So if you??re a Carlson user and running GNSS, you have probably heard their term before for averaging shots & examining them afterwards. If you have not heard about it but want to learn more, click here on this video that Mark Silver did a while back - thanks Mr Silver.
Just curious, for those using Carlson??s SurvCE or SurvPC software, who has actually enabled the ??GNSS Analysis? options under the project??s setup and played with it after several averaged shots? Or furthermore have any of you told SurvPC or SurvCE to export the GNSS Analysis raw data as a text file where it states all of the various statistical info about those observations? It??s pretty neat to look through?ÿ
So I??m mainly just curious if any here have actually opened this raw file up before and read through it all. I went through SurvPC??s user manual that I downloaded from Carlson??s website but they don??t show much on the raw output file??s format (no pics of an example of several averages shots that were analyzed & adjusted) nor does their manual don??t bother explaining the exported raw file??s contents either - some of it makes perfect sense but other abbreviations I??m unsure of. The manual for SurvPC just more or less just mentioned the term GNSS Analysis a few times, explained a few aspects of the feature, Chi-square test & blunder detection in maybe a couple paragraphs.?ÿ
I just wish there was a more in depth portion of their manual that explained all of the exported raw file??s format on the advanced averaging GNSS Analysis. Would be nice if the raw file were to spit out the ellipsoid height too somehow. If I??m missing anything somebody please correct me. Thanks guys
I'd also like to know how many people are using this feature. I just discovered it a few weeks ago and up until that point anytime we needed more accuracy on a GPS location I would just take 2 readings of the point, maybe 3 and my PLS would "average them". But since I found this analysis tool I've been experimenting in the field with it and find really neat and has helped me better understand what the rover is calculating.
?ÿ
For example, how I would set up jobs was set 2 PKs, locate with GPS, make a copy of PT1 and put the EDM on it. Then back-sight GPS PT2, store it, then back-sight that new EDM PT2. My PLS would use one of those GPS locations (100 epochs) bring the whole job up to state plane. Well now I've been hitting the nails again with GPS when I pack up the gun and I often find that the 2nd shot can vary?ÿ in elevation by .2 or .3! Any control points I can hit at job sites i try to hit them as many times as reasonable just to see the analysis
?ÿ
I know this really doesn't change much for things like a plot plan and but using this feature really helped me get a better understanding of what was going on and I too am curious how others use the averaging reports with more experience than me (Survey tech for 2 years)
It looks like the tool will evaluate single points (with at least 3 observations) one by one, using least squares analysis. As such it's not a network evaluation, but a single-point eval.
Perhaps I could see using it if all I did was RTK/RTN work and had no post-processing LSA software.
But generally I find it's easier to look at residuals and standard deviations right there in the field, store as many observations as I need to based upon the statistics, and then adjust the whole network at the same time, so I don't have to pick through point by point.
?ÿ
For example, how I would set up jobs was set 2 PKs, locate with GPS, make a copy of PT1 and put the EDM on it. Then back-sight GPS PT2, store it, then back-sight that new EDM PT2. My PLS would use one of those GPS locations (100 epochs) bring the whole job up to state plane. Well now I've been hitting the nails again with GPS when I pack up the gun and I often find that the 2nd shot can vary?ÿ in elevation by .2 or .3! Any control points I can hit at job sites i try to hit them as many times as reasonable just to see the analysis
?ÿ
This is another reason why I prefer to simply evaluate the network all together during post-processing - we can include conventional observations from total stations and levels and let them be weighted correctly against the GNSS vectors. Then we get to see how the observations compare to the entire network adjustment, rather than just against a couple of vectors.
I would consider the GNSS Analysis tool handy to have in the field, although I doubt I would spend too much time on it. I'd rather just see the weighted results and standard deviations as I average observations, and save the detailed network analysis for back in the office.
There is a bunch of new Point Averaging stuff in SurvPC Version 7 which should be released 'Real-Soon-Now'. It may be worth waiting for V7 before investing too much energy in the details because there are some neat enhancements.
There is a wealth of underlying information in this paper:
But, I don't think you will find the answers you seek there. Interesting reading though.
Another good source of information is page 422 of the SurvCE/SurvPC User Manual: 'GNSS Analysis Point Averaging'
(I got a copy by pressing the 'Download' button at this address: https://web.carlsonsw.com/files/updates/updates05.php/product=SurvCE&version=6.0?ss_email=&product=SurvPC )
I have a few customers who really live by this tool and I think that they understand it far better than I do because they use it several times per day. I am going to ask them to chime in here as they may have a good perspective.
I have been using this program for about a year now and it proves itself to me everyday. I see some difference things in the program everyday. I don't do any post processing, I need something that I can see my results in the raw file. I measure each point 5 times for 60 seconds(or more if need be) breaking the fix solution after each measurment, in doing this you can see the residuals and standard deviation change on each measurement. I would suggest to anyone interested to experiment with this program and make your on opinion. Go to page 413 (I think) in Carlson survpc/ce manuel 6.0 and make sure you read the part about un-checking the rover rod button if you use a bi-pod and to hit the run button to store the point. I haven't look at it in a txt file like Mark is talk about, but i have in the raw file.?ÿ
I start my new job by measuring one point off network and then set up my base and start surveying, allowing it to collect static information from CORS station(opus)throughout the day. Dean
@mark-silver so random question - I don't have the text file reports in front of me but I recall them saying that the "antenna freehand" feature was "on." Is this Carlson's version of tilt compensation like Trimble has? If so, I would think that feature might hurt the accuracy my repeated observations if I'm leveled up over the point & using a bipod to remain steady. I was tempted to uncheck it under the configuration tab (or where ever it might be) & try to repeat some of the same point observations without it enabled
@dean-hill thank you for your suggestion - I think I'm going to try this for myself just to see
Antenna Freehand is: handheld pole vs. BiPod/TriPod held pole.
They use different expected errors to model the estimated position range.?ÿ
The statistics that underly the computations are based on a model of observed ranges for different positions. I looked at them a few years ago and thought that the observations that they made were probably not realistic for survey use. I was going to put my robot in the trees and do 50,000 dumps to try to make a better estimate, but then I realized that their expected range was larger than anything I would expect so probably no need.
We have used the 'GNSS Analysis Point Averaging' that @mark-silver describes for 5+ years now, and it works extremely well.?ÿ All of our boundary and control ties are done with an initial tie, breaking satellite lock, and bare minimum of 15 minutes before a second or third tie.?ÿ Typical errors of <0.01' are seen in good GPS conditions and <0.05' in canopy situations.?ÿ If we can't get good repeatability, then we set control and make our ties with a total station.?ÿ?ÿ
When setting monuments or control, we stake the point out, break satellite lock and then stake again.?ÿ If we get good repeatability, the point gets set.?ÿ If we're in a less than desirable GPS area (canopy, north facing slopes, etc.) then we use a total station.?ÿ?ÿ
I have post-processed a few larger projects just to compare to our field averaged points.?ÿ You would have better luck squeezing blood from a turnip then you would squeezing a few extra hundredths out of our field ties.
As always: It Depends and YMMV.?ÿ