A description of bringing an idea from first concept to a marketable item

A few years ago when I was a Project Engineer at Honeywell Security Systems and Custom Electronics, I was asked to help the purchasing department solve a problem that they had  procuring an electromechanical component. The component was a switch, composed of a special alloy. The special alloy allowed the switch to open when the temperature was lower than 42°F, with a high degree of accuracy and low hysteresis; which means that when the temperature became warmer than 42°F, that it would close again.

The price on the switch kept increasing and the manufacturer had difficulty producing the item. This was due to the alloy supplier no longer being able to make a properly constituted alloy. The switch manufacturer elected to close the factory where the switch was being made. I glibly said that I could do the switch using solid state components for a less expensive total material price.

I developed a design and sold the Marketing Department on the idea. In addition to sensing the occurrence of low temperature, I suggested also sensing warm and hot temperatures as possible environmental alarm conditions. Marketing liked the idea and asked me if I could sense the temperature in a refrigerator or freezer. I explained that I could, if I used a remote thermistor probe. In addition, I suggested that I also could sense the presence of a flood condition with a different type of sensing probe.

I chose to use a microcomputer (MCU) as the basis for the measurement and as a mechanism for formulating a serial digital message that would be sent out over a radio transmitter link. This was done periodically from this sensor for ‘housekeeping’ functions, such as indication that a good battery condition exists and that the unit is working and a special message would be sent if there was a temperature or water alert. The MCU that I chose was capable of making a temperature measurement The MCU only measured temperature with an accuracy of ±10°F, which was not good enough. This inaccuracy was caused by an offset error variation from unit to unit. To obviate the effects of that offset, I developed a technique of calibration during the loading of the program code. The way that it worked was that after the program was loaded, a subroutine would execute that would measure a ‘control’ thermistor and estimate the ambient temperature. This measurement was ratiometric and was extremely accurate. Then I measured the temperature using the MCU and calculated the difference. Using an equation in the MCU that simulated a piecewise approximation of the thermistor curve, I was able to  develop a correction factor that ensured that the adjusted MCU measurement was accurate within ±2.5°F. When this product is used with an external thermistor, the specified accuracy improves to ±1.0°F. If the product was calibrated with a paired thermistor, considerably improved accuracy, better than the ±1.°F accuracy could be obtained. At this point, for better than the 1 degree accuracy, manufacturing facilities would have to be modified to ensure that the test and storage areas of the pre-test product are stabilized.

This product was capable of being used for six different functions versus the older product with the electromechanical switch which was only usable for one function. The price of the Bill of Materials (BOM) for my re-designed product was only 33% of the price the original product's BOM. This resulted in a definite win-win situation. This product is called the Honeywell 5821.