USB thermometer, an ambiguous result

Summer arrived unexpectedly once again, bringing the heat. Since I’m responsible for a small server room, I wanted a way to monitor the temperature remotely.

You wouldn’t really call it a full-fledged server room, but it has all the hallmarks of one, including an air conditioner that runs from around the beginning of March to the end of November, keeping the temperature at a comfortable 18–20 degrees for the servers, active network equipment, UPSs, and so on.
A year ago, the air conditioner failed. Yes, that was my oversight: I should have set up alerts for rising CPU and hard drive temperatures, but I didn’t. Fortunately, we found out almost immediately and took emergency measures to replace the air conditioner.

While browsing online stores, I came across a USB thermometer and wanted to see if I could use it to monitor the room temperature. (At home, I have a temperature sensor connected via a COM port, using a Dallas sensor.)
Anyway, I ordered one, received it, and unpacked it.
Here’s what arrived.


A small blister pack and a device that looks like a flash drive. Nothing special at first glance.


It came with a CD, presumably with the software. I say “presumably” because I don’t have a DVD or CD drive connected at home, and couldn’t be bothered to hook one up.
I googled the text on the packaging and found the manufacturer’s website, where I easily located my thermometer and downloaded the software.
Downloads page — www.pcsensor.com/index.php?_a=document&doc_id=6
Link to the latest version — www.pcsensor.com//uploadFile/APPsoftware/TEMPer%20V24.4.4.zip

The device looks a lot like a flash drive, except it has an extra hole, presumably for hanging it on a nail or something :). For now, I’m using it plugged directly into a front USB port on the computer.


The first thing that seemed unusual was its weight. The device has a metal case and feels quite hefty; the difference from a flash drive is noticeable. I got curious and decided to weigh it.


The photo isn’t great, but the device weighs 13 grams.

Obviously, nobody’s going to plug an untested device into a server, so I decided to try it at home first.
I installed the software and plugged the thermometer into a USB port.
It started right up, with no drivers to install. (The only requirement is that .NET Framework be installed; the program asks for version 3.5, but actually works with 3.0.) No configuration was needed either: I just plugged it in, launched the software, and got the temperature readings. The packaging claimed support for the major operating systems; I tested it on Windows XP. When launched without a sensor connected, the software displays two temperatures, indoor and outdoor. Most likely, you can connect a pair of these sensors and monitor the temperature in two locations.

The program’s main window.


The actual temperature was around 26–27 degrees. By the way, if you drag the mouse across the graph from left to right, you can see the selected range; dragging from right to left restores the graph to its original view.

After 6–7 minutes, the temperature rose a little higher, apparently due to the sensor’s own self-heating. Dallas sensors can do this too, especially when polled frequently.


After that, the readings barely changed. When I changed the polling interval from 2 to 10 seconds, the readings dropped slightly, which confirms my theory. You can also feel the device getting a little warm during operation. That’s not good at all; it’s unclear what inside it could be generating heat.
It’s not a major problem, though: the software has quite a few settings, including an option to adjust the readings.

Naturally, I wanted to take the device apart. I expected it to be difficult, but it turned out that pulling firmly on the connector was enough to split it into two parts: the circuit board with the connector and components, and the metal case.




The device uses an LM75 temperature sensor datasheets.maximintegrated.com/en/ds/LM75.pdf from MAXIM. Inside, it is pressed slightly against the metal housing.

I brought the sensor to work, connected it, and set it up. Below is a description of the mess shown in the screenshot.


The sensor is located near an alcohol thermometer, which reads 19–20 degrees (fluctuating with the air conditioner’s operating mode). You can see the temperature drop to about 28.38. I got curious about what would happen without the housing. In the second part of the graph (after the rise), you can see that the temperature settled at the same value. Even the magical properties of the blue electrical tape didn’t help: I insulated the sensor board with a single layer of it. I went back to the server room, removed the tape, and put the board back in the housing. In the end, we have almost the same result as before all the experiments.

So the sensor reads as much as 8–9 degrees too high (according to the datasheet, the error can range from -2 to +2 degrees). That’s very strange for a digital sensor. In the end, I added an 8-degree correction in the software. I also found that the software only allows you to set the correction within a range of -8 to +8 degrees.



The software can send notifications via MSN, email, or Skype. It also displays a graph window and a temperature measurement log window.

In summary, the device left me with mixed feelings. On the one hand, it’s reasonably well designed (with enough effort, you could even install it in a tank to monitor the liquid temperature), and the software is pretty good. But the accuracy is awful.

Pros.
The device works.
No problems during installation.
The software is convenient and intuitive.

Cons.
The sensor reads the temperature too high. Fortunately, this can be corrected. Still, that’s a pretty major drawback for a device designed to measure temperature.


Overall, I think this device has a place in the world. Time will tell how it holds up.