Analog I/O


The Story of the Little Computer That Could!


Prism, the interface card that never was

One last I/O card was designed for the HP 9825A desktop computer but it didn’t become a product. Almost, but not quite. By 1977, the HP 9825A was quite successful and was being used as an instrument controller in many data-acquisition systems. Its four key I/O cards (parallel, serial, BCD, and HPIB) could connect the desktop computer to a wide range of peripherals and instruments. However, there seemed to be a need for a lower-cost way to read analog voltages. The most common ways were to use an HPIB or BCD card to connect to a voltmeter. However, the I/O cards cost about the same as a low-end HP voltmeter. It seemed that an I/O card that could read voltages directly could further reduce the cost of data acquisition using an HP 9825A.

HP 98037A ad03

This ad for the HP 98037A analog I/O card ran once in 1980. The photo shows the I/O card plugged into an HP 9825 desktop computer. A cable leads from an I/O card to the optional HP 98637A terminal box, which made the I/O card’s analog inputs available on banana jacks. The HP 98037A analog I/O card was introduced but never produced.

This thinking led to Project Prism. Steve Leibson, now promoted to project manager after developing three successful products in less than three years, initiated the project and designed the hardware. Michaela (Mickey) Natrachevski (later Mickey Dichter) joined the project later to write firmware.

Prism was an ambitious project for a relatively low-volume product that HP would sell for less than $1000. Requirements called for fitting four channels of HP-quality voltage measurement into the confines of an I/O card case. To achieve timing accuracy, the product would require an integrated processor to direct the voltage measurements independent of the HP 9825A.

The first version of the analog voltage-measurement circuitry that Leibson designed didn’t satisfy the Desktop Computer Division’s new lab manager. That manager was Jerry Nelson, inventor of HPIB, and newly transferred from the Loveland Instrument Division (LID) where they designed “real” voltmeters. Leibson’s initial attempt at a design was admittedly weak. It required too many trim adjustments (a problem in manufacturing) and it was too prone to drift. Leibson had attempted to build an HP-quality voltmeter in an I/O card but he had not used LID’s manufacturing prowess in laser-trimmed precision resistors, which is why the first HP 98037A design needed manual trimming. After a year of development, an unimpressed Nelson sent Leibson back to the drawing board to try again.

The analog voltage-measurement board was half of Leibson’s I/O card design for Prism. A second board carried the interface logic for the HP 9825A’s I/O bus and a microcontroller to manage the overall operation of the I/O card and provide precise timing on the measurements. Selecting the microcontroller had been difficult because I/O cards for the HP 9825A could only draw a limited amount of power from the host machine. In addition, not many microcontroller chips were available in 1977.

HP’s Nanoprocessor, which had been used in the HP 98034A HPIB card and the HP 98035A Real-Time Clock card, was not fast enough nor capable enough to do the job. In addition, the Nanoprocessor lacked on-chip RAM or ROM. There was no room for additional RAM or ROM chips on Prism’s digital board. The selected microcontroller would need on-chip RAM or ROM, so a microcontroller from a commercial vendor was the only alternative. The choice came down to either the Intel 8048 or the Mostek 3870, a single-chip version of Fairchild’s F8 8-bit microprocessor. Unfortunately, both chips had unattractive architectures that made them difficult to program. Leibson finally selected the Mostek part because it drew less power than the Intel part.

Although he flew to California and took a class in programming the Mostek 3870, Leibson didn’t write Prism’s microcontroller firmware. He’d taken on a second project called Tin Roof and his expanded management duties on that project precluded him from writing the firmware for Prism. So a new HP engineer named Michaela Natrachevski joined the project to write the code.

After a second year of development using a new analog-to-digital converter chip from Texas Instruments and a custom laser-trimmed resistor network from LID, the Prism team was ready to release the I/O card to production. The new design eliminated the need for trimming the analog circuits manually. At the release-to-production meeting, Leibson connected a power cord to the Prism card’s analog input and then plugged the other end into a wall socket. All of the managers in the room winced and held their breath when he did this. Normally, you don’t connect the input of a sensitive instrument to more than 100 volts from a wall socket. Of course, it was a stunt.

Leibson had calculated that the voltage would cause no harm and, even though he hadn’t tried the stunt beforehand (which might have been smart), the program running on the HP 9825A merely displayed “overload” and returned to normal voltage readings after Leibson reconnected a voltage that was within the card’s measurement range. Leibson had made his point. The Prism card was ready for manufacture and was approved for production.

Within a month of this meeting, Leibson resigned and left HP. He saw one advertisement for Prism, now named the HP 98037A A/D card. However, the product never appeared in HP’s catalog. Prism never actually became a product. Decades later, Leibson asked Nelson why. Nelson replied that LID management had complained that the HP 98037A A/D card stepped too far into their division’s charter. The project had been released to production, but failed to ship and no production units were ever built.

Although additional I/O cards were developed for the HP 9845A desktop computer, the HP 98037A was the last I/O card developed for the HP 9825A.

HP 98037 Board 2

HP 98037 Board 1

The HP 98037A analog I/O card never went into production. Only three fully functional production prototypes were made. The two boards in the above photographs are from one of those production prototypes. You can tell from the extra wires soldered to the integrated circuits that these are prototype boards. Incredibly, after being lost and unidentified for a quarter of a century, Steve Leibson found these boards in the substantial private collection of Dyke Shaffer, another ex-HP R&D engineer.

The upper board is the I/O card’s digital board. The purple chip on the right side of this board is a ROM version of the Mostek 3870 microcontroller with Michaela Natrashevski’s operating code inside. The lower board is the analog board. The white, rectangular chip in the center is a precision, laser-trimmed, custom resistor network developed for the HP 98037A by the thin-film resistor shop of the Loveland Instrument Division.

Unlike most of the other cards in the HP 980xx I/O card series, the HP 98037A did not have a cable. Instead, a 25-pin connector exited the back of the card directly, so that the system integrator could have maximum control over cable length and routing of the analog signals.

Information on the I/O pages of this site came from interviews with: Don Morris, Geoff Chance, Chris Christopher, Mike Kolesar, John Nairn, Dick Barney, Larry Smith, Ed Schlotzhauer, and from the memories of Steve Leibson.

HP9825.COM is in no way associated with the Hewlett-Packard Company.

Hewlett-Packard, HP, the HP logos, the HP 9825, and all other product model numbers starting
with HP are all trademarks of the Hewlett-Packard Company and don’t let nobody tell you different!

This site exists strictly for educational purposes. Nothing is for sale here. Substantial material from various issues
of the incomparable Hewlett-Packard Journal appears on this site, by permission of the Hewlett-Packard Company.

All text Copyright 2004 to 2010 - Steve Leibson

[Home] [Prehistoric Times] [The 9100 Project] [Prologues] [The 9825 Projects] [Epilogues]