HP-67 DSP vs FIX

[Matt Agajanian]

Hi all. An interesting curiosity of mine.

Since the 67 was built on Woodstock technology, that would mean that if a 25/25C had FIX n already, why would the 67 need the two-step FIX DSP n?

[Valentin Albillo]

(12-01-2023 12:03 AM)Matt Agajanian Wrote:  Hi all. An interesting curiosity of mine.

Since the 67 was built on Woodstock technology, that would mean that if a 25/25C had FIX n already, why would the 67 need the two-step FIX DSP n?

Lack of enough free keycodes. FIX 0-9, SCI 0-9 and ENG 0-9 require 30 keycodes in all, not that many were available.

V.

[Matt Agajanian]

Hi.

I’m thinking it was a possibility since Woodstock had 30 keys with f & g prefix operations, that made for less keycodes as compared to 35 keys with f, g, AND h prefixed operations.

(12-01-2023 12:57 AM)Valentin Albillo Wrote:  

(12-01-2023 12:03 AM)Matt Agajanian Wrote:  Hi all. An interesting curiosity of mine.

Since the 67 was built on Woodstock technology, that would mean that if a 25/25C had FIX n already, why would the 67 need the two-step FIX DSP n?

Lack of enough free keycodes. FIX 0-9, SCI 0-9 and ENG 0-9 require 30 keycodes in all, not that many were available.

V.

[Namir]

The HP-65 uses DSP and so did the HP-21! So I guess the HP-67 uses the DISP in the same lineage of the HP-65.

[John Garza (3665)]

It's not the keys, it's the keyCODES, which are logical operations and refer to specific areas of the microcode to execute. The machines had a limited number of bits to specify these unique codes, meaning there are a limited number available.

Using 13 instead of 30 for the display ops makes more available for the designers to use in other areas.
And keep in mind these were at a premium in the early hardware. So they kept the DSP as used in the 65.
Once more space was available on later machines (29C etc.) HP moved away from DSP.

-J

[teenix]

(12-01-2023 03:03 AM)John Garza (3665) Wrote:  It's not the keys, it's the keyCODES, which are logical operations and refer to specific areas of the microcode to execute. The machines had a limited number of bits to specify these unique codes, meaning there are a limited number available.

Using 13 instead of 30 for the display ops makes more available for the designers to use in other areas.
And keep in mind these were at a premium in the early hardware. So they kept the DSP as used in the 65.
Once more space was available on later machines (29C etc.) HP moved away from DSP.

-J

This seems reasonable. The 67 has more functions to deal with than the 29C and therefore needs more of the 256 available program key codes - which both machines have.

The 67 only has 6 spare codes and could not support the larger array of display functions.

The 29C with less functionality, had more spare codes available and could make use of these with more display functions like[f][FIX][9] etc. and even with this setup, it still has another 24 spare.

cheers

Tony

[John Garza (3665)]

(12-01-2023 04:37 AM)teenix Wrote:  The 67 only has 6 spare codes and could not support the larger array of display functions.

Tony, didn't the 67/97 use some of those unused codes for the famous Word/Phrase/Graphics ?

[teenix]

(12-01-2023 03:02 PM)John Garza (3665) Wrote:  Tony, didn't the 67/97 use some of those unused codes for the famous Word/Phrase/Graphics ?

I can't help there I'm sorry to say, I don't know anything about it, any info available?

cheers

Tony

[brouhaha]

(12-01-2023 03:02 PM)John Garza (3665) Wrote:  [didn't the 67/97 use some of those unused codes for the famous Word/Phrase/Graphics ?

The display phrases were the use of non-normalzed numbers (NNNs) to get the letters C, E, d, o, and r in the displays by having hexadecimal digits A-E (and F for a blank) in a number. That by itself didn't directly use or require the extra 6 instruction codes. However, one way to get the NNNs was to glitch the calculator, using a phase 1 clock interrupt switch, or powering it through a rheostat, such that program lines could be read as data registers. Then the extra 6 codes could sometimes be useful.

On an HP-67, but not the HP-97, there were key sequences using multiple simultaneous key presses,which could insert the extra 6 scores into program steps.

WARNING: If you mess around with NNNs on an HP-97, have the printer mode set to MAN, and do not execute any printer functions. Printing NNNs can cause the PIK chip to hang with printhead drive lines active, which burs up the printhead very quickly.

[teenix]

(12-01-2023 11:48 PM)brouhaha Wrote:  The display phrases were the use of non-normalzed numbers (NNNs). . . .

Now I remember, thanks

[John Garza (3665)]

Yep, same here... now I remember. I recall using those special key-presses to generate keycodes to build messages way back then. I had the phase 1 interrupt switch on my 67 installed as a little magnetic reed switch on the inside but near the case surface - so a small magnet passed over it would activate it. I didn't want to drill holes in my 67, and it was a much neater solution that using a separate piece of hardware (the Black Box). Needless to say, I DID NOT carry the magnet around in the case with the 67 and the mag cards. That was... I think... 44 or so years ago.

-J

[Matt Agajanian]

This explanation that DSP n FIX (SCI or ENG) results in an economy of keycodes--13 versus 30 makes me wonder why HP didn’t retain this approach with calcs following the 67.

[Valentin Albillo]

(12-02-2023 01:18 AM)Matt Agajanian Wrote:  This explanation that DSP n FIX (SCI or ENG) results in an economy of keycodes--13 versus 30 makes me wonder why HP didn’t retain this approach with calcs following the 67.

Because many subsequent HP calculators either didn't have more than 256 *programmable* instructions or have multi-byte instructions, such as the HP-15C, which has 1- and 2-byte instructions and so it can have e.g. STO- and RCL-arithmetic programmable instructions for every register R0-R9 and R.0-R.9, every matrix A-E (individual elements,) etc., which totals dozens upon dozens of keycodes and matter of fact there are some 700 possible valid different keycodes which can be stored in program memory vs. 256 max for the HP-67.

The HP-67 also needed the P<>S instruction to further reduce the keycode count (it reduces the number of keycodes related to store, store arithmetic and recall essentially in half) plus there's no store arithmetic for registers A-E and I, saving another 24 keycodes.

Of course, the HP-41 has multi-byte instructions (up to 15 bytes for a single line in program memory) and doesn't even need keycodes at all because of its alpha capabilities.

In short, if the calc has less than 256 programmable instruction or 2-byte instructions (or alpha capabilities,) there's no need for separate FIX/SCI/ENG and DSP instructions, let alone P<>S.

V.

[aurelio]

thanks Matt for asking the question, thanks to all those who answered, I was curious about this topic too.

[Matt Agajanian]

(12-09-2023 01:43 PM)aurelio Wrote:  thanks Matt for asking the question, thanks to all those who answered, I was curious about this topic too.

You’re welcome! Glad to help.