HP-21S Impressions


Hello, After much looking I finally got an HP-21S, made in the US in 1989. A nice machine.

After some reasearch I found it is very similar to the HP-20S, at least in operation and programming. I also found out that user's manuals are hard to come by, apparently because this model was handed out to classrooms in lots with one manual for each. Off-the-shelf models with box and manual were also available but are much more rare.

No manual came with my 21S. I have figured out most of the functions but I have found myself wondering about these stories of many calculators being distributed to classrooms with just one manual. Are they true?




at some time there was a site which had a pdf version online for a while.

Just send me a reminder via mail, and I'll check if I find it...



Hello Raymond,

Your register does not show an e-mail address. Could you please contact me?

Thanks in advance.



Hello Juan,

my email address:






just saw that Neil had the link handy:-)

Grab it while you can.



Juan --

So, if those stories you've heard are correct, that would explain why the one I bought more than a year ago was without manual, and that I've never seen a manual or calculator with manual offered on eBay.

The HP-21S is kind of a "niche" model that -- along with the HP-20S and HP-10B -- is a low-end Pioneer-series characterized by the 7-segment LCD readout. It does have some functions that are hard to find on other RPN or algebraic models.

A really capable Pioneer-series calculator for statistics would have utilized the high-end Pioneer hardware and firmware (i.e., HP-17B, HP-17BII, HP-27S):

  • Equation editor with solver
  • 7 kB RAM
  • Statistics functionality that stores each datum

This would allow sorting, minimum, maximum, median, and mode. It would also prevent errors that occur from calculation of standard deviation from summation quantities, rather than from mean and deviations.

For example, calculate the sample standard deviation of


The correct answer is 1. However, on any RPN-based model or the HP-20S or HP-21S (which perform summations), an incorrect answer of 0 will be returned, due to roundoff error in the subtraction of summation quantities.

The HP-17B, HP-17BII, HP-27S and RPL-based models store the input data, calculate the mean and then the deviations. They give the correct answer.

The HP-21S, like the HP-20S, has those built-in keystroke programs (using LOAD) that are not intuitive to use, then wipe out program memory when they are loaded. Not the best of ideas.

The CD/DVD's available for purchase here at MoHPC include a PDF scan of the HP-21S manual.

-- KS

Edited: 8 July 2006, 4:00 p.m.


Just want to add one point:

The 21S is the *only* HP featuring 4 standard distributions (Gaussian, Student's, Chisquare and F) built in.

So it's unique in the line. I think it would be rated higher if it would have been a RPN model.


A question on "Statistics functionality that stores each datum":

I don't have a HP-21 to play with but Viktor Toth's site tells me that the HP-21S has only ten memories. Wouldn't that severely limit the capability of a statistics function that stores each datum?


Correct Palmer.

That's why he said "a really capable pioneer machine WOULD have..."

He is saying that for a statistics machine like the 21S, it would have been really nice if they had made it list-based for statistics, like the HP17b, etc.

The 21S does not do that and falls short in that regard.

Hope all is well.


Designers and manufacturers make choices when they make a product. My impression is that the HP-21 was intended to penetrate a low cost market. If so I suspect that cost drove the design and yielded a memory limitation.

You can go back and say "Wouldn't it have been nice if HP had made a really capable HP-45 or HP-80 with list-based statistics but those machines didn't have enough memory. Did the HP-65 or HP-67 come with list based statistics? I doubt it because again the memory limitations came into play. The same thing was also true with TI's SR-52. When the TI-59 came along the baseline machine with the Master Library module installed didn't provide list-based statistics but shortly after I had one I wrote such a program which mirrored the capability of the time-share computer I was using even though the TI-59 could handle only90 entries. The Statistics module of the TI-59 did provide list-based statistics. I didn't find a list-based statistics capability in the baseline HP-41 so I wrote one. I presume that one of their modules provided it.


Designers and manufacturers make choices when they make a product. My impression is that the HP-21 was intended to penetrate a low cost market. If so I suspect that cost drove the design and yielded a memory limitation.

(analogies to the HP-45 and HP-67 follow...)

Not an unreasonable statement, but I'd say that it misses my point.

There were three levels of Pioneer-series calculators released in 1988-89:

  1. High end: HP-17B, HP-27S, HP-42S

    These had 7 kB of user memory; high-resolution two-line dot-matrix display that supported softkey menus with legends; the 1000 kHz Saturn processor; and, IR printer output.

  2. Mid-range: HP-14B, HP-22S, HP-32S

    These had a coarser 7x5 pixel 12-field dot-matrix display supporting less-elaborate menuing capabilities; less memory; 640-kHz processor; and, no IR printer output.

  3. Low-end: HP-10B, HP-20S, HP-21S

    These offered a 12-digit, 7-segment display; no menuing capabilties, and even less memory.

It is clear that the three platforms had already been developed for other models. I'm saying that a really good "statistics specialist" model would have been based on the high-end Pioneer platform, instead of the low-end one. Maybe such a product would not have served a significant "features and price" market niche, but, we'll never know.

I didn't find a list-based statistics capability in the baseline HP-41 so I wrote one. I presume that one of their modules provided it.

I don't believe so. The HP-41 STAT module included 4 kB of RPN keystroke routines, providing the missing regression-analysis (curve fitting) capabilities for the HP-41's built-in two-variable summation, but no list-based statistical programs that could consume considerable user RAM if utilized.

A thoughtful list-based statistical program for the HP-41 might create a data file in Extended Memory, then store input data directly to that file. Nearly 600 single-variable values could be stored in that manner.

-- KS

Edited: 10 July 2006, 4:15 a.m.


Hi Juan,

Here is the link to the Hp-21s manual:



Thank you all for the prompt replies and useful links. I dowloades the manual and now to continue learning about this rare machines. I found it is appreciated by some medical schools and psychiatrists because of its built-in applications.

Limitations notwithstanding, the only missing thing is RPN. It would be a greater machine if it had it but, well,it is algebraic.

Once again, tanks to you all.

Kind regards,


Possibly Related Threads…
Thread Author Replies Views Last Post
  WP34S, first impressions Andres Capdevila 27 6,379 03-30-2012, 12:02 PM
Last Post: Les Wright
  HP 10bII+ vs. 10bII first impressions Kevin Colville (S. Africa) 6 2,129 10-31-2011, 04:38 AM
Last Post: Kevin Colville (S. Africa)
  HP 10bII+ first impressions Jose Gonzalez Divasson 18 4,363 04-20-2011, 03:01 PM
Last Post: Jose Gonzalez Divasson
  HP 12c+ and HP 50g - initial impressions Jose Gonzalez Divasson 44 9,750 11-19-2010, 12:45 PM
Last Post: Gerson W. Barbosa
  First impressions HP30B bt_schmidt 1 989 08-03-2010, 09:16 PM
Last Post: allen
  First impressions of new TI-NSpire CAS Don Shepherd 13 3,295 04-04-2010, 07:31 AM
Last Post: ClausB
  A pair of 21S with sticky keys Peter Klein 8 2,122 04-01-2009, 03:40 PM
Last Post: Randy
  HP 20B - 1st impressions bt_schmidt 30 6,524 09-16-2008, 04:26 PM
Last Post: Marcus von Cube, Germany
  Yet another HP35s first impressions thread Brad Davis 15 3,475 09-10-2007, 02:08 PM
Last Post: Jeff O.
  More 35s First Impressions Dan Greil 0 675 08-17-2007, 02:54 PM
Last Post: Dan Greil

Forum Jump: