My Topcon DL-502 can be set to shoot the rod 2 to 9 times and display an average. Our preferred method is to manually shoot the rod at least 3 times until the lowest reading is obtained. Our method is based on waving the rod, where the lowest reading occurs at the instant the rod is perpendicular to the ground. The other day I got something like 4.451, 4.454, 4.454, and 4.454. For all you statisticians out there, 4.451 looks like an outlier. That being said, 4.451 would be the recorded rod reading. I do not agree with our method. Clearly, the last 3 shots are grouped together and the first shot is an outlier. Assuming the level has been properly peg tested and the rod bubble has been calibrated, I would choose 4.454 as the correct rod reading.
Rocking the rod on a digital level is as useless as a camo plumb bob...
It is a clear waste of time to rock the rod on any type of leveling. That's the reason they make?ÿROD LEVELS.
JOHN NOLTON
This is Bob Stadenko, Holloman test track survey engineer, your avg for the reading is 4.4533, but hey, ain't nobody Going to beat us at flat.... Pre-digital levels and beyond.....
???? ???? ???? ?????ÿ
Disagree.
For precision work with optical level, yes, rod level is the way to go. But... I wouldn't do that. For precision work, I'd use a digital level...
Running around a construction site for production work with an optical level, I can read a rocked rod to a half hundredth way faster. Like... 10 times faster.
Have not touched an optical level since we had the digital level. The best way to use a digital level is to establish plenty of permanent benchmarks between your level runs. That way, it's easy to isolate which segment did not meet your precision. Make them short so that it's fairly fast work for the reruns if needed. I think you will realize that this method is faster than having to record 4-5 readings per tp.
Average of 3, adjust by station, limit 150 foot sightlines for typical work, 100 for precision, focus on tripod stability, use a turtle for transient work but set turn points for monitor runs, book descriptions, beginning and ending elevations with closure/adjustment but let the computer keep track of the rest.
?ÿ
I worked at an office that had established an interesting "method" for processing RTK control.
The crews were required to take six 5-epoch observations in quick succession for any point that would be used as a total station setup or was a boundary monument. That was it. No dropping SVs or reinitialization, no going back to the point under a different constellation, no changing of rod height or turning rod 180.
Back in the office, the PM or the tech would zoom in on the six observations, and arbitrarily pick a cluster of points that were "close to each other", and deem the others "outliers". The "outliers" were deleted and a simple mean of the "clustered" points was used as the final values.
No evaluation of the horizontal and vertical precisions of the observation vectors.
No weighted mean.
No minimally constrained network adjustment to evaluate possible statistical outliers.
It was just based on what the processor thought "looked right".
?ÿ
There would be much hand-wringing on the rare occasion that the crews had to go back to those points and tie them in again, and found that there was often up to several tenths of difference between their supposedly "adjusted" values and the values observed on the return trip.
?ÿ
Modern surveying is full of BS "workflows" and "QC procedures" that defy mathematics, statistics, physics, etc. What is really interesting is the response if you ask someone performing them what the fundamental rationale is for those procedures. 90% of the time it is "it's the way I've always done it", and the rest of the time it'll be a word salad that uses mathematical terms, but is devoid of any actual knowledge of what they mean and how they relate to surveying observations.
Sargent Stadenko is the cop from Cheech and Chong Up in smoke....I'm sorry, my Midwest sarcasm is sometimes hard to hear.
Was totally kidding.
Still, my question remains as to the soundness of our method. I say the outlier, the lowest rod reading in my example, occurred when the rod wasn't perfectly plumb. Therefore, the lowest reading shouldn't be recorded. I would not accept it. How about you?
@field-dog I assume a digital level reads the coded rod the same way an optical level would. If reading the rod optically when the rod is plumb that is the lowest reading that can be made. Any out of plumbness would result in a higher reading of the rod. While my eye can watch this happen and note the lowest reading of the rod, I'm not so sure a digital level can function the same way. I imagine it takes a singular moment in time and produces its reading from that one singular picture of the coded rod. If you are rocking the rod it is highly likely that this reading will take place when the rod is out of plumb and result in an incorrect reading. Again that is speculation on my part. At best the digital reading could only be mutiliple single readings and still you would be introducing bias by rocking the rod. I suppose if it were a "video" reading of the rod it would more closely approximate what a human evaluates when looking at a rocking rod, however I don't believe you are getting better information by rocking the rod when using a digital level, I suggest you are getting worse information by employing that technique.
My point is that if there's no clear mathematical sense in a workflow, or a procedure flies in the face of basic statistics, then that procedure is a farce. Doing something because it "makes sense" or "feels right" to someone is not a valid reason.
If you're not moving the rod, and the rod is plumb, then that one low reading would indeed appear to be an outlier. However, that doesn't necessarily mean you should toss it. How far out a measurement needs to be before it is rejected depends on the level specs and observation conditions.
Digital levels read a wide swath of the rod rather than a single point; waving it is practically useless. And unless that rod bubble is severely out of adjustment, as long as you're within that bubble circle, any slight change in rod verticality is going to be negligible for 95% of levelling projects.
Unless post-processing is consistently revealing a serious problem with the final values, I'd set the level to take multiple observations and accept the average rather than trying to guess at what might be an outlier. Double-run, least-squares adjust and then check it.
If your rod ends in a point, or has a flat shoe and rests centered on a point, then rocking it cannot produce a lower reading (higher rod position) than plumb.?ÿ
If your rod has a flat shoe and rests on a point off the horizontal center of the scale, then rocking it can give either too-high or too-low readings.?ÿ Rocking will raise or lower the scale as the rod pivots.?ÿ This adds in one direction and subtracts in the other direction and interacts with the lowering of the point on the scale being read.
I worked at an office that had established an interesting "method" for processing RTK control.
The crews were required to take six 5-epoch observations in quick succession for any point that would be used as a total station setup or was a boundary monument. That was it. No dropping SVs or reinitialization, no going back to the point under a different constellation, no changing of rod height or turning rod 180.
Back in the office, the PM or the tech would zoom in on the six observations, and arbitrarily pick a cluster of points that were "close to each other", and deem the others "outliers". The "outliers" were deleted and a simple mean of the "clustered" points was used as the final values.
No?ÿ
No
No!!!!
Takeaway here is 'worked' there...
Wow.?ÿ That's a pretty good story of what not to do....thanks for sharing!
Glad you escaped!
I'd set the level to take multiple observations and accept the average rather than trying to guess at what might be an outlier.
Most of the three shot readings I take are closely spaced, such as 4.454, 4.454, and 4.453. If I get readings such as these I stop after three shots. If I get an outlier, I take 4 or 5 shots.
Disagree.
For precision work with optical level, yes, rod level is the way to go. But... I wouldn't do that. For precision work, I'd use a digital level...
Running around a construction site for production work with an optical level, I can read a rocked rod to a half hundredth way faster. Like... 10 times faster.
Yup, running an NI2 level with a Lenker or Philadelphia rod or even a tubular FG rod is amazingly fast for rough work (within a few hundreths).?ÿ ?ÿRocking the rod especially when observing near the top of the rod really helps.?ÿ Don't get me wrong, I've done precise leveling with optical micrometer levels/invar braced rods/turtles (tedious) to get to sub 0.01' accuracy, and later Total Station trigonometric levels with short rods on bipods which were very efficient in steep terrain.?ÿ I don't see how digital levels improve the situation except for cancelling out some atmospherics because they observe?ÿ a span of the rod and also eliminate misreadings by the observer.?ÿ The machinery inside levels hasn't improved concerning establishing local level optically with suspended mirrors, etc., for decades.
Others may agree: