DISCLAIMER: I am not in the temperature measurement business and have no background in any hard science, so I am going way out on a limb here. I have recently been trying to put together a system for measuring my brew temperature, and have been frustrated in trying to comprehend all of the variables and how they interact. The little I think I know I have learned from materials available online at Omega and other sites of unknown provenance. The one thing I am sure of is that for anyone not truly familiar with temperature measurement there is an awful lot to learn, both in terms of how to do it and what to make of the results.
I'm writing this in the hope that knowledgeable people can (1) correct my inevitable errors - both conceptual and in the numbers, and (2) expand on the discussion as they see fit. Many of these topics have been discussed here and there in the various online forums, but some have not, and those that have tend come up in narrow contexts that cannot easily be extrapolated into general principles. I think a comprehensive, general explanation of "Thermometry for the Home Barista" would be a useful resource for many of us, and maybe our combined efforts could be worked into a reference that really would deserve the pretentious title I gave this post.
Here is a start.
FORGET ABOUT ACCURACY
The most important thing to understand about measuring temperature is that accuracy is for our purposes unattainable. By "accurate" I mean that the temperature reading is equal to the actual temperature being measured. When David Schomer says the best temperature for his Dolce is 203.5F (95.3C), that information is of no use to anyone without first taking into account the factors that affect the accuracy of all such temperature measurements. There are at least four.
SENSOR ERROR
The first is sensor error. It seems that most of us are using thermocouple sensors to measure brew temperature. The four most common thermocouples types are J, K, T and E. Schomer apparently was using a type J thermocouple, which according to http://www.omega.com/techref/colorcodes.html has a "standard limit of error" of the greater of 2.2C or .75% of the reading, meaning his 95.3C/203.5F reading could actually have represented a "true" temperature of anywhere from 93.1C/199.6F to 97.5C/207.5F. That is a very wide range, especially considering that many here report significant taste differences with brew temperature variations of less than 1F.
Type K thermocouples are comparable to Type J in terms of error. Type T is a bit more accurate, with a standard limit of error of 1C or .75% of the reading, but even if Schomer had used a T his 95.3C/203.5F could have represented an actual temperature anywhere from 94.3C/201.7F to 96.3C/205.3F. Type E accuracy falls between K/J and T, at 1.7C or .5%, but I have not encountered anyone using Type E (and would be glad to learn why).
Note from that Omega chart that sensor error can be reduced by roughly half if the sensor is made of so-called "special limits of error" material. But even with a Type T SLE sensor the range of sensor error remains in multiple degrees F for our purposes.
METER ERROR
The second factor affecting accuracy is meter error. Any temperature meter that any of us is likely to use to read a temperature sensor will add its own error on top of the sensor error. On the higher end of the accuracy continuum are the Fluke 50 Series II meters, with a stated accuracy of 0.05% plus 0.4C. Unfortunately, even with these meters and a T thermocouple a reading of 95.3C/203.5F could represent an actual temperature of anywhere from 93.9C/201.0F to 96.7C/206.1F. And many meters out there don't even come close to the Fluke 50s in terms of accuracy, some even adding multiple degrees C to the total error.
SENSOR PLACEMENT
Third is sensor placement. Many of us do as Schomer did and run a TC into the puck through a hole drilled in the brew basket. Others snake the TC over the group gasket and let it sit above the showerscreen, or attach it to the screen screw if one is available. And some have even bored into the group to place the TC at the end of the brewpath before the dispersion block, or place it even higher in the brewpath. The relative merits of these methods would require a lengthy discussion far beyond my competence and highly dependent on specific espresso machines, but for our purposes it is sufficient to note that sensor placement is critical in interpreting any temperature measurement, and that readings taken at various places on various machines cannot meaningfully be compared.
UNQUANTIFIABLE ERROR
Fourth is what might be called "unquantifiable error." Maybe you got your super-accurate temperature meter off Ebay and you don't realize that due to age and use it is now far outside factory calibration specs, and therefore the stated meter error is meaningless. Or maybe you've snipped off the TC bead and just twisted the wires together in order to fit it through the hole you drilled in your brew basket and now the stated sensor error is meaningless (how long and how tight is the twist, and is that unchanged after multiple brews?). Or maybe you routinely fail to let your meter come up to operating temperature before using it because you didn't realize that you were supposed to? Or the TC contacts on the meter are corroded to the point of influencing the signal? Perhaps your K type TC came to you through Ebay, from a seller who used it in a hostile, high temperature environment for many years? The error added by any of these factors (and I'm sure many others) could easily be very significant, and cannot be quantified.
So forget about accuracy. It is, for our purposes, unattainable.
ACHIEVE PRECISION
What *is* attainable is precision. By "precision" I mean the ability to get the same measurement each time. Both sensor error and meter error are due to material and manufacturing variables, which are (generally) constant so long as you use the same sensor and meter. As a result Schomer, using his setup, was able to repeatedly demonstrate that his blend tasted best to him when brewed at the same temperature - not because that temperature reading was accurate, but because it was always indicating the same temperature which, whatever it actually was, was "right" for that blend.
Thus our main task in measuring brew temperature is to be as precise as possible. Given the above and all things being equal I suppose I would opt for a high quality meter and T type "special limits of error" thermocouple. However, though there is no reason to prefer a less accurate meter and sensor other than maybe cost and availability, I would consider the accuracy of the meter and sensor of secondary importance to keeping any variables that would affect precision and which are within the user's control constant over every measurement. To this end I would suggest (1) being aware of and following any meter-specific requirements such as warmup time and maximum ambient temperature, (2) always using the same (not just the same type, but the very same) sensor for every measurement, and (3) finding a way to fix that sensor such that it will not migrate from its fixed position over time.
If you can achieve precision, when someone tells you that blend X brewed at "exactly 201.7F" tastes like a chocolate covered Macanudo soaked in 50/50 strained kiwi/starfruit juice, aged for 18 months in 12 year old Highland malt scotch, and then lightly toasted over a peat fire, you will be able to use 201.7F as a starting point, brew the blend at various temperatures until you achieve that flavor, then forget about the 201.7F recommendation and use your own measurement for future reference.
+++++
Please tear it up, ye of superior knowledge!
Related topics that could be explored:
TC vs. RTD and thermistor focusing on relative accuracy and response times.
Pros/cons of wirewound/thin film RTDs and possible use of latter in SS sheath.
Acceptable response times for sensors in both brewing and roasting applications.
Methods for users to compare brew temp results in any meaningful way.
Others?
