The RR is then measured as (temperature difference across reflux condenser)/(temperature difference across the product and the reflux condensers together).
The two temperature differences were then input to two channels of the analogue-to-digital converter (ADC) of the microcontroller. The RR was calculated using the converted digital values which were sent to a HyperTerminal and also converted to a pulse-width-modulated waveform, which, when filtered, produces a DC voltage. The DC voltage was then sent to an oscilloscope to watch the RR calculation in real time.
The results were exactly as predicted, within +/-2%. (This is due to the digitization of the analogue signal. I didn't provide enough gain in the instrumentation amplifers, so only 7 bits were used. Suppose the quantized level is 100, but it could be 101 or 99 with a little push. When we divide 101 by 99 as would be done in the RR equation, we get 1.0202, or 2% up. Similrly, if we had 99/101=0.9802, or 2% down. More gain would provide more accuracy.) Decreasing the coolant flow rate caused a transient that came back to the same value:
And increasing it again caused another transient:
So, the technique works very well. Q.E.D. and all that!
M
