This is only a preview of the February 1991 issue of Silicon Chip. You can view 47 of the 104 pages in the full issue, including the advertisments. For full access, purchase the issue for $10.00 or subscribe for access to the latest issues. Articles in this series:
Articles in this series:
Articles in this series:
Articles in this series:
Articles in this series:
Articles in this series:
Articles in this series:
Articles in this series:
|
A practical approach
to amplifier
output stage design
There are many requirements for a
successful amplifier design. One of the
least understood is how to design the
output and driver stages to remain
within their safe operating areas. Faulty
design here can be more than just
expensive; the smoking remains are
down right embarrassing.
By DAVID EATHER
Many cope with this task by copying from other designs. This is a limiting approach and almost invariably
leads to designs with a certain sameness about them. Also, bad design is
perpetuated. There is a better way.
This article shows a practical approach to amplifier output stage design and covers the calculation of
power supply voltages, output load
. lines, derating transistor SOAR curves
and heatsink selection.
You get nothing for nothing so before you start there is a fair swag of
calculations to be done. You also need
access to a power transistor data book.
The payoff is reliable amplifiers with
output. power levels customised for
your needs.
There are no definitive answers as
to exactly how many or what type of
transistors you have to use in your
14
SILICON CHIP
design; a lot depends on taste. At the
same time, there are certain design
rules that should not be transgressed.
My method is a simplified approach the overall aim has been to achieve a
reliable design without too much
pain. What has been shaved off one
area is generally compensated for elsewhere.
For illustration, I will be designing
the output and driver stage of a general purpose amplifier capable of 25
watts into 8Q. The circuit is a simple
10-transistor design as shown in Fig.l.
After a general discussion on each
step, I will provide some specific results, so you can check your understanding of the principles being discussed.
Assuming you already know how
much power output you want, the
first step begins with calculations to
find the peak voltage and peak current delivered to the load. Use the
following formulas.
Vmax load = ✓ (2 x P x Z)
lmax load = ✓ (2 x P/Z)
For my amplifier this works out as:
Vmax load= ✓ (2 x P x Z)
= ✓ (2 X 25 X 8)
= 20 volts
Imax load= ✓ (2 x P/Z)
= ✓ (2 X 25/8)
= 2.5 amps
Emitter resistors
At this point, I will give a quick
mention of the emitter resistors, Rl
and R3, on Fig. l. These resistors help
provide thermal stability of the output stage bias current and in designs
with output transistors in parallel they
help to ensure equal current sharing.
The higher the resistance the better
the thermal stability and current sharing but the more power they waste.
The final value is a compromise.
As a guide, you would normally try
for about 0.6 volts across the emitter
resistors at lmax load.
For this design, 0.22Q should prove
adequate. The resistance is a little
low but I would not expect any problems for the following two reasons:
(1) there is only one output transistor
for each rail so there is no current
sharing; and (2) I don't intend to set a
high quiescent current.
The next step is to work out the
required supply voltage (±Vee). You
must consider the requirements of
Vmax load, the voltage drops caused
by the driver and biasing circuitry
and the voltage drop caused by the
emitter resistors in the output transistors, and lastly the ripple voltage
(hum and audio signal) on the supply
rails (±Vee).
At this stage, refer to back to Fig, 1.
, - - - -.....- - - - - - - - - - - - . - - - - - - . . . . . - - - + V C C
R1D
R9
01
?
09
BC547
R17
07
BC549
R9=R10=2R8
C4
06
-YCC+1.4Y
BC639
RB
...__ _ _----,.__ _ _ _ _...__ _ _ _ ___.____ _ vcc
Fig.1: the circuit for a general purpose 10-transistor audio amplifier capable of delivering about
25W into 8Q. Note that the design uses split supply rails and complementary output stages.
Notice that the bias voltages applied
to the bases of Q5 and Q6 are equal.
Notice also that the outputs, drivers
and pre-drivers are mirror images of
each other (R5, Q5, Q2 and Ql vs R6 ,
Q6, Q4 and Q3). This is a common
situation and allows calculation for
±Vee by considering just the positive
side. In cases where the quiescent
biasing varies for the positive and
negative sides, the ±Vcc rails are
worked out separately. Depending on
the design, use the larger value for
both supply rails.
In this example (see Fig.1 again),
there is 1.4 volts across the base-emitter junction of Q5 and R3. We will assume that at full power, the same voltage of 1.4 volts appears between the
collector of Q5 and +Vee, 0.55 volt
peak (Imax load x Rl) across Rl, and
1.6 volts across the base-emitter junctions of Ql and Q3. This gives a total
overhead of 3.55 volts.
Power supply ripple
Next, you have to make an estimate
for the ripple on the power supply.
For this I like to use what I call
Eatber's rule of thumb. Stated as a
formula it looks like this:
Vripple = 6300 X lmax load/Cwhere Vripple is the peak to peak voltage ripple on the power supply and C
is the filter capacitor size in microfarads. The capacitors should be a
minimum of 100-200µF per watt of
output power for a class B amplifier
with a full wave rectifier. (Actually,
Eather's rule of thumb is not just a
whim of mine but is a condensation
of the maths for capacitor-input power
supplies).
For this example, I have elected to
use two 2500µF capacitors for each
power supply rail:
Vripple = 6300 X lmax load/C
= 6300 X 2.5/2500
= 3.15 volts
The value for ±Vee is:
±Vee= Vmax load+ Vripple + circuit
overhead
For my design this becomes:
±Vee= Vmax load+ Vripple + circuit
overhead
= 20 + 3.15 + 3.55 = 26.7 volts
This can be safely rounded off to ±27
volts.
Transistor load lines
OK so far? The next step is to figure
out the transistor load lines. We are
not going to bother with the load lines
for resistive loads. These are straight
lines and not really the problem for
amplifiers. We are concerned with
reactive load lines. These show the
instantaneous voltage and current
flowing through the transistors when
driving a complex load impedance
such as a speaker.
To do this you need the output
power, Imax load and ±Vee. You also
need the value of the emitter resistors
in the output stage (Rl, R2 in my
case), the power output of the amplifier and the load impedance, Zl.
Before leaping into the computations , we need to make an estimate of
the maximum phase shift caused by
the inductive portion of the speaker
load. 45° seems to be the accepted
standard in many electronics magazines and is the value we shall use
FEBRUARY1991
15
20
" "- ....
le = lmax load x sin(wt - 0)
Write down each result in turn for
the value of wt. Next is Vee using the
more complex formula:
Vee = Vee - Imax load X Zl X
sin(wt) - le x RE
The column for Ppk(W) is calculated by multiplying the collector
emitter voltage Vee by the collector
current le:
Ppk(W) = V ce X le
Table 1 shows the results for the
amplifier under discussion.
I'\
10
....
" I'--
~
1 ",
1"
I
Sn LINE
"-....
\
-........I'-
V
I
'\
I
\
\
\
\
Load variations
I
I
I
0.5
I
0.3
0.2
I
\
\ \
\
\ I
I
\ I
3
I
20
10
30
60
Vee (VI
Fig.2: the 4Q & an load lines for the
output transistors of the amplifier.
These curves were plotted using the
data shown in Tables 1 & 2
respectively. Note that the load lines
should be fully enclosed by the DC
SOAR curve of the selected transistor
as shown here.
here. However, if you intend to use
your amplifier with highly reactive
loads such as electrostatic speakers
or line transformers, 60° would be a
better choice.
Drawing up a table
Now we draw up a table with five
columns and 13 rows. The columns
are labelled: (wt - 0), wt , le, Vee and
Ppk(W). Theta (0) is the electrical
phase shift caused by the speaker.
