Build a MOSFET Citation 12

6~ NELSON P A S S
art by K IRK R A D E R

Audio Amateur has published a number of THE HARMON KARDON IG. 1 f

projects modifying popular kits, usually
Dynaco’s. Obviously it is much less ex-
pensive to use the chassis and power
supply, pots, connectors, switches, PC
boards, and heat sinks that can be had
for the price of a kit than to buy the com-
ponents individually or have them
made. Creating a one-off copy commer-
cially typically costs as much as making
MOSFE T TUBE BIPOLAR
CITATION 12 I I
/
The Citation 12, introduced about 10
years ago, is based on the circuitry
found in the RCA transistor manuals
(for example, see RCA data sheet file
#647). This circuitry was the basis for
nearly all the quasi-complementary Fil. 1. Morft, tube, and bipolar drawing.
L-

10 copies of the same item, because the _

dominant costs are design and set-up IG. * lO.OOA
time which do not increase with quanti-
ty. Small wonder then that most techni-
cians and designers who write for this
and similar magazines base their pro- /RF-I00
jects on yesterday’s readily available BIPOLAR MOSFET
kits, which, often as not, languish unap-
3.OOA
preciated in closets and basements. This
is one such project.
We will, however, do two things here
you may not have seen before. We’ll
modify the Harmon Kardon Citation
12, a popular device which has somehow
escaped the kit modifier’s attention; and l.OOA
we’ll do so with power mosfet circuitry
instead of the traditional bipolar power
transistors. This project is for experienc-
ed builders only, and Harmon Kardon
want me to remind you that it will void
any warranty coverage on the Citation
12: .3OA

PLEASE NOTE
After this article was prepared In-
ternational Rectifier withdrew the
IRF-100 from production and dis-
tribution and offered an alter- .0/A
native, the IRF-130 as a replace- GRID 2 =
ment. 3oov
Author Pass has tested these new
devices in the prototype and re-
ports that performance is as good
or better than that using the .0_3A
IRF-100’s. The change came too ov /V 2v 3v 4v 5v 6V 7v 8V
late to alter the diagrams so the text B I P O L A R - MOSFET ‘ON ’ V O L T A G E -
VS CURRENT

and drawings have been left in -35v -30v -25v -20v -/sv -/ov -5v ov
their original form. In all cases 7027A GRID ! ‘CONTROL’ VOLTAGE VS CURRENT
where IRF-100 is indicated, the
Fit. 2 Characteristics of Mosfts, tubes, and bipolars.
IRF-130 should be used.-EDITOR.

6 THE AUDIO AMATEUR 2/1%1

 =IG. 3
designs in existence, as typified by the 200
large power amplifiers of the early ‘70’s.
so --
In its time the Citation 12 was a truly 2N3055
excellent amplifier, e m p l o y i n g a VCE’4
number of concepts which have only
recently become popular: dual split 100 --

power supplies, DC output coupling, 70 --

and no active current limiting. These HFE 5 0 - -

characteristics allowed it to sonically
outperform the Dynaco 120 which was 30 --
its popular competitor. It could drive
anything with high reliability, and in the 20 --

time I spent as a repair technician I

fixed a few Dynaco, Phase Linear,
M c I n t o s h , Pioneer, a n d S a n s u i 10 -1
.2 .3 .5 .7 I 2 3 5 7 IO 2 0
amplifiers, but I’ve never seen a broken DC COLLECTOR CURRENT (AMPS)
Citation 12. Fig. 3. Bipolar current gain cwuc.
I have another, nostalgic, reason for
choosing the Citation 12: it was the first
amplifier I built myself. About eight :IG.4 20A

years ago, when I was an impecunious

student working part-time for ESS Inc. vG5= 7 V
(no, I did not design their electronics),
the home-brew power amplifier built by
a friend broke down, leaving me without
/6A--
music. Like most of you in a similar
80ySEC P U L S E TEST
position, I decided to do without some
other necessity, and shortly my kitchen
table was covered with solder blobs and
the parts to the Citation 12. 1 6V

