DIYstompboxes.com

Regarding the biasing,
https://pdf1.alldatasheet.com/datasheet-pdf/view/18878/PHILIPS/NE571.html
https://pdf1.alldatasheet.com/datasheet-pdf/view/18878/PHILIPS/NE571/+07JJ9UL.hKpRudSICEY+/datasheet.pdf

Look at figure 4.  This circuit is running on 15V.   pins 5,12 have an 8.2k to tweak the bias point.
--------------
Edit: And at the bottom of the second column on page 5:

Expandor case only:
"The output will bias to 3.0V when the internal resistors are used.
External resistors may be placed in series with R3, (which will affect
the gain), or in parallel with R4 to raise the DC bias to any desired
value."   

Compressor case is just above that:
"The values of RDC will
determine the DC bias at the output of the op amp."

Go to page 52 here:

https://www.americanradiohistory.com/Archive-Radio-Electronics/60s/1968/Radio-Electronics-1968-05.pdf

which is the article on "vibrating-wire audio filter and oscillator" and you can see how to arrange for an electrical signal from the drone wires.  It covers all the harmonics that would be present in a vibrating wire if the magnet is located properly.  You take part of the guitar output and put it through the Wheatstone bridge filter network.  The Q of a vibrating wire is quite high - over 600 at 1000 Hz.  They used two transformers but a power driver like a 386 would enable you to eliminate the input transformers and the output could be taken from a differential or instrumentation amplifier for each wire.

This is the same article I quoted for the thread on sustain generators.  The filter behaves like a pickup on a guitar - the location of the magnet determines what harmonic content you have.

Some people copy the datasheet for other voltages but the biasing is stuck on 3V and is non-optimal.

My example is also used by Boss: 9v -> 3v bias
The other example is EHX DMM and uses -15v with bias of -9,96v (and their expander to 6.26v)

Both are not 'some people", or so I thought...

Thanks Rob, but I'm afraid I still don't understand.
Gain depends on input voltage.
I'm varying the input voltage.
How can the db always be 3 while the gain is different for each input voltage?

For example:

input: 10mV peak to peak:

Vav =  0.693*0.005 = 0.003465
gain= sqrt((10000*20000*0.000140)/(2*30000*0.003465)= 11,6 (+db around 20db ?)

output= 116mV

-------------------
input= 100mV peak to peak:

Vav =  0.693*0.05 = 0.03465
gain= sqrt((10000*20000*0.000140)/(2*30000*0.03465)= 3.6 (+db around 9,5 ?)

output= 3,6V

The formula for the gain uses Vin_av.

Vrms = 0.707 * Vpk
Vav   = 0.693 0.636 * Vpk

Vp = one sided peak (0 to the peak)

I should ask. How are you measuring the voltage? Peaks?