[sdiy] 1-quadrant multiplier with 2164

David G Dixon dixon at mail.ubc.ca
Thu Jan 20 20:27:51 CET 2022

There are now 87 Freak Shifts out in the wild all over the world, and everyone seems to be impressed with their performance.  Plus, the price is a measly US $300, which is a fraction of what other Bode frequency shifters cost.  (Bear in mind that I build each one by hand, and each one takes me about 5 hours, so they are not mass produced.)
Hence, not to put too fine a point on it, but I would say that the Freak Shift is the analogue design to beat.


From: Todd Sines [mailto:sines_list at scale.la] 
Sent: Thursday, January 20, 2022 10:32 AM
To: synth-diy
Cc: David G Dixon; cheater cheater
Subject: Re: [sdiy] 1-quadrant multiplier with 2164

[CAUTION: Non-UBC Email]	
I’m not very dialed in with the math, but I would say that Harald’s analogue design is the one to beat, and Don’s was more of a quasi-kludged modular building block that had mixed results. The math was right but the sound was a bit different, from what I can tell.  

The 185 is basically a dual mixer that combines 2 phase shifters and a ring modulator that are hard wired as a hopped up ring modulator. 
Daniel (LA67) himself mentions that the Bode / Haible design is an “easier, flexible, and more modern" route to go.

As you can see even a smaller clone can be quite costly.
http://www.cluboftheknobs.com/pro_c1630.html [970€]

The 285 rev 2, all analogue, does all of the above but puts all of the above components accessible on the panel to utilize the sections on their own.


The 285 rev 1 design, based on a Spin FV1 DSP IC is “reasonable” according to Dave Brown
"While not as good as the original all-analog frequency shifter, the performance is reasonable and the sound qualities are quite nice."

Haible’s design is still available, somehow, for sale at Random Source.


On Jan 20, 2022, at 1:02 29PM, David G Dixon via Synth-diy <synth-diy at synth-diy.org> wrote:

I suppose that anything is possible, cheater.  However, not really with a
Bode frequency shifter, which is what the Freak Shift is.

In case y'all didn't know, the Bode frequency shifter is really just a
trigonometric engine.  It realizes the so-called angle sum and difference
identities, and this gives the frequency shifting.  It will shift the
frequencies accurately over whatever frequency range that the Dome filters
give accurate 90-degree phase shift, and will give inaccurate shifting
outside of that range.

-----Original Message-----
From: cheater cheater [mailto:cheater00social at gmail.com] 
Sent: Thursday, January 20, 2022 7:25 AM
To: David G Dixon
Cc: Neil Johnson; synth-diy
Subject: Re: [sdiy] 1-quadrant multiplier with 2164

[CAUTION: Non-UBC Email]

I wonder if it's possible to build a frequency shifter that shifts higher
harmonics more than lower harmonics.

On Wed, Jan 19, 2022 at 6:48 PM David G Dixon <dixon at mail.ubc.ca> wrote:

I must confess that I've lost the thread of this argument just a little


However, what I like about my approach (which I have used many times 
in many different contexts) is that, in order to build a nice linear 
VCA from 2164, you really need to have a clean 5V source anyway.  I 
keep a pile of LM336Z5 for just this purpose, and use two opamps to 
buffer and invert this to get low-output-impedance +5V and -5V 
references on all my multipliers.  If one uses precisely matched 
resistors on the inverter, then one can get those references within a 
mV of each other -- the actual voltage doesn't matter (and it is 
usually around 4.90V), but as long as they are equal and opposite, 
then they can be used for precise multiplication.  This is one of the keys

to the precision of my Freak Shift frequency shifter circuit.

I don't really understand how adding a stable DC value to a signal 
increases the noise of that signal.  I must confess that I also don't 
care at all about it.  My method is the simplest.  You don't have to 
pre-condition the incoming signals at all.  The CV signal is 
unchanged, and the DC reference levels are simply summed to the incoming


If you want to change the actual levels, you can simply change the 
resistor values.  I do this all the time.  One of the keys to my 
one-VCA four-quadrant-multiplier circuit (of which there are two in 
the Freak Shift, made from a single 2164 chip) is to lift and diminish 
the CV such that the zero point of the multiplier is at +5V and full 
+/- unity-gain multiplication occurs between +2.5V and +7.5V.  This 
gives lots of headroom
-- it essentially makes it impossible for the CV in the multiplier to 
hit zero at the 2164 control pin (because the incoming CV signal will 
never be anywhere near 20Vpp), which would give a dead zone on the 
multiplication.  I achieve this simply by bringing the CV in through 
200k while using 100k on the reference voltages.  Of course, the 
signal is now cut in half as well, so I simply double the feedback 
resistor on the I-V converter.  As long as all of these 100k and 200k 
resistors are within 0.1% of each other (and the 100k and 200k 
resistors don't need to be in a precise ratio -- they only need to be 
precise within their own values), and all incoming signals are AC 
coupled through big back-to-back electrolytics, then the four-quadrant

multiplication is very tight, which is important for frequency shifting.

-----Original Message-----
From: Synth-diy [mailto:synth-diy-bounces at synth-diy.org] On Behalf Of 
cheater cheater via Synth-diy
Sent: Wednesday, January 19, 2022 4:23 AM
To: Neil Johnson
Cc: SDIY List
Subject: Re: [sdiy] 1-quadrant multiplier with 2164

[CAUTION: Non-UBC Email]

I wonder if it matters that Dave's version will create theoretically 
more distortion on the positive swing of whatever vs the negative 
swing, whereas my version will apply distortion (non-linearity) more 
or less symmetrically... do the numbers show that it matters at all? I 
bet it would matter with some, let's say, crappy devices.

On Tue, Jan 18, 2022 at 1:57 PM Neil Johnson via Synth-diy 
<synth-diy at synth-diy.org> wrote:

This is certainly true but note also the importance of zero when

multiplying. The zero signal stays zero no matter what you multiply 
by. In Rutger's case that zero is in fact -5V, so the origin of Neil's 
graph should be at -5V signal and zero control voltage. That is why 
the level-shifting solution is so effective and it is also why I 
believe Rutger is correct to call this a one quadrant multiplier.

Yes, this is just a bit of algebraic juggling.

If we take Dave's approach:
- convert the bipolar +/- 5V input to a unipolar 0 to -10V input
- add a -5V offset to the output _after_ the VCA (so no bearing on 
the quadrantiness of the VCA itself)

With a unipolar CV and a unipolar signal ... a 1-quadrant VCA.
And don't forget that as-drawn the linearised VCA is inverting.


Synth-diy mailing list
Synth-diy at synth-diy.org
Selling or trading? Use marketplace at synth-diy.org

Synth-diy mailing list
Synth-diy at synth-diy.org
Selling or trading? Use marketplace at synth-diy.org

Synth-diy mailing list
Synth-diy at synth-diy.org
Selling or trading? Use marketplace at synth-diy.org

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://synth-diy.org/pipermail/synth-diy/attachments/20220120/34b4533b/attachment.htm>

More information about the Synth-diy mailing list