Hello People,
I needed to test other spec instruments, and in a different manner, to try and verify my earlier results in prior threads that suggested that Leica total stations are made to spec by imparting precise pattern errors on the encoder circle.
I tested our companies other 5" total station, and also was able to test a 3" TS15I VIVA. The test I devised involved the instruments mounted on a Leica heavy tripod, 6-8 m from a brick wall. Targets were stuck to the wall, spaced around a 45 degree arc. All of my preliminary tests proved the setup to be stable to probably better than one arc second in azimuth over the time it takes to read an arc, and a couple of seconds for level. I redesigned my printed targets to a type that I could reliably sight to under one second precision. Preliminary testing revealed there was no detectable difference or benefit in reading in both faces versus multiple pointings. Actually due to the stability of the test area and the precision I was able to sight the targets, single pointings gave me all the precision I needed.
Directions were read to each of the targets in each of the 3 stud positions and adjusted using least squares as 3 separate arcs to the same targets. The tribrach was loosened and twisted some amount, re-tightened and the procedure repeated. These next 3 arcs were treated as new targets. I decided not to solve for a new instrument position as there would be no way I could see to prevent the errors I was trying to detect from being hidden as variations in the new station position. The nature of the angular error prevents it from being eliminated by FL/FR pointings.
I won't bother to upload the results for the second of our TCRP1205+ instruments as, statistically, it proved to be IDENTICAL in specification to the other. You cannot tell the difference between the results above the 0.1" level. They are, for all intents and purposes, the exact same instrument. This lends weight to my argument that the encoder circles are probably all pretty much identical quality.
Here are the residuals for the 5" TCRP1205+ plotted against the circle reading modulo 180. The curve is my best fit function sin(A*8)*3.8".
For more clarity regarding the precision of these residuals, here are the same residuals plotted modulo 45.
The standard deviation, once again, for this instrument is 2.8" for a single pointing, making a STDEV for a single face angle 4". All identical to earlier tests. After applying my best-fit sine correction the standard deviation of a direction residual is 0.5". Sadly I shall never know for sure If this function actually describes the errors present. But if it did, It would be a hoot to use it in a real world application that could verify the post-processed accuracy of the instrument.
Now for the other instrument. I was very happy to have this hire instrument arrive in our office just before testing. I asked for a 1" or 2" but had to settle for a 3" Leica VIVA TS15I. I love this instrument; imaging camera, clear optics, smooth and precise servos, and quite accurate as borne out by the plot of the angle residuals and the best fit 1.4" amplitude sine correction.
The instrument returned 1.1" standard deviation for direction residuals, making it near enough to a 1.5" instrument for single face angles. This seems to me to be quite a dramatic increase in actual accuracy over the 5" instrument. Perhaps it's no wonder the 3" model is the best selling instrument around these parts. After applying the best-fit sine correction the STDEV of the direction residuals is 0.44" - virtually no different to the 'corrected' 5" instrument.
I didn't apply any rejection criteria when analysing or fitting the data. All residuals were considered valid despite there being some improvements on offer for rejecting only a small number of outliers. The residuals in each of my tests can be improved by applying a second, smaller, longer sine function. This additional, smaller harmonic may be an artefact of the testing procedure, or a mechanical effect, like instrument wobble around it's axis. I'm not convinced it actually exists though.
This is good stuff. Like and appreciate what you are doing.
But what would be really interesting is to get some comments from Leica from someone who could comment authoritatively to your tests.
>I redesigned my printed targets to a type that I could reliably sight to under one second precision.
I'd be interested to see an example of the target design. Previously, I believe you had described targets that were lines of very nearly the same angular width as the split wires in the telescope reticle. Did you continue along those lines?
> Directions were read to each of the targets in each of the 3 stud positions and adjusted using least squares as 3 separate arcs to the same targets. The tribrach was loosened and twisted some amount, re-tightened and the procedure repeated. These next 3 arcs were treated as new targets.
So, if I understand things, the direction to the first target was arbitrarily assigned and pointings were made to the targets over a 45 degree arc. Then the instrument was advanced by nominally 120 degrees in the tribrach by removal and replacement and directions to the same targets were observed. The instrument was again advanced by nominally 120 degrees in the tribrach and directions repeated.
Those three series of directions were adjusted as if they were to the same targets and the residuals in the adjustment were used to estimate the errors in each series of directions, i.e. from 0 - 45, 120 - 165, and 240 - 285
Then you rotated the tribrach by approximately 40 and repeated the series of measurements in three instrument positions in the tribrach treating them as to new targets with the first target direction nominally 45 degrees? That gave you the errors of directions 40 - 85, 160 - 205, 280 - 325?
Or was the tribrach was rotated by more like 5 degrees between measurements series? I wonder how much of the noise in residuals is just an artifact of the overlap of directions taken from slightly different centers after tribrach rotation. Wouldn't the solution bet to use more targets in the 45 degree arc and much less overlap between successive series?
Do the Leica instruments have a native "zero" direction when turned on or did your test require keeping track of the directions assigned to the circle?
