What comes after Thyme?

After completing the Spice Rack I/O expander project, Steve Leibson took on the design of yet another I/O card for the HP 9825A. Following the lyric sequence of the Simon and Garfunkel song “Scarborough Fair,” the next I/O card should have been Thyme, the real-time clock card. However, that was not the most needed I/O function at the time. An I/O card was needed to implement the serial I/O protocols associated with the RS-232-C standard. (An engineer visiting Loveland from HP’s Boeblingham Division in Germany eventually developed the HP 98035A Real-Time Clock card in 1977.)

Serial protocols were first developed to connect teletypewriters to modems. these protocols allowed digital communications over telephone lines but many of HP’s instruments and many computer peripherals such as paper-tape readers and punches used the RS-232-C I/O protocol as well. In RS-232-C terminology, teletypewriters were dubbed DTEs (data terminal equipment) and modems were DCEs (data communications equipment). The serial I/O card for the HP 9825A would need to be capable of acting as a DCE when connecting to DTEs and as a DTE when connecting to DCEs. All that was really required was two different cables and connectors to accomplish this.

This cartoon for the HP 98036A serial I/O interface card appeared with the article on serial interfacing techniques that was part of the article series called “Leibson on I/O.” It depicts two HP 9845 desktop computers talking over a tin-can telephone, a reference to one of serial I/O’s main uses back then, telecommunications for computers via analog modems. The cartoon appeared in HP’s Keyboard magazine.

Again, the most important part of the project was to develop the project name. Simon and Garfunkel’s 4-name sequence was exhausted so the possibilities were endless. Leibson finally suggested “Cap’n Crunch” for two reasons. First, it’s a really bad pun: “Cap’n Crunch” is a serial (cereal) interface. Second, a notorious phone hacker named John Draper called himself Captain Crunch (because he used a toy whistle included in boxes of Quaker Oats’ Cap’n Crunch cereal to spoof AT&T’s long-distance system and get free long-distance telephone service) and the serial I/O card would sometimes be used to communicate over phone lines. Geoff Chance rolled his eyes, approved the project name, and went off to fight other battles.

As with all of the other I/O projects for the HP 9825A, the next most important task was to decide on the features for Cap’n Crunch, which would officially be launched as the HP 98036A serial I/O card. The HP 9830A had several available serial I/O cards, none of them ideal. The HP 11205A serial I/O interface card was designed mostly as an output card to drive teletypewriters, printers, and paper-tape punches. It operated at bit-transmission speeds of 1200 bits/sec or less and could be used as an input card, but only if the HP 9830A could control the attached device to allow only one character at a time to be sent. Otherwise, the interface would drop bytes. The HP 11206A modem interface card was restricted to 440 bits/second and was suitable only for low-speed communications. The 11285A Data Communications Interface was a far more capable serial I/O card. It could operate asynchronously or synchronously at speeds to 9600 bits/sec but it required three new option ROMs and two separate interface cards joined by a cable and therefore consumed two of the desktop calculator’s I/O slots.

Clearly, there was something different about serial I/O that made it more complex than parallel or BCD I/O. That difference lay in the nature of full- and half-duplex communications. Full-duplex communications allows information to move simultaneously in both directions, both into and out of the computer. Like most computer buses developed in the 1970s and 1980s, the I/O backplane buses on HP’s desktop calculator/computers were inherently half-duplex devices. They can input data or output data, but cannot do both simultaneously. Operating full-duplex devices on half-duplex buses requires buffering and some advanced I/O software. That’s one reason why the HP 11285A required two I/O cards and three option ROMs.

You can see the difference in complexity for serial I/O versus parallel and BCD I/O reflected in the design of the two boards in each of the three I/O cards. The HP 98032A parallel I/O card and the HP 98033A BCD I/O card required a handful of smaller, simpler integrated circuits. The boards in these two cards appear roomy when compared to the layout of the HP 98036A serial I/O card.

The design of the HP 98036A required the use of an LSI (large-scale integration) chip from Intel called an 8251A. This chip contains the relatively large amount of circuitry needed to perform the parallel-to-serial and serial-to-parallel conversions and to provide 1-character buffers in each direction that aid full-duplex operation. However, Intel’s 8251A was designed to be connected to Intel’s 8085 microprocessor not HP’s hybrid microprocessor, so many small chips were needed to adapt the Intel 8251A to the HP 9825A I/O bus. The result: a lot of chips crammed into a little space.

The Intel 8251A is a USART (universal synchronous/asynchronous receiver/transmitter). It could operate either synchronously or asynchronously. However, synchronous serial I/O also requires complex software protocols not supported by the HP 9825A’s General I/O or Extended I/O ROMs. In addition, the 1-character buffers in the Intel 8251A were sufficient for low-speed, full-duplex communications but would still require additional language-ROM support for the highest transmission speeds.

Development of the HP 98036A serial I/O card required about a year. At the end of the project, Leibson’s next assignment was clear: develop an I/O ROM to allow the HP 9825A desktop computer to get more capability and performance from the HP 98036A serial I/O card. In the 21st century, it’s unusual for large companies like HP to switch designers from hardware to software and back again. Back in the 1970s, in simpler times, HP engineers were expected to be almost universal. They were often asked to handle wildly divergent jobs from project to project.

Click here for the next part of the HP 9825 I/O story.

Steve Leibson designed these two boards for the HP 9825A’s serial I/O card, the HP 98036A. This card uses 44 integrated circuits (actually four of the parts are switches, but they’re the same size as integrated circuits), nearly 50% more than the 30 that Mike Kolesar needed for the HP 98032A parallel I/O card. Consequently, Leibson followed Dick Barney’s lead on the HP 98034A HPIB card and used a similar board form factor. Notice how close the spacing between components has become compared to the HP 98032A and HP 98033A I/O cards. The spacing is so close that some of the components almost touch each other. The HP98036A serial I/O card is based on an Intel 8251 USART serial chip, although the board set in this photo uses a second-source chip from NEC instead of the Intel part, indicating a later date of manufacture.   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.