My ongoing astro endeavor has raised an issue I'm curious about. All of my observations and reductions so far, between two NGS marks appear to be off in one direction...My azimuth is greater than published. This could be technique...like being consistently late on the time for example.
I'm a ways away from introducing astro to my control network in any serious way (that is next), I'm wondering now how I would write the lines for astro azimuth in Star*net generally, and more specifically:
If one thinks a systematic error is greater in one direction than another, is it possible to put that assumption into Star*net? Something like a machine tolerance for a pin in a hole (+0, -.002)? In other words, can the one sigma curve ever be lopsided?
How are you figuring the azimuth between the two marks?
If you are using SPC coordinates have you accounted for convergence?
Which mark are you occupying? A to B is different than B to A.
Paul in PA
Paul in PA, post: 369217, member: 236 wrote: How are you figuring the azimuth between the two marks?
If you are using SPC coordinates have you accounted for convergence?
Which mark are you occupying? A to B is different than B to A.
Paul in PA
Yes. All of that. I'm using the published convergence at the setup mark; Laplace, etc. I'm not minutes off; I'm seconds off. I'm comparing my results with the Inverse3D program results. My question is not related to blunder detection, but to a methodology for ascribing error parameters to my observations.
rfc, post: 369213, member: 8882 wrote: Something like a machine tolerance for a pin in a hole (+0, -.002)? In other words, can the one sigma curve ever be lopsided?
If it were lopsided, the error wouldn't be random, it'd be systematic and therefore predictable. Survey measurements in a least squares adjustment are assumed to be free of systematic error.
As an amateur (*very* amateur) machinist, if I were going to be creating some pins that had to fit into a specified hole, I'd set my target diameter at the desired size minus my likely achievable tolerance (standard error). That way I'd be reasonably certain that 2/3 of the pins would fit the hole. I wouldn't aim for the hole size, because I'd know that a significant percentage of the pins would be too big to fit.
Jim Frame, post: 369251, member: 10 wrote: If it were lopsided, the error wouldn't be random, it'd be systematic and therefore predictable. Survey measurements in a least squares adjustment are assumed to be free of systematic error.
So you're saying there's no such thing as a "Quasi-systematic" error? Any error induced by human beings might be "systematic" but not entirely predictable.
Suppose one instrument operator always turns the screws clockwise; another turns them counter clockwise, producing different results. If both work on the same project, and all the observations are put into Star*net, are they considered systematic or random. I guess the question has to do with where does the "human element" come into play with regard to probability calculations?
rfc, post: 369296, member: 8882 wrote: So you're saying there's no such thing as a "Quasi-systematic" error? Any error induced by human beings might be "systematic" but not entirely predictable.
Suppose one instrument operator always turns the screws clockwise; another turns them counter clockwise, producing different results. If both work on the same project, and all the observations are put into Star*net, are they considered systematic or random. I guess the question has to do with where does the "human element" come into play with regard to probability calculations?
If the error is systematic, it has to be eliminated in order for the random errors to be properly accounted for in the adjustment. Otherwise, it'd be like someone adjusting total station measurements without accounting for the known error characteristics of the EDM (e.g. 2mm + 2ppm).
All azimuths are greater than expected. But not by the same amount. A systematic error would mean all azimuths are greater by the same amount. And if each successive azimuth gets closer to the expected value, that's not systematic.
Here's an source of systematic error in Astro:
If your obs on multiple days are at the same time of day. i.e. Setting sun v rising. If you only observe setting sun an instrument error, like trunnion leveling, will be systematic.
Larry Scott, post: 369530, member: 8766 wrote: All azimuths are greater than expected. But not by the same amount.
I guess I missed that. Where did you see that their not greater by about the same amount?
Jim Frame, post: 369577, member: 10 wrote: I guess I missed that. Where did you see that their not greater by about the same amount?
From my "Sun's not getting hotter so I must be" thread:
Accumulated results from several astro outings....
10 April: 175-26-13.0
15 April: 175-26-08.3
24 April: 175-26-01.2
NGS Published Az between AA8189 and AA8188: 175-25-56.7
I guess I should keep doing more, and see if they stop at the published number or keep going in the other direction.:-S
Three results is a bit sparse for a reliable data set, given all the error sources available. Also, what instrument are you pointing with?
Note how much his az change, or get better. There not some 20-30" erratic. But progressive.
(And the first is probably an ignore. So a population of like quality azimuths have to be evaluated for systematic err or bias.)
Jim Frame, post: 369585, member: 10 wrote: Three results is a bit sparse for a reliable data set, given all the error sources available. Also, what instrument are you pointing with?
Topcon GTS-255. When you say "three results", remember that's three sets of 16 observations (8 Direct; 8 Reverse).
I've read (or Larry might have said), that "in the day", they used to do 7 or 8 sessions in order to call a solar usable. My objective here was just to see if my results would be usable (in the "real" world) for my control network. I'm calling these good enough to move shop to that location.
Someone suggested picking a distant tower or landmark to use rather than setting up a reflector all the time, and I've found such a perfect mark about a mile from one of my best central control points (a Magnail in ledge) on a knoll in the center of my network. I'll be able to do AM and PM solar, and Polaris from there.
That's the ticket. (You can see a mag nail at a mile?)
When it's warmer, and you can take an Astro without much fuss, then you'll take enough to know how good any one Astro may be.
