Lots of noise

My next few Kosmodrome modules will be ones based on Music From Outer Space designs, because the boards are available from SynthCube and SynthCube had a 20% off sale Thanksgiving weekend, so I bought them then. It’s been a long time since Thanksgiving, I’ve not been doing much building, but the first one’s done now.

The first one probably should have been the VCO, but instead I got curious about a potential mod to the Noise Cornucopia and breadboarded the mod; then I breadboarded the rest of the module and, well, figured I might as well just go ahead and build it.

The Noise Cornucopia has no inputs and six outputs: White noise, high pass white noise, low pass white noise, grainy noise, high pass grainy noise, and random gates. All this starts with a white noise generator based around a reverse biased transistor with its collector cut off. The filtered white noises are what it says on the tin, white noise run through a fixed frequency filter. Grainy noise is produced by running white noise through a pair of comparators whose summed output is low, zero, or high depending on if the input is below, within, or above a pair of thresholds symmetric around zero. A “graininess” pot adjusts the thresholds. High pass grainy noise, you guessed it, grainy through a filter.

Random gates is not a noise source at all. The white noise goes into a comparator after a low pass filter. The output clocks a CD4024 binary counter, and one of the output pins connects to the random gate output. Again there’s a pot to control the comparator threshold, hence the gate frequency.

Bergman’s mod

As I mentioned in my earlier post, Eddy Bergman made a noise module based on the MFOS Noise Cornucopia, and he wrote:

I also changed the transistor used to generate the noise and I changed the way the transistor is integrated in the circuit. My way is simpler and generates 200mV worth of noise right at the emitter of the BC547. … You don’t need to choose a BC547 that gives you the best noise results. Any BC547 will work fine. That’s an other reason why I changed the transistor arrangement from the original 2N3904 with capacitor. This works much better.

For comparison here is the relevant section of the MFOS schematic:

And here is the corresponding section of Bergman’s version:

The MFOS circuit is Revision 1, about which Wilson wrote:

I added a bit of filtering to the noise transistor voltage supply, The voltage across the noise transistor is now the difference between V+ and V- whereas it used to be +V and ground. … I added a resistor between the noise transistor’s base and ground [sic] to improve the noise signal’s symmetry.

There’s a link that is supposed to be to the older version but in fact it just links back to the newer version. However, the older version can be found on the Internet Archive. Here is the noise source from that:

It turns out Bergman’s “improved” version is exactly the same topology as Wilson’s “pre-improvement” version! The only differences are the type of transistor and the size of the resistor. So what really improves on what?

(Incidentally, Wilson also writes: “The original design works very well … The changes provide a negligible improvement and while they can be fairly easily kluged into the previous PC board I don’t recommend going to the trouble.”)

I did some breadboard tests. I could not verify Bergman’s claim, “Any BC547 will work fine.” Maybe I was doing something wrong, but I had three BC547 available and tried two of them, and couldn’t get good noise with either. On the other hand, three out of four 2N3904 transistors worked, but they needed a fairly large resistance to +12 V. I decided to just keep Wilson’s (improved) version for my build.


I built the rest of the module on breadboard and it all seemed to work fine. I did feel the filtered grainy noise lost too much of its grainy character so I increased R23 to 10k.

The PCB provides pads for jumpers to connect any of the CD4024 outputs Q1 through Q7 to the random gates output. You’re supposed to choose one — Wilson preferred Q5. I briefly, like for five seconds, considered a rotary switch on the panel, but no, I chose a single output: Q4. I liked it better than Q5, which I felt wasn’t random enough. (The higher the number of pulses before getting a change in the output state, the less random spread there is in the timing.)

Module build

I went ahead and built the module. The front panel in Kosmo format is my design, available from GitHub.

And… it didn’t exactly work. The white noise seemed all right and I was able to set its amplitude to about ±5 V using the trim pot. The filtered outputs looked very low on the scope, but they sounded reasonable, so maybe? The grainy noise… well, I wasn’t sure how that was supposed to sound, but I was getting an output anyway. But the random gates output would power up either off or on and then never change.

On investigation it seemed the noise after filtering was very low amplitude, so that even at the minimum threshold setting the comparator was never firing. I tried cranking up the white noise amplitude with the trim pot and at a high enough setting I started getting random gates, but the white noise now was being heavily clipped by the op amp. Odd. It worked on the breadboard!

I ended up breadboarding it again, just the noise source and the filter and comparator, and it worked. The noise after the filter was much larger amplitude than on the PCB even though before filtering it was the same. Looking at both on the scope it started to appear the pre-filter noise on the PCB was more jagged, as if it had less low frequency than the noise on the breadboard.

Then finally I looked closely and discovered C1 and C2 were 1 nF instead of the 100 nF they were supposed to be! That meant they were passing only much higher frequencies than was intended, and then the low pass filter before the comparator left hardly anything standing. No wonder it didn’t fire.

Somehow I guess I misread the values in the build. Worse yet, I went to replace C1 and C2 with 100 nF film caps and discovered I had none. I’d used ceramics on the breadboard. The BOM said ceramics but I wanted to use film in the build for all but the bypass caps. Somewhere I missed having to convert the BOM’s 8 ceramic 100 nFs to 2 film and 6 ceramic. Well, it was time for a Tayda order anyway.

Tayda order came in, I soldered in the right caps and re-adjusted the noise amplitude, and it all looks to be working! The module had earned its knobs so I put them on, took a couple glamour shots, and put it into Kosmodrome.

One thought on “Lots of noise

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s