The Museum of HP Calculators

HP Forum Archive 13

 !7bii+ vs other HP calcsMessage #1 Posted by Brent on 23 Nov 2003, 8:31 p.m. On my 17bii+ I put PI on the stack and multiply by 100 and then divide by 100. I put PI on the stack again and subtract. I get 2.1E-13. None of my other HP calcs do this--I get zero. Why?

 Re: !7bii+ vs other HP calcsMessage #2 Posted by Fred on 23 Nov 2003, 10:09 p.m.,in response to message #1 by Brent Please explain to me how this is a useful calculation.... but it has to due with the internal significant digits the calculator maintains and round off errors during calculations in the least significant digits....

 Re: 17bii+ vs other HP calcsMessage #3 Posted by David Brunell on 23 Nov 2003, 10:58 p.m.,in response to message #1 by Brent Sheesh, even my 1973-vintage TI SR-20 desktop scientific gives an answer of zero for pi*100/100-pi. Maybe you should be happy to have a pi key on a financial calculator at all. :) Doing a quick survey here, none of the other HP financials that I have contain a pi key.

 Re: 17bii+ vs other HP calcsMessage #4 Posted by Brent on 24 Nov 2003, 12:05 a.m.,in response to message #3 by David Brunell OK, It does it even if I multiply PI by 1 and then put PI in the stack and subtract. It shouldn't do it. It should be just zero. I don't know why or how I stumbled on it, but it just seems odd to me. The other calculators (11c, 48GX) don't do it. The 17bii+ gives a non zero number (2.1E-13), and I don't care whether it's a financial calculator or not--it shouldn't do it.

 Re: 17bii+ vs other HP calcsMessage #5 Posted by David Brunell on 24 Nov 2003, 1:28 a.m.,in response to message #4 by Brent I have no idea what the internal architecture of the 17bii+ is, but I'd be willing to bet that the error is due to the calculator doing some unnecessary BCD-to-binary conversions. On hp's web site, the calculator is listed as having "memory capacity unlimited within available memory." What the hell is that supposed to mean? Maybe they should also state that accuracy is "unlimited within available significant digits." ;)

 Re: 17BII gives 0Message #6 Posted by Graham Wilson on 24 Nov 2003, 6:50 a.m.,in response to message #4 by Brent My 17BII gives 0 like the other "old" HP calcs. Does the bii+ use the BII ROM in hardware emulation?

 Re: 17bii+ vs other HP calcsMessage #7 Posted by Valentin Albillo on 24 Nov 2003, 9:22 a.m.,in response to message #4 by Brent Hi, I guess the trouble is caused by improper binary to decimal conversions and back in the emulator. Older HP calcs did use BCD, which allows for exact representation of base-10, decimal quantities such as 0.1, while 0.1 is a periodic, never-ending fraction in binary (base-2). As a simple test, just try to calculate Pi*1-Pi in your trusty Windows XP standard calculator applet (Scientific mode). You should get a 0, but actually you get instead: ``` [Pi] [*] [1] [-] [Pi] [=] -> 3,1370505526082969463696315942562e-38 ``` Close, but no cigar ! Pi + 0 - Pi gives the same result. Best regards from V. Edited: 24 Nov 2003, 9:26 a.m.

 Re: 17bii+ vs other HP calcsMessage #8 Posted by Frank Wales on 25 Nov 2003, 8:30 p.m.,in response to message #7 by Valentin Albillo Well, I infer from this that either there is a stupid bug in the low-level routines that emulate the BCD-type arithmetic operations done in hardware on the older machines, or HP have finally decided that binary floating point is good enough for financial calculators (and maybe even for all of them -- has this been tried on the 49G+?). Personally, I'm hoping it's a bug, and not a policy change, since the latter would indicate that HP had misplaced their understanding of the need for consistency of operation. Which would be sad.

 Other testsMessage #9 Posted by Tizedes Csaba on 24 Nov 2003, 9:32 a.m.,in response to message #1 by Brent Hello, my favorite test is the next: 3 [ENTER] [y^x] 27 [-] And the test for PI: 355 [ENTER] 113 [/] [PI] [-] Csaba

 Re: Other testsMessage #10 Posted by Joe on 24 Nov 2003, 9:58 a.m.,in response to message #9 by Tizedes Csaba I'm not biting.

 Re: Other testsMessage #11 Posted by Namir Shammas on 24 Nov 2003, 10:17 a.m.,in response to message #9 by Tizedes Csaba The operations 355 / 113 gives an good approximation for Pi that typically has the following error: (355 / 113 - pi) / pi * 1E+09 = 84.xxx Namir

 Re: Other testsMessage #12 Posted by GE (France) on 24 Nov 2003, 5:11 p.m.,in response to message #11 by Namir Shammas Another test would be : 355 [Enter] 113 / [Enter] SIN + SIN Which is quite close to zero. 48GX gives -2.06E-13, a nice PC1262 could probably do better as the result obtained from Maple (for instance) is about -3.16E-21.

 Re: !7bii+ vs other HP calcsMessage #13 Posted by Scuba Diver on 24 Nov 2003, 1:33 p.m.,in response to message #1 by Brent Hi Brent! Just wondering; what are your thoughts of the 17BII+? Is the keyboard the same as the 10BII, or is it better quality than that? Now that the 17BII+ is out, I find the ebay prices for 17BIIs has dropped; would you suggest that I get a backup 17BII while I can, or should I pick up a +? B.

 Re: !7bii+ vs other HP calcsMessage #14 Posted by Brent on 24 Nov 2003, 1:53 p.m.,in response to message #13 by Scuba Diver I only have a 12c so I can't tell you anything about the other business calcs. It seems like a pretty good calculator but it has kind of a hollow or tinny feel to it. The eqation SOLVE part is nice. I wish they would make the 33s equation solver the same way. You can use names for variables and input values via softkeys. The 33s will only allow you to use A thru Z for variables (I think). I would go ahead and get the 17bii+ if I were you. It works fine, but it just doesn't have that feeling that you are holding a valuable piece of equipment.

Go back to the main exhibit hall