Project Sous Vide, Part 1

Saeed Masroor (LinkedIn, Facebook) told me about a current project of his: sous-vide cooking, in which you slowly cook meat and vegetables, at moderate temperature, by putting them in sealed bags in water. He tells me that the temperature of the water is critical - apparently cooking at 62 degrees Celsius gives a completely different result than at 64 degrees.

This means that one needs decent equipment in order to do this - a good thermometer, for instance. In addition, if you don't want to have to continuously check the temperature and adjust the heat, then you also need a device to keep the temperature constant. These devices exist, but tend to be pricey.

Saeed asked me whether it would be possible to make one of these sous-vide devices DIY-style. The main spec is that the temperature variation in the water should be within 1 degree Celsius of the set temperature. I was looking for a next project to do, after Irene's heartbeat tshirt, and this sounds like a nice challenge. Especially since it has to be done on the cheap ...

My first thought was that measuring the temperature of water would be tricky, given that water and electronics don't go well together. Luckily we're not the first to face this issue, and there are many water-suitable sensors, such this DS18B20-based watertight temperature sensor from iPrototype. It claims 0.5 degree accuracy (absolute error), 12-bit resolution, working range -10 to 85 degrees Celsius, costs 8 euros, and has a clear hookup guide.

It was time for some experimenting. Today Saeed and I tried out the sensor. We wanted to know how accurate it was and whether it was easy to use, but I was also really interested in finding out how big the variation in temperature within a pan. Since the pan is continuously being heated from the bottom (at least on our stove) and cools from the side and top (especially from the top, through evaporation) there should be a gradient in temperature. Of course I could look up some numbers and work out a good guess, but doing the experiment is more fun ...

We first tested the sensor and two mechanical thermometers on the classical calibration points - ice water and boiling water. The sensor and one of the mechanical thermometers agreed that the ice water was between one and two degrees above zero (the other mechanical thermometer didn't go down to zero). I believe that the temperature really was zero, since we measured close to the ice cubes, so my current guess is that both measurement devices were off. Though it is striking that they're off by exactly the same amount ... In boiling water the sensor varied between 99.5 and 102, depending on whether it touched the bottom of the pan or not. The mechanical sensors were more difficult to use here without burning our hands. These two calibration experiments suggests that the accuracy is not too bad.

In fact, the sensor reported numbers with a resolution of 0.06-0.07 degrees, and was remarkably consistent; over time it would typically report many times exactly the same number, and then shift to a new number 0.06 or 0.07 degrees higher or lower, and then stay there for a significant number of measurements. I could see no `flutter' at all in the measurements - it looked all as if the changes in reported numbers were really created by changes in temperature.

Since the really interesting case was a big pan (4 litres) of 60-70 degrees, we moved on to that. After lots of fooling around, here's what we learned:

• If we left the water undisturbed, then the sensor would register about two degrees more lying on the bottom than held just below the surface. 
• If we tried to prevent the sensor from being heated by the bottom of the pan, by constantly wiggling it, then the temperature difference was smaller; we now think that that comes simply from the mixing that our wiggling created.
• Over the course of 5-10 minutes the temperature seemed to fluctuate by one or two degrees, even if we did nothing. We now think that that may come from fluctuations in the gas pressure, leading to fluctuations in the heating of the pan, even though the cooker setting was constant. Maybe if gas was turned on in one of our neighbour's houses, then the pressure would slightly decrease ...

What do we conclude from this? Well, for starters, it seems like the sensor has plenty resolution and  good enough accuracy - that's very good. We also learnt that in an open pan, temperature variations of two degrees happen, which is more than the specs allow for; I expect that that will only get worse if we also put things in the pan, or make the pan deeper. So we need to do something about that. In fact, the professional devices also seem to create significant circulation, as you can see in the picture to the right (they're also called `sous-vide circulators').

So we're in good shape when it comes to measurement, but we still need to find a good way to mix. That's the next step ...