Modified : october 24th, 2007

Fixed Filter Bank, FFB
En français
back to summary


Description


WARNING, considering the size of the PCB, the number of parts, and the high density of the wiring, I would not recommend it as a beginner's project. However, if you proceed with care and method this remains a feasible project.
Although this module is often missing in many modular set-ups, I think a fixed filters bank (FFB) is a really useful module and personally I could not imagine a decent set-up without one. The reference FFB is the 914 module from Moog .

The FFB makes it possible to greatly enrich the sonic palette of a modular system. It is also a key element for the recreation of "natural" and environmental sounds, or of conventional music instruments. Although it evokes the principle of a mere graphic equalizer, it has marked differences. Mainly, the filter Q is higher and the BP slopes are steeper than in a graphic EQ. Furthermore when a control is set to zero, the  corresponding spectral band is completely muted.
Complex/entangled rhythms can be created on the fly by multi-band filtering a simple sequence.

I designed this module with the Moog 914 filter bank in mind. Therefore I tried to obtain similar characteristics for the band-pass, low-pass and high-pass filter as in the Moog 914. The original Moog's design uses passive cells based on inductors and capacitors. Nowadays, finding this type of inductors is very difficult and it would cost a lot of money to have them specially wound. This is why I chose to replace these passive components by active components. In order to design band-pass cells using active components (i.e. OP amps) with the same characteristics as in the Moog 914, there are two possible approaches :
  • the first one is to simulate the inductors using gyrators or N.I.C.s (negative impedance converter). This is a very efficient solution but is very expensive because it requires at least four OP amps to recreate a single band-pass cell with the same Q and -12dB/octave attenuation slopes.
  • the second one is to replace the passive cell with an active filter. There, the choice of architecture is wide. However, considering the constraints (low number of OPAs to reduce costs, ease of calculation of component values, low sensitivity to tolerance of the components...) the best choice appeared to be the Deliyannis band-pass cell also known as Rauch multiple feedback filter. 
The other advantage of using active filters is that the output level of this filter is much higher than that of Moog's.

This module has twelve fixed band-pass filters with -12dB/octave slopes, a low-pass filter and a high-pass filter with -24db/octave slopes. All the filters are resonant with a Q coefficient around 3.7. The gain of each filter is adjustable from completely silent to maximum.

Two inputs (with different gain levels) are provided as well as three outputs : a main output with all filter cells mixed, an ouput with only odd cells and the third output with only even cells. These separate outputs make it possible to use the FFB for creating pseudo-stereo sounds (spacialization).

This module uses not less than 32 OPAs (8 ICs),  125 capacitors,  95  resistors and 14 potentiometers !
The current draw of the module is 70mA both on the positive and negative power lines.

Principle and schematics
A BIT MORE ON FILTER DESIGN
The band-pass cell
The maths of this architecture give :


setting C=C1=C2 it now reads :



In order to obtain steeper slopes (-12dB/octave) and it is necessary to chain two Deliyannis 2nd order band-pass cells. I selected fixed standard values for the resistors giving the expected gain and Q values and next I calculated capacitor values to obtain the desired center frequency. However, the capacitor values that are obtained this way are different from standard capacitor marks. As a matter of fact I used two capacitors mounted in parallel, this way by adding two standard values one obtains a value close to the desired capacitor value (see the list of cap values below). The actual schematic for a band-pass cell is :


SPICE3 simulation of the above circuit

The circuit was simulated for C=18.3n, that is a predicted central frequency in the vicinity of 1kHz.

 The red curve corresponds to the spectrum at the output of the first OPA. As it can be seen, the lowpass and highpass slopes are -6dB/octave (-20dB/decade), Q=2.7.

 The blue curve corresponds to the spectrum at the output of the second OPA. Here we obtain a narrower bandwith Q=4, as well as steeper slopes, -12dB/octave (-40dB/decade)

The  low-pass cell
The maths of this architecture give prvided that R=R1=R2 :


setting C=C1=C2 it now reads :


In order to obtain a steeper low-pass slope (-24dB/octave) it is necessary to chain two 2nd order low-pass cells. In order to increase Q the capacitors of the first cell are set to different values. Therefore, the actual schematic for a 4th order resonant low-pass cell is :


SPICE3 simulation of the above circuit

The predicted cut-off frequency is in the vicinity of 88Hz.

 The red curve corresponds to the spectrum at the output of the first OPA. As it can be seen, the lowpass slope is -12dB/octave (-40dB/decade), Q=2.7.

 The blue curve corresponds to the spectrum at the output of the second OPA. Here we obtain a steeper slope : -24dB/octave


The  high-pass cell
The maths of this architecture give provided that C=C1=C2 :


setting R=R1=R2 it now reads :

In order to obtain a steeper high-pass slopes (-24dB/octave) it is necessary to chain two 2nd order high-pass cells. In order to increase Q the resistors of the first cell are set to different values. Therefore, the actual schematic for a 4th order resonant high-pass cell is :


SPICE3 simulation of the above circuit

The predicted cut-off frequency is in the vicinity of 7kHz.

 The red curve corresponds to the spectrum at the output of the first OPA. As it can be seen, the highpass slope is -12dB/octave (-40dB/decade), Q=2.7.

 The blue curve corresponds to the spectrum at the output of the second OPA. Here we obtain a steeper slopes : -24dB/octave





Printed Circuit Board and Component Layout

NOTE : the PCB and a component kit for this module are now made available by Bridechamber
PCB design
Layout






 Download the schematic as a PDF file 
Download the PCB as a PDF file 

WARNING ! The document is formatted to be printed directly on a mylar for photo-etching or a "press & peel" paper. Make sure that when the printed face of mylar is in contact with the copper side of the PCB, the lettering can be read normally.


List of parts and building instructions
reference
 value
 number
U1,U2,U3,U4,U5,U6,U7,U8
 TL074
 8
RN1,RN2,RN3,RN4
 SIP resistor network : 8 x 100K 2% with one common  leg
 4
R90,R91
 10 ohm
 2
R78a
 470
 1
R80b,R80b
 680
 2
R87,R88,R89
 1k
 3
R2,R5,R8,R11,R14,R17,R20,R23,R26,R29,R32,R35,R38,R41,R44,
R47,R50,R53,R56,R59,R62,R65,R68,R71
 1.8k
 24
R78a
 3.3k
 1
R77 3.9k
 1
R79b 15k
 1
R1,R4,R7,R10,R13,R16,R19,R22,R25,R28,R31,R34,R37,R40,R43,
R46,R49,R52,R55,R58,R61,R64,R67,R70,R80a,R81a
 22k
 26
R83 39k
 1
R3,R6,R9,R12,R15,R18,R21,R24,R27,R30,R33,R36,R39,R42,R45,
R48,R51,R54,R57,R60,R63,R66,R69,R72
 47k
 24
R84,R85,R86
 100k
 3
R79a
 120k
 1
R73,R74,R75,R76
 150k
 4
R82
 390k
 1
C71,C72,C73,C74
 22p
 4
C45b,C46b,C47b,C48b,C50a,C50b,C53,C54,C55,C56,C57 1n
 11
C37b,C38b,C39b,C40b 1n5
 4
C5b,C6b,C7b,C8b,C21b,C22b,C23b,C24b,C33b,C34b,C35b,C36b,
C41a,C42a,C43a,C44a,C41b,C42b,C43b,C44b,C45a,C46a,C47a,C48a,
C51b,C52b
 2n2
 26
C9b,C10b,C11b,C12b,C17b,C18b,C19b,C20b,
C25b,C26b,C27b,C28b,C29b,C30b,C31b,C32b
 3n3
 16
C13b,C14b,C15b,C16b,C37a,C38a,C39a,C40a,C49b 4n7
 9
C33a,C34a,C35a,C36a 6n8
 4
C29a,C30a,C31a,C32a,C51a,C52a 10n
 6
C25a,C26a,C27a,C28a 15n
 4
C21a,C22a,C23a,C24a 22n
 4
C17a,C18a,C19a,C20a 33n
 4
C1b,C2b,C3b,C4b,C13a,C14a,C15a,C16a 47n
 8
C9a,C10a,C11a,C12a,C49a 68n
 5
C1a,C2a,C3a,C4a,C5a,C6a,C7a,C8a 100n
 8
C61,C62,C63,C64,C65,C66,C67,C68
 100nF ceramic
 8
C59,C60
 10µF /25V electro. 2
C69,C70 100µF/25V electro.
 2