One Face Sighting Does Not Account For Real World Problems
Out in the field you do not get the luxury of perfect setups.
Interesting test of instruments ultimate capability but not for surveying.
Paul in PA
Conrad; very nice charts you have been showing. Have you read any of the literature
on encoders? (just wondering, it looks like you have. If so what are they?)
There are others that have done test on the electronic Theodolites, Hennes & Witte, 1991. You can also find some information by HP and there HP 3820A from the 1970's.
In one of Dr.J.M. Rueger books (Monograph) he has the following information for the following Instruments:
Nikon DTM-750 (peak to peak Variation) 3.0"
Topcon GTS-6b 1.7"
Topcon GTS-301 3.3
Zeiss ETh3 5.2"
All this is very useful information for the surveyor as he should know what his instrument can do. Keep up the good work and thanks for a very nice presentation of your findings.
JOHN NOLTON
Tombstone, AZ.
> I'd be interested to see an example of the target design. Previously, I believe you had described targets that were lines of very nearly the same angular width as the split wires in the telescope reticle. Did you continue along those lines?
Here is my 'stepped A'
If I remember correctly those are 0.7 mm vertical line segments that open up 0.05 mm each step. At some point one pair of segments become almost a perfect width to pinch the single vertical crosshair at the distance range I'm working. and the next widest one is a good check also. It became a slight frustration in the end that the servos won't smoothly move at the 0.3" level as I came to know what around 0.3" looks like to these targets.
> So, if I understand things, the direction to the first target was arbitrarily assigned and pointings were made to the targets over a 45 degree arc. Then the instrument was advanced by nominally 120 degrees in the tribrach by removal and replacement and directions to the same targets were observed. The instrument was again advanced by nominally 120 degrees in the tribrach and directions repeated.
>
> Those three series of directions were adjusted as if they were to the same targets and the residuals in the adjustment were used to estimate the errors in each series of directions, i.e. from 0 - 45, 120 - 165, and 240 - 285
>
> Then you rotated the tribrach by approximately 40 and repeated the series of measurements in three instrument positions in the tribrach treating them as to new targets with the first target direction nominally 45 degrees? That gave you the errors of directions 40 - 85, 160 - 205, 280 - 325?
>
> Or was the tribrach was rotated by more like 5 degrees between measurements series? I wonder how much of the noise in residuals is just an artifact of the overlap of directions taken from slightly different centers after tribrach rotation. Wouldn't the solution bet to use more targets in the 45 degree arc and much less overlap between successive series?
I tried various movements in between arc triplets. The final procedure involved rotating the tribrach 22.5 degrees between triplets which gets you starting from each successive peak and trough in the 'errors' so that you will see the full amplitude several times. Those peaks and troughs being established earlier and identified by some pen markings which allowed me to line up and zero in the same orientation within about 15 minutes of arc.
I have tried all the combinations I could think of to either tease out the errors or make it go away. Rotations include 30* a couple of times to cover the entire circle and give a little overlap, 60* to give a little gap, 11.25* to go halfway between peaks, 7.5* then 15*, then 11*, and other combinations that I don't recall. I have adjustments containing no overlap to others with 75% overlap. I also tried controlling centring errors and keeping the same target numbers. I used between 8 and 24 targets in the arc at various times. It made no difference. The only time I didn't get a 'good' result was when my target arc was too narrow to fully cover one half wavelength (peak to trough), which seems obvious now. I tried expanding the arc to above 45* but it made no difference. I have a lot of data now. I tried hard to make the pattern go away but it didn't.
> Do the Leica instruments have a native "zero" direction when turned on or did your test require keeping track of the directions assigned to the circle?
Pen mark, as described earlier, to line up the upper and lower halves. The instrument will always turns on with the same circle orientation that it had when you turned it off. You can't know where the absolute 'zero' on the circle is.
> Conrad; very nice charts you have been showing. Have you read any of the literature
> on encoders? (just wondering, it looks like you have. If so what are they?)
>
> There are others that have done test on the electronic Theodolites, Hennes & Witte, 1991. You can also find some information by HP and there HP 3820A from the 1970's.
>
> In one of Dr.J.M. Rueger books (Monograph) he has the following information for the following Instruments:
>
> Nikon DTM-750 (peak to peak Variation) 3.0"
> Topcon GTS-6b 1.7"
> Topcon GTS-301 3.3
> Zeiss ETh3 5.2"
>
> All this is very useful information for the surveyor as he should know what his instrument can do. Keep up the good work and thanks for a very nice presentation of your findings.
>
>
>
> JOHN NOLTON
> Tombstone, AZ.
Hello John,
I have read some literature on the Leica encoder system specifically, and also general stuff about circles when they were scribed mechanically on a rotating circle but nothing that included data that you just mentioned. I would surely be grateful If you could link me to some of the texts you mention.
Thanks Dave,
I would absolutely love to know If what I'm seeing is what is the reality. Two different independent test procedures seem to suggest it is.
> > Do the Leica instruments have a native "zero" direction when turned on or did your test require keeping track of the directions assigned to the circle?