I decided to fire up the completed unit i2A - -

on a variac instead of merely plugging it

into the wall: and was dismayed to
discover that both channels exhibited ex-
tremely large DC offset when I applied
AC power. After some weeks of recheck- 8A--
ing my work and finding no errors, I 5v
finally decided to connect it up and plug /
it into the wall anyway?* When I did,
the woofers on my speakers plunged ful-
ly forward as if attempting to escape the 4A--
amplifier; but after a few seconds they
returned to normal and music appeared. 4 v
Thus I discovered one of the Citation’s
/
few faults: an enormous subsonic turn-
on thump. 3 v
1 I
I kept the amplifier for several years 0 - I I 1
3ov
and ultimately sold it to a friend, only to 0 IOV 2 o v 4 o v 5OV
D R A I N CURRET V S DRAIN TO SOURCE O T
end up buying it back. It sat in my closet
V L A G E
-

for a long time, a candidate for that
perfect project, until one day at
Threshold we discovered a vendor had
accidentally shipped us several power
mosfets instead of diodes, and this arti-
cle began falling into place. The power
mosfets, reasonable analogs of the
hornetaxial npn devices in the Citation
12, fit quite neatly into the original
layout and provide for a state-of-the-art
super-position upon a classic piece of
circuitry, improving the speed and
distortion characteristics in an even
more simple topology than in the
original.
POWER MOSFETS
As is the case with most electronic com-
ponents, an exact understanding of
mosfet operation is a formidable goal,
and I won’t address it here, rather we
*rctr>mmrn<lrd .‘““1 rrrommcnded
Fig. 4. Characterzsttc of IRF 1001.
will treat the mosfet as a black box
possessing certain simple characteristics.
In the power mosfet we have a gain
device which combines the best proper-
ties of tubes and bipolar transistors.
Seen in Fig. 1, the gate is analogous to
the grid and the base, the source cor-
responds to the cathode and emitter,
and the drain corresponds to the plate
and collector.
In all cases the current through the
device (terminal 1 to terminal 3) is a
function primarily of the voltage be-
tween terminals 1 and 3. As the voltage
increases between 2 and 3 and/or be-
tween 1 and 3, so does the current
through terminals 1 and 3. Figure 2
shows a typical relationship of these
parameters for mosfets, tubes, and
bipolar transistors.
In Fig. 2 we note that the rate of
change in output current versus input
voltage is greatest in the bipolar tran-
sistor, a difference resulting from its be-
ing a current gain device, where the in-
put current is to be multiplied linearly
by the gain. Figure 3 gives a better pic-
ture of the bipolar device’s gain and its
comparative linearity. As Fig. 2 shows,
power mosfets and tubes have similar
characteristics, but the mosfet has much
higher transconductance and operates at
bipolar voltages. Considered overall, the
power mosfet is like a tube but retains
the high transconductance and current
of bipolar technology. Power mosfets
can also be made complementary, with
the p and n charactertistics comparable
to pnp and npn bipolar transistors,
where the devices’ polarity makes them
mirror each other. a f e a t u r e n o t
available in tubes.

THE AUDIO AMATEUR 2/ 198 1 7

 ADVANTAGES OF
POWER MOSFETS
The power mosfets used in this project
are International Rectifier IRF 130’s, a
150 watt, 100 volt, 12 amp device in a
TO-3 style case. They are part of a
relatively new line of transistors using
IR’s proprietary hexfet process, which
combines a particularly high voltage
capability with very fast switching times
and low saturation losses. Except for the
steep price tags, this type of device has a
number of advantages over regular bi-
polar transistors in audio power
amplifier use.
First, their simple low current cir-
cuitry eliminates the need for driver
transistors. Second, they are immune to
the second breakdown phenomenon
which robs bipolar transistors of their
power rating at higher voltages. This
breakdown results from the positive
temperature coefficient which en-
courages local current hogging within a
power mosfets may not necessarily be
paralleled without the use of source
resistances to equalize the current draw
between devices. Unless the devices are
matched, the designer cannot assume
that one device will not take on more or
less than its share of the load. In more
than one commercially available ampli-
fier paralleled mosfets are accompanied
by source resistors, to ensure that their
characteristics are equalized.
One more aspect affecting these
device’s audio quality is their relatively
high intrinsic capacitance, on the order
of 500-lOOOpF, which makes special
demands on any circuitry which would
realize the high speed capability. For ex-
ample, the current required to slew
lOOV/$S i n t o 1OOOpF is .lA, a c o n -
siderably higher current than most front
end circuits are designed to source
without an additional g>in stage.
:IG. 5
Fortunately, in actual use much of the
capacitance is associated with the gate-
source, which sees only a small fraction
of the gate-to-drain voltage swing. In
our project amplifier, .003A from the
front end circuitry will drive the positive
output transistor’s capacitance at about
4OVI@ for an effective capacitance of
some 100pF. This is about five times
better than the raw capacitance figure
quoted for the device, indicating that
most of the effective capacitance is gate
to source.
I must mention one final and actual
drawback. Because the transconduc-
tance is low compared to bipolars, there
is significant loss in maximum power for
a follower unless the source voltage can
deliver several more volts than the sup-
ply to the output transistor, resulting in
either lower efficiency, or the use of
tiered power supplies to feed the

