to why someone would want an all in one on a stick. In this case, the lot was grown up in briars. I much preferred the portability of a small pole with no protrusions, such as brackets and data collectors and external antennas and extensions to tangle on the briars and limbs as I needled my way through. Also, I drove around the subdivision locating the additional corners. With the collapsible pole, I could put the receiver in the front seat next to me without much effort. I can understand why some would resist this, I thought it was strange too at first. For my purposes I prefer it.
R.J. Schneider, post: 383400, member: 409 wrote: It's a compelling post, Shawn. If there was ever a question on the test results, and the performance of any one system over the other, the idea that you can't perfectly recreate the test parameters would be it. Taking the signal dissolution through the pines as the point in question, and knowing that it is physically impossible to recreate the exact geometry, i.e you can't physically occupy the same exact position with more than one antenna for the exact same epochs, how does that factor into your testing ?
In one of your videos you demonstrated the collapsible antenna pole. It seems to be a three section collapsible, do these sections lock-out, or is it a simple compression lock ?
Yes, in canopy it's more difficult to model the accuracy of the system. Multipath degrades accuracy and can cause absolute failure (i.e. wrong integer ambiguity resolution). We can trap the failures through multiple fixes over time (I'm persuaded that this should span at least 3 minutes). The receiver models the accuracy based on the observation. I believe more time on site will provide a better estimate of accuracy. I also believe that repeating the observation under a different constellation can lend some validation to the error estimates.
I've tested quite a bit under canopy of varying degrees and feel confident that under heavy canopy a single observation of 180 seconds in duration with a demonstrable good integer fix will not exceed 0.15'. This would be akin to a 3 sigma value. My 1 sigma estimate would be 0.06'. These can be improved by a second observation. Now, consider rough, rural property. How far can you traverse by total station and be under this error estimate? What measures will you have to take to be less than this error estimate. How does this equate to money spent in time and effort? What value is this to your client and the public. The answer to these questions will vary based on the project and the project needs, so I'm not asking this rhetorically. A professional will evaluate the project and adjust his methodology accordingly on a project by project basis.
To tackle one other statement made on the metal locator thread, I don't know where in the new ALTA/NSPS standards that RTK for boundary surveying is disallowed. Depending on the conditions of the project site, RTK may be the very best tool available for locating evidence and features related to the boundary.
Worried about millimeters, just where is the middle of the stone mound? :unamused:
Shawn Billings, post: 383381, member: 6521 wrote: One of the tests I performed last year was an analysis of accuracy. I should note that in my observations, similar to the accuracy stated for OPUS, time on site plays a key role in expected accuracy. The longer the time on site, the better the precision of the results. Similar to OPUS, the improvement eventually comes to a point of diminishing returns, but I am persuaded that 4 minutes of RTK can produce an accuracy of:
5mm+0.7ppm at 1 sigma and 10mm+0.7ppm for the Vertical Precision at 1 sigma.
I based this on tests at three ranges:
at 13560' BL: HSD (estimated) 0.025', VSD (estimated) 0.041'
at 41930' BL: HSD (estimated) 0.045', VSD (estimated) 0.061'
at 50100' BL: HSD (estimated) 0.051', VSD (estimated) 0.067'To determine the 2 sigma error estimate of the horizontal, we would multiply the 1 sigma value by 1.6, which would give 8mm+1.1ppm, close enough to call it 1cm+1ppm. For the vertical, we would multiply by 1.96 to get the 2 sigma estimate, which would give 20mm+1.4ppm, close enough to call it 2cm+1.5ppm.
(I think that's great that you're encouraging RTK users to use occupation times much longer than normal).
As for the error analysis: unfortunately, you've neglected to mention how the standard errors were computed, i.e. whether they are DRMS or pooled RMS of X and Y components. The other missing element is what the ratios of the variances of the X and Y components were. Can you provide the missing information? It makes a significant difference in the analysis.
