Comparing Six Thermistors from 14° to 72° Fahrenheit

(continued from previous page)

It gets reasonably cold in Chicago during the winter. Six thermistors were placed in a cooler (or ice chest) outside overnight. The cooler should ensure that they either aren’t affected or are equally affected by indoors temperatures when they are brought inside to be measured.

Apparatus for measuring multiple temperatures at the same time, starting with the temperatures in a cooler.

Apparatus for measuring multiple temperatures at the same time, starting with the temperatures in a cooler.

The test setup consists of:

Upon bringing the closed cooler inside (with Molex connectors hanging out), I was impressed to find that the thermistors were all within one resolution of each other: 14.7°, 14.9°, 14.7°, 14.7°, 14.9°, 14.7° Fahrenheit. After warming up for a day outside of the cooler, the sensors were again all within one resolution of 72.4° Fahrenheit. Fantastic!

These results are somewhat due to precision resistors from the same lot, somewhat due to the thermistors being from the same lot, somewhat do to the precision resistors being kept at the same temperature, and somewhat due to the thermistors being read four times over the course of a second and the results averaged.

Attempting to Reduce Temperature Fluctuations

As part of the first temperature test, I wanted to find a way to even out temperature fluctuations due to the indoor heating system. That is, I wanted to measure a stable room temperature.

A took a large piece (about 6 pounds) of scrap aluminum and screwed a heat-shrink-tubing-covered thermistor to it. I then covered the sensor with a fender washer.

Big chunk of aluminum with temperature sensor.

Big chunk of aluminum with temperature sensor.

The idea is that aluminum is thermally conductive (easier to reach room temperature and the heat distributes evenly) but the block has a large thermal mass (slower to change due to random air currents). Think of it as a thermal capacitor.

Unfortunately, when I held my hand against the washer the temperature started rising. It would have been better to encase the thermistor inside of two aluminum blocks.

Temperature Results

Here is a really huge graph showing the temperature of all six thermistors from 14 degrees to 72 degrees Fahrenheit over a period of 2400 seconds (40 minutes). The chart is annotated with various points of interest.

Chart of six temperature sensors warming from winter cold to indoor heating.

Chart of six temperature sensors warming from winter cold to indoor heating.

For the first 10 seconds, the sensors remained in a closed cooler. All six sensors stayed at approximately 14 degrees Fahrenheit. This demonstrates that the sensors match well and that the cooler does a good job of insulating.

  1. At 10 seconds, the lid to the cooler is opened but the sensors are not touched or removed. Surprisingly, the sensors show an immediate response. Over the next two minutes, the sensors continue to rise, although they look like they’re leveling out. Perhaps most of the temperature rise is due to the room air and cooler air mixing during the opening of the lid.
  2. At 180 seconds, the sensors are moved to the desk shown in the photograph atop this page. Notice how quickly the temperature rises on the free-standing thermistor (not attached to anything). The sensors attached to the circuit board take longer to reach room temperature. The sensor attached the aluminum block rises much more slowly due to thermal mass.
  3. Around 330 seconds, I get bored. I decide now is a good time to take a picture of the aluminum block (literally the picture you saw earlier on this page). For the next couple of minutes, the cycling of the electrical field in the fluorescent bulb induces a change in voltage on the sensor wires. The wires on the aluminum block are nearest to the desk lamp, since that’s what I’m taking a picture of, but the other wires are also affected by the fluctuations. Notice the temperature readings go above and below the expected temperature line. That’s how you can tell it is electrical noise, rather than localized heating from the light beams or variations from the room air (both of which would only increase the temperature above the baseline).
  4. As previously stated, the sensors attached to the circuit boards take longer to heat up than the unattached sensor. At first it seems marginally surprising that the sensors on the inside of the circuit board are taking longer to heat up than those sensors on the outer edges of the circuit board. But, I suppose that’s how ordinary things thaw, from the outside in. The extra copper on the v2.0 boards seems to have enough thermal mass to cause the v2.0 boards to heat up noticeably slower than the v1.5 boards. This makes sense since the sensor without any mass attached to it takes less time than those sensors with masses attached to them. The greater the mass, the longer it takes to change temperature.
  5. Unexplained temperature blips on the v2.0 boards and the aluminum block. I’m eating dinner so I’m not in the room. Did a cat sit on them? Well, at least you know I’m not faking my data.
  6. At approximately 840 seconds, even without the glitch, the temperatures on the v2.0 boards start passing the v1.5 boards. (Note: the circuit boards are powered off during the entire experiment.) It seems logical that at some point in a thermal curve that the faster transfer of heat (thermal conductance) will overcome the deficit of the larger thermal mass.
  7. Oops. The free standing thermistor wire has drooped over the RS232 serial board that contains four LEDs (RX, TX, CTS, and RTS). The heat from the LEDs has added a degree to the sensor readings. Let me just nudge that over a bare spot on the desk. Ahhh, that’s better.
  8. The aluminum block was measured two hours later and had just reached 60 degrees. This reinforces the notion that large electronic components (such as cameras and computers) need to sit unpacked indoors for 5 or 6 hours to normalize if they’ve been delivered on cold day.

Circuit Board Tests

Now that we’ve established that the temperature sensors are consistent and follow the expected behaviors, we can proceed to testing the motor driver chips, heat sink, and copper fills on PCBs.