area on the chip, so that at higher

voltages one small part of the transistor R3
tends to do most of the work, resulting Q6
RI :
in more probable failure. In a power
mosfet the temperature coefficient is Q4
negative, and energy dissipates more
evenly across the surface of the chip,
allowing full power application at the
highest rated voltage. This characteristic
also avoids the bipolar design problem
of thermal bias runaway, and eliminates
costly compensations to maintain reli-
able operation over a range of condi-
tions.
Third, because mosfets are majority (,
carrier devices (bipolars are minority Q 3 Q 7
carriers) their intrinsic speed is much
\
higher. Rise and fall times are about ,b / ,b

150ns while similarly rugged bipolar

R2 R5
devices have rise and fall times several
factors larger.
Fig. 5. Citation 12 conceptual schematic.
DISADVANTAGES OF
POWER MOSFETS
Basically, there aren’t any. Some ad- ZIG. 6
vantages popularly attributed to them
are however, not necesarily true. For ex- P 3
ample, the negative temperature coeffi-
Q6 D
cient does not in itself guarantee the RI X

device is indestructible. While this cha- R 4

racteristic eliminates the second break-
C l E,
down mechanism, it does not offer bet- 0
ter reliability at lower voltage levels, so
that in many well designed bipolar
amplifiers (including the Citation 12)
this effect does not provide for signili-
cantly better reliability.
In many applications the power
mosfet will exhibit greater linearity than
bipolar devices. An important exception
to this is when the device is used as a
follower driven by a low impedance
source. In this case, the higher intrinsic
transconductance allows for more ac-
curate voltage following and experimen-
tally, bipolars show about one-third the
distortion of mosfets. Fig. 6. Mozfet 12 conceptual schematic.
Also contrary to popular conception,

8 THE AUDIO AMATEUR 2/1981

FIG. 7

03

DUAL DIFFERENTIAL
TRANSISTOR

4 OUTPUT

DUAL DIODE

Y U S E D lN
LATER V E R S I O N

Fig. 7. Citation 12 schematic.

‘IG. 8

Q6
/RF- 100

/NPUT RY

. OUTPUT

210 % M Y L A R

A L L RESi5TORS A RE R L - 0 7 M E TA L F I LM 2 %, 1/4 w

UNLESS OTHERWISE SPECIFIED

Fig. 8. Mm&-t 12 schematic.

