Designing low-noise

audio amplifiers
In any system, the front-end amplifier sets important limits to
performance. This article examines the limitations of low-noise
amplifers as defined by the laws of physics and the practicality
of real-world components.

WILFRIED ADAM

ince the invention of electronic ampli­ 100Hz right up into the multi-CHz range. of thermal noise increases with tempera­

S fiers, their dynamic range (the ratio

between the smallest signal just above
amplifier noise and the largest possible
The cause of thermal noise is thermal
oscillation of electrons; for example, in the
crystal lattice of a conductor. Thermal noise
ture; it is proportional to the bandwidth and
the resistance value. The so-called Nyquist
equation gives the relationship
signal at a given level of distortion) has is frequently also called Johnson noise or
increased steadily. Despite this improve­ (not quite correctly) white noise - there are
ment, the dynamic range of a modern other sources of white noise. The amplitude
microphone amplifier is considerably less
Fig.I. Comparison
than the capability of the human ear or of a
between the dB Ratio
modern microphone (Fig.I). Therefore in
dynamic range of 140 10 000 000
the quest for fidelity it is desirable to
the human ear, a 130
improve the dynamic range of electronic
modern micro­
amplifiers still further. 120 1 000000
phone and a
The main factors behind the expanded
microphone 110
dynamic range of modern amplifiers have
amplifier with a
been an improved knowledge of distortion 100 100000
gain of 1000

UJ""
mechanisms and methods of combating
distortion, together with a better under­
(GOdB) . 90

standing of noise sources in amplifiers and z

BO 10000
«
the way in which external noises are picked '" 70
up. �
L
Modern transducers general analogue sig­ « 60 1000
z
>-
nals at low level, mainly because of their 0 50
poor efficiency in converting the input ener­ Fig.2. Noise
40 100
gy into electrical energy. Analogue signal spectral density
amplification is therefore needed to match plot showing white 30
the transducer signal to the subsequent noise and flicker 20 10
analogue or digital signal processing circuit­ noise components
ry. The noise performance of this first 10
of a bipolar op·
amplification stage is of considerable im­ amp for audio, and 0
portance, because signal information from the definition of Human ear Modern Microphone
the transducer which is lost through noise in corner frequency
microphone amplifier with a
gain of 1000 {60dBl
the first stage cannot normally be recovered fe·
through signal processing later on.
One type of interference which influences
the performance of the first amplification
stage may arise from external sources such
as radio frequency signals, in all frequency
1000
ranges. Such interference may vary statisti­
cally or may be of a repetitive nature. It may
be coupled into the amplifier either induc­
tively or capacitively.
Other forms of interference are generated
within the amplifying circuitry itself. They
� 100 "
'"
are of a statistical nature and are also present '>
at all frequencies. It is on this class of c:

interference that this article will concen­ �

trate. �,
10

NOISE AND PHYSICS

In audio amplifier design, several varieties of
internally generated noise must be consi­
dered. to
0.1 fe 10 100 1k
Thermal noise. This occurs at all frequen­ FREQUENCY �
cies and is the dominant noise from around

628 ELECTRONICS & WIRELESS WORLD June 1989

where En=noise voltage (RMS); k = Boltz- En= y 4X 1.38X 10 23VA s K I x R(n) En(nV) En(dBm) En(nV/YHz)
mann's constant (l.38x10-23VA s K-I); T= 300Kx20 000s-lx50VA-1 b=20kHz b=20kHz
absolute temperature in kelvins; b = band-
width in Hz; and R = resistance in ohms. 50 125 -136 0.91
Using this equation it is possible to calculate 75 153 -134 1.12
the noise voltage of, for example, a 50n 200 250 -130 1.82
(=50VA) resistor at a temperature of 2rc Listed in the table are noise voltages for a 600 433 -125 3.15
(300K) and a bandwidth of 20 000Hz number of useful resistances common in lk 559 -122 4.07
(=20 000s-I), audio engineering.

Optional weighting
filters to CClR or NAB
r--- ...
Output for
I
I oscilloscope
I
I...... n1
Rs 20 to 20kHz
Variable bandpass filter
gain amp. with 36dB/Oct
in 10dB­ roiloff;
steps gain: OdB 30dB
Precision full-wave Meter
rectifier with two
time constants:
330ms/l0s

30Hz low pass filter Output for

ladB/Oct. rolloff; power amp.
gain: OdB with loudspeaker

Fig.3. Block diagram of a noise measurement circuit for white noise and flicker noise.

Fig.4. Circuit forthe measurement of white noise and flicker noise.

62k

Device
under test NE5534
2k
TL071
+ 1k 1k
680n 680n
.
Rs
1)J:; 47k + TL071

11
/1 1k2 Full scale
Input for power -30dBm
supply noise 3k3
measurements -100dBm
10k
10000" -110dBm

I
1.1OOOO)J
33k
-120dBm
100k
-130dBm

0/10/20/30/40dB r--1 lnsert for optional

weighting filter (( Cl R I NAB)
__
40dB

82k
82k
l
2k 1n

1k White 20 . . 20kHz
680n 680n TL071

+ +

+ TL071 +
TL071 82k
12 n 1 TL071
Flicker
0.14
to 30Hz
3k3
3k3
82k

100

30dB
10k

1}J3�1!'1I +
TL071

2k2 2k2 3k3

O.775V nominal level
/ 2k2
TL071
2k2 W 100k + R

'--+----1+

+
Time constant for:

1k
White noise: 330ms
TL071
Flicker noise 10s

�------�/
Average reading meter (ARM)

June 1989 ELECTRONICS & WIRELESS WORLD 629

 Fig.5. Optional CCIR weighting filter for the circuit of Fig.G.

In the right-hand column, the impulsive nature, consisting of

table also gives the noise voltage Use at least 1% resistors and 2.5% momentary changes of the out­
in units of nV/YHz, i.e. normal- tolerance capacitors put voltage due to fluctuations in
ized to a bandwidth of 1Hz. This Adjust lOOk pot. for OdBgain at 1kHz
the current through an active
has the advantage that it is not device. Contaminated semicon­
necessary to state the bandwidth ductor surfaces aggravate this
to which a particular noise vol­ type of noise. Other factors
tage applies. In the case of white which may increase popcorn
noise this is permissible since noise are low temperatures and
the noise spectral density is con­ high-value resistors. The "pops"
stant - the noise voltage is the occur quite randomly with the
same within any bandwidth in­ device, sometimes being absent
terval, no matter where this in­ for several minutes and then
terval is placed in the frequency appearing several times per
spectrum. second I. The precise causes of
popcorn noise, or burst noise as
Flicker noise. In taking a closer
it is sometimes called, are not
look at the low-frequency noise
known.
from an amplifier in the frequen­
cy range below 100Hz one dis­ Current flow through resistors.
covers an additional noise com­ Besides thermal noise, resistors
20000
ponent. Flicker noise is caused generate excess noise when a
by material impurities and also current flows through them or
depends on the production pro­ Fig.G. Optional NAB weighting filter for the circuit of Fig.G. when a high voltage is applied.
cess employed. The amplitude of The reason for this excess noise
flicker noise increases at lower is inhomogeneity in the resistor
frequencies, where it very much dominates one further white noise component super­ material. Metal film resistors produce less
the white noise, especially in view of the imposed on the ever-present thermal noise. excess noise than carbon film types and
small bandwidth under consideration. Since The cause of this shot noise lies in the fact small-sized resistors are noisier than large
the flicker noise amplitude is proportional to that the flow of an electric current depends ones.
the inverse frequency it is also called 1/f on the motion of discrete particles - elec­
Mechanical contacts. Inhomogeneous and
noise: trons. Shot noise occurs only in active
unstable contacts in connectors or potentio­
Enf=K1f devices. Its voltage is proportional to the
meters cause voltage fluctuations and are
where Enf is the flicker noise voltage, K a current through the device, the bandwidth
especially troublesome sources of noise.
constant of proportionality and f the fre­ and resistance:
quency. Es= Y2qx IdcxbxR Vibration. Noise can also be generated by
In expressing the amount of flicker noise where q equals the charge of an electron, mechanical vibration of components such as
present in an amplifer, the frequency at 1.6x10-9As: Idc is the current through the vacuum tubes and transistors. This causes a
which the flicker noise region starts is of device; b is the bandwidth; and R is the displacement current which makes itself felt
particular interest. This corner frequency, fc resistance. as noise. It is especially noticeable in coaxial
is defined in the noise spectral density plot of cables where vibrations cause a change in
Recombination noise. A further noise com­
Fig.2. For audio purposes a corner frequen­ capacitance. Large components such as
ponent only to be found in semiconductors
cy well below 100Hz is desirable. capacitors or printed circuit boards are also
is the so,called recombination noise - noise
Besides these two principal types of noise, affected.
which occurs when electrons recombine
several others are significant in the design of
with holes. This type of noise occurs mainly Leakage currents. Noise may arise from
low-noise audio amplifiers:
at high frequencies. Its amplitude drops in leakage currents due to contaminants such
Shot noise. Noise from active components proportion to 1/f2. as finger prints and soldering residues on
such as vacuum tubes and transistors con­ Popcorn noise. This noise component printed circuit boards or across compo­
tains, in the frequency range above 100Hz, occurs at frequencies below 100Hz. It is of an nents.

630 ELECTRONICS & WIRELESS WORLD June 1989

 NOISE MEASUREMENT
To measure noise precisely demands a spec­
trum analyser, with which to measure the
spectral noise density distribution. This in­
strument is not part of everybody's electro­
nics tool kit because of its price; but simpler
means, suitable for the purposes of compari­
son, may be used for measuring the two
principal noise components.
A circuit suitable for evaluating white
noise and flicker noise in audio amplifiers is
shown in Fig.3 and Fig.4. Its frequency
range is 20Hz to 20kHz for white noise and
O.lHz to 30Hz for flicker noise. Full scale
deflection of the meter can be set in the
range between -130dBm and -90dBm, so
that the lowest noise voltage which can be
indicated is in the region of -140dBm
(75.5nV). In making measurements it is
important to set the gain of the amplifier
under test to 20dB, because otherwise the
calibration of the range switch does not hold Fig.7. Optimum source resistance of an NE5534 op·amp (typical example without flicker
good. The device under test should be well noise in the audio band) .
screened from external interference such as
mains hum, preferably by a metal box.
Optional weighting filters can be added as
given in Fig.S and Fig.6. Note that the CCIR
standard requires peak rectification and the
NAB standard true-RMS rectification2,
whereas this circuit provides only full-wave
averaging. If gain of the CCIR filter is set to
OdB at 2kHz this will correspond to the
C CIR-ARM standard3.
When switching to the measurement of
flicker noise, wait for at least 10 seconds for
the meter to stabilize. Usually it will be
better to evaluate the amount of flicker noise
on a DC-coupled oscilloscope. It is also a
good idea to connect a power amplifier with
an old loudspeaker to the output, so that one
can listen to the noise as long as the
loudspeaker can stand it. This helps to
identify rapidly any noisy components and
amplifiers, or any externally-injected mains
hum and radio interference.

Input-referred noise. Results obtained from

Fig.B. Noise figure and optimum source resistance Rs as a function of collector current le
the instrument are the noise voltages at the
for an LM394 transistor.
input of the device under test; the 20dB gain
of the device under test has already been
subtracted. The value of input-referred noise r-----�--_.--�--- ·15V
LOW-NOISE DEVICES
is useful because it allows a quick calculation
of the noise voltage at the output of the Bipolar transistors. P-n-p transistors are
device for any gain setting. For example, the preferred in low-noise transistor circuits
input-referred noise of an amplifier is because of their marginally better noise
-122dBm. At a gain of +20dB the output characteristics at low source resistances,
noise level will be -122+20= -102dBm compared to equivalent n-p-n devices under .!. + .--!- --{
__

and for a gain of +32dBm the output noise the same conditions. The magnitude of the
will be -90dBm and so on. collector current influences the value of the
base spreading resistance Rbb and thus the
Optimum source resistance. By taking a achievable optimum source resistance of a NE
series of measurements with different resist­ particular circuit. This is demonstrated in 5534

ance values, preferably using fixed-value Fig.8 for the LM394 transistor.
metal film resistors, connected to the input The base spreading resistance is effective­
R= 250
of the device under test, it is possible to ly in series between the input terminal and
determine the optimum source resistance of the base of the transistor and generates most
--�-- ---�-- 15V
the amplifier under test. This is the source of the noise within the transistor. Unfortu­
1= El-0.6V
resistance at which the difference between nately the value of base spreading resistance R
the noise generated by the amplifier and the is not normally stated in manufacturers' Adjust Rl= R2 so that the voltage drop
noise of the source resistance is at a mini­ data sheets and it cannot be measured easily. across them is 2.5Vapprox.
mum. Figure 7 shows an example for the This leaves the designer with three op­
popular low-noise op-amp NE5534. From tions when selecting a transistor for low­ Fig.9. NE5534 op·amp with substituted
this diagram it can be seen that the optimum noise circuits: differential transistor pair for optimum
source resistance is approximately 5kD. 1. Use special transistors designed for low- source resistance matching.

June 1989 ELECTRONICS & WIRELESS WORLD 631

 ZN4 S 915, ZN4 6 01 5 , S S M 2 0 1 61 6 a n d
SSM213416. The NESS34 o r the improved
LM833 are standard devices for use in many
low-noise circuits, although they have to be
monitored for noise performance.
33 Op-amps may be paralleled by summing
their outputs, to reduce the value of the
optimum source resistance. But this techni­
que suffers from the limitations described
above for the paralleling of transistors.

Hybrid devices. There are some very good

hybrid or discrete op-amps, such as the
Transamp LZ17, the Matchamp XTX12918 or
the Jensen JE990. The latter is notable for
excellent op-amp designl9. Besides being
very low-noise devices with optimum source
resistances in the region 200-S00n, these
components have the additional advantage
of extended ±24V supply rails, thus increas­
ing dynamic range by a further 6dB or so.

LOW-NOISE CIRCUITS
In designing a low-noise amplifier, the fore­
most task is to adapt the optimum source
resistance of the amplifier (that source
Fig.IO. Microphone amplifier with op·amp and transformer for optimum source
resistance at which the amplifier is quietest)
resistance matching.
to the value of the AC source resistance
provided by the transducer. It is usually a
noise purposes such as the MAT01 (dual)\ optimum source resistances between 3n good idea to measure the resistance of the
MAT03 (dual)4, MAT04 (quad)4, LM394 and approximately 10kn. If the source re­ source, as manufacturers' data usually states
(duaI)5. This may be expensive or the devices sistance is higher it worth considering field­ only the DC resistance, which in our case
may be hard to obtain, but the advantage is effect transistors. However, bipolar transis­ does not help, or some sort of nominal value.
that one can be sure that these devices are tors might be preferable even for source Source resistance is best measured by con­
matched and will be essentially free of flicker resistances where field effect transistors necting a resistor across the source termin­
or popcorn noise, because they have been might be used, because bipolar devices have als and determining the resistance value at
screened by the manufacturer. a much lower flicker noise corner frequency which the output drops by half. This is
2. Use medium-power RF transistors. By of 1Hz to 100Hz; fets exhibit corner frequen­ preferably done at several frequencies - say
design these transistors exhibit a low value cies between 100Hz and 1kHz. 100Hz, 1kHz and 10kHz - within the audio
of base spreading resistance, 4-30n being band. The average of these values is taken as
Field-effect transistors. If the source resist­
common, and they are usually free of flicker the source resistance to which the optimum
ance is higher, than approximately lOkn,
and popcorn noise, although it can be source resistance of the amplifier is to be
n-channel field-effect transistors should be
worthwhile to check this using the noise matched. The maximum value of the three
used on account of their lower current noise
measurement circuit outlined above. A good should be multiplied by 10, giving the input
at high input resistances8. P-channel fets are
example is the BFW16A with its Rbb' of less resistance value for which the amplifier has
noisier than n-channel ones. Here again,
than Sn6. to be designed.
special low-noise devices such as 12029,
3. Use medium-power AF transistors. The There are three different ways to lower the
J2039, NFS10110, 2N4867AIO, 2N6483
familiar B0140 or medium current switch­ optimum source resistance of a given ampli­
(dual)ll may be used although the good old
ing transistors such as the 2N44037 will fier.
dirty 2N3819 or BF264A will still do the
work well because the chip is fairly large, as • Collector current through the transistor.
trick, if measured for noise. Modern low­
they are designed to handle currents of Let us consider a common dynamic micro­
noise fets can be made to work at optimum
around lA. They therefore exhibit a corres­ phone having an average source resistance
source impedances as low as soon if you
pondingly low value of base spreading resist­ of lOOn and a common low noise op-amp
dare (noise corner up to 1kHz).
ance. They are easily available, but before with an optimum source resistance of Skn
being used in a low-noise amplifier should be Mosfets have no place in the front-end of (Fig.7). As Fig.8 shows, the collector cur­
screened individually for low flicker and low-noise amplifiers because their optimum rent determines the optimum source resist­
popcorn noise with the noise measurement source resistance is well above 100Mn or ance of a transistor stage. For the given
circuit. SOI2. source resistance the corresponding collec­
If the optimum source resistance obtained tor current is around 2mA per transistor. If
with one transistor is not low enough, Bipolar, fet and c-mos op-amps. Bipolar such a diagram is not at hand, for example
several may be connected in parallel. This op-amps can be selected by the noise voltage when using medium-power transistors
method suffers from diminishing returns, quoted in the data sheet and the curve giving whose noise performance is not specified by
because the noise-free gain increases by the the optimum source resistance. The flicker the manufacturer, the following rule20 may
factor n, where n is the number of devices in noise corner frequency of bipolar op-amps be applied:
parallel. For example, 50 transistors are ranges between 1Hz and 100Hz compared to
paralleled within the LM394. Moreover each 100 to 1kHz for fet op-amps. C-mos op-amps le
_ VF
40xRs
transistor may have to be screened for low also have corner frequencies in the region of
flicker and popcorn noise. This can become 1kHz. With cheaper types of bipolar beware where le is the collector current, !3 is the
quite unpractical when using a large num­ of flicker and popcorn noise. Frequently current gain of the transistor and Rs is the
ber of devices per amplifier. Emitter resis­ op-amps are stated to have low noise; this source resistance. The collector current
tors are necessary because transistors ex­ may have been so at the time they were first must be within the limits given in the data
hibit wide tolerances, but this increases marketed, or they may be the lowest-noise sheet of the transistor.
noise because any emitter resistors are effec­ devices of a quite noisy family. One can, of course, place a differential
tively in series with the input. Low-noise op-amps by today's standards stage running at 2mA per transistor in front
Bipolar transistors may be used with are the LTl02813, LTl03i4, OP4,4, OP374, of the NESS34, but this creates stability

632 ELECTRONICS & WIRELESS WORLD June 1989

problems. These can be avoided nicely by Capacitive reactance. When using the input
substituting the input differential pair of the Rl-S = 2k2 Rf = 2 k2
capacitor and input resistor time constant to
NE5534. This is done by taking the normally cut off low-frequency signal components,
Rl
used inverting and non-inverting inputs to low-frequency noise is introduced. For ex­
-/
-15V and feeding the signal from the differ­ R2
ample, with a 1fLF capacitor and a 10kn
ence amplifier running at 2rnA per transistor -/ resistor giving a cut-off frequency of 16Hz, a
R3
into the offset adjustment pins 1 and S. --/ horribly high series resistance Xr of Skn at
These are internally connected to the collec­ R4 20Hz and still 160n at 1kHz is placed in
tors of the now disabled internal differential --/, series with the input. This produces con­
RS
pair (Fig.9). It is also necessary to parallel --/, siderable amounts of noise. Since the capaci­
the internal collector resistors with external tive resistance increases at lower frequencies
resistors so that the collector is maintained Gain for each input: Rf/R=l (OdB) irrespective of whether you are using foil
at 2.5V. Noise gain: capacitors or electrolytics, the noise spec­
Resistance values around the op-amp Rf 2k2 trum due to the capacitive reactance is
----;-:.
-- S(14dB)
similar to that of flicker noise. Designers
=

should be made as low as possible to prevent 440 Ohm

the introduction of additional thermal noise have spent many hours, happy and other­
from the feedback resistors. The NE5344 wise, finding this one. The only thing to do is
supports this, being capable of driving a to use very large and therefore electrolytic
600n load without reduced output voltage. capacitors so as to achieve a cut-off frequen­
Fig.n. Noise performance of the inverting
• Paralleling. When going for even lower cy of well below O.OlHz. Bypass the elec­
summing amplifier with five inputs.
source resistances, several transistors can be trolytics with ceramic capacitors, if you
directly paralleled as outlined above. Howev­ must. Remove any low-frequency signal
er, to avoid excessive offsets at the output, components after the low-noise amplifier
-96dB. So, surprisingly, this stage must be
the transistors should be selected so that stage.
designed for low noise by using, for example,
each one carries the same current. Spurious oscillations. Be warned that inter­
the circuit of Fig.9 as the operational ampli­
• Resistance transformation through a modulation products of RF oscillations pro­
fier. What is even worse, the more inputs,
transformer. A transformer converts a low duced by an unstable amplifier will cause
the noiser the summing amplifier will be.
impedance into a high impedance and vice unexpected noise. •
versa. If a low signal source resistance is to
TRAPS AND PITFALLS
be matched to the higher optimum source References
resistance of an amplifier the required trans­ Besides taking into account the design rules 1. Harris, Analog Pocket Application Guide No.5:
former turns ratio is given by outlined above, you should note some pit­ Operational amplifiers (Audio) 1988, p.76.
falls to avoid. 2. Rohde and Schwarz, Operating Instructions for

n= '\/&m. UPGR.
V Rs Reverse biasing of transistors. During tests 3. Ray Dolby, D. Robinson and K. Gundry,
with an ohmmeter or while the circuit is CCIR/ARM: A practical noise-measurement
where n is the turns ratio of the transformer, method, Journal of the Audio Engineering Society
being switched off, the transistors may be­
Rs the source resistance and Ropt the opti­
come reverse biased, leading to an increase Vol. 27, No 3, March 1979. p.149-157.
mum source resistance of the amplifier. 4. PMI, Linear and Conversion Products 1986/87
in flicker and popcorn noise. If, for example,
With Rsat lOOn and Roptat5kn, n=7. Data Book.
a circuit measures well the first time after it
Fig.10 shows the use of a 1:7 step-up 5. National Semiconductor, LM394 Data Sheet.
1 has been switched on and develops noise the
transformer with a rather clever circuit2 .22 6. E.H. Nordholt and R.M. van Vierzen, Ultra-low­
next time it is used, it is very likely that the
which allows the gain to be varied over a very noise preamplifier for moving coil phono car­
transistors are briefly reverse-biased during
wide range of 60dB by a single linear poten­ tridges, Journal of the Audio Engineering Society
tiometer, whilst maintaining optimum switch-off. This fault can be cured only by Vol. 28, N04, April 1980, p.219-223.
changing the discharging time constant in 7. D. Self, Design of moving-coil head amplifiers,
noise conditions (i.e. first-stage gain is
question and replacing the front-end tran­ Electronics & Wireless World December 1987,
always higher than the gain of the second
sistor. p.1206-1209.
stage). 8. John Maxwell, The low-noise JFET - the noise
Don't be put off by rumours that transfor­ Power supply noise. Noise can be injected problem solver, National Semiconductor, Ap­
mers have a bad reputation. In the days of into an otherwise low-noise power amplifier plication note AN 151.
valves with their high optimum source via the power supply. Circuits usually exhibit 9. Intersil, Data book.
resistances (30kn or more), there was no good power supply noise rejection ratios of 10. National Semiconductor, Fet databook.
other way than to use step up transformers typically 120dB at 50 or 100Hz, but values 11. Steven W. Smith, Internal noise of low­
frequency preamplifiers, Review of Scientific In­
with ratios up to 1:20 with consequently decrease significantly to 20 -40dB in the
struments 55, May 1984, p.812/813.
nasty frequency response and distortion range between 1 and 20kHz. This means, for 12. M. Hartley Jones, A practical introduction to
characteristics, and with less-than-perfect example, if there is a noise level of -60dBm electronic circuits, Cambridge University Press,
p roduction techniques and materials. on the power supply voltages (a value typical p.40-51.
However, modern audio transformers are for most fixed-value voltage regulators), a 13. Linear Technology, Linear Databook 1986.
produced using much improved materials noise voltage of -1 OOdBm will appear at the 14. Linear Technology, Linear Databook Supple­
and techniques, and excessive step-up ratios output of the low-noise amplifier, if its ment 1988.
15. Ferranti, Technical Handbook Standard ICs.
are no longer necessary, thanks to the lower power supply rejection ratio is 40dB. So
16. PMIISSM, Products Data Book
optimum source resistance of modern tran­ some care must also put into the design of
17. Transamp LZ Data Sheet.
sistors and op-amps. low-noise power supplies. Consequently all 18. Hugh Ford, Matchamp XTX 129 mic. preamp.,
types of switching power supply should be Broadcast Sound January/February 1984. p.52-
Inverting summing amplifiers have their avoided in such applications because they 55.
particular noise problems, which are due cannot be made to run sufficiently quiet. 19. Deane Jensen, JE990 Discrete operational
Zener diodes also produce significant amplifier, Journal of the Audio Engineering Soci­
firstly to the unavoidable series input resis­
ety Vol. 28 No 1/2, January/February 1980, p.26-
tor and secondly to the amount of noise gain amounts of white noise if not properly
34.
as defined in Fig.H, which shows the circuit bypassed with low series-resistance capaci­
20. A. Foord, Introduction to low-noise amplifier
of a typical summing amplifier with five tors. design, Wireless World April 1981, p.71-73.
inputs. As far as each input is concerned, the Noise levels of power supply voltages can 21. Steve Dove, Designing a professional mixing
gain of this stage is OdB. But as far as noise be measured by connecting the supply vol­ console, Part 4: The Mixer Frontend, Studio
gain is concerned there is 14dB of noise gain tage via a capacitor to the input of the Sound December 1980, p.40-48.
(2.2kn divided by the 440n of the five NE5534 amplifier (Fig.4). Note that in this 22. Steve Dove, Designing a professional mixing
parallel input resistors). This consequently case 20dB must be subtracted from the console, Part 12: The Channel Frontend, Studio
increases the noise level of, say, -llOdB to calibration values of the stepped gain switch. Sound October 1981, p.70-72.

June 1989 ELECTRONICS & WIRELESS WORLD 633