I guess I understand how the triple point of water could be used to calibrate a thermometer at 0°C, but what does that prove about the accuracy of the thermometer at 150°F?

 

I guess you need a standard for a second temperature. How about boiling distilled water in a quartz container while compensating for atmospheric pressure? Just a guess, but it's all I have. Maybe somebody else sells something that boils or changes from a solid to a liquid at 65.00 degrees C.

In Chemistry 101, we used the melting point of phenyl as a temperature constant.

 

Well accuracy is less important than precision and this research facility is top 5 in the world so I'm sure their equipment is pretty damn good if not the best money can buy. Until I can come up with a better solution, I think i'll stick with that.

As far as minimizing interference with pcb design, filters, components, etc , do you have any suggestions?

For example, in this article

http://www.scientificamerican.com/article.cfm?id=a-homemade-high-precision&page=2

in scientific america, the author is using pyrex tubes and glass beads, I believe, as insulation but I'm not totally certain of their function.

 

A correctly chosen thermistor will give a lot more mV change per 'C (much more sensitivity) than a LM35 which outputs 10mV per 'C.

 

Yep your right the linearity spec is .5C from -55 to 155.

http://www.ti.com/lit/ds/symlink/lm35.pdf

 

Here is a good read.

http://www.thermistor.com/sites/default/files/specsheets/Negative%20Temperature%20Coefficient%20Thermistors.pdf

 

this research facility is top 5 in the world

Considering your target, you're either going to have to build or borrow a couple of parts of a world class laboratory, so I'd say you're on the right track.

 

Right on ******* target, ronv! Great link, very much appreciated.

>Because thermistors exhibit a high degree of sensitivity
and a quick response to temperature changes, thermistor-based temperature controllers typically provide the best results for baths requiring excellent temperature uniformity and stability.

Well it looks like that's finally settled, I'll be sticking with the thermistor.


#12,

+/- 0.05C stability in water IS pretty much world class. That's why I said in my original post that I knew my goals were bold but if nothing else they set the tone for my project. Which is essentially "no loss in precision is trivial". From that I was hoping to learn about any methods or tricks for signal conditioning or pcb design that could squeeze out any uncertainty in the temp sensor circuit. As I said, I'm completely new to this so even fundamentals would be helpful.

 

Wayneh may have a method in mind to improve the accuracy of the LM35, or he may be planning to calibrate the circuit to the particular LM35 in use. Otherwise, an LM34 is intrinsically more accurate than an LM35 simply because one degree F is smaller than one degree C. I think that, with the relatively narrow temperature band you are trying to cover (135 to 165°C), a thermistor can be more accurate than either an LM35 or LM34. But that's just based on my experience, which is not exhaustive.

My approach, which I've used with success, is to just use an op-amp to amplify the LM35 signal before feeding it to a comparator. If you only need a narrow range of temperature, the op-amp gain can be as large as you like - you don't need to allow for large changes.

Amplifying the LM35 voltage doesn't improve it's precision or accuracy (actually it degrades both a little) but it eliminates the impact of hysteresis inherent in the comparator. That's something like a few mV for the LM339, which corresponds to a few tenths of a °C for an un-amplified signal. I think amplification also eliminates the precision advantage of the LM34, which I would have used but it wasn't as easy to find. I suppose the LM34 does not have any accuracy advantage, only the precision advantage.

I think you could achieve quite a bit higher precision if you took this same approach, but added a microprocessor to watch for the mean, time-averaged reading, and overlook the noise. Just speculation on my part.

Once you have a very precise thermostat, you'll still need to calibrate its accuracy against known standards. That's a whole science in itself, as you are learning. In the good old days, research facilities used to employ instrument technicians, and you'd do well to find yourself one.