EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: Seekonk on September 01, 2020, 09:35:22 pm
-
I've used LED for years as a convenient way to offset voltage and provide status at the same time. A while back there a really good thread about just how stable they were with temperature. I've searched and searched a variety of terms and have been unable to locate it. Did anyone bookmark this? I'm looking to compare the TC of LED vs a standard diode.
-
Somewhere I have seen a table of -millivolt/C temperature coefficients for different LED processes. Some of them are a good match for the -2 millivolt/C of a forward biased silicon diode or base-emitter junction making them a convenient way to build temperature compensated current sources.
-
In his book:
https://books.google.ca/books/about/Current_Sources_and_Voltage_References.html?id=03JmxpE39N4C&printsec=frontcover&source=kp_read_button&redir_esc=y#v=onepage&q&f=true (https://books.google.ca/books/about/Current_Sources_and_Voltage_References.html?id=03JmxpE39N4C&printsec=frontcover&source=kp_read_button&redir_esc=y#v=onepage&q&f=true)
page 118 Linden T. Harrison gives some numbers
-
In his book:
https://books.google.ca/books/about/Current_Sources_and_Voltage_References.html?id=03JmxpE39N4C&printsec=frontcover&source=kp_read_button&redir_esc=y#v=onepage&q&f=true (https://books.google.ca/books/about/Current_Sources_and_Voltage_References.html?id=03JmxpE39N4C&printsec=frontcover&source=kp_read_button&redir_esc=y#v=onepage&q&f=true)
page 118 Linden T. Harrison gives some numbers
Harrison gives the forward voltage drops based on "color" but not the temperature coefficients or materials which is what I remember, and ultimately more important. But in practice, I found that LEDs should be qualified by measuring the temperature coefficient in applications where it would matter, which is easy enough to do. I found too much variation between models and manufacturers made of nominally the same material.
-
Materials is a major issue, even before you add in varying choice of dopants and all the possible substrates. Even Wikipedia knows that many of the common semiconductors used for LEDs have variable composition (start from the materials links in the table: https://en.wikipedia.org/wiki/Light-emitting_diode_physics#Materials, and hover over any ternary or quaternary compound semiconductor). I'm not surprised that Vf temperature coefficients are highly variable between suppliers.
-
Some years ago, I have made some quick & dirty measurements
with heat gun & thermometer
I checked a few more for noise:
< https://www.flickr.com/photos/137684711@N07/24354944411/in/album-72157662535945536/ (https://www.flickr.com/photos/137684711@N07/24354944411/in/album-72157662535945536/) >
cheers, Gerhard
-
Interesting! Inspired by this I tried to measure the voltage across a LED at 5 different temperatures. The LED in question was a Hewlett Packard HLMP-DL24 (amber), material = AlInGaP.
The LED was driven at 1 mA by a Keithley 224 current source, and the voltage was measured by a Fluke 8842A. The LED was located in a VWR Peltier incubator oven, while the temperature was measured by a Pt100 sensor connected to a thermometer AOIP PN 5207.
Results:
Avg. temp. (C) / Avg. LED voltage (V)
17.447 1.80562
25.884 1.78860
31.684 1.77672
36.363 1.76703
41.698 1.75588
The voltage and temperature numbers are averages over a period of 10 minutes once the temperature had settled.
The figure shows the resulting voltage at the measured temperatures.
[attach=1]
The resulting voltage temperature coefficient for this LED turns out to be -2.05 mV/K. This is close to the temperature coefficient of a transistor's Vbe path, so this LED could be used as a "zener" bias voltage for a simple constant current generator. The question is, however, to what extent the LED's forward voltage and temperature coefficient vary within a batch, and from batch to batch...
It would be interesting to see if anyone has made similar measurements.
When I got time, I will try to measure a couple of other LEDs.
BR, Ole
-
The question is, however, to what extent the LED's forward voltage and temperature coefficient vary within a batch, and from batch to batch...
The foward voltage varies quite a bit, but the temperature coefficients are much more consistent.
-
Careful, exposing the LED to bright light temporarily throws out your reference voltage.
The big + for LEDs unlike Zener is if you want to use a ~20 megaohm pullup, most small voltage Zeners will sit at just above 0v due to leakage current, but not LEDs. The old fashioned green ones are my pick, feeding the gate of a depletion mode mosfet/J-fet to create a low power 0v dropout linear regulator with a near un-measurable quiescent current.
