Back to the front page

Stupid metrology I: Dippy bird thermometer

July 2011

I have a list of life goals, and somewhere on that list is 'use a drinking bird toy to solve a problem'. The classic application is pushing a button repeatedly or triggering a motion sensor, but somehow neither of those represent a problem for me right now. Since I haven't yet identified a clever use, today we're going to talk about a completely stupid one: using a drinking bird as a temperature sensor.

The dippy bird is a popular physics toy, because it demonstrates a lot of interesting thermodynamic principles. There is no shortage of explanations online about how the toy works, but to recap:

  1. The bird consists of two chambers (body and head) connected by a tube; the body is filled with a volatile solvent.
  2. The head is covered in felt; when the felt is wet the head is cooled through evaporation.
  3. This lowers the pressure in the head and draws liquid up the neck
  4. The bird becomes top heavy and tips over, which empties the liquid back into the lower chamber. Cycle restarts.
The traditional glass of water just serves to keep the felt wet; the bird will continue to dip without it until the head dries.

My thermodynamics classes were brief and also a long time ago, so I have no intention of invoking specific evaporation enthalpies or endoreversible heat engines here. As an experimentalist, what I see when I look at this toy is:

Dip period = f(temperature, humidity, wind)

where f is some unknown (to me) function containing the thermodynamics. In my case, it's enough to simply determine whether there is a predictable relationship between dipping period and the ambient temperature. This would form the basis of our temperature sensor. I figure I should be able to just take some measurements and work something out empirically.

Taking data

I set up the bird next to my desk for a few days with some data logging electronics. I'm measuring three quantities: ambient temperature with a DS18B20 temperature probe, relative humidity with an HH10D capacitive sensor and dipping period with a photo-interrupter ripped out of an old CD-ROM drive. (I tacked a piece of business card onto the body to trip the photointerrupter). Since this is indoors, we can eliminate wind as a variable. Here's what it looks like (humidity sensor just out of the frame).

With a glass of water to keep the felt wet, the bird is happy to keep drinking for a very long time. I left this apparatus running for about 2 days, after which I got sick of it distracting me and dismantled it. This length of time constituted something like 5000 drinking cycles. The data is available here if you want to play with it, but allow me to plot it out:

Let me first note that the relative humidity (blue trace) is not calibrated - the I2C on that unit appeared to be broken, so instead of reading out the device calibration data I had to just use typical values. That said, I would still expect it to capture trends in the humidity.

The temperature (green trace) varies periodically over 24 hours as one might expect, but the overall variation is quite small since this is a climate controlled office (19.5 ± 1) °C.

The dipping period (red trace) is extremely noisy, and over small timescales quite unpredictable. I've marked two points of significance on the plot above. At some point I briefly opened the window next to the bird, and with the rapid changes in temperature, airflow and humidity the bird went completely nuts (also reflected in the temperature). The second point happened when I topped up the glass the bird was drinking from, I'm guessing this temporarily spiked the evaporation rate.

Our first order of business is to clean up the dip period data, which I've done by running over it with a 30 point moving average filter:

Looking at the long-term trends in the data I can see something of a correlation between drinking period and ambient temperature (which is what we wanted), but the humidity doesn't look like it's having much of a bearing. To better demonstrate correlations, we can plot variables against each other:

From these plots we can see that there's a clear (albeit slightly messy) negative linear relationship between the dipping period and ambient temperature. This is exactly what we would expect; as the temperature goes up the evaporative cooling is enhanced, so the drinking process goes faster. The second plot (against humidity) is mess with no discernible relationship. In principle the local humidity should influence the evaporation rate from the felt head, so this lack of correlation is not what I would have expected. But I'm also not sure how much I trust the sensor.

To bring things back to the idea of a temperature sensor: by fitting a line to the first plot above we get a formula for converting dipping period into temperature. How good is it?

Pretty good! And by pretty good, I mean "worst case error of 0.5 °C in an operating range of 2 °C, with a response time of approximately 15 minutes". The two major deviations correspond to opening the window (vertical spike just before the second midday) and topping up the glass (dip period falls during the second afternoon).

Which, as far as temperature sensors go, is barely passable. But hey, cut it some slack, it's based on a dippy bird

A dippy bird!

Are you not charmed? Is your heart made of stone?

I tried to repeat the data collection at home where lately the temperature has been hanging around 14 °C. Evaporation is then so slow that the dipping period drops to about 10 minutes to drink - not ideal for a sensor.

Now, lest you suppose that I am alone in my ambitions, clinging desperately to the dream of a drinking bird toy serving a higher purpose, allow me to offer you some perspective from the scientific literature.

To conclude: I have successfully made a temperature sensor out of a drinking bird, provided you want to know the temperature in a stable, climate controlled indoor environment.

Back to the front page