Shawn Billings, post: 383404, member: 6521 wrote: To tackle one other statement made on the metal locator thread, I don't know where in the new ALTA/NSPS standards that RTK for boundary surveying is disallowed. Depending on the conditions of the project site, RTK may be the very best tool available for locating evidence and features related to the boundary.
It isn't disallowed. It is entirely possible that an RTK user could take a sufficient number of occupations and generate the uncertainties of the points positioned so that they can be tested in a least squares adjustment. The missing piece for common RTK usage, though, is that the spec requires that the surveyor demonstrate that all points positioned by the survey meet the relative positional accuracy specification by formal analysis of uncertainty. I don't think anyone should hold his breath on that one.
Shawn Billings, post: 383384, member: 6521 wrote: It was suggested that the error estimates could not be improved by repeat observations.
I must have missed that post. The point of contention was that one RTK user claimed that RTK is more accurate than other positioning techniques used by surveyors, which is obviously false, as was demonstrated.
A newbie here (although 22 yrs as a CE, 13 yrs PE). GPS is an incredibly productive tool. The fact one can get an absolute coordinate in minutes (or even seconds) is incredible. People don't often go to court over millimeters, even in my neck of the woods, but it obviously has its limitations. I wouldnt set a sewer main with it
Mark O, post: 383418, member: 11591 wrote: A newbie here (although 22 yrs as a CE, 13 yrs PE). GPS is an incredibly productive tool. The fact one can get an absolute coordinate in minutes (or even seconds) is incredible. People don't often go to court over millimeters, even in my neck of the woods, but it obviously has its limitations. I wouldnt set a sewer main with it
One common area where RTK falls apart is when it is used to place or locate boundary markers on small lots or similar situations where the distances between markers are well under a couple of hundred feet. A quickie shot with an uncertainty of maybe +/-0.07 ft. at 95% confidence means that the bearing and distance from it to other nearby markers also positioned to the same standard will fail some relative accuracy tests, such as that of the ALTA/NSPS specification, that could be easily met by other means.
I don't have it in me to argue with your irrational opinions about RTK for surveying. You've demonstrated nothing regarding comparative accuracies of various technologies with the exception of a vague accuracy specification provided by Trimble. More significantly you've failed to outline the practical implications of the accuracy differences between static GPS and RTK. If, for example, static processing is twice as accurate as RTK, what is the significance of 3mm compared to 7mm for most cadastral surveying needs. Even if RTK is less accurate than other methods, this is a straw man argument. There are more accurate methods of surveying than EDMI, but EDMI, when used properly exceeds the minimum tolerances required for cadastral surveying. The same can be said for RTK. Few surveyors are concerned with what is "most accurate". The question is "Can the technology be exploited to exceed minimum standards in an efficient manner?" If it fails on either point, accuracy or efficiency, then it is the wrong tool for the job. In my quantifiable experience, I have observed that RTK provides results that exceed minimum standards for most cadastral work that I do with much less time and effort than conventional or post processed static methodologies.
I don't have any variance data to share with you. I performed hours of observations at various ranges, collecting single epochs of data. I compared the radial distance of these epochs from the control coordinates and determined the percentage of those radial distances that fell within a normal distribution. Javad's values on the spec sheet that Nate linked (which I have not seen before) would seem to corroborate my findings.
Your knowledge of RTK is based on anecdotal evidence at best, and a very skewed experiential sampling at that. I've always strongly encouraged that operators use relatively long duration occupations with RTK when collecting data for boundary and control. With the ability to post process in the field, the reasons for this have only grown. In the grand scheme of things, the difference between 10-30 seconds vs. 3-5 minutes on boundary points is pretty insignificant. Surveyors typically aren't locating enough boundary or control points per day to notice a substantial difference in cumulative observation time. Most of the time spent in the field on a boundary survey will be in getting to the search area, finding the boundary points and marking the boundary points. Longer duration has the benefit of reducing error estimates by 15-20% (30 seconds compared to 4 minutes), allows more raw data to be collected for post processing (used as a check on the RTK) and provides enough time separation to more definitively prove a reliable fix on integer ambiguities. I look at longer occupation times thusly: RTK provides significant time savings in most cases compared to conventional traverse. By reinvesting a fraction of the time savings by putting extra time into slightly longer occupation times and repeat observations, surveyors can improve efficiency on the front end and improve their work product through improved reliability and accuracy on the back end. Classic win-win. For the RTK surveyor there are two cards he can play to improve reliability and accuracy: time on site and repeat observations.
