▼
Posts: 2,448
Threads: 90
Joined: Jul 2005
Recall Paul Dale's HP 35S bug list:
[link:http://www.hpmuseum.org/cgi-sys/cgiwrap/hpmuseum/articles.cgi?read=735]
Item 14 concerns an anomalous equation editor result running inside a program. I just tested this on my recent 33S and found that it too has the same problem. This is not suprising considering that other bugs from the 33S also propagated to the 35S including the trig bug close to 90 degrees.
Here is item 14:
LBL A
156.25
STO X
208.333333334 ;There are eight 'threes' in there
STO R
1.77951304201
STO Q
-R*X/(X*Q-R) ;Should evaluate to roughly -467, and it does
-R*X/(X*Q-R) ;Should (still) evaluate to roughly -467, but
calculator outputs 31.323 instead!
RTN
This is REALLY freaking disappointing.
And BY THE WAY, it fails EVEN if you run it MANUALLY in the equation editor! Even without having two lines programmed identically. Go into equation editor, hit <ENTER> to evaluate the expression, and the first time through it pops the correct value onto the stack. Go back into the editor, press <ENTER> again and voilà--the wrong answer is in X!
FURTHER, in addition to what Paul tested, I found that variables Q and R have to be exactly as he shows, however X can be up or down one (1) in the last postion i.e. 156.24 and 156.26 also fail. 156.23 or lower, and 156.27 works fine. For those who really know something about BCD, or machine address space or etc, is this a clue? Remember when someone found that M$ Excel 2007 was a DUD and gave wildly wrong results with particular numbers? The of course sent out a patch...
I ran this on my 33S serial cna 02500940 and it failed as described. I'll test my old one later, too and edit this post.
Testing on both a 32sii and a 20s shows correct operation.
Edited: 17 Mar 2012, 11:12 p.m.
▼
Posts: 980
Threads: 239
Joined: Aug 2006
Hello there.
What does all this mean for the reliability of calculations on the 33s or 35s? On another note, in what professions do you need accuracy to the 9th or 10th decimal place? Does this mean that we need to carry two calculators just to verify the results of each? Is there any way to rectify these inaccuracies or should the 33/35s come with a "Use at your own risk" disclaimer?
Edited: 18 Mar 2012, 1:54 p.m.
▼
Posts: 24
Threads: 3
Joined: Mar 2012
All these devices should come with a "flashable" ROM, so serious bugs can be fixed after the unit has been released which is not the case with the 33S and 35S. Some level of inaccuracy or minor bugs may not be a problem for most people but some of the bugs on the 35S are really bad in my opinion.... that's why I'm switching to the WP34S!!!
Regards.
Posts: 776
Threads: 25
Joined: Jun 2007
Quote: On another note, in what professions do you need accuracy to the 9th or 10th decimal place?
Not many, but some! I used to be part of a NASA program which performed world-wide geodesy to measure, among other things, plate tectonic motions (typically a few centimeters per year). This needs sub-millimeter accuracy over earth-size distances (up to 10,000 km or so), or better than parts in 10^10. Our time standards (hydrogen masers) had short-term (a few minutes) stability of parts in 10^14. Our analysis programs used quadruple precision FORTRAN (and inverted million by million matrices - which were, fortunately (!), very, very sparse).
However, for normal day-to-day calculations, I don't worry very much about more than 3 or 4 digits, and rather like my 35S (and don't worry about its accuracy or precision).
▼
Posts: 980
Threads: 239
Joined: Aug 2006
Quote:
Not many, but some! I used to be part of a NASA program which performed world-wide geodesy to measure, among other things, plate tectonic motions (typically a few centimeters per year). This needs sub-millimeter accuracy over earth-size distances (up to 10,000 km or so), or better than parts in 10^10. Our time standards (hydrogen masers) had short-term (a few minutes) stability of parts in 10^14. Our analysis programs used quadruple precision FORTRAN (and inverted million by million matrices - which were, fortunately (!), very, very sparse).
However, for normal day-to-day calculations, I don't worry very much about more than 3 or 4 digits, and rather like my 35S (and don't worry about its accuracy or precision).
Thanks for the reassurance. I was curious as to the fields that need 10 digits or more of precision. I did not mean to trivialise the professions. I was just befuddled as to who would need such detail. Although I love being a calculator user, programmer and collector, I'll admit, all this discussion about bugs even as far back as the 35 (to quote Scotty, "no bloody A, B, C or D" [or S as in HP's case]), takes the joy and enthusiasm out of calculators and brings an anxiety of how reliable and accurate my calculators are.
Edited: 18 Mar 2012, 7:10 p.m.
▼
Posts: 2,448
Threads: 90
Joined: Jul 2005
"takes the joy and enthusiasm out of calculators and brings an anxiety of how reliable and accurate my calculators are."
And that is actually a good thing.
Posts: 2,448
Threads: 90
Joined: Jul 2005
Hi Dave,
But the idea that one could put an expression into the solver and get that competely bogus result--doesn't that really sour it for you? It does for me...
