Remote Temperature Monitoring with Linux

Use a small footprint Linux with some cheap hardware to create a remote temperature monitor.

I started this project to record and access temperature readings remotely after I had a conversation with a friend who is in the HVAC business. His job is to make sure the climate indoors is comfortable—not too hot and not too cold, depending on the weather outside. He finds many new installations have startup bugs that must be worked out, because no two installations are exactly the same. The end of the job is the most stressful. A customer calls after he is done and lets him know something is wrong. The customer isn't happy, but he doesn't know where to start looking for the problem because there isn't any good objective information about what is going wrong with the installation.

We agreed that it would help to be able to record the outside temperature and log the readings electronically. This would be one way of improving the troubleshooting process. I then started the search to purchase an off-the-shelf recorder that was inexpensive, easy to install and simple to use. While looking, I found a wide range of commercial products and kits. Some are standalone and some use a PC for displaying and recording temperature data. Each of our three requirements was equally important and I found that most of the products were too expensive for our budget. Ease of installation was typically another problem. Some devices had complicated wiring or the requirement that they be placed where temperature was measured. Many people consider thermostats to be just clutter on a wall, so it wasn't going to be easy to convince them to have another box to record temperature.

The Linux Solution

Finally, to solve the problems of cost and installation, I looked at the possibility of building a system from components. Following the Linux idea of assembling and integrating tools to get a task done, I looked at using a digital multimeter, a PC and software to make them work together. The digital multimeter with an RS-232 serial port interface would measure temperature using a sensor. The PC would collect the data from the multimeter and process it for display.

I was aware of open-source utilities for the multimeter serial port interface using Linux and had purchased a multimeter earlier for general troubleshooting. We had a retired PC available, so all the components were on hand to build a prototype temperature recording system.

The digital multimeter came with a serial port cable and DOS software. I didn't use the supplied DOS program. There was no way to modify the program to allow temperature measurements with the sensor. Instead, I used QuickBasic to write new software from the ground up. I had the necessary details about the serial port interface for the multimeter, and QuickBasic had all the features I needed. I got a prototype communication program to work, but I ran into memory management issues with DOS and QuickBasic as the application grew in size, especially when I started dealing with the need to display and record data.

It seemed like a big step backward to struggle with memory management at this point. I knew Linux would provide an environment where I wouldn't need to be concerned with memory management, so I looked for a distribution to use as a replacement for DOS.

I found that the most popular Linux distributions weren't appropriate for this application. Even a minimal installation of these distributions would exceed the capacity of the retired PC. The distribution I found that overcame these restrictions was University Linux from Paul Muller. It has small memory and disk requirements. I was able to run it on the retired PC using less than 20MB of DOS formatted hard disk space and 24MB of RAM. Best of all, the distribution is tolerant to power failures. If the power goes out, the PC reboots without causing file corruption problems that need manual help. This saves money and reduces complexity, because I didn't need a UPS to keep the system running during power failures.

Once I configured everything on the PC, there was no need for a keyboard or monitor. I could use a Windows PC and Telnet, along with an Ethernet connection to communicate with the system PC for development and testing. I prefer to write and test incrementally, so I chose Perl for the language for this project. University Linux comes with Perl version 5.003. I couldn't use Perl modules, the application size was too small, so this was a minor inconvenience. University Linux also includes Acme Labs thttpd server. This allowed me to set up the system to use a Web browser for viewing temperature measurements.

Testing Hardware

I used a Tandy Catalog No. 22-805 digital multimeter that comes with an operating manual, DOS software, wire test leads and serial cable with nine pin connectors. According to the manual, the communication settings are 600 baud, seven data bits, two stop bits and no parity. Important information was left out of the operating manual, but I found what I needed on the Web. The DTR and RTS lines need special attention. The DTR line has to be set low and the RTS line set high for the meter to communicate through the serial port. It is impossible to get data from the meter without the two lines set this way.

I could use only stty for serial communication with this distribution and couldn't explicitly control the DTR and RTS lines in the script. This meant I needed a hardware hack to make things work.

