How can I build a frequency multiplier circuit for acoustic signals?

[jescriba]

I'm trying to build a frequency multiplier that will take an input signal in the acoustic range (from a guitar) and multiply its frequency by an integer value (play harmonics of the input). I was wondering how I could do this in analog. I have been looking into trying to use some sort of voltage controlled oscillator but I want to preserve the input waveform, and all the VCOs I've seen generate either sine waves or a sawtooth. Maybe I should try some sort of harmonics PLL, but I don't know anything about PLLs. Help please. And much thanks.

 

[vk6kro]

You could record the sound and play it back at a faster speed. This would retain the shape of the waveform but increase the frequency of all the components.

I've seen voice recorder chips that may be able to do this.

You have probably seen demostrations where they record a dog barking and then play it back by hitting keys on a keyboard to get different pitch dogs barking.

 

[schip666!]

If what you want is an "exact" replica of the input signal played at a higher frequency you are probably out'o'luck as that requires some kind of memory. Long ago in the pre-digital-everything world there was a company that made a pitch shifter using a CCD array for memory -- I still have one of their chips in my aged synth -- but it was really only good for creating garbage artifacts.

To get a frequency multiple you might try a phase-lock-loop like the CD4046, but that will only give you a square wave output.

To just add harmonics to your signal you can clip it with a couple diodes and add an adjustable filter to taste.

 

[mheslep]

I'm trying to build a frequency multiplier that will take an input signal in the acoustic range (from a guitar) and multiply its frequency by an integer value (play harmonics of the input). I was wondering how I could do this in analog. I have been looking into trying to use some sort of voltage controlled oscillator but I want to preserve the input waveform, and all the VCOs I've seen generate either sine waves or a sawtooth. Maybe I should try some sort of harmonics PLL, but I don't know anything about PLLs. Help please. And much thanks.

I don't know about a random integer but you can double it with a http://en.wikipedia.org/wiki/Frequency_mixer" thus: sin(2u) = 2sin(u) cos(u)

 

[schip666!]

I don't know about a random integer but you can double it with a http://en.wikipedia.org/wiki/Frequency_mixer" thus: sin(2u) = 2sin(u) cos(u)

I can't claim to understand what is being described in that web-page, but I think it's basically a multiplier which will frequency modulate one signal by another. This will give you two side-bands centered around the modulating frequency -- this kinda makes sense if you are making an FM transmitter where the mod frequency is way higher than the signal, but makes for all kinds of interesting NON-harmonic stuff happening if the two are close to the same frequency. We, of the old-timey-electronic-music crowd, used to call this a Ring Modulator.

The key idea here is that the results don't retain their harmonic relationships, it doesn't multiply 2x in the frequency domain: 100Hz to 200Hz & 200Hz to 400Hz. IIRC it's more like 100 to 200 & 200 to 300...

 

[mheslep]

I can't claim to understand what is being described in that web-page, but I think it's basically a multiplier which will frequency modulate one signal by another. This will give you two side-bands centered around the modulating frequency -- this kinda makes sense if you are making an FM transmitter where the mod frequency is way higher than the signal, but makes for all kinds of interesting NON-harmonic stuff happening if the two are close to the same frequency. We, of the old-timey-electronic-music crowd, used to call this a Ring Modulator.

The key idea here is that the results don't retain their harmonic relationships, it doesn't multiply 2x in the frequency domain: 100Hz to 200Hz & 200Hz to 400Hz. IIRC it's more like 100 to 200 & 200 to 300...

Let me clarify. Yes if the input signal is multiplied by a fixed frequency we'd have a band shift as you describe. However, in the single case where the input signal sin u1 + sin u2 + sin u3 + ... is multiplied by itself phase shifted 90deg to cos u1 + cos u2 + ... then yes you will see a doubling of all the frequency components, just as the trigonometric identity predicts. This can be repeated to 4u, 8u, etc, picking up more noise at every step.

 

[jescriba]

Thanks for all the responses. I feared that I would not be able to preserve the waveform by a simple analog circuit. Does the distortion of the harmonics from the originally signal produce a very flat undesirable tone? How about just making a integer multiplier with a VCO that puts out a standard waveform like a square wave at a frequency dependent on the original signal's frequency (is there a way to tweak a VCO to depend on the frequency of the input rather than the voltage say a FCO?). Since the square wave has more harmonics than a sine wave for instance would the tone sound preferable to do that than a sine wave since I can't preserve the waveform simply?

 

[akc185]

I'm trying to build a frequency multiplier that will take an input signal in the acoustic range (from a guitar) and multiply its frequency by an integer value (play harmonics of the input). I was wondering how I could do this in analog. I have been looking into trying to use some sort of voltage controlled oscillator but I want to preserve the input waveform, and all the VCOs I've seen generate either sine waves or a sawtooth. Maybe I should try some sort of harmonics PLL, but I don't know anything about PLLs. Help please. And much thanks.

i personally think that all the discussion going around here is out of the question ask.
Can you tell us why you want frequency multiplier? What is your PROBLEM STATEMENT?