>
> Pen mark, as described earlier, to line up the upper and lower halves. The instrument will always turns on with the same circle orientation that it had when you turned it off. You can't know where the absolute 'zero' on the circle is.
That would seem to be the hurdle to clear in making post-processed corrections. You'd have to be meticulously careful about not rezeroing the circle in the course of a setup (or larger part of a project) and to work out the map of periodic angular errors for the circle orientation assigned. One the other hand, since the form of the periodic errors is so well known, it should be possible to figure out the phase of the periodic error with respect to a circle reading of zero from just one or two 120-degree-rotated triplets on relatively few targets.
> One the other hand, since the form of the periodic errors is so well known, it should be possible to figure out the phase of the periodic error with respect to a circle reading of zero from just one or two 120-degree-rotated triplets on relatively few targets.
Yes, that is the case.
Conrad
Conrad if you do an internet search under "Hewlett-Packard Journal 1980's, then look for Sept. 1980 you will find the HP 3820a Theodolite talked about.
For Dr. J.M. Rueger work you want his Monograph 18; Electronic Surveying Instruments, A Review of Principles, Problems and Procedures. Order from;
School Of Surveying and Spatial Information Systems
The University of New South Wales
UNSW Sydney, NSW 2052, Australia
The Hennes,M, Witte,B. 1991 in German, in Vermessungswesen-Raumordnung (VR),53(1):30-45
JOHN NOLTON
Tombstone, AZ.
PS I have not tried an internet search for Dr. Rueger's monograph #18 but you might find it. I had the phone number at one time to the University. Don't know where it is now.
Conrad
> For Dr. J.M. Rueger work you want his Monograph 18; Electronic Surveying Instruments, A Review of Principles, Problems and Procedures. Order from;
Found it thanks John. It is downloadable, as are a lot of other interesting publications from the UNSW. For anyone else interested:
Conrad
Conrad, thanks for the link. Now I have the phone number again. Now you can get them for free and years ago I had to pay for them. Story of my life.
In Rueger's #18 you will find all kinds of interesting stuff; like information on bar code levels and sighting distances you want to stay away from etc.
Good Hunting
JOHN NOLTON
Tombstone, AZ.
Conrad
A company I once worked for had it; I made them pay for it. But, alas, It went missing when we moved. Lots of interesting stuff in there.
One Face Sighting Does Not Account For Real World Problems
> Out in the field you do not get the luxury of perfect setups.
However, as the distances to targets increase well beyond 8m, the effect of 0.24mm centering errors does tail off fairly quickly.
One Face Sighting Does Not Account For Real World Problems
> Out in the field you do not get the luxury of perfect setups.
>
> Interesting test of instruments ultimate capability but not for surveying.
>
> Paul in PA
Hello Paul,
I don't quite understand where your post is aimed. The purpose of the test was not to prove the accuracy of single face observations (they can be well-accurate) but to determine the angular errors affecting the total stations I use. Single face observations were accurate enough and were just part of my procedure.
Your comment about not getting 'the luxury of perfect setups' affects ALL observations and ALL instruments. It is applicable to everything everybody does 'in the field' with all manner of instrument. I fail to see why you make it, unless it is just to remind those who need reminding that the world is 'fluid' at some scale and climate.
I am a surveyor, and my testing has already affected what I do in the field and how I do it. Thus it has an application to surveying, if only my own.
part of the reason for the similarity could be that Leica only builds 1" instruments, and then once assembled puts them through a battery of tests to decide which category they fall into. That means that you could have a 5 second that was just barely not a 3, and a 3 that was just barely not a 5.
> part of the reason for the similarity could be that Leica only builds 1" instruments, and then once assembled puts them through a battery of tests to decide which category they fall into. That means that you could have a 5 second that was just barely not a 3, and a 3 that was just barely not a 5.
The more likely explanation, though is that the different instruments in the same family have the same circle but correct circle readings differently by removing different parts of the cyclical error in circle graduations that the manufacturer designed into the circle.
It's a very simple strategy to implement in the manufacturing process since the amplitude of the sine wave error pattern of circle graduations is known as is its period (angular distance from peak to peak). The only thing to be worked out would be the phase of the sinusoidal error pattern at the zero of the circle and that would be very easy to determine by a quick test.
The most precise instruments in the family probably use two different types of corrections, a simple sine wave pattern of corrections and corrections based on a lookup table to correct small irregularities that remain after the sinusoidal error pattern is removed. Those would most likely be determined by automated testing of the instrument against a super-accurate reference standard.
> part of the reason for the similarity could be that Leica only builds 1" instruments, and then once assembled puts them through a battery of tests to decide which category they fall into. That means that you could have a 5 second that was just barely not a 3, and a 3 that was just barely not a 5.
Hello mneuder,
My guess is that when people speak of the testing and sorting at the factory, it is the assembled total stations that are being tested for true running and perpendicularity of their axes to assess whether they are worthy to be corrected by software to become 1". 2" or otherwise. Just purely a guess though.
