[Revised 2023-11-17]

A couple years or so ago I built a Kosmo version of the MFOS Noise Cornucopia . Now I’m designing a module that needs white noise so I thought I’d just drop in the Noise Cornucopia source. I breadboarded it:

Section of schematic diagram for Noise Cornucopia Rev. 1.

Noise source from Noise Cornucopia Rev. 1

And… it didn’t work. I mean it did generate noise, but it also output a faint tone, up very very roughly around 1 kHz I’d say. Almost a whistle. I’d never heard that from the module I built.

Figured it had to be a mistake or bad part or something, couldn’t find it, so I built a second copy… and it had the same problem. Huh. Maybe a design flaw after all.

So I decided to try some other design and the first to come to mind was the same MFOS module in its Rev 0. incarnation . Wilson later modified the noise source design and called that Rev. 1. Rev. 0 was simpler:

Section of schematic diagram for Noise Cornucopia Rev. 0. It is a little simpler than the Rev. 1.

Noise source from Noise Cornucopia Rev. 0

In Rev. 1 Wilson changed the transistor emitter connection to the +12 V rail, adding a capacitor to filter the rail voltage, and changed the base connection: Rev. 0 went direct to ground, Rev. 1 went via a resistor to -12 V, to improve the symmetry of the noise signal. But Wilson also said “the changes provide a negligible improvement “.

Anyway… Rev. 0 worked. Noise, no tone.

Then I found this post . With the Rev. 1 source they were getting a much lower tone, 50 to 200 Hz, but otherwise it was similar to what I was hearing, and the solution seems to have been to connect the transistor base to ground rather than to -12 V. And in the Rev. 0 source the base is connected to ground. Aha.

The next morning I tried the Rev. 1 source again. The tone was gone! Mystery! But the noise signal still wasn’t good. Looking at it on the scope I could see it glitching in coincidence with a 1 Hz trigger signal on the breadboard circuit — a signal the noise source isn’t supposed to see.

Eventually — it can take me a couple weeks to see the bleeding obvious but I do get there, sometimes — it occurred to me I had an oscillator in the breadboard circuit. I turned up the oscillator frequency and the tone came back. I’d just had the frequency so low it was masked by the noise.

Meanwhile someone on the LMNC Discourse noted the curious fact in the Rev 1 source Wilson put a filter on the +12 V rail but not the -12 V. Hmm.

On the +12 V rail the 470k resistor and 1 µF capacitor form a low pass filter with a cutoff of 0.3 Hz. The resistor to the -12 V rail is 47 times smaller, so I got a capacitor 47 times larger, 47 µF, and added it to the Rev. 1 source from the transistor base to ground, making a filter with the same cutoff frequency. Turned the power on and… the tone and the 1 Hz glitch both were gone. Took the capacitor out, the tone and glitch came back. (My former boss used to say it’s not good enough to show you’ve gotten rid of a problem, you also have to show you can bring it back.) Put the capacitor back in, the tone and glitch went away. Problem solved!

Rev. 1 noise source schematic altered to include 47 µF capacitor from transistor base to ground.

Rev. 1 noise source with added filter capacitor

And in fact, the Rev. 1 improvements may not be dramatic but they’re real. So that, with the added capacitor, is the design I’m going with.