I have a temperature sensor with accuracy that can vary +/- 1deg celsius. I would like to do some form of calibration. To do this, I need a surface with a stable temperature reference. Has someone done something similar in the past? The problem I have now is how to get a surface with a stable temperature.
How about putting it, and another accurate thermometer in an ice chest for 20 minutes. Maybe with a see through window.
The two standards are the freezing and boiling point of water. For the freezing point get a container of ice water such that both ice and water are present, so you are sure it is at 0°C. Pure ice could be far below that. If you are really going for accuracy, you can find the true boiling point of the water based on the given air pressure for slightly better results.
The triple point of water is even better but that is hard to reach without a special cell or a vacuum pump.
Triple Point of Water - YouTube - Cool Triple Point Video
While we would love to be able to buy stable temperature references like those for length or mass they simply don't exist.
ProTip: To get a ° on Windows hold alt and type "0176".
jroorda is correct. Freeze some distilled water, then crush it up. Either let it melt a little, or pour a little more distilled water in it and stir, then let it sit for a few minutes to stabilize. As long as most of the contents are ice, all the water is at 0°C.
For the boiling point, again you need distilled water. Find out your barometric pressure and then find a table, chart, or calculator online to tell you what your water will boil at. Bring it to a rolling boil. Make sure your sensor is not touching the bottom or sides.
I am teaching classes in using Op Amps and Comparators for the local maker's club, OlyMEGA. The last one was a workshop, we use a silicon diode as a temperature sensor, with an Op Amp removing the offset and amplifying the signal. This is how we calibrated them.