I too find it odd.
I'm impressed at being strict enough to spend the time to go all the way back to the base and change height and restart it - before measuring all the observations again - but not spending the hour or so in the office to run LSA on all (GNSS + terrestrial) observations at once, which is far faster than running traverse loops using compass rule off of field-averaged GNSS observations, and far better than Bowditch at detecting outliers, because Bowditch is an arbitrary adjustment routine that can only return closures, which may or may not actually be indicative of a problem with the data.
Back-of-napkin calcs....for what a crew costs per hour, if we sent them back to restart the base - even only once per day - and it only took an hour each time, we could simply have them sandbag the legs, save that money, and spend half of it on 5-6 seats of StarNET or TBC or whatever, and the other half on office tech time at 2 hours per day processing the data with LSA, which likely would take less time when analyzing all the data at once.
(Edit: and that's only for one crew.)
@lukenz what brand of gear are you running. I have not done it yet with rtk data but NGS opus projects beta version will allow I think gvx vectors and you can adjust the raw data lsa and project to whatever system you want. Its free software so no cost. I am more inline with rovers mentality as far as incorporating rtk static and conventional data. All together. Just get more redundancy quicker and a better idea of the uncertainty. I did a project a while back where i had several days and used VRS and base and rover . 3 points were shot at three different times over that 3 days from vrs. All other corners and control were shot twice at different times from different base set up some not in canopy were shot back to back but one observation from vrs then switched to base. I checked these independently as a test using the robot from a different base line turning rounds. In the end i used all observations. 35 control and property corners total. Over 3 miles radius and topo and mapping as well in a 30 acre section. Closest point pairs were 1.15 feet apart. Measured with 25 ft tape . Also located with gps . Most of the work was all walking because it was quicker than all the driving and fences and access. After throwing out or disabled a few vectors from the start that only logged about 20 to 30 epochs because of multi-path in some tough locations all data was very clean in my adjustment. In areas of multi path i did some independent measurements with robot just to have a good feel before I headed back home. Me an another old crew chief worked together. He knows his stuff but wanted to better understand why 2 observation and moving the base mattered. And it was a safety issue being close to roads and railroad. We had no one to go with him. We had fun and both enjoyed the cold beer in the evening . And both were very slow the next morning getting the old joints to move again
Need to preface this with some more details it seems.
First I'm in New Zealand not USA and for boundary work we prepare a survey sheet and a title sheet which is recorded with government department. On the survey sheet we have to show vectors between boundary points and control points (physically observed or calculated from observations, missing line from observations etc.) and traverse spreadsheets (Bowditch) are the most convenient way to do this in a hierarchical manner (new survey obs first, adopted survey obs on old plans, adopted boundaries, calculated boundaries) as we upload the traverse sheet to government database, much quicker than copying it all over manually. Working from the whole to the part is what they call it at survey school here.
We actually need to use the observed base/rover vectors, not just a case of dumping all into LSA and outputting some nicely adjusted coordinates.
Second I think LSA is a fantastic tool for adjusting a control network where you have redundancy in the observations and are mixing multiple instruments (typically GNSS/total station/level). And where the desired output is those nicely adjusted coordinates.
To me redundancy is 3+ different independent vectors to a point (not just using the same base setup to measure three rover vectors one after another).1 vector is a hanging line, 2 provides a check and 3+ starts to let you determine which is best.
I am familiar with LSA in TBC and LGO (not used Infinity yet) and the government geodetic office here created SNAP, https://www.linz.govt.nz/products-services/geodetic/geodetic-software-downloads/snap-concord-downloads, which is excellent and from what I can see it's somewhat like Starnet from what I can see from marketing material and comments on here. LINZ require all geodetic contractors to do their adjustment in it.
In my experience boundary surveying rarely sees the level of multiple independent observations as in a control network, typically just observe the boundary point, resetup and check (two shots).
It seems as if many of your field procedures actually negate the benefit of LSA, despite having it in 12d (unless it's a pay-to-unlock option?)
Not sure I follow how ensuring GNSS field measurements are truely independent has any bearing on a LSA?
You can do LSA in 12d but I've never seen that module used. I'd use SNAP if I had a need for LSA.
And I could be wrong, but Bowditch/Compass Rule adjustment does not do statistical analysis, so how are you coming up with "outliers"?
Yes I agree, that Bowditch is not doing any statical analysis. I'm determining "outliers" by comparing the base to rover vectors for each mark, I'm only really looking for blunders at this point as the quality of the data relates to how well I collected it in field (given I don't consider I have 3+ independent vectors where I would get benefit from LSA)
I'm impressed at being strict enough to spend the time to go all the way back to the base and change height and restart it - before measuring all the observations again
Assuming the base is over a control point then if you don't resetup between the time separated vectors to another point then how can the subsequent set be considered to be independent? No different to moving base to another control point for subsequent set of observations isn't it?
but not spending the hour or so in the office to run LSA on all (GNSS + terrestrial) observations at once
Only worth that extra step if you have real redundancy in data (as opposed to a shot + a check shot where what you are really checking is you didn't stuff up the first shot)
which is far faster than running traverse loops
Base/rover vectors go into 12d digitally and then you trace them (click on vector and it appears in traverse sheet) into a traverse sheet you'd be doing anyway. Traverse block looks like averaged base to rover vector in first round from base out to other point and then from second round back and closed; the Bowditch adjustment isn't doing any heavy lifting here, just checking no gross errors.
