Today the team successfully modified the temperature sensing system to remove ~90% of the steady-state measurement error. Originally, internal thermistor Peltier readings differed from external thermocouple readings by about 7 degrees Celcius. This error poses a large issue for any closed-loop temperature control since the total swing is only about 10 degrees nominally.
While improving thermal contact between the Peliter and thermistor reduced the measurement discrepancy, a 5-degree error remained. The team ran a characterization test to experimentally determine the actual thermistor resistance values at a given temperature.
Experimental setup
The team developed a script that incrementally increases the Peltier power via PWM and then reports the measured resistance value once everything stabilises. Meanwhile, a thermocouple is connected to the Peltier to supply the real temperature. This test allowed for a new resistance-temperature relationship to be determined, differing significantly from the formula provided in the data sheet.
Temperature-resistance relationships, experimental and data-sheet
Note that while the Steinhart-Hart relationship is more accurate for thermistor temperature-resistance relationships the team uses simple logarithmic models instead. Using simple logarithmic relationships dramatically reduces the time complexity of the calculations due to the absence of exponential terms. Thus this model simplification offers great accuracy within the operating range without needless hardware requirements.
Implementing this new relationship successfully reduced the temperature reading error by an order of magnitude. With accurate readings, the team can now progress further into PID Peltier control!
Comentários