LED as its own temperature sensor

[kender]

led temperature sensor

Folks,

Have anyone used the forward voltage of the LED to estimate LED's own temperature?

Here's my application. The light from the LED is used periodically for spectral measurement*. The spectrum of the LED itself changes with temperature**. I would like to monitor the temperature after LED is turned on and make measurements when it stabilizes. The LED has a constant current driver, so the current into the LED is known.

I'm sure somebody thought about this. Do you think it can work?

- Nick

* Somewhat crude measurement, compared to other light sources.
** An article on relationship between LED spectrum and temperature https://www.jstage.jst.go.jp/article/jlve/32/3/32_288/_article

 

[Dmitron]

temperature sensing circuit with 9 led

What's the problem?

 

[rad20]

temperature measurement +forward voltage +led

Hi Kender,
I think you are trying to use an LED as a stable light source for monitoring the state of some equipments. Yes, it may work by the method you mentioned. However you have to ensure the environmental temperature is also controlled over time if measurements are made at different time or day. Keep room temp the same every time you use the LED. The LED will eventually come to equilibrium with room temp after a while. This settling time depends on the type of LED used. Also make sure the electrical connection to LED for Vf measurement is configured using Kelvin (4-wire) connection to compensate for voltage drop across wire length.

I hope this help to add some light to your idea.

Regards,
rad20

 

[FvM]

led stabilization forward voltage patent

The method can be used for temperature control and compensation of known spectral temperature dependency as well, I think. As far as I see, LED forward voltage vs. temperature is typically unspecified by manufacturers. It's not guaranteed, that the sum of junction forward voltage (NTC) and bulk resistance voltage drop (typically PTC) results in an unequivocal characteristic at all operation currents, but I guess you get overall NTC behaviour with most types.

 

[kender]

temperature sensor with led

Hm, this idea didn't get shot down. Here's a schematic for the circuit. http://www.prolifictec.com/filehost/LED_driver_rev01.pdf. I'll post measurements as soon as I make them them.

- Nick

 

[jonconnellnyc]

led use junction temperature

This is in fact the method that LED manufacturers use to predict lamp life in a real world lighting fixture application.
The Vf being a known function of junction temperature and LED lumen depreciation being a known function of junction temperature you can use the Vf / Temperature curve to get to the lamplife.

If you hunt around the Philips Lumileds web site you will see them describe the process. I don't think they give you access to the exact mathmatical functions they use though - presumably proprietary information - but you can see the graphs they use.
There are multiple patent filings and a lot of design work from the LED driver ic people integrating micro controllers with LED drivers which give fedback on exactly this subject. You will see graphs of frequency versus junction temperature on the Lumiled and Nichia webs sites.

Regards,
Jon Connell, IESNA
www.productswithpassion.net

 

[kender]

Re: temperature sensor with led

Hm, this idea didn't get shot down. Here's a schematic for the circuit. http://www.prolifictec.com/filehost/LED_driver_rev01.pdf. I'll post measurements as soon as I make them them.

I've built this circuit, but it has a principal flaw. The changes in LED forward voltage in response to changes in current are tiny (tens of microvolts) compared to the rest of the forward voltage (units of volts). The general approach could still work though - with AC coupling. Something similar is done in IC temperature sensors (e.g. Figures 2 and 3 here: http://www.analog.com/library/analogDialogue/archives/33-04/temperatures/temperatures.pdf ).

- Nick

 

[FvM]

The changes in LED forward voltage in response to changes in current are tiny (tens of microvolts) compared to the rest of the forward voltage (units of volts).

I don't get the sense of the statement. Do you actually mean, the response to change in temperature is small?
The response to changes in current is obviously considerably with LEDs.

Regarding your idea that the well-known dual-current method can be solution. I don't expect, it can. It's used for diodes,
that have basically a rather ideal diode characteristics, but an unknown area factor. The particular problem with
LED is their deviation from ideal diode behaviour due to the large (and possibly exemplar varying) bulk resistivity.

But of course, you get more information about diode parameters when measuring at two known currents rather than one.