[sdiy] freq shift allp filter ?

[Magnus Danielson]

From: Fernando de Izuzquiza <fdi at ran.es>
Subject: [sdiy] freq shift allp filter ?
Date: Mon, 13 Jan 2003 20:52:07 +0100

> Hi all, and happy new year.

Dear Fernando,

> I'd like to build a frequency shifter (phase type) made out of various
> modules (quadrature osc, 90deg allpass filter and two ring mods), better than
> a large circuit. I'm not looking for linearity or precision, just one more
> musical tool within my modular syntheziser.
> 
> I have the two ring modulators (Roman Sowa design, AD633 based) and I see
> there are some quadrature oscilators on the net (some simpler than others).
> But I have many questions about the filter...
> At first, I thought that using a cell from a phaser would just yield the 90
> degrees shifted signal, but I readed that it will not shift the same amount
> for all freqs... Then I understood that what I need is a short delay to shift
> the signal 1/4 of it's period, so a delay capable of very short delay times
> was the solution... but, as far as I can imagine, this will only work for one
> frequency...

So far, you've gathered the right facts.

> So, dear masters, what kind of "90deg constant phase shift filter" is needed
> for this kind of freq shifting?
> And, very important (I'm not an EE), there is any circuit ready to build out
> there?
> 
> (for those not familiar (ough! am I?) the circuit I'm talking about: input
> signal to filter / filter 0deg out to RMa / filter 90deg out to RMb / QOSC
> sin out to RMa / QOSC cos out to RMb / RMa and RMb out are the up and down
> shifted signals, QOSC freq is the shift amount)

The majority (all?) of the frequency shifters (that I know of) do not use a
filter to generate the 90 degree signal out of the 0 degree signal, but rather
uses two filter sections (really a chain of all-pass filters) to generate two
phase slopes, such that the phase difference between the output is about 90
degrees throughout the signal bandwidth. This filter is often known as the
Dome filter. You could either look at Jürgen Haibles FS-1:

http://www.oldcrows.net/~jhaible/tonline_stuff/hj_fs.html

or the Moog/Bode Frequency Shifter Model 1630:

http://home.swipnet.se/cfmd/synths/companies/moog/

(I know the schematics are momentarilly offline, but I will look into that).

I strongly recommend you to look through Jürgen's frequency shifter project.

Anyway, what you do is achieving the same overall goal, but taking different
routes to the goal. The result as delivered to the multipliers (a.k.a. mixers
or ringmodulators, but I tend NOT to use the later term) is the same as the
theory requires.

Interestingly enought, over at another list, in another languague frequency
shifter theory is just now one of the ongoing topics, and I've also scribbled
down some writing on it. Sometimes the world has this strange synchronism in
it.

PS. The 4-quadrant multipliers we use today, all some variant on Gilbert cell,
is however not ringmodulators, they just strive to acheive the same goal as the
originial ringmodulators did, namely 4-quadrant multiplications. A
ringmodulator you build out of 4 or 8 diodes hooked up in a ring, that's when
you've got yourself a "ring modulator", else it's a multiplier or more
specifically a 4-quadrant multiplier. In radio and especially suitable in a
frequency shifter, which is really nothing buth Single Side Band (SSB) radio,
the use of mixer is appropriate (you mix frequencies and get new frequencies
out), but that would confuse the audio-useage (where you mix signals).

Cheers,
Magnus