Thanks for the replies.
These are the instrument settings I used:
Project Default Instrument
Distances (Constant) : 0.002000 Meters
Distances (PPM) : 2.000000
Angles : 3.000000 Seconds
Directions : 3.000000 Seconds
Azimuths & Bearings : 3.000000 Seconds
Zeniths : 3.000000 Seconds
Elevation Differences (Constant) : 1.000000 Meters
Elevation Differences (PPM) : 0.000000
Differential Levels : 0.002403 Meters / Km
Centering Error Instrument : 0.001000 Meters
Centering Error Target : 0.002000 Meters
Centering Error Vertical : 0.00200 Meters
Adjustment Statistical Summary
==============================
Iterations = 7
Number of Stations = 307
Number of Observations = 1443
Number of Unknowns = 909
Number of Redundant Obs = 534
Observation Count Sum Squares Error
of StdRes Factor
Angles 481 131.798 0.860
Distances 481 106.647 0.774
Zeniths 481 294.563 1.286
Total 1443 533.008 0.999
Green dots are know points. If You zoom in You can see that on these open traverses ellipses are bigger (red color)then those closer to the known points. There is one more pit that will be 800m long (open traverse again), and the is no way I can avoid these open traverses. The only thing I could do now is to observe this red line i draw on network plot. Also my sights are 50 m at best.
My one sets is: BS->FS->FS(2nd face)->BS(2nd face).
PS. The sections that intersect are not connected, traverse goes from 2 known points down and left, then it goes from the left to right side to another known point.
If you have strong math skills try Kuang 'Optimal Network Design'. No fluff...
Those are some tight standard error numbers, so I'm impressed that you are meeting chi squared test - good work.
I guess you have no direct elevation difference measurements, but was surprised to see such a large standard error on that.
What I don't understand is why I seem to see red ellipses (quite hard to make out at the posted resolution) near some known points that are a sizable fraction of those at the far end of a traverse. I'd expect the near ones to be negligible and the far ones large.
Near the lower left of the figure, why are the ellipses larger on the southern route between known points than on the northern branch that has a much longer distance to a known point? That's more like what you might achieve with the red traverse than what I assumed were your results without the red traverse.
It looks like the ellipses are very narrow, being mostly due to angle error. That makes sense if you have a lot of short sights, and why I emphasized always taking the longest sights available as well as any shorter ones.
Since you are making good measurements, the available tools for improvement would be cross connections (like the solid red line), shooting all the visible points from each setup (if you aren't already), and repeating measurements more times.
Southern route has more stations and wall prisms (sights are 20m avrg) while northern route is just traverse on the surface so there are no wall prisms (and sights are longer). I am not at work now to check but I think that going from 2 known points towards lower left point there is 17 stations and 40 wall prisms (southern route), while northern route has just 10 stations.
This red line is just 3-4 stations, I would go back 2-3 station and then shoot one sight that would be maybe 600-700 m long.
Also on top right side going from that known point towards this red line I have sights that are just 3-4 m long (stations),and ellipses are larger that can also be seen on this plot.
Indeed ellipses are narrow Semi major axis goes from 1cm to 50 cm, but semi minor axis is 15cm at most (need to check this as well).
I have used only Total station for this work, so there are direct elevation difference measurements, and those numbers were set by default on starnet.
I didnt shoot all the visible points from every station, I tried not to shoot 3x times longer sights than my backsight, meaning if my backsight was 25m away I didnt go above 70m for foresight (tried to keep all neighbouring stations approximately on the same distance or doubled at most) , but there are some wall prism observed from 3 even 4 stations.
I have used only Total station for this work, so there are direct elevation difference measurements, and those numbers were set by default on starnet.
What I meant to say is there are no direct elevation difference measurements.
senci, post: 350230, member: 10869 wrote: I tried not to shoot 3x times longer sights than my backsight
I think you're cutting off a source of valuable data by imposing this restriction. There's nothing wrong with going from a short backsight to a long foresight unless you're trying to set the far point at a specific coordinate. Other than that, it's all good data that will be properly accounted for in the network adjustment.
Jim Frame, post: 350239, member: 10 wrote: I think you're cutting off a source of valuable data by imposing this restriction. There's nothing wrong with going from a short backsight to a long foresight unless you're trying to set the far point at a specific coordinate. Other than that, it's all good data that will be properly accounted for in the network adjustment.
Well they teach us to keep traverse sights aprx the same lenght, or doubled at most, also there are some rules in the law that are suggesting this. I could try to observe some more points with far sights to reduce angle error.
I need to measure another 800m long pit of open traverse, and then I will have my network ready for all the upcoming work, in the future all pits will be connected underground (now only 2 of them are) and i will have less of these open traverses.
The data doesn't know a foresight on your traverse from a backsight on a reverse traverse. The shorter leg of an angle usually dominates the angle error (assuming both legs have equal centering errors), but both contribute. Your centering errors are going to be much worse than your instrument error on angles with short legs.
The only place I think balancing lengths of sight helps is with respect to refraction and curvature for vertical control, and that is most important when you get beyond 70 meters. You could consider using larger standard errors for the vertical component on long sights than for short sights, or you could compute the refraction and curvature corrections.
The approximate calculations below indicate that going beyond 70m won't give you any significant improvement in each angle. My intuition wasn't quite right. But using longer sights can in some places give you additional angles to help the least squares average things out. Further repetitions of an angle measurement at each setup won't help much because they have the same centering errors, but new setups to remeasure the angles will randomize the errors and thus help the least squares average them down. Two independent measurements of equal quality give you 0.7 times the rms error of one. I think Star*Net assumes all centering errors are independent, as if you did a new setup each time. This is optimistic if you really do a lot of angles from one setup.
Having 2 mm rms target centering and 1 mm instrument centering error a 25 m leg gives you an rms error 18.4 seconds for that leg. At 70 m it would be 6.6 seconds. I'm ignoring the fact that the instrument centering will be correlated in the two direction readings. On angles near 180 degrees that correlation will give you slightly larger errors in the angle.
So the angle error contribution due to centering, rms seconds error:
26" for 2 legs 25 m
19.5" for a 25m and a 70m leg
18.4" for 25 m leg and a very long leg
9.3" for 2 legs 70 m
Plus instrument error in each case.
This rule is limiting me,yes, but it was imposed more because traversing in general surveyors use longer sight, and with this rule you would have to balance your sights and thus avoiding short sight in that traverse.
But I will do more observations using longer sights at least in these open traverses.
Thanks S