I'm doing cooling for a DSLR camera. The cooling is based on the peltier element. It is necessary to obtain information about the sensor temperature with high accuracy and adjust the power on the peltier element to obtain the set temperature. From the information that I found on the Internet, it is not so easy to find a device that will allow to read temperature information with an accuracy of 0.1 C. For example, the good MAX31855, which is designed for K-type thermocouples, has an accuracy of 0.25С. What other options are there? I have no particular desire to use RTDs, since they have a fairly high inertia, which is unacceptable in this system (if I'm wrong, I'll be glad to hear options that don't have much inertia). Therefore, it remains to use only thermocouples. Maybe use an analog AD8495 and some ADS1115 to get less rounded results compared to the MAX31855?
Two questions:
- what is the range you need? e.g. -128C to 500C or -40C..60C or ...
- how much time may the measurement take? or frequency of measurements
The DS18B20 is a good sensor with 1/16 degree accuracy from -55..125 C
Has good libraries, but takes up to 750 milliseconds for highest precision.
There are waterproof versions of it to be used outside, these take even longer to stabilize.
Another option might be to use four or more max31855 in parallel and average their measurements. With 4 devices you get ~2x accurate result 0.25 => 0.13
The advantage is also robustness / redundancy, if one fails you still have 3 left
Do You need absolute temperature readings or would temperature changes be useful?
One way could be to use a precision thermometer to establish the working temperature point and then have an Arduino checking the variations and control the temperature like a fine adjust.
- -20 ... +60
- Less than 500ms would be nice.
Problem is that accuracy is almost always inverse to speed.
Using five DS18B20's could schedule a measurement every 200 ms.
Hi, @quarklabs
Welcome to the forum.
Sorry but I had to edit the layout of your post to make it more readable.
For that level of accuracy, do you have a suitable calibration standard?
You need to look at sensors with stability and resolution, a calibrating procedure will take care of accuracy.
What temperature are you aiming for?
Do you have a suitably insulated enclosure.
Tom.... 
You can usually calibrate-out the error for higher-accuracy, especially if you are reading a narrow temperature range. Of course you need a way of reading accurately in order to make the calibration. And calibration won't help with noise or drift.
A basic linear calibration includes an offset (a value added or subtracted) and a slope correction (a multiplication factor).
1/10th of a degree is tricky, and often meaningless because moving the probe usually changes the reading. If you are measuring a room, different parts of the room will be at different temperatures. If you are measuring a metal plate, different points on the plate will be at different temperatures, etc.
Thermocouples are "good" because they are based on fundamental physics so they are accurate and stable. But they put-out low voltages requiring amplification so noise can be a problem.
Oh... It will be difficult to fit it near the DSLR sensor 
Thanks! Just now I went to the forum 
For calibration, I believe that the temperature of melting ice and boiling water should be enough. I also have a TM-902C, although it will have less accuracy.
I have already described the necessary parameters a little bit above.
I don't know if the body of a DSLR camera is considered insulated, given that there are many heat sources inside it.
And how about an MLX90640
could monitor your setup from some distance.
That's the problem - there's no room there at all. And the place that needs to be monitored (the side of the sensor, which is clamped on both sides) is extremely narrow. Only a thermocouple without a case will fit there, just two welded wires with thermal paste.
The insulation of the testing container you will have the camera installed in.
Have you a diagram of your testing setup.
Tom....
That's about where I need to measure the temperature. Very little space, unfortunately. The copper plate is the one that cools.
It's not a container. The peltier element stands under the radiator and cools the copper plate, which goes inside the camera directly to the sensor =) In fact, this is my attempt to refine what I have already done. The lack of proper temperature control makes everything very difficult.
Please define this as a percentage and a range.
it is not so easy to find a device that will allow to read temperature information with an accuracy of 0.1 C
±0.1C
-20C...+60C
http://ww1.reesscientific.com/brochures/Catalog/Pdfs/2_sensors.pdf You may be forced to use a small mass RTD sensor.
RTDs are generally more accurate than thermocouples. RTDs have typically an accuracy of 0.1°C
A standard K-type thermocouple as suggested typically has an accuracy of +/- 2.5 degrees Celsius (or +/- 0.75%), whichever is greater.
Dallas Temperature sensor ±0.5°C Accuracy from -10°C to +85°C
A sensor that meets your exact specifications may be on the pricey side. Consider adjusting the specifications to allow for a wider tolerance, which could give you access to a broader range of more cost-effective sensor options.
No matter what accuracy the sensor is able to provide it is quite a different matter to control the temperature to provide precise temperature control
But... In this document, none of the sensors are "small", they are all too big.