how do you patch a frequency shifter [Modmix]?

[narwhal]

Hey folks,

I was just doing some (totally unnecessary) modular window shopping when i came across something that perplexed me. I was reading up on the Hinton Modmix, which states,

"The Modulators provide classic RM effects and may be combined with the inputs for AM, or used with external sinewaves and filter for frequency shifting."

http://www.hinton-instruments.co.uk/pap ... /index.htm

I read up on a few threads to try and get my head round this but it's still eluding me. I understand that Freqshifting is very similar to ringmoding but read it uses a quadrature sine and I'm guessing the sum/difference outputs are key for the filtering? but could someone explain how to patch a frequency shifter using a Modmix (or basic modules generally)?

thanks


edit.

ah this helped. trusty nord.
http://www.cim.mcgill.ca/~clark/nordmod ... shift.html

[Navs]

It's tricky, but doable:

http://navsmodularlab.blogspot.de/2012/ ... bouts.html

trusty nord.

Better than Strange

[worker8]

Long story short, you can turn any dual ring modulator into a frequency shifter as long as you use quadrature VCOs. Both the carrier and the modulator should be quadrature VCOs.

Just plug one RM as usual, and the second one with 90° shifted waves. Output of both RM into a polarizing mixer. You'll ear up or down shifting depending on the settings of the mixer.

[narwhal]

cheers for your help guys!

great blog post navs, thanks for the heads up.

[Graham Hinton]

Frequency shifting is a technique as old as radio and that's over 100 years now. For detailed theory look up Heterodyne, Modulation and Single Sideband Modulation on Wikipedia.

To create sidebands with ModMix you need the sine and cosine of both a signal and an oscillator. If you have a sin/cos oscillator that makes it slightly easier otherwise two 90 degree phase shift networks are required. (A cosine is just a sinewave 90 degrees ahead.)

Patch these four sources into ModMix like this:



Turn down all the input levels apart from 4 and 8 so that only each Modulator output comes through to the A and B outputs. Then the sum and difference outputs have the difference and sum frequencies, respectively.

Each Modulator multiplies its two input frequencies, f1 and f2, to produce both sum (f1+f2) and difference (f1-f2) frequencies as multipliers do by the usual trig identities. However note the subtle difference of sign of the terms and when these are combined either one or the other are cancelled out resulting in a single sideband.

The trick for frequency shifting is to upshift the signal with one oscillator, separate the sidebands (usually by filtering one out) and then to downshift the frequency with a second oscillator and separate the sidebands again. The frequency shift is the difference between the frequencies of the two oscillators. In radio applications the transmitter is doing the upshift and the receiver the downshift by the same amount and restoring the signal. The maths is the same whether the frequency changes are 1Hz or 30kHz, but the bandwidth of your modules may not be. Not many synthesizer filter modules operate up in the IF range and the advantage of the above technique is that it requires no filtering, but the A/B levels may need a little tweak to obtain the best cancellation.

If you have a suitable filter then the upshift may be done by Modulator A, the sum frequencies filtered out and the downshift down by Modulator B. Phase shift networks and quadrature oscillators are not required for this method, but two filters are. Note that the lower sideband has the original spectrum reversed, but this is mirrored again when downshifted.

Here are some waveforms and spectra of a ModMix Modulator multiplying two low distortion sinewaves at 400Hz and 1000Hz producing sidebands at 600Hz and 1400Hz. The original signals are mixed in at a lower amplitude just so they can be seen together, they are not present in the Modulator output.



The lower pictures show one oscillator moved to 1200Hz and the sidebands (caused by the 400Hz) move with it. In this case the sidebands are still in the audio frequency range so a 24dB/oct low pass filter set at about 800Hz would remove about 70dB of the upper sideband. If one input is a complex signal there would be two sidebands for every harmonic present in it, if both inputs are complex signals there will be every combination of the harmonics of both of them.

[SubliminalSandwich]

You learn a new thing every day..



right, im off to mess with some sine waves..

[narwhal]

Many thanks for taking the time Graham, that's a fantastically comprehensive answer.

The Modmix certainly looks a very useful module to have in your toolkit, the sum/difference outputs in particular intrigued me. I liked the sound of this-

"if you use A-B to control the frequency of a HP filter patched into a LP filter with A+B controlling its frequency then any signal into inputs 1 to 4 controls the filters in tandem and any signal into 5 to 8 widens or narrows the frequency pass band. If 5 was an envelope and 6 an LFO you could widen the difference with the envelope, the LFO or a burst of the envelope X the LFO using Modulator B normalised back through input 8."

