Oakley Filtrex II User manual

Oakley Sound Systems
Filtrex II
PCB issue 2
Analogue Filter Rack
Builder's Guide
V2 4
Tony Allgood
Oakley Sound Systems
Carlisle
United Kingdom

Introduction
This is the Builder's Guide for the issue 2 iltrex II rack module from Oakley Sound. It
contains the schematic description, parts list and testing procedures. or those of you with an
issue 1 iltrex-II board you should ensure that you have the proper Builder's Guide for your
board and not this one.
or the User Manual, which includes a brief history about the development of the module, a
guide to the front panel controls and some notes about the power pack needed to power your
module, please visit the project webpage at:
http://www.oakleysound/filtrex.htm
or general information regarding where to get parts and suggested part numbers please see
our useful Parts Guide at the project webpage or http://www.oakleysound.com/parts.pdf.
or general information on how to build our modules, including circuit board population,
mounting front panel components and making up board interconnects please see our
Construction Guide at the project webpage or http://www.oakleysound.com/construct.pdf.
This is an early version of the documentation. If you do find any errors, even silly little ones,
please do let me know either directly by e-mail or via the orum or mailing list.
2

Parts List
or general information regarding where to get parts and suggested part numbers please see
our useful Parts Guide at the project webpage or http://www.oakleysound.com/parts.pdf.
The components are grouped into values, the order of the component names is of no particular
consequence.
A quick note on European part descriptions. R is shorthand for ohm. K is shorthand for kilo-
ohm. R is shorthand for ohm. So 22R is 22 ohm, 1K5 is 1,500 ohms or 1.5 kilohms. or
capacitors: 1u = one microfarad = 1000n = one thousand nanofarad.
To prevent loss of the small ‘.’ as the decimal point, a convention of inserting the unit in its
place is used. eg. 4R7 is a 4.7 ohm, 4K7 is a 4700 ohm resistor, 6n8 is a 6.8 n capacitor.
Please note that if you are building a stereo version of this module you will need two
Filtrex-II boards One of them will be built as a Master and the other as a Slave For the
Slave module you should not fit parts marked with a * Please see later in this
document for a suggested method on how to wire the two modules together
Resistors
All resistors 5% or better 0.25W types. Those items marked with 1% need to be 1% metal
film (or better) 0.25W metal film resistors. or the sake of clarity it may be simpler to make all
resistors 1% metal film types.
R61 is a +3000ppm/K positive temperature coefficient resistor.
22R R5, R53, R109, R4, R52, R115, R121, R96, R80, R81, R8, R122
47R R111, R116, R126
150R, 1% R51, R54
330R R10, R18, R19, R13, R17
470R R12, R20
1K R11
1K +3000ppm/K R61
1K5 R117
2K2 R9*, R91, R56, R46, R57
2K2, 1% R25, R24
2K7 R47
3K R100
3K9 R1, R15
4K7 R87, R97, R74
4K7, 1% R41, R43, R30, R34
6K8 R70, R39, R6, R38
10K R123, R3, R78, R58, R113, R55, R104, R110
10K, 1% R42, R32, R31, R44
12K R99
3

15K R40, R37, R69
22K R75, R14, R7, R108, R112
22K, 1% R28, R27, R22, R23
33K R16
39K R107
47K R85, R29, R26, R35, R71, R36, R124, R33, R82, R79
56K R59, R60
68K R65
100K R125, R101, R120, R73, R68, R88, R114, R119, R98, R92, R93, R76,
R72, R84, R89, R95, R94, R118, R66, R67, R64, R45, R83, R103
100K, 1% R48, R50
220K R21, R90, R63
330K R102
470K R62, R49, R2
680K R105
1M R86
3M3 R106, R77
Capacitors
22p low-K 2.5mm ceramic C7, C3
33p low-K 2.5mm ceramic C48, C28
100p low-K 2.5mm ceramic C26
1n , 100V polyester C43, C30, C39
68n , 63V polyester C12, C13, C14, C15
100n , 63V polyester C32, C52, C20, C21, C35, C54, C55, C41, C50*, C38*
470n , 63V polyester C5, C24
680n , 63V polyester C19, C18, C17, C44
1u , 63V polyester C16
2u2, 63V electrolytic C51, C33
10u , 35V electrolytic C37, C22
22u , 35V electrolytic C56, C53, C34, C6, C8, C4, C49, C25, C27, C45, C11,
C46, C36, C10
100u , 25V electrolytic C42
220u , 10V electrolytic C9
1000u , 35V electrolytic C47*, C40*
10u , 35V non polarised elect C23, C29, C31, C57
22u , 35V non polarised elect C1, C2
Non polarised capacitors can sometimes be called bipolar. Unlike ordinary polarised
electrolytic capacitors they can be put into the board any way around.
4

