Observed Angle Sould Have Been Close To 37°05'33"
> You're probably right. But I'm horribly messed up now, because, going through my field notes (they still are terrible), I found yet more discrepancies.
You said earlier you like to run before you walk.
SLOW DOWN.
Learn to take clear, concise, legible field notes. Be sure you have it recorded correctly and then move on to the next step.
Why run field work if you cannot understand or rely on the data in the office?
Ken
Field notes?
> > You're probably right. But I'm horribly messed up now, because, going through my field notes (they still are terrible), I found yet more discrepancies.
>
> You said earlier you like to run before you walk.
>
> SLOW DOWN.
>
> Learn to take clear, concise, legible field notes. Be sure you have it recorded correctly and then move on to the next step.
>
> Why run field work if you cannot understand or rely on the data in the office?
>
> Ken
Excellent advice. Noted. I need to see some real examples of some good field notes. Anyone who takes them anymore care to share? Should probably start another thread for that.
Making Progress
Does Starnet always expect internal angles? I put in angles to the right (going around the traverse clockwise), and I got a pretzel!
I'm getting close, though, lol. At least the syntax is getting cleared up.
It's Like Watching Paint Dry
> I'm a little confused about the discussion of point 10. I took that as the computed end of the open traverse, replacing point 2 for sighting from point 9 only. When you close the traverse, everything would refer to point 2.
>
Still can't get the final angle (9-2-3) into Starnet, without the plat blowing up. But I think, finally, I have all my OBSERVED angles FROM THE FIELD BOOK, correctly entered.
If this is the case, I can get about going after the errors...which opinion still has it in the neighborhood of 6-7-8.
I'm learning that (lousy field notes aside), you really need to think through the work ahead carefully, then have a set work flow from field to final drawing. I've wasted a ton of time measuring, adding, subtracting, converting angles to azimuths to bearing calls and back, increasing chances for errors all along the way.
I've got to believe that modern data collectors, combined with a start to finish work flow would make this easier.
Field notes?
For field note examples, get any basic surveying textbook you can find from a few decades ago. I particularly like my 1960's Davis, Foote, and Kelly, and I have a few others of various dates that aren't as good but are useful.
More recent texts by Wolf and Ghilani are good on a lot of subjects, and a not-quite-latest edition can probably be found very cheaply, but they are too modern to say much about field notes.
My own field books aren't any shining examples, and don't do a good job of following an orderly fixed format, but I make sure that I record exactly what I measured, and how, and any other pertinent info that occurs to me.
Field notes?
> More recent texts by Wolf and Ghilani are good on a lot of subjects, and a not-quite-latest edition can probably be found very cheaply, but they are too modern to say much about field notes.
>
Aloha,
I purchased a copy of Ghilani and Wolf's book Elementary Surveying. An Introduction to Geomatics The 13th Edition available online here. It is a very good book. I am learning a lot from it.
Download here:
http://www.academia.edu/3303612/Solutions_-_Elementary_Surveying_13th_ed
Field notes?
That's exactly what I'm using. See here:
While pristine in it's appearance, It doesn't really show how to deal with real, practical issues, such as:
What's the best way to name stations? How do you record multiple shots to the same station? Do you record each iteration, i.e. 33.3 degrees, 66.6 degrees, 99.9 degrees? Then where do you do the math to reduce it?
I've been back to the same site a dozen times, and if I start with station "1" each time, there may be another "1" nearby from a previous trip. Do you just increment the station number for additional shots? How about side shots? Where do you record them? Are they assumed to be turned from wherever you've back sighted to?
All simple stuff, but I guess I'm asking for some examples of field tested, best practices, rather than a perfect page from a text book. I think I might start another thread on this!
"Final" Review please
I finally have put all of my real, confirmed, verified, averaged field measurements into Starnet. I need help reading the output. As has previously been suggested, the 6-7-8 region is of concern.
