Sub Audio Noise Source and Multiplier for Random Modulation

How far have you got with it so far? Or do you want us to do it all? :)

Do it all? Nice of you to dangle the bait ;-)

I have built a few noise generators but cannot get the LP filter to work as desired.

Most of the available multiplier chips seem to be split rail rather than the single supply 12VDC I was hoping for.

If someone could provide a general design concept including IC numbers, I could take it a few steps further.

I occasionally use LTspice, but I suspect this circuit would be tricky to simulate.

Robert Martin

A 30 Hz bandwidth lowpass is possible, but 'only frequencies below' isn't (brick-wall filters take infinite amounts of hardware). As for modulation, you can use the filtered noise to modulate the gain of a (for instance) LM13700, it's crude but effective

Rather than filtering real noise, find an audio test track with a few seconds of noise and FFT-filter it, then play back with MP3 hardware in a continuous loop. Software filtering uses digits for parts, millions of 'em are cheap and never suffer from bad solder joints.

whit3rd wrote

Would it not be much simpler to just play an existing noise track at lower speed? Playback sample it slower.

MP3 player, but WAV or similar non-lossy format. Lossy formats will ruin your noise!

--
mikko

Mixing a 30-Hz noise signal with a 10-Hz carrier won't produce anything pretty. It certainly won't be Gaussian.

I'd probably use a shift-register noise source and a bandpass filter, and dump the multipliers. As long as the shift register clock is (say)

100 times the filter bandwidth, the central limit theorem says that the amplitude statistics will be Gaussian to excellent accuracy.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

Hmmm, some questions. What's the low frequency (less than 1 Hz.?) You want Gaussian (describes amplitude) and/ or white (frequency) noise? I've used zeners as white noise sources, but they are not Gaussian.

George H.

Thank you for your question. For this application, either white or Gaussian would be acceptable. The only purpose is to introduce a degree of randomness to the 10Hz signal.

Robert Martin

A few people have kindly suggested software but I am looking for a hardware solution.

I have used audio editing software for similar applications in the lab, but, from my experience, computer sound cards filter sub audio at the output.

Robert Martin

'White' and 'gaussian' are orthogonal concepts. 'Gaussian' refers to the amplitude statistics, 'white' to the spectrum. The raw pseudorandom sequence is white (up to half the clock frequency) but nowhere near gaussian. Filtered pseudorandom sequence is gaussian due to the central limit theorem, but its spectrum is that of the filter.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

Here is a shift-register schematic.

I have looked at these before, but the component count is higher than a breakdown type noise source. Is there any particular reason why you are suggesting the former? I would be willing to use white noise for the sake of simplicity.

Returning to your design suggestion above. As previously stated the desired bandwidth for the noise is 1-30Hz.

If I understand correctly, you are saying no multiplier would be needed if I run the clock (LM555) at about 15KHz. I am not sure what that output would look like, or if it satisfies my intent.

I have also never successfully produced a bandpass filter for such a small frequency range.

The purpose of the multiplier as described in my OP is to introduce random phase, amplitude and frequency modulations. In other words to make the 10Hz signal as chaotic as possible.

Your reply was relevant, but would you be willing to revise it based upon this additional information?

Robert Martin

Thank you for clarifying that.

At this stage, for the sake of choosing the least complicated option, I would be willing to consider either amplitude _or_ frequency randomness.

Robert Martin

I should correct that--the PRBS noise has a sinc rolloff due to the first-order hold. If it were composed of narrow impulses it would be white up to f_clock/2.

'Twere it I, I'd probably follow the PRBS with a 1-Hz RC highpass (160 ms time constant) cascaded with a 10-Hz lowpass. The good thing about that is you get very consistent performance that can be calculated from first principles pretty easily.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

Although it is perhaps possible to bridge - or increase the sound card's coupling capacitors, the cards hardware may not support slow sampling.

So I was thinking about some external DAC. 'How many bits?' is a question one should ask. I did that sub-audio (so to speak) playback with a PIC and PWM:

So pulse width modulation and then lowpassed, almost zero is the limit. In this case it is only 8 bits (IIRC), but more can easily be done in PWM. You can also just use an USB to serial converter, some PIC (micro processor) and drive a 16 bits audio DAC, send the data from a wave file on the PC at any speed you want.

You can even do that with an R2R network on the PC parport, so only resistors needed. For those who complain PC parports are extinct, I have a few dollar parport card in the PC, very useful to test SPI and i2c chips. Probably if you use the parport 8 data bits and borrow a few of the control bits so you can get to 10, 11, maybe 12 bits.

Anyways it all depends how experienced you are in designing and building 'tronics.

A white noise file made with 48 kHz sampling will play a l o n g time at a fraction of the speed. Seems a fun project. Some more audio related software that may show you how to process wave files:

There is a lot more on the site, also some using PC parport I/O.

Come to think of it, 8 bits parport, parport control line to select high / low byte,

8 bits buffer buffer to store high byte, gives you drive for a 16 bits audio DAC.

USB PIC with 16 bits software PWM followed by (very)low pass?

Hey, I will leave some for others... Don't want to design it 4 you and take the fun away. Have some other complicated stuff on the desk that I need to get working...

PS Use a raspberry Pi, and hang a 16 bit audio DAC on GPIO. Play [white?] noise wave file at low speed via GPIO.

Obvious cheap solution, flexible. I am not sure if / how raspberry pi audio out is AC coupled, but should be easy to fix, not sure about the Linux driver either, never looked at it, but if you can fix that, then no extra DAC is needed.

Have the sound card output a high frequency sub-carrier (say 5-8kHz) that is modulated with the LF noise spectrum. That will easily pass thru ac coupling and can then be rectified and low pass filtered.

piglet

NO! Recall that there's an oversampling/digital filter in the usual CD playback circuit, and it makes aliased trash at higher frequencies than audio. Slowing the clock rate could put that trash into the audio.

Yeah, there's wasted capability in the use of an audio player, but it's a mass-produced solution, not worth the time to re-optimize.

So, use a hardware playback; APR9600 holds a minute of nonvolatile sound recording, and I'm sure there are other options.

I am getting a lot of helpful suggestions for using a PC sound card, which I will file away for future use.

However, as previously mentioned, I am looking for a hardware solution for this present application. I need to make a few of these circuits as part of a stand-alone test device.

Is Jim Westcott still active on this list? I think I need some analog design pointers.

Robert Martin