White noise generator

After doing some research I found a handful of ways to generate white noise:
1) amplifying high value resistor noise
2) ampligying zener noise
3) using LFSRs
4) using a micro with a DAC and program a long PRBS loop

I'm currently most interested in #3, but curious where the tipping point is for LFSR length vs clock speed for a repeating pattern to be audibly (un)noticeable? And whether it may actually be more practical to use a micro (as the BOM count starts increasing)?
Anyone have experience playing with this?

I found two interesting articles:
https://www.jmargolin.com/patents/atari3.pdf

https://www.digikey.com/en/articles/techzone/2018/mar/use-readily-available-components-generate-binary-sequences-white-noise

I seem to recall a circuit using a vacuum tube and a magnet, but perhaps the MM5837 digital noise IC would work for you?

EDIT:  I did not know what LFSR was.  Apparently that is what the MM5837 uses.  Here is a link for a better noise source, designed for synths but possibly helpful: https://www.electricdruid.net/datasheets/NoiseGen1Datasheet.pdf

Thanks Mj, I'll take a look!

Ethan said:

After doing some research I found a handful of ways to generate white noise:
1) amplifying high value resistor noise
2) ampligying zener noise
3) using LFSRs
4) using a micro with a DAC and program a long PRBS loop

Click to expand...

One classic circuit uses the base and emitter of a transistor reverse biased as a zener diode. The max reverse Vbe is usually 6V, so powering through a resistor from a source greater than 6V creates the noise.  This does damage the transistor, reducing beta, but if this is the only use you're puttng the transistor to, it doesn't matter. The bandwidth of this (and a regular zener diode) is good, but the probably density function may be kind of odd (the output may be spiky in one direction, rather than continuous).

Using a microcontroller is interesting, because the more powerful ones have a built-in "True Random Number Generator" that's good enough for cryptography use. Several of the STM32 ARM chips have this. I programmed a loop to use it and got between 900,000 and 1,000,000 32-bit numbers per second out of it.

So the Big Question is always the application: What do you want to use this for?

These guys have gone on for pages; maybe not polished yet:
https://www.diystompboxes.com/smfforum/index.php?topic=122482.0

ElectricDruid, R.G., jim, Rob, et al have thrashed this field At Length in another thread there. For many practical purposes, a 98-cent PIC will do hiss FAR better than any of the fussy old techniques. But that's some learn-curve.

IMHO a cascaded TL072 will serve many audio noise purposes. Two stages of gain=100 will bring the self-hiss up to like 20mV "RMS", flat 200Hz-20KHz. Set a 200Hz low-cut in the 2nd stage to cancel 1/f rise. Since the 1/f corner is not guaranteed, if you need !FLAT! hiss, use another technique.

Ali can be your friend for stuff like this: https://www.aliexpress.com/item/White-Noise-Signal-Generator-DIY-Kit-Electronic-Kit-2-Channel-Output-for-Burn-in-Test-Therapy/32905420870.html

..works well for synth-work and such..

Jakob E.

The application is just for a bedside noise generator. I know I can buy one for $10, but where's the fun in that? ;D

PRR's suggestion of using opamp noise is pretty nifty.

MjRippe's link led me to https://electricdruid.net/noise2-white-pink-noise-source
They've managed to implement white and pink noise with a PIC12F675, impressive!
They provide the assembly code so it'll be fun to go in there and break some stuff
Thanks for all the help guys!

PRR said:

IMHO a cascaded TL072 will serve many audio noise purposes. Two stages of gain=100 will bring the self-hiss up to like 20mV "RMS", flat 200Hz-20KHz. Set a 200Hz low-cut in the 2nd stage to cancel 1/f rise. Since the 1/f corner is not guaranteed, if you need !FLAT! hiss, use another technique.

Click to expand...

Coincidentally back in the 70s when I was buying lots of TL074s for my audio kits, I tested these 100% before sending them out to customers so I could know with confidence that they "were" good parts. My functional test was pretty simple, I sent a sine wave through all 4 stages unity gain inverting, then attached small resistors from the - input to + (ground) at each stage. I was running through 4 stages in series, each with a noise gain of 40 dB or so. Looking at the output with a scope, good parts passed the sine wave cleanly with a well behaved noise floor. Bad parts failed in different ways but all easily visible with that much noise gain. I would find low single digit bad parts from a several hundred op amp shipment. One of my better investments back then was a ZIF (zero insertion force) test socket.

By the end of the 70s TI cleaned up their process/QC enough that I could stop performing 100% barrier testing.

JR

PS: Back in the day I recall making a noise source with a CMOS ripple counter... but you could hear it cycling.  Another way to make noise is reverse zenering a bipolar transistor base-emitter junction.