n
AUDIO
OUT
AUDIO OUT
L
R
By Jake Rothman
Microphone Preamplifier (for Vocoder) – Part 1
I
’m going to start an ongoing
series of top-quality modules for
studio work. Soon I’ll be doing an
analogue vocoder, to make that sound
personified for eternity by ELO’s Mr
Blue Sky and Queen’s Radio Ga Ga.
The first thing we need for this project
is a microphone pre-amplifier, which is
the topic of this month and next. I need
to buy time to get the Vocoder PCBs
together, since the original design used
the obsolete CA3080 chip. Production
was discontinued in 2005 to the dismay
of the music technology industry. The
basic mask of this chip lives on in the
LM13700, which fortunately is still
being made. To hear the Vocoder in use,
visit my son’s Soundcloud and listen to:
http://bit.ly/pe-may21-voco
Standing alone
A vocoder could use just a single inverting op amp stage for a mic preamplifier,
like the Korg and Roland designs. However, I’ve found using top quality external
pre-amps and equaliser units with vocoders gave better results. My design can
be made into a fully fledged stand-alone
unit comparable with those costing hundreds of pounds.
Input impedance and noise
Microphones require dedicated amplifiers because their output is often only a
few mV at an output impedance of
around 25Ω to 600Ω. This means
standard amplifier circuits suffer
from additional noise because they
have an optimum source impedance (OSI) for minimum noise,
which is generally much higher
than the microphone’s output impedance. For the 5534 op amp the
OSI is about 5kΩ
means that
special tricks have to be used to reduce the effective OSI. Full-noise
analysis of amplifying devices is
complex, there’s also flicker noise,
shot noise, popcorn noise, contact
noise, contamination effects, thermal leakage currents and crystal
defects. We haven’t got the space
to cover them all here, but do see:
http://bit.ly/pe-may21-noise
The oldest method to match a
low impedance to a higher one is
a step-up transformer from the microphone to the amplifier, shown in
Fig.1. This is still regarded by many
studio engineers as the best sounding approach. Naturally there are
Fig.1. Early microphone preamplifiers used a step-up problems associated with transformtransformer on the input – here, the Mu-metal can
ers: they have a droopy frequency
at top. This 1970s Astronic module used a BC109
response at the extremes, hum pickinput transistor followed by a 741 op amp. The
up, low-frequency distortion and
transformer, an OEP X187B, costs £64.00 inc VAT
cost around £60.00.
from RS (stock No. 210-6352). The swivel mounting
The most cost-effective methto find the minimum hum pick up was a great idea.
od (ie, not using a transformer) of
Practical Electronics | May | 2021
Fig.2. The Electro Voice RE20 is a
microphone that benefits from a lownoise low impedance mic preamplifier.
It’s the ‘sound of American FM radio’.
(Photo: Harvey Rothman)
reducing the effective OSI is to use a pair
of low-base-spreading resistance (Rbb) bipolar transistors run at a relatively high
collector current, followed by op amps.
Low resistance is required because all resistances make Johnson noise, a result of
thermal agitation of electrons. The higher
the resistance the worse the noise. (This
can be reduced by cooling your electronics with liquid helium!) But unless you
are into astronomical electronics with its
astronomical costs, we’ll leave esoteric
cryogenics to Jodrell Bank. It’s best just
to follow the basic audio rules for lownoise design, ‘keep all your resistors and
impedances as low as possible’ and ‘do
as much amplification in the first stage as
possible’. This is so any noise produced
isn’t amplified later.
A microphone (or any source) adds its
own noise. Microphones have Johnson
55
noise and that produced by Brownian motion of air molecules on the diaphragm.
The aim is to make the amplifier noise
at least equal to the inherent noise of
the source. Then the noise will only increase by 3dB.
All condenser microphones have an
internal amplifier, and this is often the
dominant electrical noise source for these
types of microphones. Some dynamic
(electromagnetic) microphones, such as
the American broadcaster’s mic of choice,
the Electro Voice RE20 with 150Ω output
impedance shown in Fig.2, need an especially good preamplifier. My design has
an optimum source impedance of around
200Ω, but the actual input impedance is
around 10kΩ to reduce loading. This preamplifier even managed to make my very
cheap (£5) Sony F-99A stereo dynamic
mic sound good, before it always sounded muffled and noisy – see Fig.3. I had to
snip off the F-99A’s mono jack plug cable
and rewire it for balanced operation with
XLR connectors.