The term "wt" is the instantaneous
phase of the signal frequency and is
expressed in degrees. le is the instantaneous current through the collector
of th e output transistor. Vee is the instantaneous voltage across the output
transistor. Ppk(W) is the instantaneous power dissipated by the output
transistor.
The (wt - 0) column starts at 0 and
steps up to 180° in 15° increments.
Down the wt column write the corresponding value of wt. This is the same
as adding the selected value of 0 (45°
in our case) to the adjacent value of
(wt - 0). This leaves the wt column
with values starting at 45° and ending at 225 °.
Start the calculations with le, using
the formula:
16
SILICON CHIP
Now take a deep breath. A general
purpose amplifier could drive all sorts
of speakers, some with only a very
nominal 8Q impedance. For amplifiers in this situation, it is normal to
design the amplifier so that it can
safely drive into half the nominal load
impedance. This may not be necessary if the amplifier is to drive a
known speaker impedance or if using
electronic limiting. If electronic limiting is not done carefully though, the
amplifier may produce objectionable
distortion if pushed hard into a nonresistive load.
The rule of thumb for estimating
power output into half the nominal
load impedance is that the amplifier
will produce about 50% more power.
This won't apply if the amplifier has
a well regulated power supply and
large filter capacitors, in which case
the power output will be closer to
double. Conversely, if the power supply has poor regulation and small filter capacitors, the amplifier may only
deliver a few percent more power into
half its nominal load impedance.
First, assume your amplifier will
deliver 50% more power. Then you
have to check that your amplifier will
really deliver this power into the new
load. Why? Because if it can, it will
have to dissipate a lot more power
and we need to know that the transistors can stand this extra stress.
Work out the required value for Vee
for the increased power output. This
means going through the same procedure you did before, finding the required voltage across the load, the
amplifier overhead and the power
supply ripple using the new load impedance.
A required value for Vee much
larger than that available from your
power supply means that the amp1ifier won't be able to deliver the extra
50% power, except maybe for short
peaks. If the required value for Vcc is
less than the actual supply, the amplifier will deliver a bit more than an
extra 50%.
Most times, allowing 50% gives a
close estimate of what will actually
happen. If, in your case, the value
you came up with for Vee was very
different, adjust your estimate of output power and go through the checking procedure again.
For my amplifier, half the load
impedance equals 4Q and I expect
the amplifier to deliver about 37 watts.
The calculations for Vmax load and
lmax load give 17.3 volts and 4.33 amps
respectively.
The amplifier overhead goes up by
0.4 volts to 3.95 volts due to the higher
current through the emitter resistors.
The ripple on the supply also increases to 5.45 volts. So the Vcc
needed is still about 27 volts.
The next step is to calculate another load line for the new load impedance. Use the new values for Zl
and lmax load.
My results for output into a 4Q load
are tabulated in Table 2. We can now
draw some conclusions about the
possible output transistors. The output devices must have an le rated
higher than Imax load. The Vce must
be twice Vcc and the power rating
should be at least 50% greater than
the largest value for Ppk(W) for most
designs.
In suggesting 50% more for the
power rating of the output transistors, I am assuming power dissipation is the limiting factor and not secondary breakdown. It is just a ballpark figure and may need adjustment.
Output transistors
For your design, you should be able
to make an educated guess about what
output transistors or combination of
transistors you will need to use.
For my job, I will need the output
transistors to have an le of more than
5 amps, a Vee of 60 volts or more, and
a power rating of around 100 watts.
I c;an now select some possible
devices.
For my design, transistor pairs such
as the MJE3055/MJE2955, TIP3055/
TIP2955, 2N3055/MJ2955 or MJ15003/MJ15004 could all be suitable.