Discrete Semiconductors
BC549 or BC550 Q1-15
1N4002 or 1N4004 D8, D23, D9*, D13*, D14*, D18*, D19*, D24*
1N4148 or 1N914 D1, D2, D3, D4, D5, D6, D11, D12, D15, D16, D17, D20, D21, D22,
D25
BAT-42 D7, D10
LED 3mm green TRIG
LED 3mm red/green L O
LED 3mm red PEAK
LED 3mm yellow ON
The L O LED is a bi-colour LED in a water-clear package. Do not fit an ordinary single
colour LED in this position.
Integrated Circuits
OPA2134PA U1, U7
NE5534 U8
TL072 U3, U5, U6, U9, U16
TL074 U4, U10, U14
7815 U15*
7915 U11*
THAT300P U2
4001 U13
4016 U12
Pots
All pots 16mm Alpha or Alps types.
50K linear dual gang RESONANCE, GAIN
1M log UP, DOWN
50K linear SMOOTH, REQ, IZZ, DRY/WET, THRESHOLD, ENVELOPE
10K linear VOLUME
50K log L O-RATE, L O-DEPTH
Trimmers
22K trimmer TRIM
470K trimmer BAL
5

Switches
SPDT on-on CONT, AUTO, WAVE
SPDT on-off-on MODE
Miscellaneous
Leaded ferrite bead L1
Heatsinks TO-220 clip-on 2 off for U11 and U15
2-way 0.1” KK Molex header 4 off INPUT, SIDE, CV, GATE
3-way 0.1” KK Molex header 1 off OUTPUT
2-way 0.1” KK Molex housing 4 off INPUT, SIDE, CV, GATE
3-way 0.1” KK Molex housing 1 off OUTPUT
16mm Alpha pot brackets 13 off
Off-board Parts
These are fitted to the rear panel
1/4” mono sockets 4 off Input, Side chain input, CV, Gate
1/4” stereo (TRS) socket 1 off Output
Suitable power plug 1 off Power In
I recommend plastic shrouded sockets for both the audio, CV and power sockets. This will
reduce the risk of earth loops.
Additional parts required for stereo operation on both boards
2-way 0.1” KK Molex header 2 off STV, ST-A
2-way 0.1” KK Molex housing 2 off STV, ST-A
4-way 0.156” KK or MTA header 1 off MOTM
4-way 0.156” KK or MTA housing 1 off MOTM
You will also need a power source of some kind. The recommended supply is a wall-wart
supplying 15V AC at 300mA minimum. See later for more information.
A small amount of insulated multistrand wire is needed. This will be used to connect the
sockets and power supply to the board.
IC sockets are to be recommended, especially if this is your first electronics project. You need
eight 8-pin DIL sockets, and six 14-pin DIL sockets. Choose ‘turned pin’ or ‘dual wipe’
types.
6

Circuit Description
Like many complex analogue circuits the iltrex-II issue 2 circuit can be split up in to little
bits. The first bit we will look at is the pre-amplifier stage on page one of the schematic. The
input signal comes in through a small network of passive parts: L1 acts as a filter for radio
frequency interference. C5 and R2 act as a high pass filter removing any DC from the input
signal. R3 and C26 act as a low pass filter removing any unwanted high frequency components
from the input. R3 also protects U1 from possible over voltage.
The pre-amp is built around U1. I have specified the low noise audio op-amp, the OPA2134.
The pre-amplifier is a two stage design. The first stage is a non-inverting amplifier whose
voltage gain can be varied from 1 to 12. C1 keeps the gain for DC and very low frequency
signals at near to one. This prevents any offsets within U1 from being amplified unnecessarily
C3 provides a little bit of high frequency roll-off to keep the amplifier stable.
The second stage of the pre-amp is an inverting amplifier. The GAIN pot is used in a slightly
offbeat way. It is in both the feedback and the input resistors. This way we can control the
gain over a wide range from -0.4 to -10.1. The minus in these numbers shows the inverting
properties of the amplifier.
The voltage gain of the two pre-amplifier stages in tandem can be varied from -0.4 to -122. A
gain of 0.4 means that the output of the pre-amp is only 40% of the input level. While a gain
of 122 means that the output level is 122 times bigger than the input. In audio circles this
would normally defined in dB. This pre-amplifier will give you a gain from -8dB to +42dB.
Because the pre-amplifier is made from a inverting and non-inverting stage, the overall
behaviour is inverting. This means the output is completely out of phase with the input. This is
not a problem since the inversion is corrected later on the iltrex.
Q7 and associating circuitry drive the peak LED. This is designed to light up just as the ladder
filter starts to show heavy distortion.
The main audio path continues on to the filter ladder itself via one half of the resonance pot.
However, the pre-amp also provides the signal for the 'dry/wet' circuit and the envelope
processor. We will deal with these two later.
The filter is based around the traditional ladder as designed originally by Dr. Moog. I have
used THAT300 matched NPN array for the top and bottom pairs in the ladder. This minimises
control current breakthrough to almost zero. Current breakthrough manifests itself as a copy
of the modulating signals on the output. Generally, this is not a good thing. ‘BAL’ biases the
base of the left hand bottom pair (U2, pin 9), via R21, by a small amount to even out any
differences within the ladder. This minimises breakthrough still further.
Two of the rungs of the ladder are ‘sniffed’ by a differential amplifier. Each of these is
identical, based around the classic three op-amp implementation. They are all DC coupled, and
rely on ‘close’ matching to remove any DC offset. A differential amplifier is a device that
makes larger the voltage difference between two points. In our case, the voltage across the
top and bottom filter capacitors. The gain of the differential amplifiers is set higher than
normal ladder filters to improve signal to noise ratio in the following mixer stage. C17 and
7