Because there are many circuits which will retain the shape of the waveform but will increase its frequency. e.g., modulators, tuned circuit, we can also use active filters with some extra hardware connected to it.

 

[jescriba]

i personally think that all the discussion going around here is out of the question ask.
Can you tell us why you want frequency multiplier? What is your PROBLEM STATEMENT?

Because there are many circuits which will retain the shape of the waveform but will increase its frequency. e.g., modulators, tuned circuit, we can also use active filters with some extra hardware connected to it.

Sorry if I didn't explain myself well enough. Essentially, I want a guitar effect pedal that is a harmonizer. Which is to say, a circuit that takes the guitar signal and produces its harmonics (integer multiplies/octaves) to merge with the original signal to give a chord of octaves for the note coming from the guitar.

 

[schip666!]

Let me clarify. Yes if the input signal is multiplied by a fixed frequency we'd have a band shift as you describe. However, in the single case where the input signal sin u1 + sin u2 + sin u3 + ... is multiplied by itself phase shifted 90deg to cos u1 + cos u2 + ... then yes you will see a doubling of all the frequency components, just as the trigonometric identity predicts. This can be repeated to 4u, 8u, etc, picking up more noise at every step.

Huh...it's been many years since, but I don't remember trying to mod a signal with itself. It has to be phase-shifted to work correctly? That might pose problems with complicated waveforms... Could be worth a try, no? Maybe try the AD534 multiplier chip...

I seem to remember there being "Harmonizer" guitar boxes. Perhaps this is their trick?

 

[mheslep]

Huh...it's been many years since, but I don't remember trying to mod a signal with itself. It has to be phase-shifted to work correctly? That might pose problems with complicated waveforms... Could be worth a try, no? Maybe try the AD534 multiplier chip...

Just a capacitor as differentiator gives you cos from sin.

 

[schip666!]

Just a capacitor as differentiator gives you cos from sin.

Isn't the phase shift frequency dependent, such that the higher harmonics would be shifted differently?

 

[jescriba]

Let me clarify. Yes if the input signal is multiplied by a fixed frequency we'd have a band shift as you describe. However, in the single case where the input signal sin u1 + sin u2 + sin u3 + ... is multiplied by itself phase shifted 90deg to cos u1 + cos u2 + ... then yes you will see a doubling of all the frequency components, just as the trigonometric identity predicts. This can be repeated to 4u, 8u, etc, picking up more noise at every step.

This should work. I can multiply my frequency by even integers by using the identity sin(2x)=2sinxcosx. So put the input signal (sine) through a differentiator and then through a multiplier (giving sinxcosx) and then through an op amp with designed gain of 2 giving 2sinxcosx=sin(2x). Now, I'm wondering how well would this work when the input is a guitar note which is not a perfect sine wave.

 

[mheslep]

Let me clarify. Yes if the input signal is multiplied by a fixed frequency we'd have a band shift as you describe. However, in the single case where the input signal sin u1 + sin u2 + sin u3 + ... is multiplied by itself phase shifted 90deg to cos u1 + cos u2 + ... then yes you will see a doubling of all the frequency components, just as the trigonometric identity predicts. This can be repeated to 4u, 8u, etc, picking up more noise at every step.

This should work. I can multiply my frequency by even integers by using the identity sin(2x)=2sinxcosx. So put the input signal (sine) through a differentiator and then through a multiplier (giving sinxcosx) and then through an op amp with designed gain of 2 giving 2sinxcosx=sin(2x). Now, I'm wondering how well would this work when the input is a guitar note which is not a perfect sine wave.

Arg, sorry this may not be acceptable. You're going to get a lot of other tones you don't want. That is, if the signal is sin u1 + sin u2 + sin u3 and you multiply it by its cosine self, yes you will get sin 2xu1 + sin 2xu2 + sin 2xu3, but you will also get a lot of other frequencies, something like sin u1+u2, sin u1+u3, sin u2+u3, etc, all likely inband so you won't be able filter them out.

One thing is sure about this. Frequency doubling is a non-linear operation, so it can't be done with a simple linear electronic network of any kind.

 

[akc185]

Guys, you are forgeting the delay shift occurs because of device you will use for mod or differentiate. There are ICs available for frequency domain analysis. I used it once but i don't remember the IC number now. :(
Try google

 

[JimmyAlz]

A simple solution (to double the frequency) would be to use a precision rectifier.
Wikipedia: http://en.wikipedia.org/wiki/Precision_rectifier
Just feed it through a capacitor so the rectifier has something to work with
(the 'simple' circuit in the wikipedia page will also need a say 100k resistor to ground from IN+)
You will get a lot of distortion, but that might not be so bad depending on your preferences.
Using two rectifiers in series will produce 4x the input frequency.