Components
For consistency, mic preamplifier input transistors have to be selected for
low noise. Most audio companies use
specially made test jigs to screen the
transistors and op amps. When I worked
in the mixer industry in the 1980s the
noisy ones were weeded out and used
in bass equalisation and LED drivers.
The really bad ones were put into synth
noise generators. There was a whole selection of noise grades. It is possible to
buy pre-selected, low-noise devices, but
you pay for it. More modern devices are
much more consistent.
Transistors
Low Rbb transistors are hard to find because it’s a parameter hardly ever specified
in data sheets and difficult to measure.
It was only provided for such devices as
the obsoleted Rohm and Toshiba transistors, made especially for moving-coil
vinyl pick-up preamplifiers, such as the
2SB737 with an Rbb of 2Ω. There is now a
thriving industry selling fakes from Hong
Kong on eBay, and yes, I have a drawer
full of these useless transistors.
Many engineers have their favourite low-noise transistors selected from
general devices. For years, one popular
one was the 2N4403 which had an Rbb of
40Ω. Europe’s first low-noise transistor,
the BC109 comes in at 400Ω.
Horowitz and Hill, The Art of Electronics authors measured loads of devices and
their choice for a ribbon mic preamp (the
lowest impedance device in audio) was
the Ferranti/Zetex/Little Diode ZTX751,
which was 1.7Ω.
The other method is to put lots of
small transistors in parallel, a trick used
56
in Ortofon preamplifiers. Other popular
types have been the BD139/40 power transistors at 30Ω, and the TO5 types such as
the BC461 and BC143 at 20Ω. I use the
Philips BFW16A, a medium power-RF
transistor, which because of its interdigitated structure, is effectively many
small transistors in parallel, reducing resistance. A photo of the chip is shown
in Fig.4 and the Rbb comes in at around
5Ω. Finding a surplus stash of 200 was
also a factor in deciding their use.
PNP devices are sometimes slightly quieter (also lower Rbb) than complementary
NPN types. John Linsley-Hood said this
was due to lower noise generated at the
surface of the crystal resulting from the
recombination of holes and electrons.
I’m not going to go into the physics behind this, but I’ve found the difference
to be rather subtle in practice. I’ll put
provision on the PCB for reversing polarity by links to different power rails if
required. The lowest noise transistors
are JFETs which don’t have partition
noise (where the base current splits off
from the collector current). The classic
device is the 2SK170, a real low-noise
FET, no longer made of course. There
are replacements from InterFET which
cost a fortune.
Generally, the larger the current the
devices are run at the lower the OSI and
noise up to an optimum point, specific
to the device. This is often indicated on
a special noise factor diagram shown in
Fig.5. Noise factor is the difference in dB
between the theoretical Johnson noise
and the actual noise.
Op amps
My go-to audio op amp is the 5532/4.
Strangely, you effectively get two
5534 op amps in the 5532 dual op
amp for less than the price of one!
Many more 5532 duals are made than
singles, which is reflected in the lower cost. There are small differences;
the single has an external compensation capacitor which gives better
high frequency response at high
gains. Also, the 5534 has slightly
lower noise, especially the selected 5534A.
Op amp noise is usually specified
in nanovolts per root hertz (nV/ Hz)
with the 5534A at 3.5nV/ Hz and
the 5532 at 5nV/ Hz. Good discrete transistors are typically 0.5 to
1nV/ Hz. To work out the actual voltage produced depends on the noise
current (pA) as well as the noise
voltage and circuit configuration.
The bandwidth is also important.
For audio work we usually take that
to be 20kHz, so remember to do the
square root of 20,000 (~141). Thus,
our 5nV/ Hz becomes 5×141=705nV.
