# Three-component SmartLEDs

From the MAKE Flickr pool:

The Smart LED connects a single LED to a microcontroller and power source, allowing it to do fun stuff like relate the ambient temperature via pulses of light. Jim Blackhurst’s remix dispenses with the resistors for an even smaller profile, connecting the LED directly to the coin battery and ATTiny25 microcontroller, changing the RGB and flashing as instructed by the ATTiny’s built-in temperature sensor. See Jim’s Flickr set for more views.

### 10 Responses to Three-component SmartLEDs

1. “…and flashing as instructed by the ATTiny’s built-in temperature sensor.”

Will the author confirm that the ATTiny25 has an internal temperature sensor. I did a quick look at the datasheet and didn’t see any reference to an internal temperature sensor.

2. I found the internal temperature sensor in the datasheet. It is connected to ADC4. But, before you get excited about using one like I did, you have to use the internal 1.1vRef for the adc and the accuracy of the sensor is +/-10degC (50degF)!!!!!!!!!!

3. John Baichtal on said:

Thanks, Pete.

+/- 50 degrees? In other words, just for fun!

4. jason1729 on said:

Wow…it’s +- 10 C degrees, or 18 F degrees. Yes, still not very useful, but at least get your science right.

10degC/50degF is a specific point on the temperature scale and to add or subtract that doesn’t have any logical meaning.

• Gareth Branwyn on said:

>New Math?

>at least get your science right.

Corrections, without the need to rap somebody’s knuckles for making a mistake, are preferred. Thanks.

• Yes, it appears that my Fahrenheit conversion I did at 5:30AM was flawed. The math per se however; I would not call “new math”. 10C*1.8+32 does in fact equal 50F. I should have left out the +32 bit but, out of habit (weather station conversions) I did the *1.8+32.

For the sake of summary: The ATTiny25 does have an internal temperature sensor. It is connected to ADC4. The datasheet claims an accuracy of +/-10C (+/-18F).

Meh.

• Tod E. Kurt on said:

Atmel’s Application Note AVR122 describes a simple calibration algorithm that can give up to +/-2C accuracy from the internal temp sensor. Pretty good considering it’s basically just a freebie on the chip.

• It’s a pretty obscure freebie though because you still need to set the ADC Vref to the internal 1.1v. I think this applies to all channels. So, if its just the temp sensor you want and don’t need to measure any other ADC channels above 1.1v it could work for your project.

Also, +/- 2C is huge in my book. I’m not talking about calibration quality instrumentation but 4C is a lot even for a thermostat.

• Scooby on said:

+/-2C makes it useless as a t-stat, but the precision is plenty for its (probable) intended use- for a high temp alarm in a controller.

5. Jim Three on said:

Hey Guys.

Yes the 25/45/85 do have a temperature sensor. My understanding is that the +/-10 degree range is a tolerance difference across different chips. Once you have one chip dialled in (I have a variable in my code for a temperature offset which is specific to each chip, and I also use the simply Y=Mx+C calibration) it is nice and accurate within the scope of normal winter to summer indoor temperature fluctuations.

Jim