I'm looking to make an array of sensors to record the temperatures in a house at various points

eg. Above flat metal roof,
inside flat roof cavity (above 1st floor),
ceiling of room (1st floor),
floor of room (1st floor),
floor cavity (first floor),
ground floor room ceiling,
sub-floor/basement space.

I guess I need accuracy tolerance of about ± 1ºC. More important is that once initially calibrated, the sensor do not drift relative to temperature or the data comparison will be bunk. I hope to run the sensors for weeks in winter, summer, spring, autumn to get a full profile of the thermal performance.

So which sensors are good for me. Few seem to be good for ± 1ºC and I need a way to deal with the wiring of them too. So I2C or another digital bus might be good, anything that stops the length of the wire introducing errors into the values. Digital also means I can run a cat5 daisy chain of sensors rather than a separate lead to each sensor which seems reasonable.


3 Answers 3


For such a sensor network I would absolutely consider using Dallas 1-wire components for great flexibility and reasonable price. It is well supported in Linux via owfs.

T3-R1-A picture

At www.hobby-boards.com there is a small T3-R1-A ready made board which you just connect with cat5/RJ45 priced at 18USD. According to the documentation it uses the DS18S20 sensor like already mentioned by BG100.


I've had very good results from the DS18S20. I found it to be very accurate. You can easily get a resolution of 0.5 degrees (Celsius), or if your prepared to do a bit more work you can get 0.1.

You can also string together lots of them on a single data wire, and get readings from each of them by addressing them by their in-built serial code.

  • \$\begingroup\$ I've had good luck with this series. I haven't used them at any distances over 5 feet, though. \$\endgroup\$
    – mjcarroll
    Feb 2, 2011 at 19:21

I've used TI TMP100 sensors with success.

They're IIC, and support up to 8 devices on a single bus.

Their absolute precision is only +- 2°C, but they're repeatable, and have 0.0625° relative resolution. Also, you can calibrate out (most) of the +- 2° offset, if you put the effort in.

I can also confirm that ~20' strings with sensors spaced evenly along it work without trouble.

TI Also make some higher precision, but more expensive IIC sensors, such as the TMP275.

I have a project which scans 14 strings of 8 sensors, and dumps the code out over a serial port a ~2 hz. It's built around a Propeller Microcontroller, and I could post the code and PCB layouts, if you would like.

  • \$\begingroup\$ That would be an I2C bus the size of a room. Usually, that's too large for I2C. No guarantee that a plain unassisted I2C would work at such lengths. \$\endgroup\$ Apr 26, 2015 at 1:39
  • \$\begingroup\$ @NickAlexeev - It seemed to work fine, even with 400KHz I2C. What can I say? \$\endgroup\$ Apr 26, 2015 at 1:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy