@shelby-h-griggs-pls two hours I believe is a sweet spot for what we need to achieve for sure. If that second observation does not fall within a specific range of the first you would need to wait again or use another method to decide which one was bad. It will get better once all gnss is fully functional and the kinks are worked out. There are still some timing issues which will need to be resolved but if no major conflicts happen I believe it will not be long. The other will be a few years as the other systems get enough datapoints to model the clock errors and drifts on the satellites themselves. Not enough time yet for them to do that. Which is why we are not using all signals that are available now but what the manufacturers can use safely. I noticed on B and gal that on a couple svs only 1 band was actually in the equation on one day same svs next day nothing being used and it was in the same slot. So as good as we see it now we have not even seen exactly how good it will be. Clocks on satellites are like kids each one has its own personality. Then you have the clocks at the monitor stations. This is where gps shines. IGS has more data more stations but more is not always better. I believe one day we will be solving in real-time with no corrections what we are achieving today with base and rover or network rtk solutions. Very fun times.
I take five, 30 epoch observations, pausing 30 seconds and 'dumping' (really just hitting the reset button) between each. I review the session and eliminate any clear outlier. If there are two obvious outliers, I'll kick both and take another 3 observations with the same criteria. I generally like to see a standard deviation of under two hundredths on an average with at least 4 separate observations.
If you have five observations taken within the same very short time window with approximately the same H/V estimated errors from the RTK engine, how do you determine whether an "outlier" is truly a bad measurement, or is merely the natural result of the constellation changing and/or changing signal reception (esp. in the case of a multipath environment)? If you have two groupings of three and two observations each, how do you know which set to eliminate?
I have been using the two hour rule on a different day for 33 years now.
SHG
I think it's also important to clarify that the ODOT study was really focused on how to use RTK (specifically network RTK) to obtain network accuracy results comparable to static work, specifically for control networks.
They also came up with a (really fantastic) decision matrix that employs different practices depending on the object of the work, because the goal is not always the same for each project.
But I'd wager your two-hour separation rule results in a miniscule number of required re-observations, if any at all...
If you have five observations taken within the same very short time window with approximately the same H/V estimated errors from the RTK engine, how do you determine whether an "outlier" is truly a bad measurement, or is merely the natural result of the constellation changing and/or changing signal reception (esp. in the case of a multipath environment)?
Might as well include it into one observation and give the RTK engine more fodder. That and if you're simultaneously logging you'll give your post-processing software more fodder to play with.
@steinhoff
Unless you get a bad initialisation, then breaking them up would be more reliable assuming resetting RTK 'fix' (or the equivalent from some of newer gear) each time.
Personally still sold on one bunch of shots (rotating pole 180° to mean any plumbing bias) and then another bunch at least 30 mins later.
Unless you get a bad initialisation, then breaking them up would be more reliable assuming resetting RTK 'fix' (or the equivalent from some of newer gear) each time.
Happens less and less nowadays as technology improves. If you'd like a gutcheck before taking your second observation dump satellites and really quickly stake the previous shot out after taking the initial observation.
EDIT: this is also why I instruct crews to run RTK+logging... if for some reason RTK doesn't want to play nice we can post-process to the RTK base, secondary base, or CORS.
Personally still sold on one bunch of shots (rotating pole 180° to mean any plumbing bias) and then another bunch at least 30 mins later.
Returning 30+ minutes later? Let's shake hands on that.
Bunch of shots and rotating? Plays absolute hell when running adjustments if the data is imperfect. You're essentially creating associated "pairs" of observations- a "no rotation" shot, and a "rotated" shot. If one of those observations trips up your adjustment, you essentially need to toss both observations in that pair... the reason being when you have two "pairs" of observations and you throw out a component of one of those pairs you've now created an imbalance in rod orientation- for example you'll end up with one "no rotation" shot, and two "rotated" shots. If your rod is out of plumb (which is the only reason why you'd want to do this to begin with), you're introducing a bias to the adjustment. Is the bias negligible? Probably... you're welcome to risk it, but it's made my eye twitch too much in the past.
I've personally gone back and forth with this. In practice shooting, rotating, shooting, then returning to repeat later on helps eliminate rod errors... but is it worth it when you sit down in the office and try to play surgeon when the adjustment starts flagging individual observations? I'm much happier to instruct crews to religiously check their rods and then baby those rods throughout the day (which they should be doing anyway).
EDIT 2: another thing to consider about this methodology in an adjustment is that you're flooding the adjustment with redundancy that essentially pertains to centering errors only, as the constellation is virtually unchanged in the time it takes to rotate the rod and observe again. If you're using a program like TBC, the tau factor required to flag a potentially bad observation would be stupidly high because of the amount of (in my opinion) artificial redundancy involved- which increases your chances of a bad observation making it through.
I keep taking shots until I get 2 shots in a row of close measurements.
Two in a row isn't a good criterion. Getting a majority that agree closely, and not sequential, is better.
It's as good as it's going to get for the nature of my projects and the time that I have to do them. Like I say, subsequent visits often play a role.
@bstrand I aggree. One thing i have not asked is if you are rotating the rod to try and compensate for a bubble out of adjustment is are you still keeping the antenna facing north. I know this facing the antenna orth is in the noise of rtk but lets be truthful here and truthfully understand the north antenna offset. Lets say for all intensive purposes there is only a 3mm bias in the offset. Not un heard of even between the exact same model number. So lets say we face the antenna 180 degrees out because of a blunder. The antenna file assumes you are facing north so now the 3 mm has become 6mm. So .02 ft. Are we worried.02 probably not but lets add up the 8 mm + 1or 2 ppm error in the perfect world of today’s receivers doing rtk base and rover. Now we have an uncertainty starting out of the gate of 8mm then we add we faced the antenna wrong by rotation of the pole so 8 mm + 6 mm oh wait we also need to see the we had the base at half that error. Simply observe correctly keep the rods in adjustment and lets the constellation’s change. Move the base to gain truthful redundancy. Let’s say we do this now and our rod is out of plumb. We created a systematic error because we faced the rod and antenna in the correct orientation every time. Well we missed the accuracy of the datum but our relative results are good to go. Some things we think sounds good need to be fully understood. I have always on control or corner’s tried to follow the correct orientation and i have at times spun my rod but i do this keeping the antenna oriented in correct direction. The software or firmware assumes the antenna is faced correctly period. So if you are rotating a rod thinking you are getting redundancy your fooling yourself because you are introducing an error from the start. And most people are not correcting the antenna when they rotate the rod. Truthfully your better off for a boundary survey to not rotate rod it’s relative and when we get down to it who cares if we are a tenth off absolute from the datum because of a rod out of adjustment when our relative can be better. Absolute being to the datum. This is something that i have not seen discussed. Correct. Yes shoot a edm rotate rod I understand. I have even fallen into the trap of the same with rtk and gps static. But its apples and oranges. I have a fixed height tripod out of adjustment right now. Landed on site i saw this. I rotated rod not antenna for base cked 90 degrees. I simply adjusted half the error out until it was plumb. I had no tools to adjust the bubble. Did this in 3 different set ups. Isn’t this the same exact thing we learned to do with a spirit levelon a total station. Turn 90 180 take half the error rotate do again even if the bubble didn’t fall within the graduations marks. We new we were level. No matter what happens when it comes to gnss time is your friend. Time gaps time collecting. The whole system is time. Great post and it shows you get it. Wait till we have no need for a base or network rtk. Time will still be the answer. A gap in time and collection time. We use to do PPP positions requiring 2 different 24 hour sessions. Now we can understand that we can get same results by two less collections but a certain time gap. Between those sessions. Take any 3 antennas same model. And do a antenna calibration and you will see they will most likely not match by 3mm at best. If you send in 9 and find 3 withen 3mm use those. Because thats a very good relative set.
If your rod is out of plumb (which is the only reason why you'd want to do this to begin with)
No, I find all poles have some degree of banana/run-out from new, as an aside I've had several come from dealer not even plumb! With fixed height poles this is not such an issue as can set the bubble so tip to tip is plumb and banana is negated but on adjustable height ones find it varies. Only talking 1-3mm here so not a major bias but if you observe two 'faces' with pole it averages out (more accurate, less precise). I use rotating plummet under GNSS base so I can average out any difference when I set up for same reason.
Then you have the antenna internals that @olemanriver refers to above.
In theory the pole is always perfectly adjusted for plumb and where I am the only user of a pole that is usually the case but one that sits in the truck used by many, experience tells me they never stay as well adjusted as I'd like so my proceedure accounts for the human factor. Agree they should take care of them but some crews not wired that way or even good crews have a bad day/rough terrain hard on gear.
Regarding adjustment I'm not bothering with least squares for RTK GNSS where I'm targeting 10mm in horizontal at 95% confidence; I just average the first fix pair/s and second fix pair/s. LSA is for static GNSS where I'm chasing sub 5mm and for urban boundary/the sort of control you do with GNSS where I work, it is rarely called for. But yes you do have to toss the pair if doing two faces.