THE AUDIO AMATEUR 211981 9

 amplifier’s front end several more volts
than the output stage.
CIRCUITRY
The IRF 130’s selected for this project
are sufficiently similar to the original
HK-12 devices that they lit into the
amplifier’s quasi-complementary
topology with a minimum of modifica-
tion, as seen in the simplified schematics
of the HK-12 and the project circuit in
Figs. 5 and 6.
In Fig. 5 we see the standard differen-
tial input pair (Q, Qz) driving an npn
voltage gain transistor (Q) where Ri
and Rz are used to properly bias the
three transistors. The output of Q sees
a bootstrapped current source (Rs, Rh,
and C,) and a bias voltage source for the
output stage. Q provides a level shifting
device which elicits complementary ac-
tion from Q, the negative output tran-
sistor, and Q+ provides the follower cur-
rent gain to drive Q, the positive output
transistor. The Fig. 6 circuit is basically
the same; but Q and Q have become
power mosfets and Qt has disappeared
because the additional current gain is
not required to drive Q.
THE ACTUAL SCHEMATIC
In Figs. 7 and 8 we see the actual
schematics of the two power amplifiers;
for clarity the part numbers correspond
to Figs. 5 and 6. In Fig. 8, Rii and CT
filter the power supply line to prevent
ripple and transients from modulating
the bias, causing noise and distortion.
RI2 and C2 form a low pass filter to pre-
vent spurious high frequency input
signals from being amplified.
Resistor RI feeds 2mA of current
FIG. 9
3 A
BRN/
I
I
Fi’. 9. Power supp(p schematic.
L
10 THE AUDIO AMATEUR 2/1981
equally to Q and Qs. The 1mA of cur-
rent through Q then goes through RZ to
provide the approximately .65V bias be-
tween the base and emitter of Q. This
DC current with an AC component
modulates the current through Q,
which, passed through the high im-
pedance of R3 and R+ (bootstrapped by
C,) produces a large voltage gain at the
collector of Q. This voltage gain is
followed by Q, and on the negative half
of the circuit, by the Q,, Q combina-
tion. In this scheme of things, Rs’s ac-
tion is similar to Rz’s, and Rg limits Q’s
gain.
The combination of Ri6, Ris, and Rio
damp out the resonances which would
develop from the combinations of gain
devices and internal capacitances, and
prevent parasitic oscillation. CS is in
series with RI0 to ensure that this
resistor is not driven by audio frequen-
cies, which would overheat and destroy
it. C*, a damping capacitor, enhances
the circuit’s stability by providing a
secondary high frequency loop, causing
the feedback loop to ignore the output’s
high frequency operation in favor of the
front end’s low-order output at frequen-
cies above 800kHz. This allows the
amplifier’s input stage to dominate the
high frequency response with a less than
two-pole characteristic, providing for
high stability under transient condi-
tions.
The bias network of Figs. 5 and 6 has
been replaced by Qp, Ri7, Ris, and Cg,
forming a constant-voltage source trim-
mable by R17 and frequency-stabilized
by Cs. DS ensures symmetric clipping,
and D,, D2 provide a current path for
the flyback energy which an inductive
r’eov
RIGHT CHANNEL
BLK
4 :,
BLK/WHT I --IOOV _-- SO”
I,
I,
I 9 COM
WH T
L E F T iHANNEL
load may produce. Re, RT, and C3 form
the feedback loop, with a low frequency
rolloff of .7Hz.
A word is appropriate here to explain
the operation of the bootstrapping cir-
cuitry of RB, R+, and Ci. The idea is to
cheaply achieve the effect of a constant
current source load for Q, where the
AC impedance of the load is very high,
giving maximal gain for that stage, and
w h o s e D C i m p e d a n c e i s RS + Rh
(9.4kG). This provides about 3mA of
DC class A bias current for Q without
loading the gain stage.
This effect occurs because the point
between RS and Rd is bootstrapped by
C 1, which is connected to the amplifier’s
low impedance output and causes the
node between RS and R+ to follow the
output. Because the output voltage is
nearly identical to the voltage at the col-
lector of Q, the voltage across Ra is
nearly constant, resulting in effectively
constant current through R+, the action
of a constant current source. In this
way, Q can swing full output into an 80
load with only about 10 percent varia-
tion about the bias current, as opposed
to the f 100 percent variation that
would occur without the bootstrapping.
This contributes greatly to the system’s
gain and linearity and achieves the
highly desirable effect of the constant
current source with great simplicity and
little cost. The effect is good across the
AC band, rolling off at about .14Hz on
the low end and at about 300kHz at high
frequencies.
CONSTRUCTION
Start by removing the PC board and
stripping off all components except the
T l +42v
__z t_ 6 0 0 0
~+42V
connectors. Use solder wick or a solder
sucker, and take great care not to FIG. 11 FIG. 12

separate the copper foil from the PC

board. You will also have to remove
many of the components from the
underside of the chassis (see Fig. 13), in-
cluding the heat sinks. While you may
elect to retain the original input and out-
put connectors, I replaced mine with
higher quality gold-plated types.
You may also elect to retain the
original amplifier’s output breakers
and/or thermostats; however, I removed
them as unnecessary. Personally, I don’t
trust the quality of connection offered by
the output transistor sockets (especially
old ones) and I decided not to use them,
preferring solder connections to the
TO-3 pins. Remove and discard all the
capacitors on the underside of the
chassis except the four computer grade
electrolytic power supply capacitors.
Figure 13 shows the primary AC wir-
ing; it lacks the thermostats but is other-
wise virtually identical to the original.
Mount the power mosfet output devices
on the heat sinks, using mica TO-3 in-
sulators, silicone grease, plastic shoulder
washers, 6-32 screws/washers/nuts, and
a solder lug for electrical connection to
the transistor case (drain).
Figure 14 shows the output transistor
wiring with the lead wires soldered to
the pins. The wires from the outputs
feed down through the holes in the