C18 provide AC coupling of the outputs to remove the slight DC offset from the top
differential amplifier.
The resonance in a moog ladder filter is controlled by the level of audio feedback applied to
the bottom pair of transistors from the output of the differential amplifier at top rung of the
ladder. The level of is typically controlled by the resonance pot, wired as a variable resistor
and a trimming resistor, TRIM. It is traditional to use a 50K reverse log pot for the resonance
control in the classic moog ladder circuit. When set to a value of 50K this pretty much allows
the feedback path to be ignored, and no resonance can be heard. The ‘reverse log law’ is
needed so that as you turn the pot the resonance increases smoothly. An ordinary linear law
pot would do nothing for most of its travel, and then all the resonance would be introduced in
the last quarter of a turn. Not very ‘smooth’ or musical. Unfortunately, 50K reverse log pots
are difficult to find and quite expensive.
In the iltrex-II, we use an 50K linear pot, but we wire it as a potential divider and not as a
variable resistor. The drawback of using a pot on its own like this is that it has a variable
output impedance and the ladder is therefore unbalanced, the degree of which depends on
where the resonance control is set. We get around this by using a buffer on the pot's output.
U3, a unity gain voltage follower, or buffer, 'sniffs' the voltage at the wiper of the resonance
pot and provides a copy at its own output that has a constant impedance.
The resonance pot is a dual gang type too. One gang controls the feedback loop as we have
seen, and the other sets the gain of the input. Normally the passband gain of a moog ladder
decreases as you turn the resonance up. In other words the volume drops as you increase the
resonance. This can be quite a problem for a post processor like the iltrex. So in the iltrex,
the input level is automatically turned up as resonance increases. Thus the overall effect is of a
constant volume at all values of resonance. I thought this was quite clever of me to invent a
way of doing this without increasing noise levels. However, I found out later that the very
same principle was used in the Roland SH-2000 as long ago as 1974.
The other half of U3 acts again as a voltage follower on the wiper of the gain control part of
the resonance pot. This ensures that the ladder's input is fed from a constant resistance no
matter where the resonance pot is positioned.
C11 provides the appropriate decoupling and is sufficiently large so as to allow the filter to
oscillate above 200Hz or so.
our inputs control the filter cut-off frequency via an exponential convertor based around U9.
The filter frequency can be directly controlled with the REQ pot via R63. R65 sets the
sensitivity of the envelope processor’s output. While R67 does the same job for the L O. The
CV input provides a nominal 1V/octave response, and would normally be accessed via a jack
socket on the rear panel.
The exponential convertor is temperature compensated. R61 is the positive temperature
coefficient resistor providing an approximate cancellation of the exponential convertor’s
inherent temperature coefficient.
Note the two pin header, STV, and R60. These allow two iltrex-II modules to be ganged
together to form stereo processing. There is more about this later in the document, but we can
8

