Here’s a 1N5240 10V zener diode hooked up with a 100K resistor and 47uF cap like a simple P48 mic. The noise is about equivalent to a ribbon mic’s output level. High frequency rolls off more than I had hoped for use as an ultra- wideband white noise generator. But it’s clear why zeners need to be filtered.

Blue - 10V zener noise spectrum, simple P48, 100K & 47uF

Yellow - green LED, same circuit

Green - same circuit 1K resistor

Red - 1K resistor, P48 off

For comparison, I replaced the zener with a green LED. Zener is of course reverse biased, avalanche conducting. LED is forward biased, majority carrier conduction. I also measured the noise of a 1K resistor in place of the diodes for a baseline on the mic preamp of the interface used.

Conclusions:

LED noise falls off with frequency much faster. (Green LED, pink noise?)

Forward conducting diodes are not silent. Quieter than zeners, tho.

The circuit wasn’t well shielded, just stuck inside a tin can. My guess that the dynamic impedance of the diodes was around 1K was way off. The diodes impedance was much lower, so hum canceled much better than for the 1K resistor.

The Scarlett 2i2 interface’s phantom power circuit has some inaudible LF noise.

I hoped the simple circuit would make a white noise source flat to around 1MHz, but it rolls off more quickly than that. We need something fancier.