# Accounting for temperature changes in circuit design

In my last post about a simple circuit:

+12V -- R1 -- LED1 -- LED2 -- LED3 -- ground

A user said:

...you should account for temperature changes and variation between parts. Not all parts (even of the same part number) will have the same Vf, even at the same temeprature, and your circuit design needs to be robust enough to account for that, not carefully tuned to the exact nominal behavior of a part type.

Practically speaking, using my simple circuit as an example, how does one account for temperature changes and variation between parts? Would simply using a 150Ω resistor in series with the three LEDs not work?

Edit: The circuit will be operating in temperatures of around -10°C to +150°C

• Are you sure those LEDs are going to be operational at 150 degrees C? Check their maximum operational temperature in their datasheet... – stefandz Aug 13 '15 at 12:56
• Okay, maybe 150°C is exaggerating a little realistically. The circuit is inside a car headlamp assembly so will get quite hot (not sure how hot exactly yet). The LED datasheet says operational up to 150°C though, which is why I chose that number as the upper limit for the example. – parrowdice Aug 13 '15 at 13:05
• Is this for a high-powered LED headlamp? – stefandz Aug 13 '15 at 13:45
• @stefandz It will be for Angel-Eyes / Halos – parrowdice Aug 13 '15 at 14:27

The forward voltage of the LEDs will drop significantly with a temperature increase of 125°C, perhaps 0.7~0.8 volt for a string of 3 (check the datasheet for the particular part). That will mean that the current will increase by maybe 25% at high temperature. Since the allowable current probably drops to zero at 150°C Ta this is the opposite of what is desirable (will tend to kill the LED even more readily at high temperatures). On the other hand, 25% will not be all that visible a change in light output.

Bottom line is that temperature is not too important to visible light output if you have at least 1 volt per series LED across the resistor and temperature change isn't insane. Light output also changes with die temperature, so the current isn't the only factor.

You can typically ignore the resistor change in temperature - even a crummy resistor will not change more than 100ppm/K which is 1% for 100°C change.

• Thank you. So if I understand, it will be better to just use a slightly higher resistor value so that when the temperature increases, the effects will be within the desired limits still? – parrowdice Aug 13 '15 at 13:42
• Yes, you should stay within the limits, whether that's a slight increase or something else has to be calculated. – Spehro Pefhany Aug 13 '15 at 14:37

As so often: it depends...

Your circuit is NOT robust if the sum of the forward voltages of the diodes is close to the supply voltage.

Two examples:

1. Eg. 3 blue LEDs:
$V_f$=3.7V, sum is 11.1V. Current is (12V - 11.1V) / R = 0.9V / R.
An increase of $V_f$ by only 0.1V would cause a current of (12V - 11.4V) / R = 0.6V / R.
I.e. a decrease of 33%. That's a lot.
2. Eg. 3 red LEDs:
$V_f$=1.8V, sum is 5.4V. Current is (12V - 5.4V) / R = 4.6V / R.
An increase of $V_f$ by 0.1V would cause a current of (12V - 5.7V) / R = 4.3V / R.
I.e. a decrease of only 6.5%. That's not so much.

(note everythin in degrees celsius)

I searched for LED Vf over temperature, and some can get 30% reduction of forward voltage over $100^\circ C$ 5,56 V instead of 7,98 V (2,66V nom. Vf) gives 26 mA for normal temp and 43 mA for 100 degrees rise. That is worst case. (assuming no temp change or constant temp in resistor). Resistors have tempcos in the order of 3000ppm/C and less (many have 250ppm/C) so say in the case of a resistor with 250ppm, the resistor is going to increase with 3,75 ohms over $100^\circ C$, basically nothing compared to the Vf.

So for me 100% difference equals no regulation at all. I would use a thermally compensated current mirror for best regulation