jup - yes, and no.
A thermocouple has a resolution of just over 0.04mV per degree C. If the analog input is designed to take a full scale TC input (around 60mV) then the ADC would have a resolution of about ~1.5 degrees for a 10bit ADC, ~0.4 degrees for a 12bit, ~0.1 for 14bit... 10bit chips are cheap, 12bit is a few dollars per chip in quantity and anything more than that becomes expensive - most applications opt for the cheapest they can get away with.
To 'find' extra resolution the system might then average successive readings within a threshold and look for step changes which will reset the averaging, hence the reason why you may see readings on a device which appear to have better resolution than they actually do. Systems with damping & step thresholds are also hard on feedback control loops and cause oscillations when the system isn't steady.
ah, a digitally minded person.
but i have to argue yes and yes, k-type thermocouples are reasonably linear over about 1500degC range but are only
accurate to about +-1.5degC at the temperatures we want to measure (which is FA in terms of their full scale which makes them good for that range). T-types, although have a much smaller linear range, are much more
accurate across the range we want to measure here.
http://www.microlink.co.uk/tctable.html
alot of logging/measurement equipment allows adjustment of the analog gain (setting the range) to the input of the ADC to increase
precision of low-scale measurements. also, alot of cheap ADCs are logarithmic, so precision is much higher at lower ranges. number of bits and
resolution is not the underlying factor in thermocouple
accuracy here.
wide temperature range measurements, k-types are more suitable, for room temperature measurements go t-types (or just go platinum RTD)
Bottom line: a feedback control system will never be able to control to better than +/- input resolution, and just looking at your graphs I reckon that might be +/-0.4 - so in reality it's doing pretty well.
the switching on of the compressor was set at 12.5DegC, the 0.4 deg above the 12degC point had little to do the digital resolution of the loggers DAC but the actual set point i had programmed to switch the compressor on.
i been seaching around the house for the manual for the logger to look up the specs but can't find it. if anyone cares, it's a
www.datalogger.com DT50. i can almost be sure that 1) the precision/resolution of the logger using a K-type thermocouple is a shit load better than +/-0.4DegC 2) the accuracy of the logger using a k-type thermocouple is crap (about +-3DegC).
edit: found the manual. the DT50 quotes a resolution of 0.01% over a temperature range of -250 to 1800DegC. it uses a programmable gain stage into a voltage-to-frequency converter feeding a 'Programmable time-base and frequency counter' to interface with the CPU for its ADC routine. all data is stored as 24bit (16 bit mantissa) floating point values.
) but it looks as though Zizzle is keeping his wort with half a degree and I know he uses Brissy's controllers.
