Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!thunder.mcrcim.mcgill.edu!snorkelwacker.mit.edu!hsdndev!wuarchive!zaphod.mps.ohio-state.edu!cis.ohio-state.edu!ucbvax!agate!bionet!ames!xn.ll.mit.edu!ll.mit.edu!news
From: wjc@llex.ll.mit.edu ( Bill Chiarchiaro)
Newsgroups: comp.sys.handhelds
Subject: The First (?) HP Calculator (long)
Keywords: Not the HP-35
Message-ID: <1991Jun6.153021.10025@ll.mit.edu>
Date: 6 Jun 91 15:30:21 GMT
Sender: news@ll.mit.edu
Organization: MIT Lincoln Laboratory
Lines: 69


Seeing the recent discussions of HP calculator history, I've decided
to add a couple of postings which might be of interest.

So far as I know, the first HP calculator was the Model 9100A
introduced in 1968 at a price of $4900 (this is from the HP Journal of
May 1974 -- the one that described the HP-65).  It was a desktop
machine, considerably larger than, say, an HP-97.

The 9100A was quite an interesting machine, and it clearly showed the
design philosophy evident in the later HP calculators.  It was
programmable and had 16 registers.  Fourteen of those could hold
either a real number or 14 program steps.  The remaining two registers
could not be used for program storage.  Thus, the 9100A could hold a
maximum of 196 program steps.  There was a built-in magnetic card
reader/writer which used credit-card sized mag-cards.  The cards could
be fed in either of two orientations and thus could contain two 196-step
programs.  Provisions were made for convenient overlaying or chaining
of programs.

Until the HP-28/HP-19 series, this was the last machine to show more
than one level of the stack at a time.  The 9100A had three stack
registers called X, Y, and Z.  All three were simultaneously shown on
the CRT (!) display.  Of course, the 9100A used RPN.  It could display
in either fixed point or floating point.  The range of floating-point
exponents was -99 to +99.  The available precision was 10 digits,
internally it was 10.  The stack operated a little differently than we
are used to.  Entries from the keyboard went into the X register, but
results of dyadic operators were returned to Y.  This made repeated
operations easier (X was left unchanged).

The built-in functions included: the four basic functions, square root,
log, ln, and e^x; sin, cos, tan, and their inverses; sinh, cosh, tanh,
and their inverses; polar-to-rectangular conversion,
rectangular-to-polar, and vector addition and subtraction; absolute
value and extract integer part.  Trig functions could operate in
either degrees or radians.  A PI key was also included.

The implementation and realization of the 9100A architecture was quite
fascinating.  The user memory was 368 words by 6 bits of magnetic
core!  There were essentially two levels of microcoding with a 64 word
by 29 bit "Control" ROM and a 512 word by 64 bit "Program" ROM.  The
Control ROM was a wire braid toroidal core memory.  The Program ROM
was a 16-layer PC board which used inductive coupling from its drive
lines to the sense lines.  HP was proud that they had achieved a
density of 1000 bits per square inch.

The logic was discrete transistors, diodes, and resistors.  The memory
addresses and some other state information were kept in 40 J-K
flip-flops.  The internal operations were performed by hard-wired
logic gates.  The total number of semiconductors was, I believe, only
a few hundred.  The clock period was 825 ns.

The 9100A had an output port for driving printers or other
peripherals.

I had the opportunity to use one of the machines a few years ago and
it was actually quite nice.  The speed wasn't bad.  A trig function
would take about 280 msec and an addition or subtraction would take 2
msec.  A sample factorial program supplied by HP took less than half a
second to compute 69!; a similar program on my HP48SX to just over
half a second.

Now for a question:  Does anyone know if HP made any other calculators
between the 9100A and the HP-35?


Bill Chiarchiaro
wjc@ll.mit.edu