As the header says: since a few versions the WP34s shows RUN instead of STOP (for the R/S key) when entering or listing a program.
Is this intended? Looks quite strange ...
BTW, I've found a small bug in my TVM solver - I'll release a new version today after a few more tests.
Franz
Edited: 10 May 2012, 5:02 a.m.
Quote:
As the header says: since a few versions the WP34s shows RUN instead of STOP (for the R/S key) when entering or listing a program.
Not sure where this is coming from -- it is wrong nonetheless. I can duplicate in 2939 firmware and 2975 in the emulator.
Quote:
BTW, I've found a small bug in my TVM solver - I'll release a new version today after a few more tests.
We await the new version.....
- Pauli
Quote:
We await the new version.....
Well, it was not really a serious problem, i.e. it didn't return any wrong results. It was just so that the solver SLV didn't get the correct (intended) 2 initial guesses but some other meaningless values, so it may fail in some cases when solving for I.
But it seems that bug-fixing isn't always a goos idea. :-(
Now that I use the 2 correct initial guesses, _sometimes_ the solver fails because of a different problem: it produces values <= -1 for the internal interest rate during the SLV process, and so the financial equation stops with an error (because ln(1+x) is of course not defined for x<=-1.
Now my question: should I implement a test for i<=-1 and return NaN in this subroutine for SLV?
Can SLV handle such NaN 'results' and what would SLV do in this situation?
Franz
Quote:
Can SLV handle such NaN 'results' and what would SLV do in this situation?
It fails immediately on a NaN if the user has flag D set. Without flag D set, the command interpreter fails when a NaN result would occur (with a Domain Error).
- Pauli
Ok, then NaN is definitely no option in the TVM equation.
What I still don't understand is how SLV could ever reach such values i<=-1 if I use I1=0% and I2=1% as initial guesses!?
I guess I'll have to make more tests, and maybe I'll have to even completely rewrite this (solving for I) part of TVM ... :-(
Franz
If the initial guesses don't bracket the solution, solve looks elsewhere.
0% and 100% might be better guesses.
- Pauli
Quote:
If the initial guesses don't bracket the solution, solve looks elsewhere.
I always thought that SLV would build a quadratic function from the initial guesses and use the zero(s) of this parabola as new root-estimation!?
Well, then I'll try 100% as 2nd value ...
Franz
Quadratic interpolation is one of the methods uses. That doesn't help if a NaN result appears.
What's: 3 x2 + NaN x - 2???
- Pauli
Quote:
As the header says: since a few versions the WP34s shows RUN instead of STOP (for the R/S key) when entering or listing a program.
Is this intended? Looks quite strange ...
It used to show STOP in any case, even when RUN was meant. I changed it when I added the printer trace function to make the printout more meaningful. I just forgot about program mode. :-(
Fixed!