HE10 14 male PCB connector
 1
P1,P2,P3,P4,P5,P6,P7,P8,P9,P10,P11,P12,P13,P14
 10K log potentiometer
 14
Jk1,Jk2,Jk3,Jk3,Jk4,Jk5
 6.5 mm jack socket
 5

BAND-PASS CELL CAPACITORS / Polyester 5%
BP Filter
 ref.
 value
 ref.
 value
125 Hz C1a,C2a,C3a,C4a 100n
 C1b,C2b,C3b,C4b 47n
175 Hz C5a,C6a,C7a,C8a 100n
 C5b,C6b,C7b,C8b 2n2
250 Hz C9a,C10a,C11a,C12a 68n
 C9b,C10b,C11b,C12b 3n3
350 Hz C13a,C14a,C15a,C16a 47n
 C13b,C14b,C15b,C16b 4n7
500 Hz C17a,C18a,C19a,C20a 33n
 C17b,C18b,C19b,C20b 3n3
750 Hz C21a,C22a,C23a,C24a 22n
 C21b,C22b,C23b,C24b 2n2
1 KHz C25a,C26a,C27a,C28a 15n
 C25b,C26b,C27b,C28b 3n3
1.4 KHz C29a,C30a,C31a,C32a 10n
 C29b,C30b,C31b,C32b 3n3
2 KHz C33a,C34a,C35a,C36a 6n8
 C33b,C34b,C35b,C36b 2n2
2.8 KHz C37a,C38a,C39a,C40a 4n7
 C37b,C38b,C39b,C40b 1n5
4 KHz C41a,C42a,C43a,C44a 2n2
 C41b,C42b,C43b,C44b 2n2
5.6 KHz C45a,C46a,C47a,C48a 2n2
 C45b,C46b,C47b,C48b 1n


Wiring
Because of the high wire density, I had to split the wiring diagram into two parts.
wiring diagram part #1
middle lugs of potentiometers
 wiring diagram part #2
clockwise lugs of potentiometers

Front plate
Panel design



 Download the silkscreen mask as a PDF file 


Download the silkscreen mask as a  JPEG file 



Measurements
Below are some spectrum measurements made on the prototype. The measurements were obtained using a "white" noise source and measured with the SignalScope software (evaluation version, on Macintosh MacOSX)




White noise reference signal Low-pass 88Hz Band-pass 350Hz Band-pass 1kHz High-pass 7kHz





LP 88Hz + HP 7kHz
 All BP
 All
 BP-odd (red) BP-even (green)
 BP-odd+HP(red) BP-even+LP(green)


References

Moog 914 Fixed Filter Bank at Moog Archives 
Texas Instrument Application Report : Filter design in thirty seconds
General information about 2nd order filters by Bill Bowden
Deliyannis filter : 

The DIY builders' gallery
Here are the photographs of the yusynth FFB modules built by other synth geeks around the world.
Thank you  guys for sending me these nice photos.
  


Name : David Brown
Modular project : Modularsynthesis.com
Location :  USA
Website : www.modularsynthesis.com/yusynth/ffb.htm 		Name : Alison project
Modular project : Alison project
Location : Winnipeg, Canada
Website : www.thealisonproject.com
Name : Tobias Schilly
Modular project :  funk machine
Location : Germany
Website : www.mdz.de, www.myspace.com/schrumpfschlauch

Name : DJthomaswhite
Modular project :
Location :  USA
Website : naturalrhythmmusic		 Name : Dave Wright
Modular project : Notbreathing
Location :  USA
Website :www.notbreathing.com 		Name : Julien
Modular project :
Location :  France
Website :



Name : Zarko
Modular project :
Location : Gardanne, France
Website :		 Name : David Ingebretsen
Modular project : Digembre
Location : Salt Lake City, USA
Website : digembre

Name : Newandrewthal
Modular project :
Location :  USA
Website :



Name : Baronrouge
Modular project :
Location :  France
Website :		 Name : Kevin Kissinger
Modular project :
Location :  Kansas City, Mo USA
Website :

back to summary