Designers of the IBM PS/2 learn the hard way the importance of noise reduction in all phases of electronics
By Radcliffe Cutshaw, Contributing Writer
When IBM was developing the PS/2 computer, I was developing a frequency agile broadband modem that was supposed to be able to plug into the same adaptor card that used a series of fixed frequency broadband modems being built by Motorola.
A broadband modem is a radio frequency modem which uses a cable network and the transmit frequency is converted to another frequency for reception. As part of the development of the frequency agile modem, I had to define the connector, taking into account the pins that the Motorola design used, the hardware interface.
In fact, I fully defined the interface to the modem from the connector to the detailed software formats needed to drive the modem. Since, I had defined the hardware and software interface for the frequency agile modem, I was involved with the design of the adaptor card for the PC and the PS/2.
The digital portion of the design of the cards was simple: about a dozen off the shelf TTL parts that populated a short PC buss card and populated only a small section of the longer PS 2 defined card. The rules were followed for good noise and EMC design: digital ground plane, separate and different sized ground traces to the digital and analog grounds, proper bypassing of each ic, analog supply, digital supply, and, importantly, the connection of all the grounds at a single point where the various grounds entered the adaptor card.
The adaptor card designed for the PC buss worked the first time with no problems, and in EMC testing only increased the radiated noise by a few db-standard for the notoriously noisy broadband modem card. The PS/2 computer was still in testing and had not been released, nor had any of the cards designed for the new PS/2 buss. The PS/2 was having a problem meeting FCC Part 15 requirements and, of course, a problem meeting the tougher VDE (now EC) requirements.
When the PS/2 modem adaptor card was tested, the modem worked as expected, but the PS 2 design group was thrown into a panic when the EMC emissions were measured. When the adaptor card was in the machine, the emitted noise dropped by over 20 db, improving the poor noise performance of the system up until that time; there were immediate questions about the design of the adaptor card.
The reason that the card caused the noise to decrease was that the grounds were tied together. The design group had to redesign the entire grounding system of the PS/2. I never found out about the details, but my guess is that the chassis ground and, probably one other ground was left floating, resulting in the poor emissions performance.
However, the problem goes beyond the poor design of the grounding system in the computer. Since the PS/2 was an entirely new design, adaptor cards were designed along with the computer. These cards were all the common types of adaptor cards used before most of the input/output ports migrated to the motherboard. None of these cards had followed the accepted design rules for noise reduction and EMC reduction. No one on the design team had considered the grounding system of the PS/2 computer or the proper way to handle the grounding of any of the adaptor cards.
The rules for noise reduction for both on board and EMC reduction were well established even then. This goes to show how important that noise reduction is in all phases of electronics.
Contributing Writer Radcliffe Cutshaw is a serial entrepreneur, is currently a private consultant specializing in RF and analog design and development. He has been involved in many areas of engineering throughout his career.