Do LEDs generate audible noise?

[ccaudle]

A diode would generate 2.5nA shot noise current (six times greater).

How does the spectrum compare? If the integrated noise of the diode is higher because of a 1/f rise in the low frequencies, it may not be as audible, or at least not as audibly noticeable in a typical environment that tends to have air handler noise, structure transmitted noise that is primarily low frequency, etc.

 

[ccaudle]

How does the spectrum compare? If the integrated noise of the diode is higher because of a 1/f rise in the low frequencies, it may not be as audible, or at least not as audibly noticeable in a typical environment that tends to have air handler noise, structure transmitted noise that is primarily low frequency, etc.

Not sure if it is bad form to reply to my own post, but I did a little homework and the info I found is that diode junction noise should be white.
The article linked earlier showed a measured noise voltage of 0.26uV for a red LED at 5mA. Assuming that voltage is measured over a 20kHz bandwidth that would be equivalent to the Johnson noise from a 200 Ohm resistor at room temperature. A resistor would presumably be above room temperature while sitting next to a tube, but so would a diode, so probably washes out for the most part.
So it seems that if the cathode resistor you would use is higher value than 200 Ohm using the diode would be lower noise.
That is my reading of the situation.

 

[Mr. Moscode]

Leds act like zeners, a fixed voltage drop and they are noisy like zeners, just maybe not as much. It would probably depend on the current. However they can and should be bypassed with a cap when used for cathode bias. Also blue and white leds have a higher voltage drop than red or green. Having experimented with this idea I didn't like the sound quality so I let it go.

There are some very quite and expensive resistors available from Texas Components, which is a second source for the same Vishay part used by the military.

 

[Mr. Moscode]

Light up a LED and put a scope on it.

 

[JohnRoberts]

Leds act like zeners, a fixed voltage drop and they are noisy like zeners, just maybe not as much. It would probably depend on the current. However they can and should be bypassed with a cap when used for cathode bias. Also blue and white leds have a higher voltage drop than red or green. Having experimented with this idea I didn't like the sound quality so I let it go.

all semiconductor junctions have "electrical" noise

There are some very quite and expensive resistors available from Texas Components, which is a second source for the same Vishay part used by the military.

There are different types of resistor noise. Johnson or thermal noise is a simple function of total resistance and doesn't care if it's wire wound or carbon comp. Excess noise in resistors does vary with resistor construction but is not always significant in audio paths.

JR

 

[Tubetec]

Maybe a bit more detail on how the sound was effected Mr Moscode , did you find a way to measure the difference you were hearing or was it something based on listening , In any case for me is the ear is the final arbiter .

Thanks Merlin ,
you seem to have swept out a few commonly held misconceptions like magic ,

 

[Mr. Moscode]

Maybe a bit more detail on how the sound was effected Mr Moscode , did you find a way to measure the difference you were hearing or was it something based on listening , In any case for me is the ear is the final arbiter .

Thanks Merlin ,
you seem to have swept out a few commonly held misconceptions like magic ,

Everything measured was good. Just didn't have any mojo. Didn't make me want to listen. After you get all the measurements to a working spec, it still has to sound good. You may have the world's greatest dog food but if the dog won't eat - well, there you jolly well are, aren't you?

 

[JohnRoberts]

Everything measured was good. Just didn't have any mojo. Didn't make me want to listen. After you get all the measurements to a working spec, it still has to sound good. You may have the world's greatest dog food but if the dog won't eat - well, there you jolly well are, aren't you?

Have you considered more measurements? Back in the 70s/80s I modified my old smpte IMD analyzer to use 19kHz:20kHz for two tone IMD. It was very revealing of HF linearity. When tweaking out phono preamps the RIAA playback EQ was a LPF attenuating the THD making that measurement too easy at HF.

JR

PS: I've seen dogs eat some disgusting stuff.

 

[merlin]

How does the spectrum compare?

Both Johnson and Schottky (shot) noise are white, yes.

The article linked earlier showed a measured noise voltage of 0.26uV for a red LED at 5mA.

Assuming that voltage is measured over a 20kHz bandwidth that would be equivalent to the Johnson noise from a 200 Ohm resistor at room temperature.
So it seems that if the cathode resistor you would use is higher value than 200 Ohm using the diode would be lower noise.

I haven't read the article but when it comes to noise measurement, most people get it wrong. Measuring the noise voltage can get really confusing when you're trying to compare with something like a resistor, because you have to take into account the different effects that each would have in an AC circuit, to get a fair comparison. In a tube stage a bias resistor causes degeneration, an LED doesn't, for example. Voltages can trick you, but currents tell the truth.

An LED may well produce a lower noise voltage at the cathode while still injecting a bigger noise current into the tube (because the LED internal resistance is low). This will still usually give you a worse overall noise figure, because you're injecting something like six times more noise current (say), but you're probably not getting six times more gain compared with using a resistor.

The LED does give you the option of running at higher tube current without changing the bias voltage, which a resistor can't do, of course. But that's not comparing apples with apples. Even if you have a lucky situation where a bias LED nets you a better noise figure than a resistor somehow, adding a bypass capacitor to the LED will still likely buy you a few dB improvement. But you could have added the capacitor to the bias resistor too so...

The more I learn the more I have become convinced that it's hard to beat a simple, resistor loaded, RC-biased triode stage. Every time you think you've made an improvement somewhere, you find something different has just got worse, and your grandfather is winking at you. Anyway thanks for coming to my Ted talk.

 

[Mr. Moscode]

Have you considered more measurements? Back in the 70s/80s I modified my old smpte IMD analyzer to use 19kHz:20kHz for two tone IMD. It was very revealing of HF linearity. When tweaking out phono preamps the RIAA playback EQ was a LPF attenuating the THD making that measurement too easy at HF.

JR

PS: I've seen dogs eat some disgusting stuff.

I didn't go the IMD analyzer route. It just wasn't that pleasant a presentation and didn't have a real production upside. It might have been useful to look at IMD at 3K or 5K, though. My phono pre is passive RIAA, too. No leds. The discomfort could very well have been caused by the noise of the led hetrodyning the music creating spurious overtones. I have tried this with Zeners and diodes for the voltage drop. I have used led based constant current plate loads with pnp high volt transistors in tube 350 volt voltage regulators, which are bypassed at the out put. Gets great line rejection, about 60+ db in a cascode circuit.

RE:dogs. Me too. My analogy holds, though.