Analyzing the output data from Star*net, are there clues that can be used to help determine whether one or more observations are just bad, vs. there not being enough redundancy in the number of observations? The statistics look bad, especially for distances.
Carefully recorded temps/pressure; no tree branches in the way, plenty of signal etc. Unless something's gone terribly wrong with my EDM, I'm scratching my head. Should I just take the adjustments and move on? Ideas?
post your listing file... .lst That would be a good start....
Rankin_File, post: 346592, member: 101 wrote: post your listing file... .lst That would be a good start....
This is the data from the now corrected (Thanks Scott Zelenak!), plot in my other thread ( https://surveyorconnect.com/threads/need-help-troubleshooting-an-rw5.324471/ ):
# Coordinates
C 800 427882.5150 1618086.7140 1086.8100 ! ! !
C 1000 427612.9060 1617967.5080 1144.2440 * * *
C 700 428231.323 1618143.184 1040.46 ! ! !
C 1100 427615.2280 1618167.7150 1116.6430 * * *
C 14A 427402.3560 1618040.0610 1161.0790 * * *
C 1200 427720.6650 1618204.8350 1092.5830 * * *
# Observations
D 800-1000 294.1440
V 800-1000 60.7040 5/5
A 800-700-1000 194.385200
D 1000-1100 200.2200
V 1000-1100 -27.6020 5/5
A 1000-800-1100 65.292900
D 1000-14A 222.6890
V 1000-14A 17.1000 5/6
A 1000-800-14A 137.073000
D 1100-800 279.3290
V 1100-800 -29.6380 5/5.08
A 1100-1000-800 73.483900
D 1100-14A 248.2110
V 1100-14A 44.4480 5/6
A 1100-800-14A 227.481100
D 1100-14A 248.2140
V 1100-14A 44.4360 5/6
A 1100-1000-14A 301.365300
D 1100-1200 111.7810
V 1100-1200 -24.0530 5/5
A 1100-1000-1200 110.042100
D 1100-1200 111.7800
V 1100-1200 -24.0600 5/5
A 1100-800-1200 36.145700
D 1200-800 200.3700
V 1200-800 -5.7730 5/5.08
A 1200-1100-800 124.285600
Listing file attached. Seems to be a couple of tenths floating around in there, here and there. Your measure ups being nominal is part of the problem.
rfc, post: 346595, member: 8882 wrote: This is the data from the now corrected (Thanks Scott Zelenak!), plot in my other thread ( https://surveyorconnect.com/threads/need-help-troubleshooting-an-rw5.324471/ ):
# Coordinates
C 800 427882.5150 1618086.7140 1086.8100 ! ! !
C 1000 427612.9060 1617967.5080 1144.2440 * * *
C 700 428231.323 1618143.184 1040.46 ! ! !
Well, the first thing that jumps out at me is that you have TWO points with fixed coordinates. I'd run the adjustment with only ONE fixed and see if you still have as large an error factor in the distances as your statistical summary presently does.
If I do that, I get the "Only one fixed coordinate and no azimuth orientation....weak geometric yada yada" error. Besides 700 and 800 Are fixed because they've been adjusted from a previous portion of the network...I'm running into the 10 station limit, so trying to analyze the network in segments...not exactly ideal.
Another option is loosen up your centering error for zenith to be 0.10 and zenith angle to 10".
add a bearing between 2 points but don't fix it...
Here are my error settings.
The verticals are still my Achilles heal, so I tried loosening them up (Elevation Differences, and Centering Error Vertical).
But why would small errors in the HI/HR's contribute so much to horizontal distances being so whacked out?
rfc, post: 346613, member: 8882 wrote: If I do that, I get the "Only one fixed coordinate and no azimuth orientation....weak geometric yada yada" error. Besides 700 and 800 Are fixed because they've been adjusted from a previous portion of the network...I'm running into the 10 station limit, so trying to analyze the network in segments...not exactly ideal.
well based on the data so far the fixed cords you're using for 700 and 800 are about whammie and rag tape quality... fix 700 and let 800 float-
C 800 427882.5150 1618086.7140 1086.8100 .1 .1 .1
that should get it to run.
Good thought. I'll try that next. I'm still not understanding how those two points, and the distance and azimuth between them, have anything to do with the adjustment. 700 is just my back sight. I'll try it though.
No dice, Commander. Not sure I understand what you mean by the comment though. 700 and 800 could be ANY two points on earth. They're not part of this adjustment at all. It shouldn't matter WHERE they are. I have only one angle that uses 700, and that's just for orientation. What am I missing. Here's the output:
My brain is fried for the evening though, so I'll start over in the morning. Thanks for the ideas.
This isn't a convergence problem, but after you lick that problem, basing the orientation of the network on two points relatively close together is going to give you large ellipses at the other end of the network.
I note that if I take out the measurements 800-1000 everything else fits a whole lot better. Recheck those.
For the V code in vertical difference mode, does your version of the software use the HI HT values? The (old) pdf manual I have seems to indicate they would be ignored. Thus when you change the rod height that difference would show up in the adjustment as a discrepancy.
clues that can be used to help determine whether one or more observations are just bad, vs. there not being enough redundancy in the number of observations?
The most important clues usually come from comparing the residuals of your measurements with the standard error. When there is a StdRes value of 5 or 50 that usually means there is a blunder in the neighborhood of that measurement, although of course the adjustment does spread the error around.
Lack of redundancy lets the errors get spread around more, so gives you a lot of moderate residuals instead of a few big ones. Low redundancy gives lower Error Factor values when using the Star*Net philosophy of accepting your standard errors and not re-estimating them.
OOPS. I re-read the manual and it appears it does use the HI/HT. The example didn't.
starnet 6.0 always treated the HI and HT as gospel fixed, except for the centering error. I must have spent weeks chasing strangeness in zeniths when the crew (not always me) busted a measure up.
Making the coordinates "free" is a lot free-er than leaving them unweighted (no punctuation after). Although that can get weird also.
Try not fixing anything at all, rather, substituting ".1 .1 .1" for the "! ! !" or "* * * " or ".01 .01 .01"
Have done many networks where nothing is fixed, coordinates get their 1 sigma NEZ std errors, whether from RTK or static or published.
Columbus has some interesting brute force X! stuff that can help decide what to truly "fix".
Bill93, post: 346623, member: 87 wrote: I note that if I take out the measurements 800-1000 everything else fits a whole lot better. Recheck those.
I think you may be on to something. I can't explain it, but I did a "what if" with the distance measurement between 800 and 1000 in the data file. When I got the "measurement" up to 294.814, the residual drops to .0057' and the Chi Squared passes.
It's still somewhat above freezing around here at times, so I'll head back and take a slew of observations backwards and forwards between them.
Your vertical standard errors are enormous.
Yes, I know. I raised them that high to eliminate measure ups as a factor. I know I still have a problem with the verticals, but in this case, I think I'm chasing down some errant distances. The vertical errors are not going to contribute much to that. Too much to think about; too little time.:excruciating: