HP Prime Unable to solve system of equations that 50g easily solved

[losfrijoles]

I have two equations that are very similar to many equations I will be expected to solve in my classes and future career:

1.05=1.8*y/(y+x)

20000=1/(1/x+1/y)

Using MSLV this was easily solved on my HP 50G in about 60 seconds (slow, I know, but I put bad guesses in).

The Prime, on the other hand, can't seem to do it. The solve() function just plain crashes the calculator (crash = screen turns off + calculator restart). The zeros and cZeros functions are able to solve it, but give me alternate answers which are completely useless (although they are correct): x=19047.6, y=-400000 (they are useless because neither X nor Y can be negative).

The correct useful answer (verified by 50g and wolfram alpha) is X=34287.71, Y=48000.

I guess what I am asking is: Is there an interative solver or something that could solve this approximately (it doesn't even have to be exact...this is electrical engineering)? Or, is there a way to enter a "region" that I expect the answer to appear in to the zeros and cZeros function? (I know the solve function and its friends takes guesses, but they crash and burn...probably because there are multiple valid solutions or something).

[Joe_Horn]

Try this:

fsolve( [1.05 = 1.8*y/(y+x),20000 = 1/((1/x)+1/y)], [x,y] )

This yields the solutions 34285.7142857 and 48000 in 0.0035 seconds on the Prime emulator, and 0.124 seconds on the physical Prime (may differ depending on which firmware version you're using).

[losfrijoles]

I tried that with curly braces (making a list), and that crashes my calculator. With brackets (making a matrix) it gives back undefined.

My firmware is 2013/08/13(5106) with CAS 1.1.0. Perhaps I have an old version?

[Joe_Horn]

Yes, the problem you're seeing is due to your obsolete version of the firmware.  Run the Connectivity Kit, and use its Help / "Check for update" facility to update the version of the Connectivity Kit itself.  After you have the most recent version of the Connectivity Kit running, tell it to download the most recent firmware, and upgrade your Prime (the current version is Version 2014 7 2, rev 6031).

EDIT: Uh oh, I just tried it on that firmware version (rev 6031) and it seems to return "undef".  It worked ok under rev 5447.  😞

Edit 2: Good news: rev 6031 works if you specify guesses close enough to the desired solutions, e.g. 10000 and 10000.  It doesn't work without any guesses (like rev 5447 did) but at least it does work if reasonable guesses are supplied:

fsolve( [1.05 = 1.8*y/(y+x), 20000 = 1/((1/x)+1/y)], [x,y], [10000, 10000] )

[losfrijoles]

Yeah, just updated to latest and got the undef as well.

However, solve() now gives the correct answer (and very rapidly) where it used to crash. So I guess the solution is, update the firmware and use solve() rather than fsolve(). 😕

EDIT: I tried fsolve with the guesses and it still gave me undef.

Kinda sad that they are having such issues with the firmware for something with such promising hardware. I just got it today after being unsure I would get my HP50g back before a test tomorrow (it was lost, but somebody turned it in to the local lost-and-found and somehow the wheels turned fast enough to get it back today).