Is the Error Propagation table telling me that? Along with the bigger ellipses there? It looks that way.
Finally (before I head out and do those again, as well as 6-8 directly....
What does "failing the Chi-Square test" mean? is there a numerical target? Does passing it mean I can move on, lol?
"Final" Review please
Regarding Chi Square test, see my post above at August 13, 2014, 21:44
The 6-7-8 region has large error ellipses for two reasons. One is the short leg of the traverse, but the main one is that the likely error accumulates (propagates) from your "known" point as you go in either direction around the traverse, so that you don't know as accurately where those points are relative to the starting point.
The "Relative" ellipses you can turn on will tell you that, for instance, the distance and direction of the 6-7 leg is better known than the large ellipses for where those points are relative to your fixed 1-2 leg.
If you re-did it holding point 6, for instance, as the known point you would see similarly large ellipses on the other end that is now tight. This says you don't know very accurately where points 1 and 2 are relative to 6.
And realize that the program doesn't know where the errors actually are, but says there must be this much error somewhere, so the most likely guess is to distribute it around. Without more redundancy and cross-connection measurements it doesn't have any better data to work with.
And before somebody dismisses least squares as being useless because it distributes the error around, note that it is still making a better guess than compass rule or other adjustment methods.
getting interesting
I set up on 6,7 and 8 and shot the others, 3L, 3R then averaged both.
I think my original distance between 6 and 7 might have been wrong for some reason. It was 96.09. Now I've measured it 2X3 (six times total, averaged) and it's 96.665.
If I "cheat" and change 96.09 to 96.665 on line 8 , the ellipses all get a lot smaller, although still not enough to pass Chi Squared.
Next I'll try to get through the woods from perhaps 4 to 8 or 4 to 9, with only a single station in between.
"Final" Review please
Failing the chi-square test may also indicate your a priori error estimates are too small or large. I think the Star*Net manual speaks to this with possible explanations if you are failing the upper or lower bounds of the Chi-square distribution. I don't remember off the top of my head... Sorry.
Failing chi-square doesn't necessarily mean your data is bad.
I have also seen network adjustments.... Say in TBC run afoul when I ran a 3D adjustment on the crews traverse. It blew up. Turns out they lost a tripod and prism to wind gusts and were running angles only back sights. So the foresight point had a valid 3d position, but TBC did not know how to handle the 2d back sight combined with a 3d foresight.
My solution was to make the crew run a set of levels through the network. I the. Held those elevations and ran a 2d adjustment of their original traverse. Worked like a charm.
Even though Star*Net is flexible regarding data, I would make sure you strive for the same field techniques, always.
Good luck! You have a fun problem to solve.
getting interesting
If you get rid of the 96.09 and the angle at 8 of 274-56-30 then the rest of the measurements fit reasonably.
They also pass the chi-squared test for giving it reasonable error estimates. The Error Factors say your angles are a little better than your estimate and the distances not quite as good as estimated.
I get confused by the T and M formats, so I broke it into A and D records. This allows me to * out any individual angle/distance measurement while keeping the corresponding distance/angle from the traverse. This is what I meant earlier about using a network format instead of a traverse format. Just easier for me.
This example also shows how I put the "known" point at 6 instead of 1 to get the different set of ellipses.
#C 1 1000 1000 ! !
# alternate "known" point
C 1 1000 1000 * * # for initialization only
C 6 324.5554 552.1820 ! !
B 1-2 N56-23-13W # 303-36-47
D 1-2 130
D 6-7 96.665
D 6-8 122.125
D 2-3 329.69
D 3-4 170.195
D 4-5 175.12
D 5-6 271.084
D 6-7 96.09 *
D 7-8 43.158
D 8-9 384.65
D 9-2 376.702
A 3-2-9 37-07-04
A 6-7-8 116-37-57
A 7-8-6 44-59-37
A 1-2-3 94-57-47
A 2-3-4 177-08-18.5
A 3-4-5 142-58-33
A 4-5-6 282-08-17
A 5-6-7 264-14-20
A 6-7-8 116-34-41
A 7-8-9 274-56-30 *
A 8-9-2 218-54-50
Still Not An Adjustable Traverse
You have not closed the traverse, since there is no angle at 10.2 including 9 and 3.
