hello again to all. i am currently revisiting a LSA package called CG Adjust. the cg part refers to charles ghilani, of penn state. he and paul wolf published a book called adjustment computations, and this software was included in the purchase.
anyhow, i am working on a '3d geodetic network', and am getting an error message 'invalid floating point operation'. anyone know what this means?
i have a suspicion that it might mean there is not enough measurements to fix the stations in the network, but i believe i actually do have enough. i have noticed that other error messages in this software don't really point directly to the problem, or give me much detail of what could be wrong.
all in all, i like the software, and the book. cg adjust reads a text file similar in format to a star net file, but is a bit more cumbersome to write.
CG Adjust software, clarification
i am working on a '3d geodetic network', and am getting an error message 'invalid floating point operation'. anyone know what this means?
maybe no one wants to think that hard on a sunday, i will try again tomorrow
CG Adjust software, clarification
thanks so much
Hi,
to me it sounds as a program bug and probably not a problem with the data ...
floating point refers to a type of data and the error states that an invalid operation is to be performed on this type of data. (could be an integer division operation)
I should try to find a link to their website and search the FAQ or update section ...
Chr.
Is there an example or tutorial file which comes with the software?
If so does it process correctly?
Which Edition Of The Book ?
I have several editions of Wolf and Ghilani's adjustment book and several different versions of Adjust. I have an early DOS version and a newer Windows version. I have run Adjust and StarNet side by side on the same adjustments and never saw any differences in the first 2 decimal places. I have never seen that floating point error, so it may be something with a newer version.
If you want to send me the file I can try and replicate your problem, but believe that intervention from Chuck may be required. I just fired up my older computer and it has Adjust Version 4.1.0 (CG Consulting) on it from the second adjustment book I bought. I have just downloaded Adjust Version 5.10 from the Penn State Wilkes-Barre site and will install that also to run your file.
How big a file are you trying to adjust?
Paul in PA
> Is there an example or tutorial file which comes with the software?
there is an example for every network type
>
> If so does it process correctly?
they do appear to process properly
Which Edition Of The Book ?
> I have several editions of Wolf and Ghilani's adjustment book and several different versions of Adjust. I have an early DOS version and a newer Windows version. I have run Adjust and StarNet side by side on the same adjustments and never saw any differences in the first 2 decimal places. I have never seen that floating point error, so it may be something with a newer version.
>
> If you want to send me the file I can try and replicate your problem, but believe that intervention from Chuck may be required. I just fired up my older computer and it has Adjust Version 4.1.0 (CG Consulting) on it from the second adjustment book I bought. I have just downloaded Adjust Version 5.10 from the Penn State Wilkes-Barre site and will install that also to run your file.
>
> How big a file are you trying to adjust?
>
> Paul in PA
i am currently using adjust 5.0. this network is only about 20 stations total, and geographically very small, fits in about 1.5 acres.
'Invalid Floating Point' is probably being caused by a division by zero operation. Since you are performing a 3D adjustment, verify the following:
1) Do you have vertical observations (e.g., a zenith angle with slope distance, etc.) to every station that is not fixed in height?
2) Do you have horizontal observations (e.g., horiz angle, directions, etc.) to every station not fixed in Lat/Lon?
3) Are any observation standard deviations set to ZERO?
4) Do you have one network or 2..N disconnected networks?
5) Do two or more stations have the same starting (approximate or fixed) coordinates?
6) Does your software indicate the 'step' being processed when the error occurred (e.g., computing weights, computing inverse, etc.)?
Normally, it is up to the software to detect these problems and alert you. Software is never bug free, but many severe bugs (crashes) can be pinpointed (and user alerted) by adding numerous check in the code.
> 'Invalid Floating Point' is probably being caused by a division by zero operation. Since you are performing a 3D adjustment, verify the following:
>
> 1) Do you have vertical observations (e.g., a zenith angle with slope distance, etc.) to every station that is not fixed in height?
>
> 2) Do you have horizontal observations (e.g., horiz angle, directions, etc.) to every station not fixed in Lat/Lon?
>
> 3) Are any observation standard deviations set to ZERO?
>
> 4) Do you have one network or 2..N disconnected networks?
>
> 5) Do two or more stations have the same starting (approximate or fixed) coordinates?
>
> 6) Does your software indicate the 'step' being processed when the error occurred (e.g., computing weights, computing inverse, etc.)?
>
> Normally, it is up to the software to detect these problems and alert you. Software is never bug free, but many severe bugs (crashes) can be pinpointed (and user alerted) by adding numerous check in the code.
1 might not, have to check
2 yes
3 no
4 one network only
5yes, but they have a large 'float', appeared ok to do this based on instructions and samples. i should vary the coord by a hair, as a check
6 no, i don't think it is that cooperative. this software seems to be good, but not so good when something is wrong. it gives vague error messages
thanks so much for the food for thought.
there has been no response by the author, but given the date, i don't expect anything from him for as much as a month
I should clarify item #1 in my previous post. For a 3D adjustment you need some sort of vertical data at or to every non fixed (1D or 3D) station. What I omitted earlier was the AT or TO part.
If station AAA (not fixed in 1D or 3D) does not have any vertical data coming into or leaving it, there will be a problem. Some adjustment packages will recognize the problem and fix that point in height (automatically during adjustment). Others will alert the user to either add more observations or manually fix the station in height.
In situations where you are using conventional measurements (non gps) and are having problems with the 3D adjustment, try performing a 2D adjustment with the same observations. Often, you will be successful and you can assume there is something wrong with your vertical observations (zenith angles, height difference observations, etc.) This greatly narrows the problem space.
It's also quite likely the problem is not related to any of my checklist items. In that case, simplify the network any way you can (remove stations or observations) until you can get an adjustment to succeed (maybe focus on 2D adjustments first). Then, add each station or set of observations back in gradually, adjusting after every data change until you get the failure again. That way you will isolate what data caused the problem. This sort of what-if analysis can be tedious, but I believe it pays off in the long run.
Thanks Scott,
That's actually how we learned to trouble shoot it in school. I had a problem where I was presented an adjustment that wouldn't converge in Star*Net and I had to figure out why. I reached out to Kent, Peter Lazio and Doug Bruce and I picked their brains. It was a great exercise and I feel blessed to have been able to have access to such great minds via this board ad the original RPLS.com.
Cheers,
Ralph
Agreed. Those guys are very analytical.