[sdiy] Fixed filter bank questions
Bernard Arthur Hutchins, Jr
bah13 at cornell.edu
Sun May 12 06:02:05 CEST 2019
It seems to me that this matter is being way over-complicated.
First of all, it is a PERFORMANCE parameter of a circuit block; such as center frequency, Q, or peak gain, (NOT any particular component value or values) that needs to be fine-tuned. Secondly, in order to “correct” a particular parameter or component (actual resistor or capacitor picked out of the parts drawer of nominal conscripts), you have to make a MEASUREMENT (perhaps of an actual component, or likely, better, of the initial performance parameter).
Given these TWO considerations, can we develop a procedure for a systematic and simple near-comprehensive fix (like one “tweak and done”)? Yes we can, and it works extremely well with the filter banks based on the MFIG with positive feedback (“Deliyannis”) bandpass.
With such a filter bank, we absolutely need to get the gains all set quite
uniformly. This is essential if we want to use an overall positive feedback
around the whole bank to simultaneously manipulate the Q of all the
Say we design the nominal peak gains to 1.00. We expect any particular instance to have a spread of gains (perhaps 0.97, 1.06, 0.98, etc.). [ Gain is MUCH easier to measure than frequency or Q! ] We could imagine a very tedious (and often frustrating) trimming task.
Instead, suppose we design for a gain of say 0.92 so that we end up with a spread, but with all less than 1.00. In our circuit board, we allow a space for the addition of a small series resistors added to the lower legs of the feedback divider (temporally we make these wire jumpers). We are looking to have the original resistor in the lower leg of the divider about 1000 ohms and the “series correction” less than 100 ohms  (to be determined – a simple CALCULATION based on the measured (with the jumper) gain.
The “formula” for the correction comes through the magic of “classical
sensitivity”) [3, 4], with  telling exactly how to do the Deliyannis.
References  and  show my 39 channel filter bank which also
illustrates the method.
Getting all the peak gains the same was our essential goal, but rather
automatically the center frequencies and Q are completely acceptable.
And the tune-up is even fun (under 2 hours for 39 channels) and you get
to watch the gain settle at 1.00 as the soldered resistors cool!
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Synth-diy