[sdiy] Anyone built the ELEKTOR VOCODER?

[Scott Bernardi]

The advantage of the state variable design is that the Q and fc can be
adjusted independently. A simple resistor ratio sets the Q, and the fc
is determined by an RC product. 
My design called for a Q of 7 to get the maximally flat response with
two bandpass stages. I chose this response to have the response curves
at least -12db down where they overlap the next channel. Lower Q would
have resulted in too much overlap. A Q of 7 put the MFB configuration
into the area with high sensitivity to component tolerance.
So for low Q, say a maximum of 3, I would agree with your conclusions. I
found that with computer simulation or on a breadboard, it was almost
impossible to get acceptable results with the MFB configuration, but it
was very easy with the state variable configuration. 

Going to a 3rd order maximally flat response with 3 stages per filter
only requires a Q of 3, 1.5, and 3. It might be realizable with MFB
configurations. It would also allow for 1/3 octave spacing between
channels. I'm going to look into this a bit more.
I'm sending this reply to the whole list because I think your comments
are very worthwhile. 


Bert Vermeulen wrote:
> 
> If you read the cookbook the tollerance is depending of the q of the
> filters. A low Q you can use 5% components.Not only the tollerance of the
> resistor is important but also the caps are important!
> 
> If you use 5 or 10% components it WILL WORK FINE but the Q, Fc and gain
> could be different as you had calcullated. It depend of what you want as
> result. I used 5 and 10% (cheap) components and match them for each channel.
> Today are 1% resistors cheap. I suggest use 1% resistors and match the
> capacitors. I used the for each channel 8 capacitors  i selected from the 25
> i bought.
> 
> State variable filters are not the best choice. If you have build them and
> compare them the multiple feedback bandpass circuit you will know that. You
> need more components and
> the need of "better"components is the same.
> 
> bv
> 
> ----- Original Message -----
> From: "Scott Bernardi" <sbernardi at home.net>
> To: "Synth-DIY list" <synth-diy at node12b53.a2000.nl>
> Sent: Sunday, January 28, 2001 4:53 PM
> Subject: Re: [sdiy] Anyone built the ELEKTOR VOCODER?
> 
> > Before anybody tries to build my design, let me mention this caveat! The
> > circuit looks good on paper, but when I breadboarded a couple of stages
> > I found the filters and WAY to sensitive to component tolerance! All
> > those calculations I did to find the nearest 5% resistance just WILL NOT
> > work. The multiple feedback bandpass filter design requires around 0.1%
> > tolerance. All the parameters are interactive - Fc, Q, and gain, so if
> > you want to use trimpots its a real pain to trim.
> > I picked up Don Lancaster's Active Filter Cookbook to delve into the
> > theory more, and came to the conclusion that I needed state variable
> > filters.
> > These can be tuned nicely, but the parts count increases dramatically (3
> > opamps per filter stage). I was even looking into doing third order
> > filters (3 cascaded state variables) and whereas it looks like it would
> > work quite nicely (I've done some computer simulations), we're talking
> > about 100's of opamps here.
> > I am debating whether I even want to even tackle building a vocoder. I
> > mean, you can get a WarpFactory by Electrix (24 band DSP vocoder) for
> > under $300 on ebay nowadays (there's two listed right now). I'd be
> > spending near that just for parts on a 14 band analog vocoder.
> >
> > Kevin Peek wrote:
> > >
> > > Just wondering if anyone has built the ELEKTOR Vocoder before?
> > >
> > > I downloaded the files from : http://omega.tellus.vallentuna.se/anders/
> > >
> > > It looks rather more complicated than what I had in mind. I was
> > > wanting to build something like Scott Bernardi's design.
> > > But I am willing to try the ELEKTOR if it not too difficult.
> > >
> > > Thanks.
> > > Kevin
> >
> > --
> > Scott Bernardi
> > sbernardi at home.net
> >

-- 
Scott Bernardi
sbernardi at home.net