Re: Replacement for HP41CX Message #6 Posted by Les Wright on 9 Feb 2007, 1:53 a.m., in response to message #5 by Jeff O.
I have revisited the 33S in recent weeks, mostly out of interest in porting some HP41 programs in hopes they will run a bit faster. I am not so troubled anymore about the Cosine Bug (i.e., some glitch in the algorithm gives less than full 12 digit accuracy for cos and tan for angles close to 90 degrees), but must admit that even though it is not a full blown abomination it is a wee bit of an embarrassment for the calculator.
What continues to be a source of lament for me is the incongruity of having a whole bunch of memory for programs and equations (almost 5x that of the 42s and 15x that of the 41CV and CX), yet limitations on programmability that make it hard to take full advantage of that. There are 26 lettered registers and 6 statistics registers that can be used for storage only by indirection, plus the index register. This is a hard cap--there is no movable partition between program memory and storage memory like there is in the 41, 11C, 15C, and 34C (to name a few). Most reasonable programs don't require more than a few registers, but anything that involves matrix computation with matrices of any size is essentially ruled out. (That said, someone here did share recently that the 33S got him through some tough matrix computations on an engineering examination.)
I particularly lament the serious restriction on available program labels. There are only 26 lettered labels available. There are no multicharacter string labels or numeric labels, and there is no distinction between local and global labels. This means that if I have a program labelled A that uses within it labels B thru E, those five labels cannot be used elsewhere throughout the calculator for another program--i.e., if my next program is F, I can't re-use the label D again as a branch point or I will get an error. The difference between global alpha, local alpha, and local numeric labels in HP41 programs is great organizing feature and I miss its replication in the 33S.
I also really miss the lack of stack based register arithmetic. For example, in several of J-M Baillard's excellent advanced math routines for the HP41, he often uses something like "ST+ X" as a quick way to double the contents of the X register or "ISG Y" to add one to the Y register, and much of his computations in iterative routines like series summations and continued fractions involve clever manipulations of the stack and minimal use of storage registers. Indeed, trying to duplicate some of this ingenuity is proving to be my biggest barrier as I try to transfer some of this amazing work to the 33S.
On balance, the 33s does offer RCL arithmetic, like the 42S but unlike the 41 series. Let's say on the HP41 I have something in R15 that I want to leave unmolested but I want to multiply into the what's in the X register. On the HP41 this is a two step task--RCL 15, *. On the 42S this could be done in one step: RCL* 15. Likewise with the 33S, only in that case we are dealing with a lettered register.
For 50 bucks, the 33S is a really good calculator, but it could be a great one. I think if I spend a little more time with it trying to work around its limitations (at least as I see them) I will overlook some the flaws and be impressed by its potential. For example, I already know that it is fast--the built-in integrator with return a six or seven digit result for the famous Kahan integral in well under a minute, whereas in the original 1980 the wonderful 34C took several minutes, even after appropriate transformations were made.
If there was an improvement to the 33S, I would like to see them loose the chevron layout, add numeric and multicharacter labels, and implement some way to partition memory so as to make more storage registers available, even if one could only access the additional registers by indirection (indeed, this is the case on the HP41 and 42S--registers above 99 can only be accessed indirectly).
Finally, I should mention the display. The decimal point on the corrected display is still kind of small, but for the most part I think it is a very clear two line display--clearly superior to that of the much praised 42S.
Happy shopping!
Les
|