Shawn Billings, post: 383421, member: 6521 wrote: I don't have it in me to argue with your irrational opinions about RTK for surveying. You've demonstrated nothing regarding comparative accuracies of various technologies with the exception of a vague accuracy specification provided by Trimble.
I don't have any variance data to share with you. I performed hours of observations at various ranges, collecting single epochs of data. I compared the radial distance of these epochs from the control coordinates and determined the percentage of those radial distances that fell within a normal distribution.
All right, so I take it that what you're saying is you don't really want to have how you generated the above numbers reviewed for consistency with standard methods. Determining "the percentage of those radial distances that fell within a normal distribution" doesn't exactly inspire confidence as a statement of method, though.
Manufacturers like Trimble and Leica do actually spend a fair amount of energy on testing the performance of their equipment to support claims for accuracy. So the RMS values on the Trimble data sheet are hardly something pulled out of someone's garbage can. The fact that the Javad literature doesn't specify what "accuracy" is being stated or the confidence level at which it is quoted isn't confidence inspiring, to say the least.
Ok. You've started in with the "what you're saying is" already.
Monte, post: 383407, member: 11913 wrote: Worried about millimeters, just where is the middle of the stone mound?
The standard practice familiar to me in Central and West Texas is to set markers in a rock mound to give the corner a definite position. Unless there is something obviously wrong about how prior surveyor has picked the center of the mound to mark the corner, if the marker at the "declared" center of the mound is permanent and stable, that is pretty much the end of the hunt and the task is to survey the center as marked, which means a punchmark on a rod or aluminum cap.
Shawn Billings, post: 383426, member: 6521 wrote: Ok. You've started in with the "what you're saying is" already.
Well, when someone wants to present a claim of a test result, but doesn't want to actually provide any details of the test results, what other conclusion is reasonable?
Kent McMillan, post: 383412, member: 3 wrote: (I think that's great that you're encouraging RTK users to use occupation times much longer than normal).
While I'm not for sure how other RTK systems work. But if you follow the Javad manuals as well as what Shawn and the rest of the five team recommend you are going to spend at least three minutes on every point were accuracy is important. I would suggest you look at some of the videos than Shawn and some of the others have made as well as log on to the Javad board and read some of the post. It may surprise you on how many folks are trying to use RTK the right way.
Why start injecting "reasonable" into the argument, Kent? What's "reasonable" about quibbling over 3mm compared to 8mm in the context of boundary surveying? I've not deceived anyone regarding my test. The vast majority of surveyors don't care about the accuracy estimates I did on my own, nor how I did it. I've explained the general concept, but I'll add this:
I began by observing individual epochs over a period of hours at several baseline lengths. I imported those epochs into a spread sheet and averaged epochs into groups creating groupings of 10 seconds, 20 seconds, 30 seconds, 60 seconds, 120 seconds, 180, 240, 300, 360, up to 15 minutes. I then compared the radial residual of these clusters to my control coordinates (which were determined by various post processed techniques and compared to the average). I then calculated the standard deviation of these residuals. I compared the standard deviations to the sample set to see that they coincided, which they did. Then I graphed the accuracies over time loosely on graph paper to arrive at an inflection in the accuracy to time. Per my graph, after about 4 minutes the accuracy improved but only very marginally. Armed with this information, I then went back out and tested the actual 4 minute observations and found that my accuracy estimate held up at distances of 2.55 miles and 9.5 miles from the base. The improvement of 240 seconds over 30 seconds was about 15% reduction in standard deviation which may not be terribly significant to most, but as I mentioned earlier, there are other reasons to extend the observation time.