[macs4music]

[quote="Navs"]It's tricky, but doable:

http://navsmodularlab.blogspot.de/2012/ ... bouts.html

[quote="narwhal"] trusty nord.[/quote]

Better than Strange :tu:[/quote]


Apologies for resurrection of this thread but I'd rather that than create a new one. I am trying out this patch using the A-143-9 as my modulator and the A-125 phase shifter to give me a 90 shifted signal. Am I on the right track? Certainly the results are sounding similar to the effect in logic and live. What settings are key to the traditional effect?

Neil

[mskala]

In order to really get a frequency shift with the "Bode" approach, you need quadrature signals; that is, signals split into two parts where every frequency component differs by 90 degrees between the two part. A quadrature sine oscillator will give you that, but a non-sine quadrature oscillator (such as an A-110-6) won't, and neither will an A-125, which gives different phase shifts at different frequencies. It's critical that the phase shift has to be 90 degrees on every frequency component; not just 90 degrees on the overall waveform (which would be 90 degrees on the fundamental frequency but other amounts on the harmonics).

If you want to frequency shift an arbitrary signal instead of a sine wave using the Bode technique, you need something called a Hilbert transformer, which normally consists of two fixed-frequency all-pass filters set up so that their phase responses differ by 90 degrees across a wide range of frequencies. Complete "frequency shifter" modules have one of these built in; note it is not a "transformer" in the usual sense of magnetically coupled coils, but it's called a transformer because it does the mathematical operation called a "Hilbert transform." Signal to be modulated goes into the Hilbert transformer, it comes out as a quadrature signal (two channels, differing by 90 degrees at every audio frequency), then you combine it with a quadrature sine wave from an A-143-9 or similar and away you go.

Nonetheless, even if it doesn't give you a clean frequency shift, you can surely get some interesting sounds by combining the A-125's output with ring modulation.

[mskala]

In order to really get a frequency shift with the "Bode" approach, you need quadrature signals; that is, signals split into two parts where every frequency component differs by 90 degrees between the two part. A quadrature sine oscillator will give you that, but a non-sine quadrature oscillator (such as an A-110-6) won't, and neither will an A-125, which gives different phase shifts at different frequencies. It's critical that the phase shift has to be 90 degrees on every frequency component; not just 90 degrees on the overall waveform (which would be 90 degrees on the fundamental frequency but other amounts on the harmonics).

If you want to frequency shift an arbitrary signal instead of a sine wave using the Bode technique, you need something called a Hilbert transformer, which normally consists of two fixed-frequency all-pass filters set up so that their phase responses differ by 90 degrees across a wide range of frequencies. Complete "frequency shifter" modules have one of these built in; note it is not a "transformer" in the usual sense of magnetically coupled coils, but it's called a transformer because it does the mathematical operation called a "Hilbery transform." Signal to be modulated goes into the Hilbert transformer, it comes out as a quadrature signal (two channels, differing by 90 degrees at every audio frequency), then you combine it with a quadrature sine wave from an A-143-9 or similar and away you go.

Nonetheless, even if it doesn't give you a clean frequency shift, you can surely get some interesting sounds by combining the A-125's output with ring modulation.

EDIT: Sorry, I misread the thread, the stuff from 2013 above is not talking about the Bode technique but the Weaver technique, where you use a high intermediate frequency, filters to separate the sidebands, and no Hilbert transformer. Nonetheless it sounds like macs4music is trying for the Bode technique, and my comments are applicable to that.

[macs4music]

I was trying to make use of Nav's instructions, but it was the 90 degree shifting of the source signal that I was t sure about. I had thought the phase shifter might not work as it is fixed. Oh well it was worth a try and yes it does produce interesting results.

Thanks,

Neil

[Graham Hinton]

You can find out more about 90 degree phase shift networks here.

For a complex signal every harmonic needs to be shifted 90 degrees. This is accomplished by two chains of staggered fixed frequency all pass filters such that the two outputs differ by 90 degrees across the audio range, but the phase shift relative to the input signal will vary.

There is a tradeoff between the number of stages and accuracy and noise. Obviously a dual 12 stage network is a going to be a card full of electronics so you won't find a large number of stages built in to a dedicated module.
The way to test these networks is to plot the outputs on an oscilloscope in XY mode and sweep the frequency of a sine wave input. The display should show a circle.

[macs4music]

Thanks Graham, that was the missing link. I was trying to work out if the phase shifter would 'track' a signal but this makes more sense.

Neil

[Llouwelyn]

In the patch Frequency shifting, I use VCO-2RM and Modmix Hinton.
- https://soundcloud.com/cflmodular/fsq1