I won't use the TIP and MJE pairs
because the packages are not pin
compatible even though they are of-
HEAVY DUTY TV/SPEAKER
WALL-CEILING BRACKETS
Table 1
rot-0
0
15
30
45
60
75
90
105
120
135
150
165
180
rot
45
60
75
90
105
120
135
150
165
180
195
210
225
Vee
le
12.858
9.537
7.406
6.611
7.205
9.148
12.308
16.469
21.347
26.611
31.901
36.858
41 .142
0.000
0.647
1.250
1.768
2.165
2.415
2.500
2.415
2.165
1.768
1.250
0.647
0.000
Ppk(W)
0.000
6.171
9.258
11 .687
15.600
22.091
30.770
39.769
46.218
47.042
39.877
23.848
0.000
rot
45
60
75
90
105
120
135
150
165
180
195
210
225
Vee
le
Ppk(W)
14.753
11.754
9.794
9.006
9.445
11.080
13.800
17.420
21 .692
26.326
31.007
35.414
39.247
0.000
1.121
2.165
3.062
3.750
4.182
4.330
4.182
3.750
3.062
2.165
1.121 3
0.000
0.000
13.172
21.204
27.576
35.418
46.343
59.755
72.858
81.344
80.605
67.129
9.687
0.000
ten sold as being interchangeable. The
mounting tab is on opposite sides
when placed into a PC board. This
has the potential for mistakes during
construction or repairs. Also, the
power rating for these transistors is a
bit low (only 90 watts). They could
work but I will look around for something else.
The 2N3055/MJ2955 pairs have a
higher power rating for only a few
cents more. This higher power rating
could-lead to cost and size savings by
enabling the use of a smaller heatsink.
The MJ-15003 /15004 pairs are nice
but relatively expensive. Compared
to the 2N3055/MJ2955 pair, they do
offer a better current gain-bandwidth
product (ft) and would give slightly
berotated
360
degrees
~:~
~1H
i1
li1
_,
1
as well as being swivelled
up or down to any viewing
or listening angle. The
metal platforms have
predrilled holes for
mounting and are easily
adjusted with a large alien
key supplied with the unit.
~~
Imparted 1111d distributed by:
Table 2
rot-0
0
15
30
45
60
75
90
105
120
135
150
165
180
The M83 and MB5 are
heavy duty two platform
mounting brackets
designed for securing
small TV's and speakers
to walls, ceilings, desks or
bench tops. When
mounted both units can
lower distortion. Also the much
higher power rating means I could
possibly get away with quite a small
heats ink.
Overall, the 2N3055/MJ2955 pairs
should perform satisfactorily in my
circuit so they are my first choice.
At this stage, the choice of output
transistors is no more than an educated guess. It is possible that the
transistors may not be suitable. The
graph of the load lines, transistor SOAR
curves and the heatsink calculations
will confirm the final choice.
This is the time to check your transistor data book closely. Enlarge or
redraw the safe operating area (SOAR)
graphs for the transistors until they
are a convenient size. Then plot the
~
ELECTRONICS
Available through the fol/owing retailers:
David J Reid
All Electronic
Electronics.
Components.
127 York Street.
118 Lonsdale St.
Sydney. 2000.
Melbourne. 3000.
NSW. (02) 267-1385.
Vic. (03) 662 3506.
SOUND AUSTRALIA
Your P.A. Accessory Specialist
Electronic Component Specials
555 Timer
4001
4007
4011
4013
-15volt regulator
LED's 5mm
Red, Green, Orange,
the same price
V28 7MHz
$0 .1 0ea
$0.15ea
$0.15ea
$0.15ea
$0.15ea
$0 .15ea
Yellow All
$0 .1 0ea
$7.95ea
HEAVY DUTY CARPETED ROAD CASE
4 Unit $100.00 6 Unit $115.00
8 Unit $135.00 10 Unit $155.00
12 Unit
$175.00
Rack cases can be made to order in
any size and any carpet colour.
