Great traverse accuracy but fails least squares

[r-s-mayer]

Make sure your instrument constants are correct. They shouldn’t too loose. Look in Options, Project, Instrument. 

[ken-salzmann]

Can you post your data and settings?  Something may jump out to the long time Star Net users.  It is a great program, I use it all the time, but once in a while it can get cranky.

Ken

 

[rover83]

Posting the data will help a lot, if you can.

Failing the chi-squared test does not necessarily mean that your network is blown, it usually just means that your standard error estimates are off a bit. (This, of course, is contingent upon removing any major blunders beforehand.)

Check your standard errors for centering and height measurement, for both instrument and target. Make sure your angular and EDM instrument standard errors are correct.

It failed under, not over, the minimum bounds, and not by much, so as Mr. Mayer says, it might be a good idea to tighten up those standard errors and re-run. Change a single setting and re-run instead of changing multiple settings, it will help you get a good idea of how they affect the adjustment.

 

---

“…people will come to love their oppression, to adore the technologies that undo their capacities to think.” -Neil Postman

[rkinnie]

With a short leg, you will need to add a centering error – this will open up the angle error to match the short distance

under options – instrument – make the centering error like 0.003 for your target – may also need a vertical if it this is a 3d traverse

at 21 feet the 5 seconds is too tight – needs more like 60 which is 0.006 at 21 feet

you could also add a single error for that measurement in the dat file on that line   

remember starnet treats each measurement the same which is good for balanced shots – you have to allow for the adjustment to open up when they are not

[fobos8]

Hi guys

Have attached a copy of the listings which at the start has the settings. Also have attached the data file as a .txt as a .dat won’t upload – you can change the file extension back to .dat

Thanks for offering to look at the data for me.

Cheers, Andrew

Listings.txt

 

[lee-d]

As Rover83 points out, you failed under the minimum bounds, not over; this means that your initial error estimates exceed the adjusted errors. Check your standard errors and look for something that’s overly pessimistic, like 10″ on the angles when you’re using a 5″ gun, more than about 0.006′ of centering error, and more than whatever your EDM specs at for distances.

[lee-d]

 Project Default Instrument
        Distances (Constant)              :    0.009144 Meters
        Distances (PPM)                   :    0.000000
        Angles                            :    3.000000 Seconds
        Directions                        :    3.000000 Seconds
        Azimuths & Bearings               :    4.000000 Seconds
        Centering Error Instrument        :    0.000000 Meters
        Centering Error Target            :    0.000000 Meters

So looking at this your EDM error is 0.00914 meters, when your EDM – unless your gun is really old – is probably closer to 0.002m. That will cause you to fail the way you did. But you should also put in realistic centering errors, I usually go with one or two millimeters. All of these settings should be correct for your instrument and procedures.

 

 

[lee-d]

Also if you go down and look at your residuals most of them are sub millimeter and the highest one is 1.7mm. I don’t think you’ve got a whole lot to be concerned about here.

[fobos8]

Apologies for the earlier listing file. I hadn’t changed any of the instrument setting from the default ones. I just went onto the starnet calculator and entered the data for my instrument and it generated instrument data.  After entering the “correct” instrument data I get worse results – total error factor of 0.601, bounds were 0.917/1.082.

My instrument is a 3″ Trimble S7.  Standard prism distance accuracy is 1mm (0.003 ft)  + 2ppm. I always do 3 sets of f1 and f2.

http://support.microsurvey.com/convert/instrumentsettings.html

 

Listing_corrected.txt

[jaro]

Were you doing this robotically ?

I don’t see a 21 foot traverse shot done in robotic mode being much good for anything really.

James

[jeran-hopfe]

As some users have said here, this is a matter of your “expected error” based on project settings being higher than the “actual error” in the survey.  If the instrument settings are correct, etc, you’ve got a great survey, not a bad one.  Unfortunately it’s easy to hide a lot of mistakes by simply bumping up the tolerances in the instrument settings, making the chi-squared test artificially low.

[Norman_Oklahoma]

This is a loop traverse, no cross ties, which chances to have a spectacular closure. But here is very little redundancy. Hence it is virtually impossible to dial the error estimates down enough to get the Chi-square up.  The data is good, but probably not as great as the adjustment report says it is. Throw in more redundancy in the form of cross ties, and the result will likely be a little more “real world”.   

[rover83]

It could also be the result of forced centering, although I don’t think it was mentioned here. OK for one-time loop traverses adjusted by Bowditch, not good for networks and least squares.

---

“…people will come to love their oppression, to adore the technologies that undo their capacities to think.” -Neil Postman

[true-corner]

When running control with classical instruments you have to follow the procedures outlined in the FGCC Manual.  Cross ties on short legs are part of it.

https://www.ngs.noaa.gov/FGCS/tech_pub/1984-stds-specs-geodetic-control-networks.htm

[fobos8]

For cross ties/braces if the main traverse stations are not visible to each other directly can you put intermediate station in and “traverse” from one to another. The brace would have a station in the middle of it.

[Norman_Oklahoma]

“Have to”? This standard is the one we have, but I’m not sure that standards developed in the 1970’s have to be applied to the instrumentation we use today. The standard references minimally constrained least squares adjustments regularly – nobody outside of government or academia was doing LS in 1984. 12 sets of angles? Who does that? 

It’s the concept- of performing the work in a certain way to achieve desired results – that is important. 

[Norman_Oklahoma]

Absolutely. That’s one of the beauties of LS. You can stir almost any combination into the mix. 

[fobos8]

Okay – great thanks