Ultra Fidelity Amplifier Design.pdf

(299 KB) Pobierz
for so long - even the most complex_ music signal can be
represented by a Fourier analysis.
This mathematical equation lists separately each frequency
making up the signal, (together with its phase and amplitude).
However, a Fourier analysis is only complete in the case of
simple waveforms, with more complex waveforms it becomes
only a convenient approximation.
To make a Fourier analysis of a signal the components of
that signal have to be analysed over a period of time such
that complete cycles of the lowest frequency can occur. Thus
we take consideration of the time domain.
Where steady-state signals are concerned, the time domain
is not normally considered, as the signal is of a continuous
unchanging nature between any two periods. If the “time
window”, during which the signal is Fourier analysed, is
reduced progressively it becomes apparent that an accurate
spectral analysis becomes less possible. It can then be seen
that the important characteristics of the signal are amplitude
and rate of change. In other words its envelope.
WHAT DO WE WANT
What is required is the amplification of an audio waveform
in such a way that the ear can detect no degradation.
Let us consider ways in which such degradation can occur.
The waveform envelope can be distorted by amplitude
changes of any component or by changes in the phase
relationship of the component harmonics.
Experimental work has established that changes in the
relative amplitudes of the harmonic structure of the wave-
form are readily detectable.
Other work has shown that the qualitative. characteristics
of a complex sound depend upon the phase relationships of
the component harmonics. It would seem that as a phase
difference must be interpreted as a time delay between the
component parts of the signal, then a sufficient phase shift
in a system must eventually become audible as these com-
ponent parts are moved in respect to each other in time. In
practice large phase shifts are very audible and indeed tele-
phone lines are often phase and delay corrected to render
speech intelligible. However, establishing an acceptable
degree of phase shift is extremely difficult.
Following the arrival of “linear phase” loudspeakers great
controversy has raged over whether phase shifts affect sound
quality. A study of the experimental work performed to
date shows that
1. It seems to be very difficult to replicate someone else’s
experiment.
2. It seems, on balance, that where recurrent waveforms
(steady state) such as sine-waves (and instruments producing
a “continuous” although decaying tone) are concerned; then
quite large phase shifts, between the extremes of the
frequency band, have no identifiable effect on sound quality.
However, a phase non-linerarity on the leading edge of a
true transient appears to be audibly more perceptible, Par-
ticularly on speech and percussive sounds.
commonly occurs when an amplifier, with overall negative
feedback over several stages, is driven by a large enough
signal whose frequency (or equivalent rise time) is above the
open loop bandwidth of that amplifier.
Because the feedback loop is fed from the output of
the amplifier, there is no effective feedback until signal
current flows at the output, i.e. during the open-loop rise
time of the amplifier.
Very large signals occurring in the intermediate stages of
the amplifier cause those stages to distort or even to clip.
With some amplifiers this clipping can cause the stage to
latch-up for a time until the operating conditions restabilise.
Thus not only is the leading edge of the signal severely dis-
torted - in some cases it is removed completely.
TID is therefore a form of overloading that is dependent
upon both amplitude and time. It is audibly (but at a higher
signal level) similar to cross-over distortion, as both effects
cause phase and amplitude modulation of the signal due to
momentary change in gain. (Remember that at the cross-
over point zero, there is no current flow in the output stage
and hence no feedback current and so the amplifier is
momentarily open-loop.)
Circuit diagram of a typical amplifier circuit which employs
lag compensation techniques - provided by C.
B A N D W I D T H
AND TlD
Transient signals cause many problems of which phase
linearity is but one. Other problems include; instability
and ringing, clipping,’ slew-rate limiting, and transient inter-
modulation distortion.
Transient intermodulation distortion (TID or TIM)
is much in vogue but is often misunderstood. TID most
Lead compensation: components R end C provide the time constant.
ETI
CANADA-NOVEMBE R 1979
45
Ultra Fidelity, Part II
rough out a design specification before the circuit hardware
is considered. The following sequence is mandatory:
1. What parameters are important to prevent audible degrad-
ation of the signal?
2. Detail a performance specification that meets the require-
mentsof ( 1 ) .
3. Decide upon the circuit technology necessary; Bipolar;
MDSFET; ‘Tube; Class A; Class B; Switching; etc; etc.
4. Undertake a development programme to produce a
prototype.
The effects of slew-rate on a
signal
passing
through an amplifier prone to this fault.
Top: a squarewave, note the slight over-
shoot. Below that, a sineweve. In both cases
the dotted
line represents the
input.
0
L
I
0.1
1
I ”
I
10
100
*
WATTS
HARD LIMITING TRANSISTOR AMPLIFIER
% THD
t
rate would be about 5 V/p. This is, however, the absolute
minimum figure and experience suggests that such an amp-
lifier would have a hard, gritty high-frequency sound. Such
an amplifier should have a slew-rate greater than 20 V/jrs
to be certain of avoiding the increase in distortion caused by
the gradual onset of slew-limiting.
Unfortunately the higher the power output of the amp-
lifier the greater the required slew-rate as more volts swing
at the output in the same period of time and so as our 100 W
amp needs 20 V&s an otherwise identical 50 W amp needs
14 V/p and a 20 W amp needs only 9 V&s. But these forms
of distortion tend to give subtle audible effects compared to
the most common amplifier problem - that of clipping.
0.3 -
0 .25 -
0 2 -
015-
0
.05
t
CLIPPING
Clipping occurs when an amplifier is overloaded by high level
signal peaks. Such peaks occur frequently in much music
material and so the manner in which the amplifier clips
determines its audibility. A soft, clipping effect where the
distortion rises gradually (typical of valve amplifier circuits)
is audibly preferable to the hard clipping typical of transistor
circuits.
Worse still, some amplifiers tend to suffer saturation
effects on clipping and take a time to recover; thus
artificially extending the length of tirna the signal is clipped.
The use of overall negative feedback to reduce distortion un-
fortunately makes things worse. Overall feedback effectively
linearises the clipping - the distortion changes from 0.01%
(say) to 10% and quite suddenly too.
A comparison of the limiting characteristics - in general - of both
transistor and valve amplifier types. There is a body of opinion which
holds these curves to be the whole truth as to why valve amplifiers
are preferred by many musicians.
DESIGN PROCEDURE
We have covered just ‘a few of the requirements a designer
must consider when working upon the design of power-
amplifiers. There are many more to be considered to even
ETI CANADA-NOVEMBER 1979
At this point the designer has to accept that it’s a real
world and that his performance specification cannot be
achieved in a way that is acceptable to accountants, salesmen,
customers, customer’s wives or whoever else is around.
Tradeoffs are necessary and much of the “art” is in deciding
which defects and degradations are more acceptable than
others.
As an illustration of the changes in design approach over
the years we will briefly illustrate three designs for which
the author has been responsible:
1. Cambridge Audio P60 (P80)
Lecson AP3 Mk II
Mission Electronics Voltage Amplifier
::
47
1. A resistor is inserted between. Q10 collector and the
negative rail to give better balance between Q8 and
Ql
0.
2. A cascade transistor is fitted to Q13 collector to reduce
“early effect” distortion due to the collector-base capacit-
ance of Q13.
3.
An emitter resistor
is
fitted to Q13 to provide local
negative feedback.
8 ohm
The Lecson AP3 Mk II incorporates much of the thinking
in this article and is representative of the latest types of high
performance amplifiers. It is a directly-coupled Class 8 design
Illustrating the ,d using a fully complementary output stage of series connected
line conditions for
transistors and gives a power output of around 150 watts per
channel.
output stages
VOLTS
The New Mission Voltage Amplifier represents an attempt
to produce an amplifier that performs well irrespective of
load. The circuits cannot be described at this stage as they
The P60 is capable of good mid-band performance (THD
are the subject of patent applications. However, a brief
0.01% at 1 kHz is 30 W) but its high frequency distortion is
description will illustrate the philosophy behind the design.
poor because of the limited open-loop bandwidth. Generally
The casing contains two completely separate mono
this amplifier performs well at low and moderate levels but at
amplifiers, each with its own power supply. A separate
high levels its sound quality becomes hard and aggressive.
module carries the dc-voltage offset protection circuits;
Some improvements to this circuit can be quite simply made
the delay switched-on circuits; and the thermal protection
as follows:
Showing how some of the im-
provemen ts mentioned can be
added to the P60 basic design.
-
M
Q10
Full circuit diagram of the
Cambridge
P60
power
amplifier design.
) I
R55
15.6
#HOW
IT’
WORKS-Cambrid.ge
P60
from the long-tail pair by an emitter fol-
lower (Q11) to prevent any loading of that
stage worsening the distortion characteris-
tics.
Capacitor C33 gives lag compensation
which defines the dominant pole of the
amplifiers. The open-loop bandwidth is
quite high (for this type of circuit) at 12 kHz
but none the less this amplifier is prone to
TlD effects. The protection circuit is very
unusual in that the output is limited by an
FET (Q7). Ql9 and Q20 each ferm conven-
The P60
power
amplifier is of a conventio-
nal design but with care being taken to
optimise each stage. Q8 and Q10 form a
long-tailed pair with Q9 as their emitter
current source. Q8 and Q10 must be very
closely matched for minimum DC offset and
for maximum common-mode rejection to
avoid H. T. ripple appearing at the output.
The next stage is the Q13 voltage amplifier
which is loaded by a current source (Q12)
instead of the more common “bootstrap-
ped” resistors. Note that Q13 i s buffered
tional
V-l
summing circuits which monitor
the loading on the output stage.
If either Q19 or Q20 turns-on, the gate of
the FET Q7 (normally biased-off by R54 to
the negative HT) is biased positive and it
starts to
turn-on. It then acts as a potential
divider with R52 and thus attenuates the
audio signal. This protection only turns on
at the equivalent of 50 W into 2 Ohms load
and
when it turns on it only adds moderate
distortion (0.2% typically) as distinct from
clipping.
ETI CANADA-NOVEMBER 1979
Ultra
Fidelity, Part
II
circuits. Particular attention has been paid in the design to
achieving:
1. Low distortion with a very low order of overall feedback
2. Wide open-loop bandwidth with an excellent slewing rate
3. Minimum time and phase distortion
4. A high transient power capability with virtual freedom
from clipping effects.
The output stages have a very high current capability but
have no protection circuits, the output transistors being
designed to sink the full energy of the power-supply into the
load. A patented form of voltage feed to this stage gives the
amplifier a short term power delivery capability of about 600
watts (compared to the rated 150 watts 8 ohms). This
represents a 8 db increase in power availability over the
rated figure. The voltage amplifing stages are designed to clip
softly and this combined with the low-overall feedback
gives overload characteristics similar to those of an equival-
ent tube amplifier.
CONCLUSION
This feature has discussed just some aspects of modern audio
amplifier design. At present much attention is still given to
whether an amplifiir is designed around bipolar transistors,
FETs, valves, or switching transistors. However designers are
beginning to appreciate that the major stumbling block is not
designing a circuit using any of these technologies but in
deciding upon what is the performance specification required
that
will give faithful reproduction of the sound source.
Until
this problem is solved there will continue to be an element of
uncertainty in amplifier design.
Full circuit diagram for
the Lecson AP3 power
amplifier design,
producing wound 1 0 .
5W
I
I
HOW IT WORKS-Lecson AP3
Transistors Q1 and Q2 form a long-tailed pair differential amplifier
with Q3 as the emitter current source. Local feedback is applied in
the form of emitter resistors R5 and R6. The base of Q2. instead of
being grounded, is connected to a potential divider RVI which
permits the DC offset at the output to be set to zero. The input
signal to Q1 is passed through a low-pass filter (R1, C2) which sets
the bandwidth to 22 kHz (i.e. below the open loop bandwidth for no
TID effects). The bi-phase outputs of the long-tail pair feed a
second differential amplifier Q5 and Q7. Transistor Q5 has a
constant current load (Q8) whilst
is terminated by a current mirror
(Q9 and Q10). Transistor Q10 will always deliver the same current
as transistor Q9 hence the term “Current Mirror” and the excellent
symmetry and balance this stage achieves. Functionally, however,
Q10 can be considered
as an
active load whilst Q7 is a voltage
amplifier from whose
collector the
drive to the output stage is
taken. Note that Q5 and Q7
both have local emitter feedback (R17,
R24) and
that both are
buffered from the long-tail pair (Q4 and Q6
emitter followers).
Transistors Q12, Q13, Q16 and Q17 each form conventional
Darlington emitter follower stages. Each stage is series connected
to a further power transistor (Q14, Q15 and Q18, Q19 respectively)
which is permanently biased ON. Their emitter potentials are
determined by the ratio of the base potential dividers. This ratio
was chosen such that Q13 and Q15 each has half the supply rail
across them.
The whole amplifier is in the inverting mode with overall shunt
feedback through R4 and C3.
This amplifier is quite fast having an open-loop bandwidth of
about 27 kHz. The circuit is stable without the usual compensation
capacitors within the loop. THD is low being typically (at 100 W
into 8 Ohms) 0.004% at 1 kHz and 0.02% at 10 kHz. The HF
distortion can be further improved by selection of transistor Q7 foi
a device with a low collector-base capacitance.
No conventional protection circuits are used as extremely high
power transistors are fitted and these can survive a short-circuit
condition in the time taken for the power supply to shut down.
ETI CANADA-NOVEMBER 1979

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika:

Zgłoś jeśli naruszono regulamin