I am new to this board; you guys are a total benefit to your fellow surveyors.
Maybe someone can offer some insight here:
I have been paying a programmer to restore an old Husky FS3 and also upgrade it with his own software that enhances or supports the original Tripod Data Systems program.
He seems to think that it is important that survey calculations be based on 18 significant digits rather than 12 significant digits. His trying to solve or program for this has been causing him difficulty and delay in getting the program completed.
Does anyone know if it is important to be able to do calcs based on using 18 vs. 12 significant digits?
I may be over-simplifying this, but all I know is that in the end result, when we manually enter a coordinate value into a handheld, we seldom enter it beyond 3 or 4 decimal places. And the same goes for transferring data back and forth, most people choose 3 or 4 decimal places when configuring the transfer.
Is number crunching based on significant digits an entirely different concept that what I am describing above relating to decimal places ??
The rule of significant digits for multiplication is that you count the number of digits in the complete number irrespective of the location of the decimal place, and the multiplier must have that many digits (plus one preferably).
The significant digits for a coordinate value of, say, 1,234,567.89 would be 9 sig. digits, and if you are converting that from meters to feet, for instance you would want to use probably at least 3.280833333 regardless of the fact that the decimal is behind the first digit.
The law of significant digits for adding and subtracting is the number if digits behind the decimal. I don't think I ever would worry about 12 vs. 18 digits, but I let my calculator carry it out as far as it will go.
I hope that's helpful.
Consider State Plane Coordinates or ECEF X,Y,Z values, 12 digits limits the precision after the decimal point. Because of unseen step calculations involving these values some precision can be lost with 12 digits. As I recall HP calulators work with 16 significant digits.
Paul in PA
It sounds like he wants to use double precision calculations by default. That should easier than rounding to 12 digits for each calculation. This should be separate from any display of values which should allow a significant figure setting that he'll have to handle.
Using double precision may help particularly if you convert between feet and meters.
Melita
The only case I can think of off the top of my head would be with very large State Plane coordinates.
> It sounds like he wants to use double precision calculations by default. That should easier than rounding to 12 digits for each calculation. This should be separate from any display of values which should allow a significant figure setting that he'll have to handle.
>
> Using double precision may help particularly if you convert between feet and meters.
>
> Melita
:good:
I stated all my information about significant digits, but I agree. We used to have problems with old computers that didn't have double-precision capability. There could be issues going from latitudes and longitudes to state plane coordinates, or scale factors, or in some other calculations, if his machine makes those kinds of calculations. He might want to let the technician have it until he can make sure it is doing it right.
I appreciate all the comments -
The main interest is in doing routine traverse or inverse calculations. If this could be done with 12 significant digits, I think the precision is secondary.
Bhiker-
Not my area, but you might try Jacob Wall who is a surveyor and programmer
and is bright !
Try: JACOB WALL
He is the principal of Simple Geospatial Solutions and has recently programmed a HP 50G with surveyor friendly software that I hope to get to shortly.
Cheers,
Derek
Why try to support this old junk
Seriously had a FS3 husky geodimeter 650 robot both long dead but now getting on two years the Carlson surveyor plus.
Its really hard to believe the point in this good but old 386 junk
Carson Surveyor 1 (dos program) my current cogo and the same that I've been using since the early 1980s keeps up with calculations to 27 decimal places.
The last upgrade was in 1987.
I set it to display 4 decimal places so it doesn't wrap around the screen.
The only downside is that it and the newer versions of Carlson will not run on the same computer as they share the same start program commands and several same file directory locations. There may be a workaround, I just do not know it.
😉
Prove it to yourself.
Calculate a curve longhand (all facets of the curve) truncating the decimal at two spots then evaluate the same curve against your computer. If those discrepancies aren't a big deal, then you may not worry. If you work in a state plane environment regularly, then you want as many as possible.