My new instrument, a Topcon GTS 255 was set up on a flat field, without soccer players; The initial plan was to place targets at 150' +/- .1 feet in a decagonal pattern, but I ran out of targets, so populated only 9 of them. Instrument was set up over a target within .003'. The horizontal circle was zeroed ONLY ONCE at the beginning of the setup. Between sets, I rotated the Tri-brach counter clockwise about 30 degrees, and re-centered. 5 sets were taken, but threw the last one out because I was getting a bunch of near blunders (3" or more deviations). I chalked it up to the "almost lunch time, starting to pour rain" error inducement factor.
The question is: How accurate is the instrument?
Here's the Star*net data file:
Here's are the final settings:
Here are the deviations in direction:
And here is the output with the Chi Squared results:
I adjusted the Default Instrument Standard Error Factor for Directions up and down until the resulting error factor approached 1.
I got it to 1.081, with a 3" direction accuracy. It drops to about 1.015 at 3.2" or 3.3" (can't remember).
As to the question in the thread on "Total Station Check and Adjust routines":
"Does measuring on both faces matter if the instrument is calibrated"...All I can say is that I had as much as 10"-25" of difference between the two faces. I recently went through the calibration routines, and I could have screwed them up, but as far as I'm concerned, this is too much of a difference to ignore, so I'm going to continue to measure on both faces.
If someone knows how to calculate the standard deviation between the sets of Left and Right Face readings, I've got all my readings in an excel spreadsheet I could send along,. I can only take so much math in one day.
In the mean time, if I've messed something else up in this analysis, I'm all ears.
Why not rerun the adjustment with a standard error of centering of +/-0.001 ft. instead of 0.000?
Are Sets 3 and 4 really from the same setup as would appear? If so, choose one to use in the adjustment and turn the other off with
.DATA OFF
inline command, followed by
.DATA ON
at the end of the data to be turned off.
The key test value is the error factor for directions which should be nominally 1.00, rather than 1.22, if the correct value for the standard error of a direction is used in weighting observations. So increase the s.e. of the directions until the statistical summary reflects a value of nominally 1.00.
> Why not rerun the adjustment with a standard error of centering of +/-0.001 ft. instead of 0.000?
Done.
> Are Sets 3 and 4 really from the same setup as would appear?
BUSTED! As we went into Set 5, I said to my wife "It doesn't seem like I rotated the Tri-brach 4 times so far..." She also kept commenting (while writing down the numbers), that they kept coming up exactly the same! I should have realized that something was up. I see the value of thhe .DATA OFF/ON command, if you have suspect data, you can pretty quickly check it out. Also see the pros and cons of having pretty wife around while trying to concentrate on the task at hand.
> The key test value is the error factor for directions which should be nominally 1.00, rather than 1.22, if the correct value for the standard error of a direction is used in weighting observations. So increase the s.e. of the directions until the statistical summary reflects a value of nominally 1.00.
Here are the new settings:
And the updated results of the changes as well as the s.e. of directions:
From that data, I'd say that the standard error of a direction taken as the mean of F Lt and F Rt with your instrument is 4" (slightly better than the manufacturer's specification of 5" per ISO 17123-3:2001). That means that the standard error of an angle taken as the mean of measurements F Lt and F Rt (obtained by subtracting two directions, each the mean of F Lt and F Rt pointings) is SQRT(2) x 4" = 5.6".
> If someone knows how to calculate the standard deviation between the sets of Left and Right Face readings, I've got all my readings in an excel spreadsheet I could send along,. I can only take so much math in one day.
The simplest way to do it would be to just make separate rounds of directions out of the same three sets just used in computing the standard error, taking care to use equal numbers of F Lt and F Rt directions to end up with six different half-sets of directions. That is "Set 1" would generate "Set 1 - F Lt" and "Set 1 - F Rt".
Then, once you have the F Lt and F Rt directions in a format ready to run in Star*Net, repeat the same procedure just used, varying the s.e. of a direction until the error factor of directions is 1.00.
Not clear from statement.
Did you shut down the gun when you rotated the tribrach and then restart it?
There may be auto correction errors amplifying through your data sets
> Did you shut down the gun when you rotated the tribrach and then restart it?
> There may be auto correction errors amplifying through your data sets
I think what the poster did was to rotate the instrument on the tribrach, recenter, and carefully relevel it before taking the next set. His results are somewhat better than the manufacturer's claim for the accuracy of the instrument, so there isn't anything obviously wrong in that respect.
> Not clear from statement.
> Did you shut down the gun when you rotated the tribrach and then restart it?
> There may be auto correction errors amplifying through your data sets
No. Did not shut down gun. Just loosened clamp, rotated, re-leveled and re-centered.
Not sure what "auto correction errors" are.
