******

This thread is not about which is the best calculator.
Its about programming the HP35s with all its limitations.

I don't doubt that an HP70, and HP50, an HP426-and-a-half or a Cray Bitz would be better but I don't have any of those and I'm not planning to buy one. I have an HP35s and I am interested in programs for THIS model. Please everyone, don't bang on about how much better it would be to use a different one!

******

STRUCTURAL FRAME ANALYSIS FOR AN HP35s

--------------------------------------

I would very much to write such a program.

But I can't.

(Explained below).

The HP35s seems to be limited to 26 direct and 100 indirect storage locations, which limits the maximum size of structure. I've just looked at this a little more closely and I estimate that the calculator could possible solve a structure with eg 4 lines of columns and 3 upper storeys, which is 16 nodes and 21 members.

This would be a seriously useful tool for structural engineers.

It would be possible only by three tricks.

(1)Use a simple member numbering notation

that eliminates node numbering.

(2)Assuming infnite axial stiffness

and analysing only shears and moments.

(3)Storing only the equations halfbandwidth.

This would eliminate lengthy input routines (making for a short program). It would also need a clever equation solver with a simple algorithm to retrieve stiffness matrix terms from packed halfbandwidth storage. The solver would also need to solve the equations in-situ, by which I mean that the original coefficients become gradually over-written by the solution, and to do this entirely in the half bandwidth of a banded symmetrical matrix. Such solvers do exist in eg Fortran or c but would need porting to HP35s machine code (not hard to do).

All the above seems very tempting... but just when can I ever find the time to do it? Answer: I can't!

Would anyone else be interested in having a go at it? Probably we need a structural or aeronautical engineer who enjoys numerical modelling to tackle it. I would be happy to help with suggestions - in particular with the member numbering convention referred to above, which facilitates direct assembly of stiffness equations.

----

John

John, I think its 26 direct and 800 indirect.

You're right. I've no idea where I got 100 from. Really must glance at the manual a little more often. This adds greatly to the need for a powerful program like frame analysis, which will really USE all that storage.

Even if, after allowing space for the code for this program and a few others, we only have 400 indirect storage registers remaining, this would still be ample for anything one would ever need to do on a calculator.

For example:

Plane frame

INPUT

No of members : 54

No of nodes : 30

DoF : 36

Halfbandwidth : 6

OUTPUT (for each loadcase)

30 node rotation angles

30 nodal sway displacements

108 member-end bending moments

108 member-end shear forces

5 support reactions

THe above problem would need 222 storage locations for the stiffness equations.

---

John

Good luck! I've written 2D and 3D linear and nonlinear analysis programs in Mathcad before and THAT was a chore. I can't imagine writing or using a frame analysis program on a HP50g, much less a HP35s. I saw one of these programs for the 49/50 on a website a while back, FWIW.

Do you do these types of calculations in your current employment? What do use to do the calculations, something running on a PC, perhaps? If you had the software to run on a 35S, would you actually use it for work, or is this just for the fun of it?

I understand "for the fun of it", but I'm curious to know which it is for you.

Quote:

---

If you had the software to run on a 35S, would you actually use it for work, or is this just for the fun of it? I understand "for the fun of it", but I'm curious to know which it is for you.

---

Fair question.

It is for both your suggested reasons, plus a third one.

All of these are significant contributors to my motivation.

In no particular order or priority:

Reason (1)

For the fun of it. Its a challenge, to try to do something complex and difficult with the very limited computational resources of a programmable calculator. Success is very rewarding, I suppose in the same way as some people (not me) find it rewarding to complete a large jigsaw or a Times crossword.

Reason (2)

Programming a complex theoretical procedure is a learning tool. Like when one is studying or revising, it is only when asked to explain the theory to someone else that I find out whether I really understand it myself. Programming it is the same kind of self-test. I finally learned in full how the stiffness method works only by writing a 7500-line Fortran program to solve structures using it. In a similar vein, I have recently RE-LEARNED how to do a rigorous parabolic rectangular analysis of reinforced concrete beams (which I only rarely need to do nowadays because there are more junior staff around to do it for me!).

Reason (3)

Yes, I really do have a use for such programs in my work. As I say above, it is not something I do frequently at work because I am 'too senior', and we employ many graduates in the firm to perform such basic analysis work on an everyday basis. Paradoxically, though, this means that I no longer have easy access to the necessary software, so when, sometimes, I actually do need a quick analysis of a simple frame or a quick design check of some reinforcement, the software isn't on my computer and its a pain in the arse to have to find someone else in the firm to do it for me because I do not have the time to do such canlculations by hand. I and other senior managers have a need, therefore, for some nifty analysis tools like this that we can use occasionally to solve such problems ourselves. Programs I have written for this purpose do get used in anger from time to time. How often? In my case, I'd say about once or twice a month as a long-term average. If that doesn't seem very much use, I agree, but the usefulness of having such tools available to me adds very greatly to the value of such occasional use.

Does that answer your question?

---

John

It is do-able. I did it on a TI-59 years ago. The program could solve structures with up to around 10 members and 9 nodes. Not a lot, but really useful from time to time. -- John

Yes, thank you.