A Zero Feedback Power Amplifier, for Audio and
perfect amplifier is wire with gain."
This design originated in 1994 as I needed an
amplifier to drive my spheroidal enclosure loudspeakers.
I had followed the 'valve - solid state' debate for
quite a while, and could hear a clarity from valves which I felt
was often lacking in solid state.
Why was this, when in most respects most solid state
amplifiers can achieve far better technical performance than even
good valve designs? Is it that the commonly used specifications
for measuring audio quality are missing something?
The following table illustrates the fundamental
operational differences between the two technologies.
- Transformer Output
- Both output devices same 'polarity'
- Single Supply Rail
- Few individual active devices.
- Low or zero overall feedback.
- Benign clipping when overdriven.
- Direct Coupled Output
- Output devices N & P ( or pseudo P )
- Dual supply rail
- Many active devices.
- High negative feedback; to linearise
- Harsh clipping when overdriven.
Table 1. Valve vs. Solid State
Having considered the matter and listened to various
systems and mixed and matched different loudspeakers the one
aspect of amplifier design that didn't seem to be given much
attention is Phase Distortion. I happen to have hearing that is
particularly sensitive to loudspeaker type and discovered that
co-axial or point source speakers give a far better stereo image
than conventional drive arrangements. Further study led me to
conclude that this "muddy" sound was caused by phase
issues, particularly noticeable in the crossover regions.
Loudspeakers by their nature are a "nice"
reactive load for any amplifier, particularly when using multiple
drivers of different characteristics and passive crossovers. The
higher the negative feedback from the output the more susceptible
to this load an amplifier will become. Valve amplifiers often use
far less output stage negative feedback and some can be
configured to use none at all. Whilst this increases the harmonic
distortion many valve amplifiers are still considered to sound
better than a nominally equivalent solid state counterpart.
This line of thought ended in the ideal of an
amplifier that had no overall negative feedback whatsoever, yet
was robust and easy to construct from readily available
components and accessible suppliers. The design present here is
the embodiment of that thought.
What is an audio power amplifier?
In it's most basic form an audio power amplifier is an
electrical circuit that drives current derived from an input
signal into a voice coil.
I had read somewhere the phrase that the perfect
amplifier is "Wire with Gain". By using two
transformers, the first as a voltage amplifier, the second with
MOSFETs as a current amplifier, I believe that this is achieved.
Figure 1. Amplifier Schematic
A transformer T1 coupled input which produces a
differential drive to a pair of power semiconductors Q1 and Q2,
in this example power MOSFETs. The resistor R3 across the
secondary produces a roll off at high frequency and is chosen to
filter the input signal so that it is below the resonance point
of the transformer. The centre tap of the secondary is offset
from ground by a voltage reference / regulator to bias Q1 and Q2
(e.g. by a zener diode with resistor to the supply and a
smoothing capacitor to ground, or an active voltage regulator).
V-BIAS is typically between 3.5 and 4.5 volts,
depending on semiconductor manufacturer. Start with this voltage
under the voltage drop specification of the MOSFETS and then
slowly increase it whilst monitoring the quiescent current. Only
a few hundred milliamps are required for the amplifier to be
Q1 and Q2 are used as 'gate followers' which alter the
voltage on, and thus the current flowing through, the output
transformer T2 primaries, the centre common of which is to
ground. The output transformer's secondary is matched to drive
the load, in this instance a loudspeaker. The secondary, although
not necessary for the basic operation of the amplifier, is shown
as centre-tapped to drive to the load differentially, thereby
reducing the radiated field from the interconnecting cable.
In a no signal state both Q1 and Q2 are at the same
potential and equal current flows through both halves of
transformer T2's primaries, thus cancelling out the magnetic
flux. Any input signal on T1 is magnified by the turns ratio and
causes Q1 and Q2 to follow the voltage on their gates, one device
rising whilst the other falls, and visa versa.
NOTE: The transformer effectively generates the
negative rail, so the transformer primaries WILL swing the same
amount negative as it does positive - less the bias voltage. I.e.
if one has a 45 volt supply the MOSFETs must be rated at a
MINIMUM of 100 volts.
The total output power available to drive the load is
determined by the design of transformer T2, with higher powers
requiring a bigger transformer with lower impedance primaries,
higher current semiconductor devices with larger heat sinking, and
a higher current power supply. However the operating voltage of
the amplifier does NOT need to be raised to increase the power
output, unlike a direct drive semiconductor amplifier.
N.B. The input needs to be driven by a proper balanced
line driver from the pre-amp as the standard phono outputs
are not sufficient. (For demonstration purposes the amplifier may
be driven from an ordinary headphone output such as found on a
portable radio, CD or tape player.)
Input transformer input impedance:
Parallel = 165 Ω.
Series = 667 Ω.
See Input Transformer specification for
- R1-2 are 200 ohm non-inductive 1/4 watt
- R3 is selected by tuning first with a 470K pot
depending on the transformer specification, then is
replaced with a fixed non-inductive 1/4 watt resistor.
