[sdiy] Filter frequency cutoff with linear response?

[Tom Wiltshire]

On 20 Feb 2012, at 17:29, Justin Owen wrote:

> On 20 Feb 2012, at 13:19, Florian Anwander wrote:
> 
>> Am 19.02.2012 21:32, schrieb Olivier Gillet:
>>>> But is there any
>>>> of these synths that does this VCF modulation linearly?
> 
>>> The Doepfer Dark Energy does it.
>>> ???
>>> The DE is based on a CEM 3394, so most of the 3394 based synths should do that: AX60, SixTrak, MAX, M1000...
> 
>> I was just having a look at the CEM3394 datasheet, and they did it that way for exactly the reason predicted - linear FM of the filter. The cutoff CV is exponential (3/8V per octave) but the chip also includes a filter modulation VCA from the VCO triangle output, and that input is linear.
> 
> So assuming that CV control of Frequency Cutoff on the 3394 (I haven't looked at the data sheet) is handled via a VCA - it appears they've made a decision to use a Linear response for the other cutoff modulation source...
> 
> ...is that a correct assumption?
> 
> Curiouser and curiouser.

Filter Freq CV is done the usual way with an exponential convertor. The VCF modulation is taken from the VCO triangle output, via a Modulation Amount VCA, to the exponential convertor's "Iref" - current reference, presumably. At least, that's how it's shown on the datasheet diagram.

It's definitely a choice to use linear FM for the VCO-to-VCF modulation path rather than just doing things the simple way. On this chip the VCO outputs aren't available directly, but they could have provided a triangle output (at no cost bar one pin) and left you you to add a mixer to feed it into the Filter Freq CV input if you wanted that feature. But instead they went to the trouble of adding both a VCA and a linear FM input. They must have thought it was worth it.

Part of the reason for that is that the voice is based on a single VCO, so doing this allows you to get both filter FM and two "oscillator" FM (one is really the filter making a sine wave). Hence it adds a fair bit of versatility to an otherwise simple synth architecture. And since it was intended to be controlled by a uP, they didn't have to worry about filter tracking so much - it'll be compensated in software tuning tables - making decent-sounding filter FM a realistic possibility.

T.