[sdiy] Fixed filter bank questions

David G Dixon dixon at mail.ubc.ca
Fri May 10 21:31:21 CEST 2019

Hey Tom,

Thanks for the nice comments!

To answer your question, no, they don't need to be linearized, but it makes
their voltage control a whole lot more convenient.  In fact, I'd probably
opt to try it exponential first, just to save a whole bunch of circuitry.

-----Original Message-----
From: Tom Wiltshire [mailto:tom at electricdruid.net] 
Sent: Friday, May 10, 2019 2:14 AM
To: David G Dixon
Cc: rsdio at audiobanshee.com; *SYNTH DIY
Subject: Re: [sdiy] Fixed filter bank questions

I like this, David.

Taking the design and simplifying it to this point seems like a great idea.
As you say, the E6 series already gives you nicely distributed frequencies
across a decade, so you can just add the same set of resistors each time and

Might not be *absolutely* perfect, but it's more than likely good enough,
and often it's the imperfections that makes things interesting anyway, so it
can be good to leave a few in! These are musical devices we're building,
after all, not test equipment, so there's no requirement for technical

One query, if you do add VCAs to each channel: Why do they need to be
linearised? It'd be an audio mixer, wouldn't it? In which case, exponential
control wouldn't be a bad thing.


       Electric Druid
Synth & Stompbox DIY

> On 10 May 2019, at 08:23, David G Dixon <dixon at mail.ubc.ca> wrote:
> Brian said:
> "I don't know why Yves would have simply doubled every cap. Maybe it was
> purely for the improvement to tolerance."
> Yves didn't just double the caps.  He used two caps in parallel (where
> values add together) to make capacitors with specific values, within the
> confines of the E6 series.  Hence, if he needed an 18.3n cap, he combined
> 15n cap and a 3.3n cap in parallel.  Of course, there is a limit to how
> precisely one can derive values this way, but he came very close to his
> target capacitances doing that.  He also ignored tolerances.
> What I did instead was to select a single cap value from the E6 series,
> then design the resistor networks to give the desired frequencies.  There
> a much larger selection of readily available 1% resistor values, so this
> an easier, more accurate, and much cheaper way to get specific frequencies
> than Yves' method.  I can buy 100 of any 1% resistor value I need from
> Digikey for about $1.70, and from that I can build a few dozen fixed
> banks.
> However, as I said a little while ago, I'm thinking now that I'm just
> to use the same resistor network (100k-51k-2.7k) for every filter, and
> value of E6 cap between 100n and 1.5n, and just take whatever frequencies
> get -- they will be almost evenly distributed since the successive E6 caps
> have value ratios which are between about 1.42 and 1.5.  If this doesn't
> give pleasing results, then I can always change it.  My layout for all 12
> bandpass filters is on a board which is only 2.1" x 6", and it will take
> about 30 minutes to print, etch and drill the board, and about another 30
> minutes to stuff and solder it so I can try this circuit several different
> ways to see which version I like the best -- it will just be a matter of
> swapping out the filter board.  All the other circuitry will be on a
> separate board which is also 2.1" x 6".  If the "quick and dirty" one
> good, then I will go with that.
> I'm going to add a few other little improvements and innovations to the
> design.  First of all, I'm going to make a panel PCB to eliminate pot
> wiring.  Also, Yves and a few others have created separated outputs for
> odd and even filters.  I'm going to do that as well, but I'm also going to
> give the option of separate inputs (the odd and even inputs will be
> to each other).  Also, I'm thinking of putting a white noise source in the
> unit, and normaling that to the inputs so that the bank is also a
> stand-alone wind machine.  Finally, I'm going to lay out the circuit so
> I can retrofit voltage control to each channel later.  I will do this by
> inserting a linearized 2164 VCA between each attenuator pot and input
> resistor and the summing output amps (that's 14 linear VCAs requiring 7
> chips and associated circuitry).  I will rig it so that this VCA network
> be inserted into the circuit simply by unplugging and replugging a few pin
> headers.  I'm thinking that the basic FFB will be 2U wide, and there will
> a 1U panel next to it for the voltage control option.  I'm very curious to
> hear what it sounds like when the filters are driven by the octature sine
> waves from my Dr. Octature filter.  Finally, I might put a crossfader on
> odd-even outputs.
> _______________________________________________
> Synth-diy mailing list
> Synth-diy at synth-diy.org
> http://synth-diy.org/mailman/listinfo/synth-diy

More information about the Synth-diy mailing list