I'm not lucky enough to have the ppk type option either, just RTK or static so we try to maximise the RTK but not push it too hard in challenging conditions as often breaking out the robot just provides that piece of mind it's right.
Also I need to learn how to do multiple blocks of quoting like you did, couldn't work out out on my phone so grabbed something to link the posts!
In theory the pole is always perfectly adjusted for plumb and where I am the only user of a pole that is usually the case but one that sits in the truck used by many, experience tells me they never stay as well adjusted as I'd like so my proceedure accounts for the human factor. Agree they should take care of them but some crews not wired that way or even good crews have a bad day/rough terrain hard on gear.
On the other side of the coin: some crews aren't wired to actually rotate the rod and shoot and just take a second shot and say "good enough." The human factor is always a wildcard. I've always carried a spare rod level bubble to check throughout the day to backcheck, though.
Regarding adjustment I'm not bothering with least squares for RTK GNSS where I'm targeting 10mm in horizontal at 95% confidence; I just average the first fix pair/s and second fix pair/s. LSA is for static GNSS where I'm chasing sub 5mm and for urban boundary/the sort of control you do with GNSS where I work, it is rarely called for. But yes you do have to toss the pair if doing two faces.
So you recommend to shoot, rotate, shoot and return to repeat... and then don't adjust? What are you doing with the "rotated shot" and second round of observations? Are you just averaging your point?
If you're taking the time to try to be careful with field procedures, why throw methodical office procedures to the wind?
I'm not lucky enough to have the ppk type option either, just RTK or static so we try to maximise the RTK but not push it too hard in challenging conditions as often breaking out the robot just provides that piece of mind it's right.
Out of curiosity, what gear are you using that doesn't allow for PPK work?
@scott-bordenet
https://www.oregon.gov/odot/Programs/ResearchDocuments/SPR304-821_UpdatedSurveyStds.pdf
In theory the pole is always perfectly adjusted for plumb and where I am the only user of a pole that is usually the case but one that sits in the truck used by many, experience tells me they never stay as well adjusted as I'd like so my proceedure accounts for the human factor. Agree they should take care of them but some crews not wired that way or even good crews have a bad day/rough terrain hard on gear.
On the other side of the coin: some crews aren't wired to actually rotate the rod and shoot and just take a second shot and say "good enough." The human factor is always a wildcard. I've always carried a spare rod level bubble to check throughout the day to backcheck, though.
Regarding adjustment I'm not bothering with least squares for RTK GNSS where I'm targeting 10mm in horizontal at 95% confidence; I just average the first fix pair/s and second fix pair/s. LSA is for static GNSS where I'm chasing sub 5mm and for urban boundary/the sort of control you do with GNSS where I work, it is rarely called for. But yes you do have to toss the pair if doing two faces.
So you recommend to shoot, rotate, shoot and return to repeat... and then don't adjust? What are you doing with the "rotated shot" and second round of observations? Are you just averaging your point?
If you're taking the time to try to be careful with field procedures, why throw methodical office procedures to the wind?
I recommend two bubbles on the rod.
Such a simple thing to do.
I recommend two bubbles on the rod.
Such a simple thing to do.
You mean you think I should have ANOTHER way to backcheck myself at a glance? Really twisting my arm with that one...
When using the double-bubble as a way to check rod plumbness, should they be at 90° or 180° to each other. Does it matter?
When using the double-bubble as a way to check rod plumbness, should they be at 90° or 180° to each other. Does it matter?
If you do 180 it's difficult to look at them, so we do it at 90. Technically, it doesn't matter. I suppose you could simply stack them, but then it's also hard to look at them as you take a shot. As the bubbles drift out of plumb with each other you will see it. Time to adjust them.
Geomax, but I don't recall ever doing it with Lecia either, only Trimble. Hardly anyone here buys TBC or Infinity though. Raw observations go straight into 12d software https://www.12d.com/
What are you doing with the "rotated shot" and second round of observations? Are you just averaging your point?
If it's topo use the point averaging in controller in field with all four shots prior total station setups. Also restart the base between rounds at a different height, mainly so you actually relevel it again.
For boundary everything goes into traverse sheets using Bowditch. You can see the outliers quickly. Here chasing the last 1cm not last 1mm.
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?)
And I could be wrong, but Bowditch/Compass Rule adjustment does not do statistical analysis, so how are you coming up with "outliers"?