eprom to generate pink noise

I think it should.

Pink noise is low-pass filtered white noise. White noise is noise with equal amount of noise at all frequencies.

I recommend an active filter, to have a higher frequency selectivity. The order of the filter depends on the bit rate. The frequency response of the filter should be such that 20 kHz passes by with little attenuation (less than 3 dB) and your switching frequency passes with as much attenuation as possible (at least 40 to 60 dB). First you choose your bit rate (according to your noise shaping scheme) and then the filter appropriate for that bit rate. It would be best if you could run some simulations (Matlab, Simulink, Simplorer...) before building anything.

BTW, do you sleep well? :-) I say this because I tried pink noise a couple of nights, long time ago, to mask noise from my neighbors (human and not human), but couldn't stand it. However, _if_ I could get used to it, I think it would help. I used a CD, though.

Best,

That was provided that the 1-bit modulator can run at the very high frequency required to generate baseband signals which are free of noise up to 20 kHz. This will probably limit you, because you may end up needing some GHz.

Also, ask yourself if you really need 20 kHz, or whether 10 kHz is enough.

I haven't tried listening to pink noise that has suffered an MP3 compression. MP3 was not optimized for that. If it sounds sufficiently well for you, use a dirty cheap MP3 player (it will not need a high SNR :-).

Pink noise is white noise that has been mildly low pass filtered. You can take advantage of this in the design of the EPROM playback system. You didn't say how low you wanted to go with the bottom end of the frequency responce. Just for fun, lets assume you want to go down to 20Hz. Lets also assume that you want to use only a single RC filter.

I'd place the RC pole at:

F(rc) = sqrt(20 * 20K) = 632Hz

Assuming R on the order of 10K, I get C=0.025uF. If this is surface mount, it would be time to stop and re-think because that is too big for a good NPO capacitor.

Now comes a question you need to answer. How accurate do you want the top end of the frequency responce curve? There is an extra noise in the system from quantization. This noise has an RMS value of about

To make the math simple, lets assume that this noise is uniformly spread up the Nyquist. You need to move Nyquist up until the amplitude of this added noise is low enough that it doesn't mess up your amplitude curve at the top end.

Yes, white noise has a constant amplitude over frequency. Pink noise is white noise passed through a 1/f filter.

Very interesting! But how do I figure out what the bitstream needs to be and what bitrate I need to use?

Get ahold of Don Lancaster's "TTL Cookbook" and build a PSRG. :-) or do a google search:

Cheers! Rich

You have to follow that with a "pink filter" to make the white noise into pink noise. I explained how to make the contents of a PROM and a single RC work for the job. The extra complexity of the PROM saves quite a few parts on the PCB.

Another way to do it may be to use chaos to fill in for the randomness. There is a circuit that uses all 4 parts of a quad op-amp like the LM324 to make chaos. I wonder if anyone has done a spectrum on its output.

On Mon, 20 Mar 2006 21:35:50 GMT in sci.electronics.design, Rich Grise wrote,

I like the one in the "CMOS cookbook".

That will depend on what kind of D/A converter you use and your sample rate.

I think the OP is long gone. He was hoping to use a single RC filter on a "one bit ADC". Since he is only trying to make pink noise, it is likely this would work. I think he found he was in over his head.