FIG. 10
PARTS LIST (one channel)

Rs, Ris 1on

RN 3303
R7 4703
RB, Ru, RIS 68OQ
RI3 8203
Rs, R,I lk
Rl 3.3k
R3, R1, RIG 4.7k
Rs, RIO, Rlr 10k
RI 18k
R20 1OQ 1W 5% carbon
camp.
R17 5k CTS vert. PC mount
Pot
All resistors are % W f 2% metal film
RL-07, or * 1% RN-55D or equiv.
unless otherwise specified.
c1 470/5OV Radial Electrolytic
390pf f 5 % Dipped Silver Mica
c”: 20pF f 5% Dipped Silver Mica
470/16V Radial Elect.
c”: 4.7125V Tant.
CS 47l5OV Radial Elect.
C7 .068/1OOV Mylar
Q, Q, Q MPS-L51 M o t o r o l a
MPS-LO1 Motorola
Z:$ International Rectifier
lRF-130
D1, Dz, Ds lN4004
2 .01/1.4kV CAPS IN
POWER SUPPLY
Fix. Il. Component side of PCB. Fix. 12. Foil side of PCB
4 .47/1oov

THE AUDIO AMATEUR 2/1981 11

 chassis to the connector pins as shown in
Fig. 13. After mounting the output
devices to the heat sinks, check first for
possible connection between the tran-
sistor case or either of the two pins and
the heat sink. Use an ohmmeter with
one probe connected to an unanodized
portion of the heat sink, (possibly in the
tapped holes) and the other probe
touching the various parts of the tran-
sistor. The transistor must not connect
with the heat sink.
When you have mounted and wired
the output transistors, reattach the heat
sinks to the chassis and solder the wires
to the PC board connector pins. Note in
Fig. 13 that some of the pins are hard-
wired together to achieve the best possi-
ble connector contact by paralleling con-
nections to the PC board. Also shown in
Fig. 13 are the .47pF/lOOV capacitors
which I placed in parallel with the power
supply capacitors to achieve lower
dissipation factor at the highest frequen-
cies. These are optional and the
ZIG. 13
INPUT
GND
‘~_.
Fig. 13. Underside of chassis.
12 THE AUDIO AMATEUR 21198 1
amplifier will run quite well without
them. I also placed .OlpF capacitors
across the dioded bridges to suppress rf
emission by the diodes; these too are op-
tional.
Figures 11 and 12 display the compo-
nent layout on both sides of the PC
board and are self-explanatory. Please
note the 18 gauge jumper wires on the
back. Two replace the coil/resistor in
series with the output of the original cir-
cuit (a standard feature on 99 percent of
today’s solid state amplifiers), and to
reduce the distortion inherent in the
original PC board. The original PC art-
work, of which more than one version
apparently exists, takes the feedback
from a point separated from the actual
output node by a short length of copper
which carries an asymmetrical portion
of the output current. The feedback thus
operates off a voltage several milivolts
removed from the actual output,
generating even-ordered distortions.
As always, take special care to orient
FUSE FUSE
OUTPUT 3A P I L O T 3A OUTPUT
t. t-
_- _ _..
LINE CORD
the polarities of the components and
avoid damaging the components by
overheating the leads when soldering.
As this project is for the more advanced
home-brewer, I will not mention that
the use of a solder gun is forbidden.
Leave the TO-92 transistor leads as long
as possible. Q and Q require press-fit
heat sinks.
TURNING ON THE AMPLIFIER
When you first switch on the amplifier,
turn the bias pots so as to exhibit max-
imum resistance. This minimizes the
amplifier’s initial bias current when first
fired up, a valuable safety precaution
which should be checked with an ohm-
meter. The pot’s wrong extreme setting
should read about 0 ohms and the right
setting will read 5000 ohms, depending
on the polarity of the probes (note the pn
junction of the bias transistor).
The following test is important, and
should be carefully performed, first for
Continued on page 14
INPUT
.- ---.__._.__-./
(L3

