[sdiy] Fwd: Shruthi 4PM (was re something else...)

rburnett at richieburnett.co.uk rburnett at richieburnett.co.uk
Fri Nov 29 13:00:10 CET 2013

That's the nightmare of working with filters made from real-world 
analogue components (>.<)

You could always trade component tolerance problems for those of 
coefficient rounding and state rounding instead! ;-)

That multi-mode filter technique works very well in the digial domain 
where you can specify the pole positions and the mixing ratios with 
excellent accuracy and repeatability.

Regardless of whether the implementation is analogue or digital, the 
resulting band-pass and high-pass responses might not be quite the shape 
you would initially expect though.  Due to the way the 4 poles split in 
that x-formation in the "4-pole cascade with feedback" the bandpass 
response becomes a little lopsidded as resonance is applied.  And the HP 
response has a dip in the passband to the right of the resonant peak!

These are surely not show-stoppers musically, but since we are obsessed 
about accuracy of the filter shapes it's probably worth mentioning :-)


On 2013-11-29 10:25, cheater00 . wrote:
> Roman,
> On Fri, Nov 29, 2013 at 11:10 AM, Roman Sowa <modular at go2.pl> wrote:
>> For example in notch filter the notch comes from summing 2 signals at
>> certain frequency shifted by exactly 180 degrees. So in theory this 
>> one
>> frequency (Dirac spike in spectrum) should attenuate to zero. If there 
>> are
>> mismatches between stages of the filter in amplitude, this will never 
>> go to
>> 0, because at the frequency when there is 180 degrees difference the
>> amplitudes are not the same, and at the spot when they are, relational 
>> shift
>> is not 180 degrees.
>> Even worse if it's 4-pole filter. Then you have to match 4 amplitudes 
>> and
>> corner frequencies.
>> A few simple calculation can give you exact attenuation of notch 
>> frequency
>> in dB if you asume 1% deviation in parameters. Why not make a 
>> simulation of
>> ideal filter with altered values?
>> Let's go further - 0.1% resistors are not a big deal nowadays, I buy 
>> them
>> for about 0.25EUR, but precision capacitors is a whole different 
>> story. OTOH
>> I never needed them so far, so I may be wrong. Also you can always add 
>> a
>> trimmer cap.
>> Roman
> If the capacitor is wrong, the frequency at which the filter is tau/2
> out of phase will shift, however the amplitude at that point will not
> change. That is, the amplitude at the new point is the same as the
> amplitude would have been at the old point. So it's not an issue for
> two-pole filters, while it might be an issue for higher order filters.
> Selecting capacitors might be a chore, especially if you need to later
> send them out to manufacturing. However, you could put say 8
> capacitors in place of one, with an unpopulated solder bridge which
> would put one of them in circuit. Then when the boards come back from
> manufacturing, you hook up a test harness which characterizes that
> pole. If separate poles are on separate boards you can then select
> them out of a database. If they're all on the same board, it'll tell
> you which set of capacitors will be the closest. My point is that if
> you're manufacturing with pick & place, it'll likely be cheaper to
> have more capacitors than to select them yourself, send them in, and
> make sure the contractor knows how to place them by using elaborate
> packaging and documentation.
> Cheers,
> D.
> _______________________________________________
> Synth-diy mailing list
> Synth-diy at dropmix.xs4all.nl
> http://dropmix.xs4all.nl/mailman/listinfo/synth-diy

More information about the Synth-diy mailing list