next topic »

[on:]

Hi all. I have been trying to make a temperature sensor from a diode and my arduino uno. I have done a lot of googling and have got no where. I have tried two different approaches. The first approach with no prior research involves one diode and a resistor. The diode's cathode is connected to the arduino's ground and the diode's anode is connected to a resistor which connects to an arduino digital pin. I then have an arduino analog pin connect to the node between the resistor and the diode to read the voltage. So the equation for temperature is:
T = [ q Vd ] / [ k ln( (Id+Is)/Is ) ]
Where
Id = (Vs + Vd)/R
q = 1.61 x 10^-19
k = 1.38 x 10^-23
Vd is the voltage measured from the arduino analog pin
and lastly Is is the reverse saturation current.

current "Is" is my issue since itself depends on temperature which is what I am trying to solve for. That is where I got stuck on that approach (is the equation a transcendental equation?).

After some research, I found valuable:
page 338 of "The Art of Electronics 2nd ed." and
http://en.wikipedia.org/wiki/Silicon_bandgap_temperature_sensor

For the art of electronics page 338, it shows circuits using transistors, is it possible to NOT use transistors for this project (op amp ok though)?

also looking over some forums online I read that an op amp is needed, is that true?

For my second approach I made two (almost) identical circuits with different resistors. The circuits resemble the one described above. Each circuit has it's own Vd connected to a different arduino analog pin. The resistors differ for each in order to have different currents flow through the diodes. I plan to use the second equation in the wikipedia link above, although I have a feeling that my second approach is super wrong.

Please help me out here. At least direct me where to go. Thank you.

[Reply #1 on:]

The common 1N4148 diode can be used as temperature sensor (but almost every diode will do).
The resistor can be connected to +5V.
For the 1N4148, I used 10k.
Rule of thumb is: 2mV / C    (2mV less voltage drop for every degree increase of temperature)
   0.4 V at +100 degrees Celcius
   0.6 V at 0 degrees Celcius
   0.8 V at -100 degrees Celcius
But you have to test for yourself to determine a more precise calculation. I think that 2 degrees accuracy is possible.

You don't need an Op Amp. The analog input of the Arduino is high impedance, it will never influence your circuit. The variation of the voltage drop is also large enough for a good accuracy. But you have to use the internal reference of 1.1V
http://arduino.cc/it/Reference/AnalogReference
Here 3k3 is used : http://www.openimpulse.com/blog/2012/07/low-cost-diode-temperature-sensor/#more-261

[Reply #2 on:]

http://www.openimpulse.com/blog/2012/07/low-cost-diode-temperature-sensor/#more-261
"Step" voltage arduino depends on REFERENCE ADC voltage, to measure 2 mV w/o OPA you should switch reference to internal 1.1V (UNO or ArMega328 board)

[Reply #3 on:]

Thank you so much for your help and thanks for the links, all the time spent on google and I was not able to find that link. So it seems my initial hardware connection is correct (so happy!). It also seems I do not need to use the Diode equation (so my software part is wrong). I don't quite understand this "internal reference" thing you guys talk about. So say if i change the internal reference to 1.1V, would that mean that analog max voltage would be 1.1v? If so, why must I do that? Why can I not keep it at 5v and use a higher resistance for my resistor? And lastly, in the link you guys gave me, it says the current has to be arround 1mA, why is that?

Also Krodal said:

The resistor can be connected to +5V.

So, once again, why not keep the internal reference at 5v instead of 1.1v?

Magician said:

to measure 2 mV w/o OPA you should switch reference to internal 1.1V

What does OPA stand for (op amp?) ? If OPA = op amp then why would I have to use an op amp for 5v and not for 1.1v?

Thanks again.

[Reply #4 on:]

So say if i change the internal reference to 1.1V, would that mean that analog max voltage would be 1.1v? If so, why must I do that?

