LEDs as a low noise shunt regulator?

[EE54]

Hi everyone, recently I came across this Reddit post: https://www.reddit.com/r/electronics/comments/ro9h5b/voltage_regulator_from_1_led/
In it, someone in the comments claims that an LED offers better noise performance ( but worse regulation) than a zener diode when used as a shunt voltage regulator.

How true is this?
Has anyone tested this before?

[RoGeorge]

I didn't measured that, but it could be true.

Zener diodes are reverse polarized, working in some sort of a controlled breakdown, similar with a continuous avalanche, while a LED is direct polarized, so no avalanche noise from the LED.

[SeanB]

They do make reasonable voltage references, but have a high slope resistance, and are very light sensitive as well. But if you want a simple low voltage reference that is better than 2 or 3 silicon diodes in forward bias, they do work well enough. In non critical applications they do work well enough, but do not expect them to replace a reference voltage IC at all, though they are rather better than low voltage zener diodes.

I have used them as voltage references, but varying the current from 100uA to 1mA increased forward voltage drop by around 0.2V for a red 5mm LED, and going to 10mA was about the same again. If exposed to ambient light they will have additional noise from the light hitting them as well, making them useful as a light sensor on a MCU pin, as both ambient light sensor and power indicator. Just turn off every minute for 2ms, and sample the pin voltage with the ADC with a weak pull down, or 10k parallel resistor, then turn it back to an output and turn it high, so the high output drives the LED again, via a 1k series resistor.

[2N3055]

Google this : "walt jung led regulator"

[EE54]

Thanks, that was quite an interesting read. Never knew you could use LED for that.

[Gyro]

They do make reasonable voltage references, but have a high slope resistance, and are very light sensitive as well....

I remember from one of Morgan Jones's books that good old fashioned Red LEDs (the glow-worm at 20mA type) have a lower slope resistance than a 1N4148 at 10mA (4.3R @ 1.7Vf vs 6R @ 0.75Vf) IR comes close but only has 1.2Vf. I keep a few in my drawer for low noise bias purposes. Other colours and all high brightness types have way higher slope resistances though.

Just 'random' information that lingers at the back of my mind.

[mawyatt]

@2N3055

That's a very clever concept by Walt

Best,

[2N3055]

@2N3055

That's a very clever concept by Walt

Best,

Mike,

I agree, over the years Walt had quite few interesting ideas..

Best,
Sinisa

[langwadt]

They do make reasonable voltage references, but have a high slope resistance, and are very light sensitive as well....

I remember from one of Morgan Jones's books that good old fashioned Red LEDs (the glow-worm at 20mA type) have a lower slope resistance than a 1N4148 at 10mA (4.3R @ 1.7Vf vs 6R @ 0.75Vf) IR comes close but only has 1.2Vf. I keep a few in my drawer for low noise bias purposes. Other colours and all high brightness types have way higher slope resistances though.

Just 'random' information that lingers at the back of my mind.

some other "random" related info, https://www.dropbox.com/s/d4ntmq7fdzah69a/LED_Isrc_data.JPG

[mawyatt]

@2N3055

That's a very clever concept by Walt

Best,

Mike,

I agree, over the years Walt had quite few interesting ideas..

Best,
Sinisa

Yes Walt did develop many interesting ideas, he did lots of great work, but wasn't as well known as Bob Pease and Jim Williams. He sent me a note long ago (think he was still at Analog Devices) and told me my Current Source had been published in Current Sources & Voltage References by Harrison (I had no idea it was published outside EDN), also later asked if he could use the Equal Valued 3rd Order Butterworth active filter he had seen in EDN in a book he was writing.

Best,

[RoGeorge]

He sent me a note long ago (think he was still at Analog Devices) and told me my Current Source had been published in Current Sources & Voltage References by Harrison (I had no idea it was published outside EDN), also later asked if he could use the Equal Valued 3rd Order Butterworth active filter he had seen in EDN in a book he was writing.

You mean this EDN article, about Peaking Current Source?
https://www.edn.com/peaking-current-source-has-high-rejection/

Interesting circuit I didn't know before, thank you.
Wow, and better stabilization than the Widlar Current Source?!   

Wyatt Current Source vs. Widlar Current Source   
First round score:  1 - 0   

[mawyatt]

Thanks, yes that's it 

The story behind this current source is in late 80s to early 90s we were developing the first fully integrated MicroWave Receiver in a new silicon Bell Labs process (CBIC V2) and based upon a unique "Synchronous Homodyne" architecture, which is a variation of what was later called Direct Downconversion or Zero IF and culminated in patent 5603111.

During the course of the development a requirement arose which demanded operation from a single +5VDC source which had significant noise, both low and high frequency content (from a noisy switcher) and a size requirement which wouldn't allow any decoupling capacitors or other external components other than 1 resistor and capacitor (used for offset correction)!! With this in mind we set out to develop on-chip regulators which had good LF and HF rejection, and also a curvarture-corrected bandgap reference. The regulators and bandgap reference required various current sources and we needed very high overall PSRR, so we came up with this Peaking Current Source Reference. The bandgap design with this current source got some interest and was published within a Bell Labs Journal/Paper before later we got permission to release to EDN.

Classic example of a stiff requirement being the "Mother of Invention" 

Best,

[TimFox]

The Morgan Jones references to using "cheap" LEDs were originally for cathode bias voltage in vacuum-tube preamplifiers, to avoid large bypass capacitors.  In that application, Merlin Blencowe (in "Designing High-Fidelity Tube Preamps, 2016, pp 210-211) points out that in forward-biased diodes (rectifiers or LEDs), the noise current riding on the DC value is given by shot noise calculations, but the noise voltage depends on that noise current and the slope resistance, which is why IR or red LEDs (with lower forward voltage and slope) are often preferred.

[David Hess]

Using an older LED with -2mV/C temperature coefficient to bias a single transistor current source/sink makes a low noise temperature compensated current source/sink.

The transistor Walt Jung recommended is a Zetex "super e-line" type with very low base spreading resistance which results in low noise.  Others make these now for their high gain and low saturation voltage but they are not always easy to identify.

I now wonder if Zetex named those transistors "super e-line" because they were exposed using the 546 nanometer e-line.  The way I remember it, they bought obsolete photo-lithography equipment originally used for ICs and made transistors with it, so they could support a much finer feature size.