Hey everybody, once again I have a question that's only tangentially tied to electronics, but I'm starting here cause you all have been so helpful and knowledgeable in the past.

I'm about to restart a project I tried a few years ago related to RTD temperature sensors. We use platinum RTDs for temperature sensing in espresso machines and want to calibrate them to fairly tight tolerances (a few tenths of a degree F, the closer the better.)

Years ago I built a partially enclosed hot water bath which held the water at 200F (the center of our target operating temperature range, so the temperature where accurate performance is most critical) and had openings sized to allow one reference thermometer and four RTDs to-be-tested to be suspended in the water very close to one another. I recorded the differences in temperature readings and used those to determine the calibration offsets to be used in our machines.

This calibration method proved to be an accurate predictor and is much, much faster than our current calibration method, but my boss didn't trust it. He insisted that since metal is a better thermal conductor than water, we had to build a block of aluminum or copper with holes sized to accept the RTDs and then find a way to control and monitor its temperature.

It seems to me like a block of metal will act like a heat sink with massive temperature gradients between the heat source and the outer edges, and like water will be easier to hold at a uniform temperature, in part thanks to naturally occurring convection currents.

It also seems to me like our RTDs, which are encased in 1/4" diameter (6.35mm) stainless steel probe shafts, will have far better thermal conductivity in contact with water than they would just touching a piece of metal (unless lots of thermal paste was used.)

I've tried to find info on heat transfer in various scenarios to back up my ideas, but I'm having a hard time. I've made some headway on stainless steel immersed in water, but SS touching aluminum (with imperfect contact and tons of little air gaps) has proved trickier.

What do you all think? If you needed to get a bunch of temperature probes all to the exact same temperature, would you prefer a water bath or a block of metal? And why?

Sorry I've written so much (again) and thanks to anyone who will share their insights.

 

If you needed to get a bunch of temperature probes all to the exact same temperature, would you prefer a water bath or a block of metal? And why?

If I needed to get a batch of temperature sensors all to the same temperature very quickly, I'd opt for a stirred water bath or oil bath. Otherwise, if I had plenty of time for the whole lot to achieve thermal equilibrium, I'd probably choose the metal block-- or, just loosely tape the sensors together in a bundle and let them sit in a closed box at a constant air temperature, provided RTD self-heating isn't an issue (i.e., the RTD circuit doesn't apply too much voltage to the sensors).

 

You can model any solid with two temperature faces as a set of gradients, and those gradients are fixed in space. The only way to completely stop that condition is to have the solid at the same temperature in all places. If you have a fluid, the temperature gradients can be reduced to almost nothing by simply stirring the liquid or gas. You can't stir a block of metal.

 

You can't stir a block of metal.

...unless it's REALLY hot!

 

One other note: make sure to include a generous length of the RTD leads in the isothermal environment along with the RTDs themselves, to avoid errors caused by thermal conductivity of the wires.

 

I would recommend against the metal block in favor of a stirred liquid. I'm pretty sure heat transfer from moving water, which is the final application after all (a valid point of argument in favor of the liquid method), is superior to the conduction and radiation transfer you would see between the block and your probes, which may vary from trial to trial with the precise dimensions of the probes, and between the heat source and the bulk of the metal block. The heat in the block will move more slowly and show gradients, and so the temperature of the block at the probe would be expected to be harder to control to a given precision.

You can buy NIST certified temperature probes. How do they do their calibration?

 

I would recommend against the metal block in favor of a stirred liquid. I'm pretty sure heat transfer from moving water, which is the final application after all (a valid point of argument in favor of the liquid method), is superior to the conduction and radiation transfer you would see between the block and your probes, which may vary from trial to trial with the precise dimensions of the probes, and between the heat source and the bulk of the metal block. The heat in the block will move more slowly and show gradients, and so the temperature of the block at the probe would be expected to be harder to control to a given precision.

You can buy NIST certified temperature probes. How do they do their calibration?

I could be wrong, but I believe they use instruments like this one (or maybe fancier versions of the same idea.)
http://www.omega.com/pptst/CL1000.html

We're looking for something that splits the difference on accuracy, price, and throughput. If we can get within a few tenths of a degree, but test 4 or more probes at a time, that would be ideal. Also, I think I can build a very nice water bath with parts we have on hand plus maybe $100 tops (as opposed to $1000+ for the Omega products.)

 

Many thanks to all of you! I feel more confident now that I was on the right track in the first place. I'm definitely glad I asked though - I appreciate the extra insights you all have offered.

(which isn't to say that I wouldn't welcome more perspectives if others disagree.)