AUDIO
OUT
AUDIO OUT
L
R
By Jake Rothman
Analogue Vocoder – Part 3: Driver Amplifier
L
ast month, I mentioned
Circuit description
that driving all the filters in the
Vocoder needed a small power
amplifier because a single op amp has
insufficient output current capability.
This month, we’ll discuss the design and
next month the build of a special currentboosted op amp circuit. For a smallscale manufacturer like me, it’s rarely
cost-effective to design a circuit and
PCB for just one job, so this is a versatile
audio module that can double as a highquality headphone amp, mixer, springline driver, output transformer driver
and even a high-impedance speaker
driver. It can have a balanced input and
two boards can be used in bridge mode.
It’s used as a virtual earth mixing amplifier – it mixes all the Vocoder filter
outputs together, along with the synthesiser and vocal inputs. Also, its high-current
output can drive transformers to give
isolated balanced XLR outputs. These
options are shown in Fig.1.
An effective way to boost an op amp’s
output current is to put a couple of complementary transistors in a push-pull
emitter-follower configuration on the
output. These are included within the
overall negative feedback loop, as shown
in Fig.2. This configuration is very tolerant of Hfe mismatching between the
transistors. To retain the low distortion
of the NE5534 op amp, a few additional
tweaks and refinements are needed. The
full circuit is shown in Fig.3.
Output stage
The output stage consisting of TR1 and
TR2 is class AB. This allows high current peaks of around 500mA. A small
idling or quiescent current (Iq) with zero
signal of around 5 to 15mA through the
transistors is necessary to avoid crossover distortion. In most power amplifiers
the bias voltage to set Iq is achieved by
a pre-set resistor in conjunction with
Non- inverting
X L R
input
1
O utput
–
3
–
Bridge driver
(Two boards
req uired)
+
2
+
O utput
load
–
Ch assis
a)
b)
+
a Vbe multiplier transistor. In this circuit’s case, since the output is expected
to drive loads upwards of 50Ω, we can
use a fixed bias voltage to set Iq. This is
possible because loads of 4Ω or 8Ω do
not have to be driven as in normal power amplifiers and larger emitter resistors
can be used without significant power
loss. If R11 and R12 are around 10Ω to
33Ω, rather than the typical 0.22Ω, sufficient DC negative feedback voltage (Vre)
is dropped across them to stabilise the
bias current. These high-value resistors
will also provide a degree of short-circuit
protection – but not for long.
Biasing
The extra emitter resistor voltage drop (2×
Vre) is added to the base emitter voltages (Vbe) of the transistors – therefore, the
bias voltage needs to be higher than the
normal 1.2V (two diode drops) used. A
fixed bias voltage of 1.5V is required. I’ve
done a bit of research on ways of achieving this; one is to use a BZV46 ‘Stabistor’
diode which is now hard to source, and
another is an infra-red LED. In this circuit, I’ve chosen to use two Zener diodes
(D1 and D2) which because of their high
doping, exhibit a higher forward voltage of typically 0.75V, rather than 0.6V.
(Note, Zener diodes are normally run in
reverse breakdown and this is the device
voltage specified). Whatever the Zener
reverse voltage is, say 5.6V, 12V or 3.3V,
0V
Inverting
Summing amplifier
–
+
Negative feedback (NFB)
Input
–
O utput
+
3
0V
Mix bus
c)
d)
L ow- value
Sense resistor
(on board)
–
Input
+
O utput
PNP
0V
Fig.1. Variations using the same board: a) differential/balanced input amplifier; b) summing/
mixing amplifier; c) bridged driver and d) distortion-cancelling transformer driver.
58
NPN
1
2
Inverting
Inputs
V +
Balanced
output
Push - pull
emitter follower
Fig.2. The output current of an op amp
can be boosted by adding an emitterfollower to the output.
Practical Electronics | January | 2022
C7
470pF
+
+
1
L 1*
1mH
J5
C4*
22µ F
C2
470pF
R3
22
Ch assis/
metalwork
C1
470pF
4
–
C12
22pF
D1
2. 7V
8
IC1
3 +5534
R4
22
C3
470pF
R10
7
2
2
R2
1
+ 7. 8V
R5
22
C5*
22µ F
5
V re
0. 15V
C9
D1
2. 7V
6
7
R11
1
V re
0. 15V
T1
BC143
R6
22
R14
22
R13
7
0V
+ 10µ F
7
C6
470pF
+ 15V
Iq
15mA
+
Balanced
line input
(X L R)
R1
1
3
R9
7
C8
22µ F
R7*
22
T2
2N2219
C10
150µ F
R12
1
Iq
15mA
+ 3V
+
R15
C13
470µ F
C11
150µ F
C14
100nF
O utput
+
R8*
22
+
n
Mixe r
input bus J2
0V
L 2
10µ H
R16
1
+
C15
470µ F
1
0V
Fig.3. Full circuit of the high-output-current amplifier. Note that the main circuit is for a differential input amp. Note, R8 is not normally
used. (Components marked ‘*’ are for the mixer amp.)
its forward voltage will be around 0.75V. One advantage of
using two diodes, is that each can be mounted adjacent to
its associated output transistor’s heatsink to provide a degree
of temperature compensation (see Fig.4). Alternatively, transistors can be used as diodes. The base-emitter junctions of
small-signal types, such as the BC549, give around 0.7V. It’s
easier to couple transistors to heatsinks than using diodes, as
shown in Fig.5. The fastest tracking thermal compensation
can be achieved by mounting the bias transistor on top of the
can – see Fig.6. A small bit of stripboard provides rigidity and
allows a 500Ω preset to be fitted in series with the ‘diodes’ to
increase and trim Iq if wanted.
The diodes themselves need a bias current to generate the
bias voltage. This is provided by R9 and R10 connected to the
15V power rail. This can be thought of as a single resistor of
9.4kΩ with 13.5V across it. (15V power rail minus the 1.5V
diode bias voltage). Using Ohm’s law (I = V/R) we can calculate the current as 1.4mA. The exact value is not critical, but
this is the minimum. The current is sunk into the op amp’s
output pin. This sounds like a bad idea, but audio designers
have found that a DC bias current can reduce op amp distortion. It was a popular trick to reduce the distortion of cheap
op amps, such as the LM324, 741 and LF351. It was usually
done by connecting a 3.3kΩ to 10kΩ resistor from the output
pin to the negative rail. Douglas Self later discovered that the
popular audio op amp – the NE5534 used here – gave lower
distortion with the resistor going to the positive rail. Thus, the
diodes’ bias current should be beneficial. Noise from the power rail is suppressed by the op amp’s low output impedance.
Clipping
There is a little loss of headroom with the diodes connected
in this way because the op amp’s output pin sits at −0.75V
with the output at 0V. This means the lower cycle clips first.
This is not a problem in practice, since it is only a reduction
from 24V peak-to-peak to 22V into 50Ω with ±15V rails, a
loss in headroom of −0.75dB, which is inaudible. It can’t be
fixed by the normal offset adjustment pins on the chip because
Fig.4. Each Zener bias diode (D1 and D2) should rest against their
associated heatsink/transistor to provide thermal compensation.
Fig.5. An alternative to Zener diodes is to use the base-emitter
junction of small-signal transistors as diodes – here clamped to
the driver transistors’ heatsink.
Practical Electronics | January | 2022
Fig.6. A superior but rather complex way of ensuring good
thermal coupling between transistors is to use a small board
wired as shown. A bit of heatsink compound helps the coupling
from the top of the transistor cans to the biasing devices.
59
V +
IBIAS
7
Bootstrap
+
22µ F
7
+ 0. 75V
–
10µ F
+
10µ F
+
2. 7V
+
1
O utput
1
2. 7V
– 0. 75V
7
22µ F
+
7
capacitor (C12) on the board for this upgrade, since there is no
connection to its pin 8.
IBIAS
Fig.7. A symmetrical bias chain with a dual bootstrap minimises
offset and current into the op amp’s output.
the range is not sufficient. If it’s good enough for Focusrite
(a high-end audio manufacturer), who use a similar output
circuit with 5532s, it’s good enough for me.
For those who are driven by ‘scopemanship’, the problem
can be eliminated by making a symmetrical biasing-plus-bootstrapping network, as shown in Fig.7. This needs two extra
electrolytic capacitors and two resistors.
An alternative, for the differential amplifier is to inject an offset current into one of the op amp inputs. In Fig.3, add a 180kΩ
resistor from the −15V rail to the op amp’s inverting input, plus
change R5 to 10kΩ to provide a little extra gain. Power rail noise
injection could be a problem, raising distortion. However, this
method does avoid any DC voltages appearing on the inputs.
Alternatively, the non-inverting input can be jacked up by
one diode drop with filtering (Fig.8). This works well for
the summer amplifier. Another trick once used by Quad in
their pre-amps, is to make the power rails slightly asymmetrical. For this design +14.4V and −15.7V works. It’s
confusing though, if someone has to fault-find later.
Another useful little tweak is to bootstrap the mid-point of
resistors R9 and R10 with capacitor C8. This prevents the resistors loading the op amp’s output and boosts the maximum
positive output swing a little. C9 bypasses the diodes to provide better AC drive to the base of TR2. R13 and R14 bias the
output coupling electrolytics C10 and C11 at around 3V. This
also reduces distortion, especially with tantalum types. Again,
the current from these biasing resistors is sunk into the op
amp’s output. These capacitors are dimensioned to feed 600Ω.
For lower impedances, a noticeable increase in low frequency
distortion will occur. R16 is a pull-down resistor to hold the
output at 0V to avoid pops when leads are plugged in. Without this resistor, the output will float up to the capacitor’s bias
voltage. Input coupling capacitors C4 and C5 form a bi-polar
unit which also reduces distortion. I found there was no further benefit in biasing these.
High-frequency stability
C1, C2 and C3 are RF filtering for balanced inputs. C6 provides
RF filtering on the non-inverting input and C7 is the normal
feedback phase-lead capacitor needed for stability. R15 and L2
allow loads with significant capacitance, such as long cables,
to be driven without oscillation. This problem often affects
emitter followers. Normally, R15 is sufficient to prevent this,
but if a very low output impedance is necessary, it can be bypassed at audio frequencies by adding L2.
Since this is a high-current op amp circuit, decoupling capacitors C13 and C15 have to be much bigger than usual to
cope with the high peak-current demands. This means that
they can inject high ripple currents into the ground rail, possibly causing distortion and hum. The PCB layout is designed
to deal with this by placing the capacitors where the power
supply comes into the board. There are two pins and wires for
the earth on the power connector to give minimum impedance.
Another decoupling capacitor (C14) is necessary for high-frequency stability, and this must be placed close to the op amp.
The 5534 in particular, will oscillate without this capacitor.
V +
IBIAS
NFB
Rf
High-voltage upgrade.
+
Blocki ng
The 5532 is unusual among op amps in that it can withstand
capacitor
1
±22V power rails. This is the absolute maximum stated
22µ F
on the datasheet, but in practice I’ve found the Signetics/
V BIAS
1. 2 to 1. 4V
+ 0. 6V
Texas NE5534P remains reliable at this level. (Note: the
Blocki ng
O utput
1
capacitor
+
0.
6V
dual version, the NE5532 gets too hot). This will give 35.5V
Rin
pk-pk output into 80Ω (1.97W RMS). With the diode mod
–
0V
Input
5534
(Fig.8), 36V pk-pk (2W RMS) will be obtained. By replac+ 0. 6V
IBIAS
+
ing the 5534 with a Texas Instruments OPA604, the power
rails can be increased to ±24V, giving 38V pk-pk output
2
into 80Ω. Interestingly, the OPA604 does not exhibit the
asymmetrical clip, so it needs no diode bias. However, this
is only because it clips earlier on the positive cycle than
1
the 5534. On ±22V the output is 35V pk-pk, half a volt
+
+ 0. 6V
22µ F
less than the 5534. So, the OPA604 upgrade is only worth
it if you use ±24V rails or if you need a FET input device
0V
for high impedances. It also costs three times more. (currently, the 10-off price is £0.77 vs £2.27 + VAT at Mouser).
Compensation isn’t needed with the OPA604 since it is Fig.8. A diode voltage applied to the non-inverting input cancels the
unity-gain stable. It’s fine to leave the 5534 compensation offset created by the output bias network.
+
60
Practical Electronics | January | 2022
with unity gain,
the effective gain is
increased in proportion to the number
of inputs, so C12 can
also be reduced for
this reason. This is
assuming the output transistors used
are fast planar types
with an F t of over
50MHz. If slow ( Ft of
1-4MHz) then types
such as BD436/7 are
used, and greater
compensation may
be needed, with a
Fig.9. The completed differential input board.
consequent increase
in high-frequency distortion.
R8 is another bootstrap component for
use on summing and inverting amplifiers. This is a circuit trick developed by
Uses and configurations
Calrec to multiply the input capacitors
I explained in the introduction that this
(C4 and C5) value by 10 times, avoiding
board can be configured in several ways
bass loss when the total input impedance
– here are the Vocoder-specific examples.
is low. This situation occurs when lots
of inputs are summed together, as is the
Balanced input Vocoder driver
case for the Vocoder and mixing desks.
The amplifier is a standard differential
If, for example, 10 4.7kΩ input resistors
configuration to allow for a standard XLR
are used to sum 10 inputs, the effective
balanced input. The input resistors can
impedance will drop by a factor of 10 to
be adjusted to give a different range of
470Ω. The expected −3dB point would
input impedances. Often, 10kΩ is the
be 33Hz, but with the ‘Calrec mod’ this is
standard value, but this can give a sigreduced to 3.5Hz. Note that the op amp
nificant noise contribution. I often use
feedback resistor is 100kΩ to maintain
much lower values, such as 1.5kΩ, on the
a DC path, while allowing sufficient AC
basis that most op amps can still drive it.
feedback around the capacitors. InterestThe assembled board is shown in Fig.9.
ingly, the bootstrapping also appeared to
eliminate distortion from the capacitors.
Summing and Vocoder output amplifier
Here the amplifier has to be configured as
a virtual earth summing amp. The input
Compensation
resistors are on the Vocoder channels and
C12 is a compensation capacitor for the op
the input on the board is a virtual earth.
amp. If its value is too low, the amplifier
I have added a ‘Steve Dove inductor’(L1)
will oscillate, if too high, the distortion
to decouple RF from the long mixing
around 10kHz will increase. If the gain of
buses sometimes encountered in mixers.
the amplifier is set to unity, the capacitor
His early 1980s Studio Sound series on
specified in the data sheet is 22pF, but it
mixer design is well worth downloadcan often be reduced to 15pF in practice.
ing. I’m sure it formed the basis for much
For gains over five it can be omitted. For
of Douglas Self’s later publications (see:
a gain of six, 6.8pF gives lowest distorhttps://bit.ly/pe-jan22-ao). The assemtion. If used for mixing multiple inputs
bled summing board
is shown in Fig.10.
www.poscope.com/epe
- USB
- Ethernet
- Web server
- Modbus
- CNC (Mach3/4)
- IO
- PWM
- Encoders
- LCD
- Analog inputs
- Compact PLC
- up to 256
- up to 32
microsteps
microsteps
- 50 V / 6 A
- 30 V / 2.5 A
- USB configuration
- Isolated
PoScope Mega1+
PoScope Mega50
Next month
In the next article
we will look at some
more board configurations, the components
and assembly.
Fig.10. The
completed summing
amplifier board.
Notice the green
inductor (L2) which
is needed for use
with long mix buses.
Practical Electronics | January | 2022
- up to 50MS/s
- resolution up to 12bit
- Lowest power consumption
- Smallest and lightest
- 7 in 1: Oscilloscope, FFT, X/Y,
Recorder, Logic Analyzer, Protocol
decoder, Signal generator
61