-
In his book:
https://books.google.ca/books/about/Current_Sources_and_Voltage_References.html?id=03JmxpE39N4C&printsec=frontcover&source=kp_read_button&redir_esc=y#v=onepage&q&f=true (https://books.google.ca/books/about/Current_Sources_and_Voltage_References.html?id=03JmxpE39N4C&printsec=frontcover&source=kp_read_button&redir_esc=y#v=onepage&q&f=true)
page 118 Linden T. Harrison gives some numbers
That's an excellent reference book on Current and Voltage sources. There's a current source in 4.5 (pg 87-89) that I've used often and particularly fond of :)
https://doc.xdevs.com/doc/_Metrology/Print%20Ver.%20-%20Current_Sources_and_V%20_Refs.pdf
It's in a number of old chips I've been involved with.
Best,
-
Don't you need to paint them black so they don't work as solar cells?
-
Careful, exposing the LED to bright light temporarily throws out your reference voltage.
The big + for LEDs unlike Zener is if you want to use a ~20 megaohm pullup, most small voltage Zeners will sit at just above 0v due to leakage current, but not LEDs. The old fashioned green ones are my pick, feeding the gate of a depletion mode mosfet/J-fet to create a low power 0v dropout linear regulator with a near un-measurable quiescent current.
LEDs also have a lower slope resistance than low voltage (true) Zeners. Old fashioned low brightness Red ones are particularly good.
-
Don't you need to paint them black so they don't work as solar cells?
In an application where the leakage matters, yes. The same applies to glass encapsulated diodes.
-
I can't seem find the link, but at an audio forum (pro audio, the people that push many K watts into large stacks, type of pro), someone that was an engineer for well known maker of of gear talked about using a LED for bias. Not sure if it was in an amp or elsewhere in in the chain. Cheaper than the other ways of doing the job. One day purchasing talked to him about saving costs on that LED and was going to sub in a different LED that was claimed to be just as bright and great on other specs. He had to stop them because the drop on the new LED was completely different. I assume the the newer vendor didn't look at how the LED was being used, just that they wanted the sales.
Bottom line it has made its way into commercial products. Sorry I can't find more and that is as close as I remember it.
-
I recall a data sheet or App note from National Semi or LT where a couple LEDs were used with series resistors in an Op-Amp Vcc and Vee lines to bias a PNP and NPN transistors respectively which were connected with their collectors to the Op Amp output. As the load current increased the PNP would help supply additional sourcing current to the load directly from Vcc and the NPN would assist in sinking additional load current directly to Vee. This additional circuitry increased load current handing capability of the Op Amp. I've never used this clever concept tho, so can't say anything regarding how well it works.
Best,
-
The original Microchip ICD programmer/debugger used a LED for the reference voltage for its nom. 13V Vpp programming voltage supply. See pages 63,64 for the schematic [here] (http://ww1.microchip.com/downloads/en/devicedoc/51184d.pdf), and this Piclist article (http://www.piclist.com/Techref/piclist/freeicd/index.htm) and comments, for how fussy and sensitive to Vf people cloning it found it to be. |O
Later Microchip programmers and debuggers *DID* *NOT* use LED voltage references . . . :horse:
-
I recall a data sheet or App note from National Semi or LT where a couple LEDs were used with series resistors in an Op-Amp Vcc and Vee lines to bias a PNP and NPN transistors respectively which were connected with their collectors to the Op Amp output. As the load current increased the PNP would help supply additional sourcing current to the load directly from Vcc and the NPN would assist in sinking additional load current directly to Vee. This additional circuitry increased load current handing capability of the Op Amp. I've never used this clever concept tho, so can't say anything regarding how well it works.
I have used that configuration many times in various forms and it goes back to at least 1978 where Tektronix used it in their 7A29 vertical oscilloscope amplifier, as shown below. Instead of using LEDs or a resistor, I prefer current mirrors.
Update: I also remember seeing that configuration in Fairchild application notes from the early 1970s so engineers were well aware of it which makes sense as it is a variation on cathode feedback.
-
Just finalized a measurement of another LED, a cyan Avago HLMP-CE26-WZ000, material = InGaN.
Results:
Avg. temp. (C) / Avg. LED voltage (V)
20,463 2,76333
25,888 2,74719
31,672 2,73049
36,405 2,71664
41,711 2,70043
Again, the voltage and temperature numbers are averages over a period of 10 minutes once the temperature had settled.
The resulting voltage temperature coefficient for this LED turns out to be -2.95 mV/K.