The Adventure of the Cranky Gas Chromatograph

Engineers working on an early multiplexing computer system are flummoxed by spurious results

By Richard Rowe, Contributing Writer

I was senior chemist for Amoco Chemicals at Texas City in 1969. My primary responsibility was developing analytical procedures, technician training, and troubleshooting for a Styrene Unit startup while bringing an existing unit up to current specifications. One task encompassed setting up gas chromatographs for stream analysis. The manual analysis for each stream required as much as two hours instrument time and another hour or so for hand calculation using a planimeter. I designed systems that reduced the combined instrument and calculation time to less than five minutes for each analysis.

When more than one chromatograph was online, the “new” IBM Mainframe computer with 16K RAM and a five meg. hard drive outputting to Teletypes produced spurious results. The results were extremely accurate for one input, which was encouraging because it indicated that the basic chromatograph setup and computer software were well designed. We required ongoing simultaneous accurate results for a dozen chromatographs doing real-time, routine stream analyses.

Each eluted compound was calculated from a separately charged capacitor; crude but effective. The programs were routinely installed with paper tape and IBM cards–many years before computer monitors or floppy drives were available.

The software company startup techs were unable to solve the problem with many software modifications. Time was getting short because the new unit was nearing completion and lab manpower was at a level consistent with computerized assist. I discussed our problem with the computer startup techs and they welcomed my help.

A scope on the computer input revealed a lot of noise, which precluded analytical accuracy. I looked over blueprints for the system and found the installer had used mechanical relays to perform system multiplexing for the chromatographs. It became clear a probable cause for the unwanted noise was relay point bounce.

Suitable mercury wetted relays were found and installed, which solved the problem. It turns out we successfully installed the first, functional multiplexed laboratory computer system. The multiplexing aspect had been a major stumbling block for real-time multipoint control and eliminating the problem enabled rapid successful installation of multiplexed laboratory and process control systems nationwide. I had expected plus-or-minus one percent analytical accuracy when the system stabilized.

The long term absolute analytical accuracy was better than 99.8% overall confirmed by analytical standards and gross weight of unit product output. The company financial reward amounted to several millions of dollars a year as a result of the laboratory computerization.

Of course, had we been privy to the low-priced computers and sophisticated electronics available today the problem wouldn’t have existed. However, we were in prehistoric times regards electronic development and had to work with materials available at the time, bootstrapping every step of the way.

It was challenging to participate in the early days of chemical and computer development when everything was new and untried. Now everything in the field is pretty much done cook book style.

Contributing Writer Richard Rowe has a BS in Chemistry and Physics from West Texas State. He did a stint at White Sands Proving Ground with G. Harry Stine as his electronics mentor.  Sixteen years experience in the Petroleum/Petrochemical business 1956-1972 encompassing a multitude of job titles. In 1976 he decided to change occupations and obtained a doctor’s degree in Optometry from the University of Houston, 1976. He was in private practice until his retirement in 2005.