> > If someone knows how to calculate the standard deviation between the sets of Left and Right Face readings, I've got all my readings in an excel spreadsheet I could send along,. I can only take so much math in one day.
>
> The simplest way to do it would be to just make separate rounds of directions out of the same three sets just used in computing the standard error, taking care to use equal numbers of F Lt and F Rt directions to end up with six different half-sets of directions. That is "Set 1" would generate "Set 1 - F Lt" and "Set 1 - F Rt".
>
> Then, once you have the F Lt and F Rt directions in a format ready to run in Star*Net, repeat the same procedure just used, varying the s.e. of a direction until the error factor of directions is 1.00.
I'll set about doing that, but in the mean time, isn't there a simple answer to the question? I mean, if I'm seeing 10", 15" or even 25" difference (all in the same direction, btw...that is, FR is always greater than FL", isn't it a no-brainer that if you just use one face under such circumstances, you're going to be off by half of the average difference?
My understanding is that the calibration routine is supposed to do this, and (as previously stated) I may have done it wrong, but that said, It now seems to me more than ever, that reading both faces is a the best way to be sure to eliminate errors due to the trunion (or whatever causes FL and FR errors). Not so?
That +/- 3' window for tilt compensation in the manufacturers specs.
I don't know why it matters or even if it does. I'm watching you test and a 30º rotation is not the same as a tripod leg warming in the sun.
> I'll set about doing that, but in the mean time, isn't there a simple answer to the question? I mean, if I'm seeing 10", 15" or even 25" difference (all in the same direction, btw...that is, FR is always greater than FL", isn't it a no-brainer that if you just use one face under such circumstances, you're going to be off by half of the average difference?
No. It sounds more like a botched collimation correction to me. It would be possible to work the problem from just the differences in F Lt and F Rt, but using Star*Net is something that should be easier for you to do.
> That +/- 3' window for tilt compensation in the manufacturers specs.
> I don't know why it matters or even if it does. I'm watching you test and a 30º rotation is not the same as a tripod leg warming in the sun.
partychief:
Not sure I understand. Is a tripod leg warming in the sun better or worse than a rotation of the tribrach (for all intents and purposes a completely new setup on the same point).
Are you saying that, given the tilt compensation in "modern" instruments, the kind of errors that might be introduced by such procedures are moot?
> Is a tripod leg warming in the sun better or worse than a rotation of the tribrach (for all intents and purposes a completely new setup on the same point).
Differential heating (as from combination of shadow and Sun) of a tripod with aluminum legs can move the instrument mounted on it out of level over a period of time. It's a much different effect than rotating the tribrach, which is simply intended to measure directions over different parts of the horizontal circle.
As suggested, I used "raw data" from the TS, both Face Left and Face Right; Then, adjusted the Project option for TS direction accuracy until the s.e. approached 1.0
Not sure I understand how to interpret the results. Compared to using both faces for measurements (as in the original analysis), this seems to suggest that I've lost about .7" of accuracy (4.7" vs. 4") by relying on one face measurements. Is this correct?
Project Options Settings:
Results:
And the deviations:
> As suggested, I used "raw data" from the TS, both Face Left and Face Right; Then, adjusted the Project option for TS direction accuracy until the s.e. approached 1.0
>
> Not sure I understand how to interpret the results. Compared to using both faces for measurements (as in the original analysis), this seems to suggest that I've lost about .7" of accuracy (4.7" vs. 4") by relying on one face measurements. Is this correct?
I forgot that one complication is that both the F Lt and F Rt directions for each set would be from the identical center, so the centering error for them would be 0.000, but there would be a centering error for the three different pairs of F Lt and F Rt directions. One way to get a rough upper bound estimate would be to reduce the instrument centering error to 0.000 and rerun the adjustment.
>One way to get a rough upper bound estimate would be to reduce the instrument centering error to 0.000 and rerun the adjustment.
That moves the direction accuracy to 4.9", maintaining the error factor for directions at 1.007.
Is the fact that the number is approaching 5" using a single face coincidental? Could it be that the manufacturer's spec calls that out for single face readings?
I still haven't figured out how to run the instrument compensation adjustments, but when I do, I'd be curious to know whether they would enable single face readings to approach the accuracy of measurements with both faces (regardless of compensation settings).
> Is the fact that the number is approaching 5" using a single face coincidental? Could it be that the manufacturer's spec calls that out for single face readings?
The DIN specification that the newer ISO spec superceded called for directions to be taken on both faces. I've never been curious enough about the ISO specification to pay $140 to buy a copy of the few pages that comprise it, but I'd think it most likely is also for a direction taken as the mean of two faces as a way of simplifying the test procedure. That is, if one wants to derive the standard error of a direction taken only on one face, he needs to include targets well above horizontal to include the contribution of the compensation for tilt axis errors.