Yes, max input voltage would be 1.1V . I don't think you need more with a Si diode. Lowering reference you increasing ADC resolution. For example, with default 5V reference, one "count" ADC is equal to 5 / 1023  = 4.88 mV. Which is high to measure 2 mV raising voltage per degree of temperature.  In other words, ADC would report next value only when changes in temperature about +-2.5 degree. Usually,  amplifier (OPA) required to overcome this issue. 

Why can I not keep it at 5v and use a higher resistance for my resistor?

Higher resistor value produce lower voltage across a diode.

And lastly, in the link you guys gave me, it says the current has to be arround 1mA, why is that?

I think it compromise value, from one side (lower current end) there is a leakage:

For instance, many monolithic temperature sensors have an output of only 1 μA/ ̊K. This leads to a
1 ̊K error for only 1 μ-Ampere of leakage current

  "Application Note 460 LM34/LM35 Precision Monolithic Temperature Sensors"
on the other side - self heating, P = V x I.

[Reply #5 on:]

Some basic explanation:

The analog-to-digital converter (ADC) inside the microcontroller converts a signal from 0...5V to a value of 0...1023. To use the ADC, the function analogRead() is used.

The ADC is however capable of more than just that. It uses a voltage reference for the conversion. Normally that voltage reference is +5V. But it can be set to +1.1V.

With a voltage reference of 1.1V, the ADC converts a signal an analog input pin from 0...1.1V to a value of 0...1023. That's roughly 1mV resolution. So the accuracy increases. Nice!
http://arduino.cc/it/Reference/AnalogReference

This is all happening inside the chip.
The voltage reference is connected to the outside world with a pin. On your Arduino Board, you see "AREF", that's the reference.
If you switch to another voltage reference, it is best to wait for 20ms, to let the voltages settle.

Code:

void setup()
{
  analogReference (INTERNAL);   // 1.1V reference
  delay(20);                                // let voltages settle
}

void loop()
{
  int rawADC;
  rawADC = analogRead(some_pin);   // 0...1.1V -> 0...1023
}

[Reply #6 on:]

Thank you so much. That makes perfect sense. I understand! I will give it a try now (can't wait for it to work, i hope).

[Reply #7 on:]

Sorry to break in, but I read these posts this morning and made a diode temperature sensor just now.  I had no idea you could do this!  I don't know the number of the diode I have, but it works great.  I'm watching the temperature drop a couple of degrees F when the air conditioner comes on, then slowly creep back up.  Thank you all for the information!

(edit - added my code)

Code:

// Diode Temperature Sensor, diode plus 10K resistor
// Written by Tom Fangrow, August 11, 2012

int hotVal=86;      // value with boiling water, 100 degrees C
int coldVal=256;    // value with ice water, 0 degrees C
float slope;

void setup() {
  slope=(coldVal-hotVal)/100.0;     // value per degree C
  analogReference(INTERNAL);        // 1.1 Volt reference
  Serial.begin(9600);
}

void loop() {
  int val = analogRead(0);
  float c = (coldVal-val)/slope;    // Celsius
  float f = 32+c*9/5;               // Fahrenheit
  Serial.println(f);
  delay(1000);
}

[Reply #8 on:]

Good stuff JavaMan. I would like to report to Krodal and Magician that I have got my sensor up and running, thank you guys.

[Reply #9 on:]

For consistent results ensure the resistor is 1% or better (lower temp-co), the 5V is regulated and be careful to avoid errors due to self-heating - use low currents.  With the dual pin arrangement you can apply power to the diode for just long enough to take a ADC reading then power it down till the next time - very little power dissipation in the diode then to self-heat it.

Note that different individual diodes will have different calibrations.

[Reply #10 on:]

For consistent results ensure the resistor is 1% or better (lower temp-co), the 5V is regulated and be careful to avoid errors due to self-heating - use low currents.  With the dual pin arrangement you can apply power to the diode for just long enough to take a ADC reading then power it down till the next time - very little power dissipation in the diode then to self-heat it.

Note that different individual diodes will have different calibrations.

Thanks. Yes, I did my own temperature calibration. I haven't considered powering on and off.