Sounds like this is fundamentally different approaches for the different jurisdictions.
and far better than Bowditch at detecting outliers
Only with sufficient redundancy is LSA head and shoulders above. With only two observations to a point you are only really looking to see they are within your accepted repeatability tolerance.
Hopefully I've covered these off clearly and the additional background provides the necessary context. I had several runs at it!
Well, it sounds like you are passionate and knowledgable about your craft and the requirements for your jurisdiction.
However, based upon your description, it sounds like your jurisdiction didn't really understand the technology at the time they were deriving their methods and standards. And it sounds like they haven't learned or updated anything since.
Government has a way of doing that...
We are an observation based cadastre with (I think) standard hierarchy of evidence (natural boundaries, original monuments, undisputed long standing occupation, abuttals, mathematical calculations) not a co-ordinate cadastre which I think is same as USA. And yes government departments move like molasses here as anywhere else I suspect.
When digital data is uploaded to government database, https://www.linz.govt.nz/products-services/landonline, it does run LSA but find given the low degrees of freedom and typical boundary ppm specifications (Rule 27 in https://www.legislation.govt.nz/regulation/public/2021/0095/latest/whole.html) it will mask blunders whereas a hierarchical built traverse sheet shows them up. With 3+ independent observations to a point you would have more confidence but not required by survey rules (or likely economic).
There was a proposal a few years back to move to a survey plan system where instead of mucking around pretending to show 'traverses' from mixed data you would just provide a table of coordinates for all points. These would just be relative to each other in a legal sense not absolute (although practically would be on our circuit projections which are similar to an LDP I believe). Would have made LSA much more useful. Was a really good idea I thought but tradition won out!
While LSA is great for new observations it still leaves a challenge in how to weight the old survey/boundary vectors over the years. Would need to be able to separate old vectors out by age and apply various different precision tests/weightings in adjustment.
Enjoy what I learn on here just didn't want anyone thinking surveyors away in this colonial outpost were overly ignorant of modern methods
I use base and rover all the time and my suggestion would be to set it up and see what it doing. I work by myself every day, so I know my equipment and I have settled on 5, 60-sec burns and reset the receiver each time in the canopy and three times in the wide open. I use x-pad software and it will do an average for me and allow me to set my average error to what I want( h=2cm, v=3cm), it shows me each shot and the error so I can decide what I want to do. Remember GPS is always going to be GPS and you can do something today and it will be good and tomorrow it may have a few hundred errors in it. Consistently is getting a lot better, but you still need to have your check and balances and keep your poles checked. Deano
Here is a good video showing how to create subsets of constellations so that you can get independent data from different satellite constellations without having to wait any longer than necessary.
@bstrand hey when rotating rod 180 are you keeping the antenna facing north or do not worry about that. Remember that every signal has its own north and East signal offset and that means L1 phase center L2 phase center are all different than mechanical phase center and mechanical phase center is different than the point of reference on rod and so once you spin that rod and antenna 180 you are doubling the error of phase center period. Yes it’s usually small but you are introducing that error each time if the antenna is not facing north. Some antennas do not agree with each other even same exact model by 3mm so you just added 6mm error in addition to the 8mm rtk Horizontal. Some phase centers are much tighter and i do not concern myself in topo and such about that but for control and property corners i do my best to face north and do my best to meep pole in adjustment. Even if my pole is off and out of plumb. By always facing north and bubble on rod facing the same my relative is better my absolute would have a systematic error.
Here is a good video showing how to create subsets of constellations so that you can get independent data from different satellite constellations without having to wait any longer than necessary.
This is a great method, but I would add that it's critical to make sure you have plenty of birds to use in your solution i.e., running full constellation, and are watching geometry during observations, since there are four different orbital periods in play and what used to be a good subset a couple of weeks ago may not give decent DOPs today.
Even if my pole is off and out of plumb. By always facing north and bubble on rod facing the same my relative is better my absolute would have a systematic error.
I think the differences in phase center due to antenna orientation are so far below typical RTK accuracy that it's a wash.
Even for static work I've not seen any noticeable difference between big networks where everything was facing north, and where the antennae weren't oriented the same way.
It has to be some pretty high-precision work (NGS level, bluebooking etc) for me to worry about that sort of thing. Not saying we don't make the effort, but it's rare that we really need to.
Honestly (and I hate to say this) but I hesitate to discuss this sort of thing with certain PMs because they'll zero in on this one tiny thing and start trying to micromanage it rather than worrying about whether the crews are measuring in both feet and meters and actually writing down where they measured on the antenna.
hey when rotating rod 180 are you keeping the antenna facing north or do not worry about that.
The north facing antenna is something I've heard a bit about in passing but it's not something I've looked into in great detail myself. It also hasn't been a concern expressed by anyone I've worked with so far so I haven't been taking any steps to account for it in my workflow, no.
By always facing north and bubble on rod facing the same my relative is better my absolute would have a systematic error.
Sure all rover points will be relatively tight but have to disregard the base point. Personally not a fan of just setting up base randomly as harder to check hasn't gone off level/moved when you get back.
@lukenz yeah when i set a base up the first thing after being level is I rotate spin the rod and check bubble. Had one i was given at last second for a recent job was out of adjustment. I could not adjust the bubble because i had no tool. So I simply took half the error out until i was comfortable it was level even though bubble was not. I do same thing with rover rod at start of each day inside bipod x’s 2. Not hard to do and such.