I do not like the way your angles are referenced, I am used to the BS/Occupy/FS shortcut, beginning with 1-2-3 and ending with either 9-(10/2)-3 or 9-(10/2)-1.
Paul in PA
"Final" Review please
See this explanation of the chi-square test...
"Final" Review please
The results of a minimally redundant least squares adjustment are nearly identical to a compass rule adjustment.
Redundancy fuels the power of a least squares adjustment. If one is not interested in taking the time to add network redundancy they are missing out on the most valuable characteristics of LSA.
Still Not An Adjustable Traverse
I put that angle back into the picture. Good eye; somewhere along the line I pulled it out (I think when 10 and 2 were "the same" point (they weren't...they were 2' apart.
I'm also trying to change to Bill93's method of input data...I hate the "At-From-To" method myself...it's completely un intuitive.
But I can't get it to work...I cut and paste his stuff into Starnet, and get the "could not compute approximate coordinates" error.
If I can get this method to work, I can go back to the field, take more shots, and keep adding them in.
Still Not An Adjustable Traverse
I think the big problem is that I had my project options set for angles FROM-AT-TO or Backsight-Occupied-Foresight and you had been using AT-FROM-TO. You either need to translate the order or change the option. I was trained in geometry class a half-century ago to use the FROM-AT-TO format and can't get used to the other way.
Other changes are needed:
Line 3 and Line 24 are conflicting on distance 1-2 and redundant with line 22 on the bearing. One set of numbers should be removed. I was still using one of your old diagrams for distance 1-2.
Line 17 should be deleted because it repeats line 2, and Line 19 should read
#C 1 1000 1000 * * # for initialization only
so that it is properly set up if you take the # off and try making point 6 "known" instead of 1.
By the way, get rid of that .MAP ON command unless you have read all about it and need it for something you are doing - you don't need it right now so it is a confusion factor in reading the file.
Still Not An Adjustable Traverse
Changed format to FROM-AT-TO. That did it! Thanks.
Got rid of .MAP ON
Added some cross sights, all at least 3X, including 4-11-8 and 4-11-6, which I thought would make the ellipses at 6,7,8 smaller, but they're not. Here's the input and plot:
And Here's the output:
I thought that getting across the network with additional sights was going to tighten things up, but I'm not seeing it yet. I'm really confident in the latest shots, too...Prisms on Tribrachs, or directly on the ground in the grass over the station. Maybe I just don't know how to read/evaluate the data.
Thoughts?
You Have A New Point 10
Do not worry about the cross ties , etc. until you get your traverse close.
Your first point 10 was only 0.3', not a big problem. Use an observation From 9 At 2 To 3 and/or 1. Once that works add your cross ties one area at a time.
Paul in PA
You Have A New Point 10
Use an observation From 9 At 2 To 3 and/or 1. Once that works add your cross ties one area at a time.
>
> Paul in PA
Yes, the old 10 was actually 2, after traversing around.
But I have the angle already:...Line 24 from 3 at 2 to 9: That's the 37 07 04. I tried changing it (per your advice) to 37 05 33, but that seemed to make the elipses slightly bigger at the other end, so changed it back. Not sure why that angle is suspect.
I did add the cross ties one at a time, butI think I need help interpreting all the output from Starnet, although, in all honesty, I haven't gotten to page 200 of the manual yet. That's probably where interpreting the output is.
What's the process? Add a cross tie, analyze which angles, distances have the largest errors? Then do you change them on the fly? Like manually change them per Starnet's advice?
I haven't been changing anything to this point, other than to correct obvious errors.
