AW: [sdiy] Dual Frequency Shifter update

[Magnus Danielson]

From: jhaible at debitel.net
Subject: Re: AW: [sdiy] Dual Frequency Shifter update
Date: 19 Jul 2006 13:02:34 +0200,Wed, 19 Jul 2006 13:02:34 +0200
Message-ID: <1153306954.44be114a4a162 at www.debitel.net>

>  > Even if you have a nice non-linearity which is bipolar (like direct
> > compression
> > without timeconstant) the distorsion it causes will probably be similar to
> > what you have simulated. You can only compand a small amount with such a
> > system
> > before the distorsion exceeds the S/N gains that you hoped to acheive. 
> 
> Magnus, I'm completely with you.

Good. I wanted to make sure my point was received with full signal strength
(i.e. 10) and full clarity (i.e. 10).

> Predistortion (before the FS, i.e. before the dome filter) causes a lot
> of harmonics, which won't be harmonics after the FS process anymore,
> so a static waveform won't be static after the shift process anymore
> (the now non-harmonics will beat against the original signal), and I
> can't see how this could possibly be made undone with another nonlinear
> function.
> Maybe with the predistortion before the multipliers (after the dome
> filter? - Haven't thought this thru yet.)

That will work, but you will have another hell to live with. Since multipliers
are nothing else than log, add, exp in series it makes sense to consider how
to make that work optimal, adding another log and exp seems a bit strange.

The dome filter is probably a contributor to bad S/N so you really want to
replace that with a filter having better S/N properties. An orthonormal ladder
or for that matter a polyphase filter is probably your best bets. The
traditional dome may be neat, but I suspect them not to be the best choice from
a noice perspective. The stability of these filters in order to keep mirror
suppression down is another issue and the proposed solutions is better.

> I just say sometimes I'm blind, and as I understood Tim has a solution
> that has worked in practice.
> 
> Anyway. Compander will work. I have a compander in my old FS-1.

A traditional compander works, yes. It is still a violation, but a smaller
violation so it works.

> A very simple one, too, with long release time constant.
> All that I'm fighting with now is the not-so-optimal behaviour of
> a NE572-based compander. I tried 1uF attack and 10uF release caps
> frm the datasheet. This works very clean on some recorded music I've 
> tested, but thumps ridiculously with a fast-attack synth singnal.
> Attack time constant is way too high. Tried a shorter attack last
> night, now I got distortion, but the thumping is gone. Gotta increase
> the release time constant next. (Thumping: I mean compresor overshot.
> Before the compressor can react, the output goes up until the zener
> diodes clip it. See NE572 data sheet for reference.)

OK. You might need to overlook your design to avoid it. I'll check the
details tonight. It seems like your damping of the loop goes out the window!

> Maybe I'll try a capacitance multiplier for the attack time constant
> (old NE570 trick). Will act fast with sudden change of dynamics, and
> slow on steady levels.
> If everything fails, I throw out the whole companding board and 
> build something I know will work. (Like my old FS-1 compander,
> or the vactrol-hungry Moog 12-stage phaser compander)

Mmmm, Vactrols! :-9

Cheers,
Magnus