BUILD A MOSFET CITATION 12
Continued from page 12
one channel and then for the other.
Remove the AC line fuse of the channel
not being tested and install a 1A fast
blow fuse in the channel being tested.
Drive the channel at a low level (. IV) at
1kHz by an oscillator whose ground is
TABLE 1
Distortion Figures of the Two Units
floating with respect to the oscilloscope
(you can do this by using a 3-to-2 prong
cheater adaptor on the oscillator’s AC
line if it has an earth connection). The
oscilloscope itself should be earth
grounded. Measure the output without
a load, connecting it and its ground to
the oscilloscope input.
Set the scope vertical scale at 5V/div
and the horizontal at .2mS/div. U s i n g
FIG. 15
2608
only a variac, slowly raise the line voltage
to the channel while watching the out-
put. The output voltage should exhibit
DC offset at first and then settle into a
*2V sine wave; if it does, continue to
slowly raise the voltage until the full line
voltage is achieved. If you manage this
without blowing the fuse, the next step is
to slowly increase the input voltage from
Continued on page 16

STOCK HK 12 MOSFET HK 12
Left Right Left Right
1w
20H ,025 ,027 ,010 ,010
1OOH ,024 ,023 ,009 ,009
1kH ,018 ,022 ,006 ,007
5kH ,023 ,024 ,016 ,019
20kH ,045 ,060 ,050 ,064
SMPTE
IM ,012 ,015 ,020 ,024

low
20H ,025 ,028 ,008 ,008
1OOH ,022 .023 ,006 ,007
1kH ,015 ,020 ,005 ,006
5kH ,021 ,025 ,015 ,018 2 2 LIB --

20kH ,095 ,120 ,050 ,058

SMPTE
IM ,034 ,040 ,012 ,015

3ow
20H ,022 ,028 ,007 ,008
1OOH ,025 .028 ,006 ,007
1kH ,022 ,024 ,004 ,005
5kH .027 ,031 ,013 ,016
20kH ,190 ,240 ,044 ,052
SMPTE
IM ,060 ,065 ,007 ,009 _
FIG. 16

60W
20H ,075 ,050 ,007 .008
1OOH ,090 ,070 ,007 ,007
1kH ,085 .080 ,004 .004
5kH ,075 ,076 ,013 ,015
20kH ,390 ,540 ,041 ,055
SMPTE
IM ,140 ,130 ,007 ,007

FIG. 14

DRAlN

RED LEAD

GATE
-l
0

GRAY LEAD

SOURCE - 3o”--
BLACK LEAD i

INTERNATIONAL RECTIFIER /RF-/00

/1 1 I 1 I
Fig. 14. Mosfet wiring diagram. I /O /co ,K IOK IOOK

14 THE AUDIO AMATEUR 2/l%? 1

 FIG. 17
BUILD A MOSFET CITATION 12
Continued from page 14
/ --
the oscillator until the output reaches
-60 W- RT clipping (at about f 3OV), watching for
-6OW- LT oscillation or severe distortion.
Having accomplished that, repeat the
.3 -- /
test for the other channel. When both
Continued on page 18
- - - - _ _ _ _
./ -- -
_--- -- _ --- _ i ,-/W-RT

.03 --

.0/ --

,003 - -
RESIDUAL DISTORTION
,002 ____________________--______-___---

,001 I , 11 4 I 1I I I 1I i
IO 30 100 300 lK 3K IOK 30K Fig. 21. 20k THD wauc HK-12
Fig. 17. HK-I.? distortion curues.

FIG. 18

I --

.3 --

./ -- &&k x: Fig. 22. 20k square waue HK-12

.03 -- p / z

/ /
/ y
.Ol --
=_z_-_~.=Y,_ _ _ _ _ _ ’ 2
-1 - - - - - - I ’
,003 - -

.OOZ R E S_____~~_-_~~.
_ _ _ _ _ _ _ _ ID U A L DIS T O R T IO N
_._~~~.___

/ I
.oo/ I I II I
I I
I II I
IO 30 IGC 3CO lK 3K IOK 70
Fig, 18. MF-12 distortion cumes.
Fi’. 23. 20k square ulaue MF-12

Fi,. 19 1kHz T H D uxue MF-12. Fig 20. 20k THD W~UE MF-12 Fig. 24. Current through outpuf stage MF-12.

