You guys are a bunch of Nerds!B-)
Thank you 😀
Targets
Hello rfc.
These targets are run off from a laser printer, 10 mm apart. The horizontal line is to help 'level' the targets as any amount off vertical makes it difficult to 'pinch' the target accurately at these ranges. They were spaced wherever I could fit them between obstacles in the garage. If you try it just get them spaced as evenly as you can. Changing the instrument orientation will make sure you cover the circle enough.
Hello Bill,
About 52 targets. Groups of 5 so as to easily distinguish individual outliers on a plot as opposed to a real anomaly on that little part of the circle. If a target group of 5 shows consistent residuals of the same sign and size then hopefully I imagined it was a good sample of that small part of the circle. I hoped that enough samples over a large enough part of the circle would show real differences.
Single face pointings to each of the 53 targets, so my derived SD for a pointing is for one face which looks like a good result. Then the instrument was rotated 120* and set back down in the tribrach. Repeat. Once each 3 orientations was used up the tripod was moved to another spot and the 3 orientations were done from there. The circle was not reset during the test so as to be able to plot where exactly the readings were taken. depending on where I was the angle between target groups could have been as little as 3*-4* or as large as 90*. Here is a plot of how much of the circle got sampled.
Let's jump in the deep end!
> I'll spare you the details of the least squares input stuff...
>
Unless you think I'd be getting in over my head (something I'm prone to do quite well on my own....)
I'd love to see some of the input data. Did you use Star*net?
> Single face pointings to each of the 53 targets, so my derived SD for a pointing is for one face which looks like a good result.
Ah, I'd forgotten about the directions being taken on just one face of the instrument. You'd expect that the cyclic errors due to eccentricities of the circle reading points would nearly vanish if the directions were taken as the mean of both faces, wouldn't you? I know I would.
That should be very simple to test considering how gross the cyclic error pattern is.
That is, if you have targets spaced at about 22.5 degrees, wouldn't you expect to see the cyclical pattern of errors repeated in the differences between F1 directions and directions taken as the mean of F1 and F2?
Hello Kent,
It depends on the phase of the error and whether is out of phase with itself or in phase 180* around. That's why for the plot I reduced all angles to between 0-180 to see if it would strengthen the pattern or create patterns on the opposite side. If it were that easy to cancel the error this wouldn't be a 5" instrument, it would be a 1". I just don't think Leica are selling us gear that can be turned into 1" gear with simple FL/FR readings. I could be wrong about everything I've posted though.
The differences of the magnitude that my test is uncovering is still present in the averages of multiple FL/FR sets. Did you read my reply to one of Paul's posts? Here it is again:
I had a 10" moment just recently when turning sets. In good conditions I had the robot turn 2 sets. reduced angle was about 120*00'00". Just for kicks I turned the instrument studs around and did the sets again. 120*00'10". Turned the studs. 120*00'00". Turned them back to the 'bad' postion: 120*00'10", again. Turned to a 'good' position: 120*00'00", again. Turned the tribrach to 10* off 'bad' orientation: 120*00'06". Back to 'bad' orientation: 120*00'10". Turned it by eye just to check.
I was seemingly getting a systematic error from the circle, and it was perfectly repeatable.
I marked the instrument orientation after this reading in the field so I could specifically look for this error in my tests on the same part of the circle. My tests agree well with the circle differences in amount and location.
In theory anyway.
I had a 1" instrument a while back that passed every test the manufacturer threw at it, but had an incredibly eccentric horizontal circle. It wasn't able to close traverses to better than around 1:60000 (My old 3" instrument was able to achieve 1:150000 quite easily).
In the end I developed my own test that proved the instrument was bad and got it swapped out for a new one. IIRC I think I determined it was acting more like a 10-20" instrument, depending on which part of the circle was read.
Let's jump in the deep end!
Trust me, it just looks like the data you've provided in previous threads, just heaps more. Your data may reveal similar patterns if you plotted the angular residuals against an absolute position on the circle. Don't reset your backsights to zero every time you change orientation and you'll end up with the same sort of information.
I don't use Starnet.
> It depends on the phase of the error and whether is out of phase with itself or in phase 180* around. That's why for the plot I reduced all angles to between 0-180 to see if it would strengthen the pattern or create patterns on the opposite side. If it were that easy to cancel the error this wouldn't be a 5" instrument, it would be a 1". I just don't think Leica are selling us gear that can be turned into 1" gear with simple FL/FR readings. I could be wrong about everything I've posted though.
I'd think there are two main contributing causes for the cyclic error pattern:
- net eccentricities of opposing pairs of circle readers, and
- eccentricities of the center (the turn axis) of the instrument itself.
The H1 and H2 phase and amplitude settings strongly suggest that the manufacturing process doesn't rely upon extreme high precision in the placement of the opposing circle readers, but uses a software correction to sort that out.
If your instrument has two opposing pairs of circle readers, then each circle reader would have its own cyclic pattern of errors with a period of 360 degrees, nulls at 180 degrees and minima and maxima midway between nulls, if I have the situation correctly envisioned. The puzzle is that the function representing the sum of the periodic errors of each circle reader should itself have the same period as the individual inputs. That is, an error function that is the sum of two cyclic error functions with a period of 360 degrees should also have a period of 360 degrees, if I remember that detail correctly.
The shorter period of your test data is puzzling.
If you "unfold" the pattern of residuals to cover the full 360 degree arc, does it still show the same periodicity?