Fig.4. The BFW16A chip. This transistor
is composed of interdigitated strips of
small transistors. This reduces the basespreading resistance Rbb to around
5Ω. An ideal choice for low-impedance
amplifiers. (Photo by Dr Joe Botting)
There are very few op amps that are a
useful upgrade from the 5534/2 in audio
work. The only one I have found is the
LM4562. This is only available in a dual,
so for future upgrades we need to design
the PCB for dual devices. Note also that
the maximum supply for the LM4562 is
±18V, rather than ±22V for the 5532/4.
Resistors
As well as Johnson noise, resistors also
generate additional noise caused by micro
arcs between the conducting particles.
This noise is proportional to the current
flowing and is called ‘excess noise’. The
more homogenous the resistive material,
the better. The best are Charcroft Vishay
tantalum foil resistors, and the worst are
old Erie carbon composition resistors,
which are basically soot particles bound
together with glue. For cost-effective professional audio work, axial-leaded 0.25W
Fig.3. The cheap Sony F-99A 200Ω
microphone (left) was previously unusable with
standard mic preamplifiers and even the good
AKG D7 LTD 600Ω live vocal mic required
extra high-frequency response. These are both
transformed with a good preamplifier, such as
the one we will be building next month.
Practical Electronics | May | 2021
full-size MR25 vapour-deposited
Nichrome metal-film resistors are
generally used. Watch out for the
cheap metal-film SMT types that are
really thick-film cermet types, which
are noisier. Increasing the volume of
resistive material increases the effective number of particles, so higher
wattage types are generally quieter.
Going to pot
The gain control pot is a difficult
component to obtain because it has
to be anti-log (C taper), in that the
rate of change of resistance gets less
as the pot is rotated clockwise. This
is the opposite to a normal log volume control (A taper). It’s necessary
to have the right control law or taper in these situations to make sure
the effect is evenly spread over the
entire rotational range. The anti-log
Fig.5. Noise figure map for the Toshiba 2SA1312, C taper is still not enough to avoid
a jump at the high-gain end of the
a modern SMT PNP low-noise audio transistor.
rotation. Alps make a special taper
It can be seen that to get the lowest noise factor
called ‘RD’, designed specifically for
of 1dB at the lowest source resistance of 40Ω
this application, but you have to buy
a collector current of 7mA is needed. Note the
specification for this noise factor can vary between thousands. All pots are noisy, most
using carbon composition film. There
0.2db and 3dB, so selection is still required.
is the moulded track type which is
thicker but still a composition device, or the more modern conductive
plastic, which is epoxy bonded carbon. These types are quieter.
One thing I often do to reduce pot
noise is to use a dual one wired in
parallel. This does halve the resistance, but reducing the pot value
down to 1-2.5kΩ gives a less abrupt
jump at high gains at the expense
of a higher minimum gain. I use
a switched dual-gang 5kΩ pot and
switch it out altogether for minimum gain – see Fig.6. If you are very
Fig.6. A special Blore Edwards pot for the
fussy, you can go the route of API or
gain control (VR1). It is a CTS 45 series, 5kΩ
Neve and have a gold-plated Elma
reverse-log dual-gang with switch type.
switch with 1% metal-film resistors
calibrated in equal dB steps.
Capacitors
Fig.7. Tantalum capacitors offer high values
with low leakage currents (10-20µA) – although
manufacturers won’t guarantee it. (You should
check your devices!) Getting values high enough
(over 100µF) can be difficult. The metal capacitor
to the right is a bolt-mounted Castanet wet type
and the axial capacitor is a standard metal-case
solid European TAA equivalent to US CTS13.
Practical Electronics | May | 2021
Turning to capacitors and noise,
plastic film types are the quietest
because they have very low leakage
currents. Where high capacitance is
required and electrolytics have to be
used, tantalum types (shown in Fig.7)
have lower leakage than standard wet
aluminium, especially where they
have not been used for a while and
they have to reform. X7R ceramic
multilayer types are very poor; they
actually act like piezo microphones
themselves; ie, they are microphonic.
Tap them, and they go ‘ding’.
Next month
That’s all for this month, next month
we’ll design and build the circuit.
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
- 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
57