- Z1-2 are 12V Zeners to protect Q1-2
- Q1-2 are IRFP150N MOSFETs, mounted on a heatsink
of 300 x 75 mm with 40 mm fins.
Output Impedance c. 2.78 Ω @ 1 kHz
... and this is what it looks like
There is no overall negative feedback. The only
feedback mechanism is within Q1 and Q2 as they operate in voltage
follower mode and regulate the voltage across the source / drain
to match that of the gate ( less the semiconductor voltage drop
For a given power output, three times the supply watts
are needed. I.e. For 50 Watts output use a 150 Watt rated supply
(for continuous full power operation). The power supply (not
shown) for the amplifier does not need to be closely regulated as
Q1 and Q2 take their reference from the input transformer centre
tap regulated supply. Ripple on the main supply is not a problem,
and a standard bridge and capacitor on the output of a mains
power transformer is all that is required. (With a 35 volt supply
I have used a 10,000uF reservoir capacitor.)
The bias voltage is generated from a voltage regulator
mounted between the two MOSFET output devices. This ensures that
should the output become excessively hot the bias voltage will be
automatically removed due to the regulator's thermal shutdown
The amplifier input being transformer coupled presents
an isolated low impedance input which prevents any ground 'earth'
current loops between the power and pre-amplifier stages.
Additionally it matches the cable impedance providing a better
termination characteristic and reducing or eliminating cable
reflections, and allows a long interconnect between the pre-amp
and the power amp which may then be sited close to the load, e.g.
loudspeaker. The low impedance input also has the benefit of not
producing loud hums from mains etc. pickup if the input is
touched by hand.
The amplifier output being transformer based is also
isolated and will not be subjected to 'earth' current loops
should the loudspeaker need to be ground referenced remotely from
the power amplifier.
The amplifier input and output both being
transformers, i.e. inductors, provide good RFI/EMI shielding.
Driving input and output differentially ( balanced in audio
parlance ) again minimises radiated emission, and any noise
pickup on the input cable is cancelled out.
The amplifier if starting to overdrive as the input
signal is increased does not immediately hard clip to the supply
rails on the peaks but produces, when used for audio, audible
harmonics. This warns that the input should be reduced before
excessive current flows in the loudspeaker's voice coil(s). As
the output is transformer coupled no true DC can be generated.
The amplifier's signal to noise performance is very
good. I have measured down to -130 dB, at which point I gave up
as I was unable to shield the test set-up from mains and radio
interference below this level. Basically this translates into the
practical result that the amplifier itself is totally silent.
When connected to a loudspeaker most amplifiers produce a
background "hiss" which can be heard by placing an ear
very close to the speaker (be careful to ensure no audio signal
can be applied by shorting out the input signal).
Another benefit of the amplifier's design is that no capacitors are
used in the audio signal path. There is much debate as to the effect (or
otherwise) of capacitors so used, but it is now beginning to be recognised that
some types of capacitors may indeed produce audible distortion. The detailed
causes and which types are best or worst is still very much under discussion -
see C. Bateman's series of articles in Electronics World for further
information. Volume 108 (2002) July (part1), September (part2), October (part3),
November (part4), December (part5), and February 2003 (part6).
Finally, if the amplifier is driven by the pre-amp,
but without it being itself powered, a signal can still be heard
from the loudspeaker, although feint and distorted. This
demonstrates that there is a direct electrical path between the
input and the output - a feature which I believe is unique to
I have bread boarded a single ended version of this
design, reconfiguring T1 for a single secondary to drive one
power semiconductor (Q1) and connecting both T2 primaries in
series . The principle stays the same but the output transformer
must be capable of taking the DC current without saturating, and
the voltage reference to T1 secondary must be better regulated as
any noise here will be coupled to the output by transformer T2
rather than cancelled out by it. It works at a quarter of the
power of the push-pull version, but I didn't make any
measurements as to distortion, etc.
The output stage ( Q1, Q2 and T2 ) could also be
driven by a discrete semiconductor or valve stage, or by an
op-amp, configured for balanced drive, where the isolated low
impedance input is not required. E.g. in an integrated pre and
power amplifier. The preceding paragraph comments on single ended
operation also apply.
- Zero feedback.
- Balanced and isolated input and outputs.
- Output configurable to drive any impedance load,
from electrostatics to sub-ohm parallel linear array
- Good EMI/RFI performance.
- Low voltage operation ( sub 60 volts ), no
potentially lethal HT.
- Only two power semiconductors required.
- Minimum components, little opportunity for noise
generation (-130 dB signal to noise ratio).
- No capacitors in the audio signal path.
- "Direct" signal connection from input
Design by: Susan Parker, MIEE.
The information contained here may be
used to construct one set of power amplifiers specifically for
personal NON commercial use only.
N.B. Personal liability disclaimer
This page last modified on: 23rd October
All information, drawings and images Copyright © 1994 -
2005 Susan Parker unless