> The differences of the magnitude that my test is uncovering is still present in the averages of multiple FL/FR sets. Did you read my reply to one of Paul's posts? Here it is again:
>
> I had a 10" moment just recently when turning sets. In good conditions I had the robot turn 2 sets. reduced angle was about 120*00'00". Just for kicks I turned the instrument studs around and did the sets again. 120*00'10". Turned the studs. 120*00'00". Turned them back to the 'bad' postion: 120*00'10", again. Turned to a 'good' position: 120*00'00", again. Turned the tribrach to 10* off 'bad' orientation: 120*00'06". Back to 'bad' orientation: 120*00'10". Turned it by eye just to check.
>
> I was seemingly getting a systematic error from the circle, and it was perfectly repeatable.
I missed that detail. So that was out in the field and the error was evidently entirely dependent upon the part of the circle used for the measurement.
> I marked the instrument orientation after this reading in the field so I could specifically look for this error in my tests on the same part of the circle. My tests agree well with the circle differences in amount and location.
Am I right in thinking that when your instrument is fired up, that zero is assigned to the identical point on the circle, or does it use the last assignment in memory?
> If you "unfold" the pattern of residuals to cover the full 360 degree arc, does it still show the same periodicity?
Yes, but obviously a bit lighter as would be expected.
> I missed that detail. So that was out in the field and the error was evidently entirely dependent upon the part of the circle used for the measurement.
Yes.
> Am I right in thinking that when your instrument is fired up, that zero is assigned to the identical point on the circle, or does it use the last assignment in memory?
The last orientation is held in RAM I think, and you get it again when you turn the instrument on. So changing batteries in the field does not pose a problem for orientation. You orientation persists after any kind of power off event in my experience.
I marked my backsight orientation on the instrument with a paint pen for that field event and it was assigned 0* for my test so I could see the relationship between 0* and ~120* after my tests.
> In the end I developed my own test that proved the instrument was bad and got it swapped out for a new one. IIRC I think I determined it was acting more like a 10-20" instrument, depending on which part of the circle was read.
Hello CSS,
I'd be grateful if you'd share your test procedures here, and very happy if I learnt something new from them.
Cheers.
> > If you "unfold" the pattern of residuals to cover the full 360 degree arc, does it still show the same periodicity?
>
> Yes, but obviously a bit lighter as would be expected.
So, the key to the puzzle has to be in the period of the errors, which isn't 360 degrees. Presumably, the instrument "knows" the absolute position of the circle and just adds a constant to that when zeroing or setting on some particular value. So, the instrument could apply circle eccentricity corrections after calibration.
I wonder if the circle has various index points at intervals that correspond with the period of the error you've observed.
Have you ever found any more detailed descriptions of how your instrument is constructed, as in numbers and arrangement of circle readers?
> Why do you suppose there are four periods in th graph?
I wouldn't have a clue!
I'm going to go waaaaaaaay out on a limb here and posit, without proof, that this is how you make a 5" instrument. I don't like the grading of raw circles theory now for the sole difference in different accuracy instruments of the same family. It's too random for mine.
And, surely, nowadays circle printing can be done every day to around 1"? Introducing a known, wave pattern on the circle certainly makes for a very simple and elegant solution to making different grade instruments. Want a 2" instrument? have one with correction values that have correction for the amplitude of 80% or so. It guarantees you will get the manufacturers spec with statistical testing every day.
I offer the comments in this post as nothing more than ideas to be kicked around and possibly torn apart from those with real knowledge of these instruments and manufacturing techniques.
Targets
If you try it just get them spaced as evenly as you can. Changing the instrument orientation will make sure you cover the circle enough.
>
Let's say it's a square garage, with the instrument in the middle. If the targets are on the walls, they're not perpendicular to the instrument. That's probably fine for all the measurements taken within a target because you're comparing them to each other.
But wouldn't the varied angle of the targets to the instrument around the garage impact the results? Or doesn't it matter that the targets are "apparently" 10mm apart in one place, 9.99mm apart on the next, 9.68 on the next, etc.?
> Have you ever found any more detailed descriptions of how your instrument is constructed, as in numbers and arrangement of circle readers?
There is a white paper for this instrument but it has more info on how hot the EDM is and none that I recall on the angular system.
This one on the TS30 is better and it's probably like this but scaled back a bit, like 2 circle readers instead of 4. TS30 white paper
The TS30 is from a different family of instruments but the principles are the same I think.
Targets
Hello rfc,
I'm not relying in any way on comparing the absolute distance between the targets compared with the apparent distance at different viewing angles. They are just individual targets that happen to be co-linear with their mates because they are stuck to a brick. They each have a unique spatial position which (hopefully) doesn't change, and can be determined. When I move around the room intersections and distances will determine where they actually are.
The key bit may be this on Page 6 of the above report:
"By using two encoders for the angle measurements, the periodic error of the eccentricity of the coded glass circle compared to the standing axis of the total station is eliminated. Another two encoders remove further minor pi-periodic errors which are determined by the system"
Just so you know, Keith
And I know (hope) you're out there watching, although maybe not at this point.
ANYWAY, this is a mathematics thread, NOT a land surveying thread.
But you knew that didn't you, my old reactionary buddy:-)
Don
Just so you know, Keith
> And I know (hope) you're out there watching, although maybe not at this point.
>
> ANYWAY, this is a mathematics thread, NOT a land surveying thread.
>
> But you knew that didn't you, my old reactionary buddy:-)
>
> Don
Too bad there's not such a category...there should be though.
Also, I think there should be a "hardware" category...for the discussion of instruments, electronics, prisms, tripods etc.