see that the output of the CV summing amp is present at pin 1, while a second CV input is
made available at pin 2. The master iltrex-II will be sending the CV signal via pin 1 on the
STV header to pin 2 on the slave's STV header. This way the master iltrex-II board will also
control the slave's cut-off frequency. It would be expected that the connection be done via a
switch so the linking could be turned on or off from the front panel.
The audio path continues from the filter’s differential amplifiers by going on to the mixer
stage. This is shown on page two of the schematics at the top of the page. The mixer stage is
designed to combine the three audio signals created by the iltrex so far: The first and fourth
rung ladder outputs, which will make up the ‘fizz’ and ‘smooth’ outputs respectively. And the
main pre-amplified signal which will go to the 'wet/dry’ balance control.
The two filtered signals go to identical ‘reversible attenuator’ circuits. U5 is wired as two
inverting amplifiers, each with a gain of -1. Each inverted signal is applied to one end of a pot,
while the non inverted signal is applied to the other. By simply moving the wiper of the pot to
one side or the other determines which signal is dominant. When the wiper is in the middle
position the two signals cancel each other out and the voltage is zero. U7a (pins 1, 2, 3) is
configured as an inverting summer. This circuit combines the voltages at each wiper. R38 and
R39 are deliberately low in value when compared to the 50K of the pots. This loads the
relative high impedance output of the pot and warps the law of the pot. This allows the central
null point to be more easily found.
U7b (pins 5, 6,7) in conjunction with the dry/wet pot and R45 form the output balance circuit.
This block allows the user to select between the output of the filter mixer or pre-amplified
signal. It is easy to see how it works when you consider the action of the wiper moving
between one end of its track connected to U7 pin 1 and the other which is connected to the
output of the pre-amplifier.
The output of the mixer stage is fed via a DC blocking capacitor to the ‘Volume’ pot. The
wiper of this is connected to the final output amplifier stage. This is a low noise inverting
amplifier circuit, based around U8, with a gain of -3.7. The op-amp chosen is the 5534 which
is capable of driving larger currents than most ordinary devices. R54 reduces any chance of
high frequency ringing in long cable runs and protects against short circuits. Non polar
capacitor, C57, and R50 remove any small DC offset voltage present on the output of U8.
Now let us look at the little network R48, R51 and C29. This mimics the output impedance of
the main audio output – although the small differences between C29 and C57 and the non zero
output impedance of U8 do not allow it to be an exact match at all frequencies. This network
connects to pin 2 of the output header which goes to the ring of any connected jack plug.
Now if the inserted jack is a mono one the ring terminal will simply be shorted to the sleeve by
the action of the jack plug. Thus the output of the iltrex-II behaves in much the same way as
a traditional unbalanced output. However, if you use a TRS plug (a stereo jack plug but
carrying only a mono signal) then the audio output of the iltrex-II appears as a balanced
impedance output. This can be used as you would a traditional balanced output signal; with
the hot signal being present on the tip, the cold signal on the ring and the ground on the
sleeve.
Many balanced output circuits present the audio signal on both the hot and cold connections;
the signal on the cold connection being an inversion of the signal on the hot. This alone is
9

often considered to be a balanced signal but in truth the receiving end of the balanced
connection cares not for the absolute signal levels on the hot and cold connections but only the
difference between them. And because all balanced input circuits do not have an infinite input
impedance the source impedance is very important to this differencing process. By keeping
both the hot and cold's source impedance as close in value as possible you maximise the ability
of the balanced input's circuitry to reject any unwanted interference on the audio signal.
The advantage over the impedance balanced circuit over the full balanced circuit is two fold:
Simplicity – you have less circuitry in the audio pathway to either go wrong or alter the signal
in some way. And compatibility – the impedance balanced circuit can be used for both
balanced systems and unbalanced systems with no possible problems with stability or short
circuits. Now there are EBOS (electronically balanced outputs stage) circuits around that will
automatically compensate for unbalanced connections but they do add complexity and may be
prone to instability under certain load conditions.
The disadvantage of solely impedance matching is a comparative loss of headroom. The full
balanced system has an additional 6dB of output signal to play with. It is the author's opinion
that this does not make a great deal of difference in the real world.
If the venerable 5534 is replaced by a more modern single op-amp you should remove C28 as
additional compensation will probably not be required. You may also be able to link out the
coupling capacitor, C57 and its twin, C29, if the output offset voltage is negligible. R50 and
R48 may also be removed in this case.
That completes the description of the audio path. Now let us take a look at the processing
circuitry of the iltrex. Staying on page two of the schematics, the second set of circuits down
the page belongs to the full wave rectifier and threshold detector. This takes its input from
wither the pre-amp output or the external side chain input. The choice is determined by the
‘cont’ switch.
The switch's wiper leads straight into some amplification based around U14a. This is an
inverting amplifier with a gain of -3.3. The main pre-amp is expected to produce a maximum
output of around 5Vp-p in normal use, so U14a boosts this signal up to 15Vp-p. The
amplified signal is now full wave rectified by the circuitry based around U14b and associating
circuitry. ull wave rectification can be described by the mathematical ‘absolute’ function. In
other words, the output of the full wave rectifier ( WR) is always positive. If you present
+10V to the input, you will get +10V. But if you present it with -10V you will also get +10V.
Likewise, -5V turns into +5V, -3V into +3V. Now if you put an audio signal into this circuit,
you will get a series of positive bumps that correspond to the up and downs of the audio
signal.
U14d forms a special buffer circuit. This configuration, allows the op-amp to drive medium to
high capacitive loads without instability. We also saw it being used in the output amplifier of
the iltrex. The output of the full wave rectifier is therefore protected by the odd load
presented by the next set of circuits.
Now, no real time system can recover envelope information without some disadvantage of
some sort. Some systems employ the peak and droop method. These are fast to respond to
sudden changes in loudness or envelope. They work by simply charging a capacitor as fast as
10