MM5437 white/pink noise generator

[R.G.]

For now, I have m yhands full with "random" appearances by a squirrel that has entered our home, can't seem to find its way out (despite finding plenty of places to hide), and occasionally makes appearances atop the living-room curtain rods, staring at me like Chris Griffin's "evil monkey".

Live trap and peanut butter. I'm a veteran. Catch and release any stray kids that get in there. 

Just so that folks are clear about this, before attempting anything, these ought to be a collection of parallel 2-stage paths.  Running 8 stages in series, with each pair of successive stages controlled by a different LFO just gets you the same 4 notches moving in ganged fashion, in unpredictable patterns.  Remember that it is the sum of all phase-shift that counts to produce the notches.  What you want is 4 different notches produced, and moving around independently...at least if I've understood RG correctly.

Yes, you want four notches moving independently. But you can't really do this in parallel because of the nature of notches. If you get four separate notches in 4 parallel paths, then add them, the frequency content that's been notched out in one path will be added back by the other channels.

What you have to do is run them as separate two-stage phasers by adding dry to delayed every second stage, then feeding this completed "notch" into the next two stages. Once gone, the frequency content in a previous notch will remain gone in later stages.

[PRR]

How hiss was done in 1949:

http://downloads.bbc.co.uk/rd/pubs/reports/1949-20.pdf  {1MB PDF}

Note nifty symmetrical pink-filter.

[Lurco]

google is cheap: http://electronotes.netfirms.com/AES4.PDF

[armdnrdy]

Thanks Lurko,

I performed a search for this and couldn't find it!

I just lightly breezed through this paper.....there is some very useful information!

[Mark Hammer]

How one could read anything from Bernie Hutchins and NOT find useful information, I have no idea.

My deepest gratitude Lurco.  After going through 4 binders (out of several dozen), and not finding it, it occurred to me this morning that the article I could "see" in my head as a low-contrast photocopy was not this article, but another article from Hutchins in the same journal on making a fundamental extractor....sitting somewhere in one of those binders.  I'm glad I remembered enough for you to be able to locate it again. 

[Nasse]

If my memory does not cheat me I have seen few times pseudorandom noise generator done with few 4000 series cmos logig chips (after that NS chip disappeared from catalogs), pinking filter was quite same

[R.G.]

google is cheap: http://electronotes.netfirms.com/AES4.PDF

I get:

Server not found
Firefox can't find the server at electronotes.netfirms.com.

Is the link good?
Edit: Never mind. Tried it again and it worked OK. Must be a low-bandwidth server or something, or my internet link is fritzy again.,

[R.G.]

If my memory does not cheat me I have seen few times pseudorandom noise generator done with few 4000 series cmos logig chips

Yes. In general, to do a pseudorandom generator, you need a sequence of shift register stages and a few XOR or XNOR gates.

These setups can generate sequences of pseudorandom noise with a repeat length of (2^N)-1 states, where N is the number of shift register stages.  So for eight stages, you get 2^8, or 256, minus 1 or 255 states.

For good sounding noise, you need to put out states at a rate that's a multiple of the highest frequency you want to hear as flat noise. I've read different versions of how fast is enough, but between 4x and 10x is what I generally see.

If you want noise flat to 20kHz, you want 80kHz to 100kHz for your state-clock. If you're OK with lower-max frequency noise, you can do with less. Call it 20kHz to 50kHz. At a 20kHz rate, (i.e. one change every 50 uS) you run through 255 states in 255*50uS = 12750uS, or every 12.75mS. Ooops. That's got a strong  beat.

How many states do we really need? Well, let's say you want to have non-repeating sounding noise that is 30 seconds long. The number of stages then has to be bigger than 30/50uS= 600,000. Two to the 19th power is 524,288, and 2^20 is 1,048,576 so you need 19 or 20 stages, 20 being plenty enough and 19 being not quite enough.

The CD4021 is an eight stage shift register, so you can get 16 or 24, and so you will pick 24. There are some longer shift registers, but to make a pseudorandom shift register work you need access to some of the internal bits in the register, and these are different for different lengths, so the longer registers are not particularly useful for this.

So you need three CD4021s and a CD4070 quad XOR. Mouser sells the 4021 for $0.56 in a plastic DIP, and the CD4070 for $0.59. You'll also need a clock, which you can make from a CMOS 555 or from a CD4069 hex inverter, also about $0.50. So you need five chips for a chip total of about $3.00, plus some jellybeans in the form of Rs and Cs.

You can make a 32 to 64 bit shift register inside an 8-pin PIC like the 12F615 for $1.15, and it has its own internal oscillator, but needs a 5V zener or a 78L05 to make 5V for it.

This is so familiar to me because I recently got interested in PRSRs again and programmed up a noise generator on a PIC.