How could I get a 'Random' Vinyl Crackle Sound?

[~arph]

Fantastic! I thought it wouldn't work when I thought of using the SHO originally but that sounds perfect especially if ~arph experiences simmilar results with his experiment.
Cheers Guys.

I tried it, it works. I used a 4093 with four schmitt trigger oscillators to get a pattern. Sounds really noisy and scratchy, so definately usable.

[birt]

just use a led candle (cheap!) circuit and replace the filckering led with an optocoupler. instant randomness to modulate the SHO

[~arph]

Lol,   I'm working on an LFO modulated grainy noise section of the noise cornucopia from page 1.  Thought I'd do a forum search first..  Look here. looks like I experimented with a modulated SHO over four years ago!  

BTW, The LED candle trick will not work as usually the controller is inside the LED, but it does produce a pseudo random voltage (I believe I've seen it used at evil mad science) , or it can be used with an LDR.

[R.G.]

I've done a little looking at noise.

By far the simplest noise generator is the reverse-broken base-emitter of a normal transistor feeding an amplifier. There are many schemos of that. However, the noise is not consistent from device to device, leading to many magic mojo discussions of the best noise device, and all the mess that entails.

But that doesn't get you the noise you want. As noted, a vinyl record has a few types of noise. One is best modeled as thermal noise, the kind of hiss you get from high gain noisy amplifiers. The other, as noted earlier, is randomly timed impulse noise from dust and such stuck in the grooves. You have to model both to get a good simulation of vinyl record noise. This is a version of burst noise or flicker noise, and is different from thermal hiss.

The thermal noise is easy. I would do it with and 8-pin uC these days. The notes about humans hearing a pattern in noise are accurate. Digital circuits cannot do real random noise, they do pseudo-random noise, and the pattern of bits repeats after some time. The more stages of shift registers, the longer the repeat time. It has to be long enough for humans to "forget" the pattern for people not to hear an annoying repetition.

National Semi did a chip for digital noise. It produced good-enough-for-measurements noise, but the digital shift register length was only about 1-2 seconds. Humans clearly hear it repeat.That was before uC's. I recently did a noise generator program for an 8-pin PIC that used the internal registers for very long patterns. If I did the calculations right, the pattern repeats every few thousand years, so the humans can't remember a pattern, all of them being dead before it could repeat. Not bad for a $0.75 chip.

The randomly timed, randomly sized impulses from dust are harder. This has the same problem as hiss in that you need the pseudorandom sequence to be long enough so humans don't remember the repetition rate, it has to be doubly random - timing and amplitude - and you have to generate pulses, not hiss. More programming is the simpler conceptual answer.  Run the "hiss" program, look for specific values of bits in the sequence, and make a "tick" when you hit those bits, which will be randomly distributed in time. Then from that timing info, generate a random height pulse from another pseudorandom process. This could be the sequence of bits before or after the trigger for the pulse.

I'm guessing that although both hiss and ticks could be done in one uC, it might be simpler to use two. Fortunately, this isn't hard.

[~arph]

Thanks for the insight R.G.  There already are some uC based vinyl simulator pedals out there. I am just doing this for fun/excercise and it does not have to resemble actual record noise per se. I haven't breadboarded anything yet ( apart from the SHO crackling four years ago apparently ). I threw a schematic together what would be my first attempt on the breadboard. It has three types of noise mixed in with the original signal.  White noise and grainy noise from the noise cornucopia (the grainy noise really sounds promising based on the samples at the MFOS site), plus now that I rediscovered this thread, scratchy noise from mosfet biasing. I have added an LFO so we can have some repetition of noise intensity at the speed of the record. Here it is:

(my apologies for the huge image, but I wanted it to be readable)

[Mark Hammer]

I think one needs to also consider what vinyl crackle noise IS, since that may direct what one does to mimic it.

It is impulse noise.  That is, it is not simply intermittent bursts of gated white noise.  It is something throwing your cartridge stylus way off for an instant.  So what one really wants is something like a 555-timer one-shot, that is triggered intermittently for varying durations.  One pulse could be 20usec, and another could be 400usec.

Actually, recordings of single brain cells firing sound very much like vinyl surface noise...except without the underlying music.  So do Geiger counters.  And both are registering pulses.

[~arph]

Well I am going to try it just for fun, as I said it does not have to be exactly like a record. i'll consider a neural input too  

[stonerbox]

Hi!

When I stumbled upon the demo vid of HEXE Melusine II this thread came to mind. Not too sure if the schematic is available online though.