I found that DTR and RTS change from a low to high state when I call stty in the script. This works out okay for DTR, but RTS has to remain low. I realized that the second serial port on the PC has RTS low as it isn't being used. If I connected the multimeter serial interface cable RTS to the RTS pin of the second serial port, the multimeter would be faked into seeing the correct line setting. I simply removed the RTS line from the multimeter and connected it to the second serial port.

With that problem solved, I powered up the multimeter and put together a short test script (, as follows:

# !/usr/bin/perl
# Script for reading Tandy Model 22-805 meter
# through serial port.

$port = "/dev/ttyS1"; # set to COM1

system ("stty 600 cs7 cstopb clocal -ixon -echo < $port");

open (SERIALPORT, "+>$port") or die "can't open $port. ";

print SERIALPORT ("\n"); #  take a reading

$R = <SERIALPORT>; # read returned string

print "$R" ;

close (SERIALPORT); # close port

exit 0;

If the script ran successfully, I would get a string with the same reading shown in the multimeter LCD. I set the multimeter to the resistance measurement range and ran the script. The result was:

OH   O.L MOhm

A good start! The hardware hack worked. Now it was on to measure a temperature sensor with the multimeter.

I chose an NTC (negative temperature coefficient) thermistor for the temperature sensor. Despite the fearsome sounding name, this is just a small two-wire electronic component that changes electrical resistance with temperature. With a multimeter, the resistance measurement provides information to tell temperature. The thermistor is impossible to wire backward, because it isn't voltage-polarity (+ or -) sensitive. This means one less thing for the technician installing it in the field to worry about.

The thermistor isn't fragile, but the leads to the body can be broken with excessive tugging or bending. I used a two-position terminal block to solve this problem and make the connection to the wiring simple. I placed one thermistor lead and a wire under a screw terminal and then tightened the screw to make a solid mechanical and electrical contact.

With the thermistor connected to the ends of the test leads and the test leads plugged in to the multimeter, I powered up and ran the test script again. The result was a resistance reading:

OH  34.23kOhm

The numeric portion of the reading is 34.23 with a k after it. The k is an abbreviation for kilo or 1,000. Because the multimeter LCD doesn't have enough characters to display large numbers, it uses a multiplier. In this case 34.23k is 34,230 Ohms.

I found that this reading was very close to 0 degrees C by referencing a table of resistance-to-temperature values supplied by the manufacturer. It matched the temperature reading of another thermometer with a sensor in the general area, so I was confident that this assembled system would work and provide accurate readings.

Now it was time to create a script to to use the data and display the temperature value.



Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

Try ipmitool

Anonymous's picture

Try ipmitool - read the man page.

Linux Temperature Sensor

Jeff Pile's picture

I know this is an old post, but if your looking for a temperature sensor to use with Linux, try a google search for DirecTemp or contact Quality Thermistor.

They have a virtual serial version that works well.

Alternative solution

Anonymous's picture

Nice article about connecting digtal multimetr to Linux box. In the case you just want to measure temperature, you can find several other solutions on the web. One of them is
That one uses network of digital temperature sensors (DALLAS DS1820). From my point of view it is cheaper and more better solution for your needs and you don't need to solve "ugly" DTR/RTS hacks.

Other similar projects to connect sensors with I2C or Dallas interface to serial or prallel port exists as well...


Anton's picture

If I connected the multimeter serial interface cable RTS to the RTS pin of the second serial port, the multimeter would be faked into seeing the correct line setting.

Why not just tie RTS to ground to pull it low?

You can't assume that everyone has a second serial port or that it won't be initialized, sending that RTS line high too.

DTR, RTS & Linux

Anonymous's picture

DTR & RTS signals, are you kidding? Yes, I know these signals are not supported well in Linux. You selected nice hw workarround. I selected hardware workarround in the past too.

Windows users don't need such workarrounds. Why Linux should? Why everyone only uses hw workarround and doesn't fix it in proper way in driver and user utilities? Are developers in Linux world just ignorants? I understand why Linux don't have (open source) NVidia and ATI drivers, and drivers for special hardware (like WiFi cards, etc). But I don't understand why problem with RTS and DTS on serial ports wasn't solved long time ago...