HANDLES
6" Strap $1.95 s· Strap
10• Strap $4.95 s· Nylon
Briefcase Style
Chest Handle Metal
Recessed Spring Handle
[Z]
$ 2.45
$ 2.95
$ 2.95
$ 5.95
$12.95
-
Please Call for a catalogue
SOUND AUSTRALIA
28 Walker St. Dandenong, VIC, 3175
Telephone: (03) 791 1622
FEBRUARY1991
17
10
wt-e
wt
0
15
30
45
60
75
90
105
120
135
150
165
180
45
60
75
90
105
120
135
150
165
180
195
210
225
Table 3
Vee
12.858
9.537
7.406
6.611
7.205
9.148
12.308
16.469
21 .347
26.611
31 .901
36.858
41 .142
I
le
Ppk(W)
0.000
0.032
0.063
0.088
0.108
0.121
0.125
0.121
0.108
0.088
0.062
0.032
0.000
0.000
0.309
0.463
0.584
0.780
1.105
1.538
1.988
2.311
2.352
1.994
1.192
0.000
'-
'\
'\
\
~
0.1
~
\
'\\
'
I
I\
I
I
\
.01,
1
\\
10
100
Vee (V)
wt-e
wt
0
15
$0
45
60
75
90
105
120
135
150
165
180
45
60
75
90
105
120
135
150
165
180
195
210
225
Table 4
Vee
14.781
11 .789
9.834
9.048
9.486
11.117
13.831
17.442
21.705
26.328
30.997
35.394
39.219
load lines for the amplifier using the
points you calculated above.
Fig.2 is my plot for the 4Q and 8Q
loads using the data in Tables 1 & 2.
You may be used to seeing these load
lines in text books on graphs with a
linear scale and think mine look a bit
strange. Don't worry, they are the same
type of graph only the scales arf) logarithmic.
Notice how the 2N3055 SOAR curve
fully encloses the load lines. If this
were not the case, then the amplifier
may die on the first occasion it is
required to give a big burst of power.
Make sure the load line~ are fully
enclosed by the DC SOAR curve of
your transistor. The fully enclosed
load lines show that the selected
18
SILICON CHIP
le
Ppk(W)
0.000
0.056
0.108
0.153
0.187
0.209
0.216
0.209
0.187
0.153
0.108
0.056
0.000
0.000
0.659
1.062
1.382
1.774
2.319
2.987
3.639
4.060
4.021
3.348
1.979
0.000
power transistor can work in the design.
Driver transistors
Now we come to the driver transistors.
The first step is to calculate the
load impedance presented to the
driver transistors. In my case, it is
simply the speaker impedance multiplied by the minimum beta of the
output transistor over the range of
currents of interest (0 to lmax load).
For the 2N3055, the minimum beta
is 20.
Calculate lmax for the driver by
dividing Imax load by the beta of
the output transistor. If your circuit
calls for it, make sure you add in any
Fig.3: these curves show the load
lines for the driver transistors & were
plotted using the data shown in Tables
3 & 4. As with the output devices, the
load lines must be fully enclosed by
the DC SOAR curve of the transistor.
other currents the driver transistor
must supply.
Now calculate points for the load
lines for the driver transistors in the
same way as for the output transistors . Remember to do this for half the
nominal load impedance if applicable.
My results are tabulated in Tables 3
and 4 while Fig.3 shows the plotted
load lines. Note that, as with the output devices, these load lines must be
fully enclosed by the DC SOAR curve
of the selected transistor.
Go through the same selection procedure as you did for the output transistors.
For my case, a look down the tables
on this page shows a peak power dissipation of 4.06 watts and a maximum current of 216mA. Notice that
even at these modest power levels,
small signal transistors like the
BC546/7 /8 aren't able to cope. Be wary
of designs that suggest they will. I
will try a BD139/BD140 pair for the
driver transistors.
Next month, we will see if a pair of
BD139/BD140 transistors is up to the
task of being driver transistors in the
circuit of Fig.1. I think that they will
but we'll find out for sure, next month.
See you then.
SC
|