▼
Posts: 776
Threads: 25
Joined: Jun 2007
No - but that's because I never use the solver.
I guess my attitude is that one should have an everyday calculator for simple use, resting somewhere that it can be grabbed in a hurry. We all have our favorite(s) that we grab for simple calculations. Prior to the 35S, I grabbed my 41CX for such things. Given its relative scarcity/value, I am comfortable with grabbing the 35S instead for everyday calculations!
▼
Posts: 2,448
Threads: 90
Joined: Jul 2005
I tried that with the 33s for quite a while first, and later tried that with the 35s too. But then I would get discouraged by occasional unregistered keypresses etc and I'd realize that I had a range of cool old machines to rotate through that duty such as the 20s, or a battered 15c, a battered 27s 9I lost it on a plane!!!) or a 32sii and that I shouldn't have to wonder about the accuracy--because the old machines are *known* (this happens to be a very real hurdle for HP. This is one reason they would be fools to throw away say, the 48 series (50G) and replace the whole line with new firmware. Heck, I even had a 10b doing that for some time. I think the 20s is great. Algebraic but great nonetheless. To think I bought a slew of them for $15 way back when, from Samson! Sold a number on to friends!
I really do like the 33s in many ways though. And the 35s has so much potential, if only it wasn't so stupidly screwed up with some common simple stuff (base arithmetic, rect to polar, and the stupid bugs).
I think I am rambling late on a Sunday...
Edited: 18 Mar 2012, 11:43 p.m.
▼
Posts: 3,283
Threads: 104
Joined: Jul 2005
HP knows very well the advantages of proven algorithms and code. All the new devices (after the 33S and 35S) are based on a common math library which is a C version of the code used in the Saturn based calculators. As long as the current team is in charge you can count on the results the same way as on those of the "glorious" past because they insist on the application of this library. This is one reason why we will never see an HP sanctioned 34S with our firmware because it breaks this rule.
▼
Posts: 125
Threads: 9
Joined: Oct 2011
The 20b (and 30b) had a few bugs and the automatic calculation of each item as one moved though the probability menu options was slow and flaky.
Hopefully, HP will keep their calculators open enough to allow alternative firmwares like your's to co-exist. Further, considerable portions of the CAS in the HP-49/50 series did arise from user contributed code.
Nick
Edited: 19 Mar 2012, 4:41 a.m.
▼
Posts: 2,448
Threads: 90
Joined: Jul 2005
"Further, considerable portions of the CAS in the HP-49/50 series did arise from user contributed code."
Metakernel was user (J-YA)
The CAS was also, yes? ALG48 and or ERABLE?
Didn't HP hire J-YA for the 49 project,because of his MetaKernel? When I boot my 49G, it flashes a "Metakernel" startup screen.
Has anyone put together a Bug List or Quirk List for the 30b?
Edited: 19 Mar 2012, 10:58 a.m.
▼
Posts: 3,283
Threads: 104
Joined: Jul 2005
Bill, ERABLE is from Bernard Parisse, Professor at the University of Grénoble, France. He may get payed by HP for his work but he's definitely not an HP employee.
▼
Posts: 2,448
Threads: 90
Joined: Jul 2005
Yes, I wasn't clear in my post--Erable I know Parisse not J-Y.A.
I was just pointing out examples of user-generated content leading to an HP product. What I don't know is just exaclty how the 50G CAS relates to Msr. Parisse's work (and that of ALG48 as well).
Posts: 2,448
Threads: 90
Joined: Jul 2005
Hi Marcus,
That is interesting. I was wondering about the current situation--a situation that seems very promising.
Posts: 980
Threads: 239
Joined: Aug 2006
Quote:
HP knows very well the advantages of proven algorithms and code. All the new devices (after the 33S and 35S) are based on a common math library which is a C version of the code used in the Saturn based calculators. As long as the current team is in charge you can count on the results the same way as on those of the "glorious" past because they insist on the application of this library. This is one reason why we will never see an HP sanctioned 34S with our firmware because it breaks this rule.
This concerns me as yes, I am very fond of my 33s and especially my 35s. But, just how common, severe and numerous are the bugs that I realistically come across?
▼
Posts: 2,448
Threads: 90
Joined: Jul 2005
One of the great features of those machines is the solver. But it can't be trusted. See bug #14.
▼
Posts: 980
Threads: 239
Joined: Aug 2006
Please direct me to the bug list. Thank you.
▼
Posts: 3,229
Threads: 42
Joined: Jul 2006
Posts: 1,089
Threads: 32
Joined: Dec 2005
Quote: But, just how common, severe and numerous are the bugs that I realistically come across?
If you use and need all the new features of the 35s, I'd say you have a good chance to even discover new bugs ;-(.
▼
Posts: 3,229
Threads: 42
Joined: Jul 2006
I'd have to agree. I'm fairly confident that the bug list is tiny compared the bugs that actually exist in this calculator.
- Pauli
|