16 THE AUDIO AMATEUR 211981


BUILD A MOSFET CITATION 12
Continued from page 16
channels operate successfully without a
load, repeat the test for each channel
with a 4A fuse and an 89 load, still
watching for waveform problems or fuse
blowing. You will see crossover distor-
tion at this point, which is normal for an
unbiased amplifier.
When both channels can successfully
drive an 83 load at full power, it is time
to bias the amplifier’s output stage. Put
most simply, we want to adjust the bias
potentiometer so as to remove crossover
distortion, a condition which cor-
responds to about lOOmA of idle current
through the output stage. We can do this
as follows.
Using the test setup shown in Fig. 2.5
(see following paragraphs) and with the
drive channel at +8V into the 8h2 load,
look at the notch distortion on the other
channel and adjust the bias poten-
tiometer until the spike just disap-
pears--and not more. Be careful here: it is
easy to set the bias too high in search of
perfection. Disconnect the load and
monitor the operating temperature of
FIG. 25
SHORTING
PLUG
Fig. 25. Distortton hookup.
18 THE AUDIO AMATEUR 2/1981
the channel for 15 minutes by placing
your hand on the heat sink. Ideally the
channel should operate slightly warm at
idle. If it becomes hotter than this, ad-
just the bias current down. If it stays
cool, you may with care, slightly in-
crease the bias.
After 15 more minutes recheck each
channel’s temperature and distortion,
readjusting as necessary. If you are for-
tunate enough to possess a current pro-
be, this entire procedure can be ac-
complished by simply setting each chan-
nel to draw lOOmA from the supply,
again rechecking after 15 minutes.
Alternatively, you may monitor the bias
current by inserting 1Q in series with the
positive power supply lead and measur-
ing 1OOmV DC across it.
USING THE AMPLIFIER
TO ANALYZE ITSELF
It may surprise you to know that one
can quite easily perform cursory distor-
tion analysis on this (and many) power
amplifiers using an 83 power resistor,
an oscilloscope, oscillator, and the am-
plifier itself. The oscilloscope need not
be of particularly high quality and the
oscillator need not have low distortion.
Figure 25 shows how one amplifier
channel (A) uses the oscillator’s signal to
source current into the other channel (B)
through an 83 resistor. At the output of
channel B we will find that channel A’s
voltage output has been divided by
channel B’s damping factor. In this
amplifier the broadband output impe-
dance is .0640 (83/.0643 = 125 damp-
ing factor), resulting in a condition
where channel A’s voltage output is
reduced to .8% at the channel B output,
plus the distortion of channel B caused
by sinking the current.
At channel B’s output we therefore
see the channel’s notch and other distor-
tions in response to channel A’s drive
conditions, but much less obscured by
the fundamental voltage. Because in this
and most other amplifiers the distortions
are dominated by current fluctuations (a
concept addressed b y t h e S t a s i s
amplifiers) this serves as a useful and in-
expensive bench technique and for the
most part it is easy to see the distortion
added to a .8% fundamental. This
procedure is useful in biasing the
amplifier as You can clearly see
crossover notch and other effects.
Continued on page 46

BUILD A MOSFET CITATION 12
Continued from page 18
This test will also reveal distortions
caused by poor connections between the
PC board and the connectors (some-
thing I encountered in the actual unit
and which I cured by cleaning and
reseating the connectors). To evaluate
the percentage of distortion remember
that the voltages seen by the ‘scope
divided by the output voltage of the
other channel will equal the fraction of
distortion. For example, if channel A’s
output is at + 28V (20VRMS) through
the 8 Ohms into channel B, then chan-
nel B’s output will show a signal com-
prised of channel A’s fundamental
divided by the damping factor (for
POWER MODIFICATIONS
for the ST- 150-BJ- 1,
DH-200, and ST-400
Continued from page 45
signal across the power supply caps when
driven hard; this signal is the combina-
tion of both channels of load, and it
couples the two channels together,
resulting in a variety of distortion pro-
ducts. The stock regulator reduces this
signal by 60dB (a l-volt signal is reduc-
ed to 1 millivolt at the regulator output);
Walt’s additions to the circuit reduce it
by 80dB (1 volt is reduced to 100 micro-
volts at the output). Either way you are
going to see a substantial reduction in
problems from this source.
IMPLEMENTING CHANGES
Pat used the Old Colony kit for the Pass
f 32V supply as the basis for his
regulator; he wanted a 50 volt output, so
he substituted a 33V zener for the 15V
zener supplied at Z2 (see Table 1). P a t
changed the value of Rs to 0.333 to give
the regulator a 9.5A max current (see
my TAA l/80 article for details). He also
added a 2200pF cap to the output to flat-
ten the impedance curve at a low fre-
quency. The Old Colony kits turned out
to be quite a bargain, as Pat ordered two
kits and received two complete plus and
minus regulators in each kit. The kits all
contained LM340T regulators, however
this is the result of my own drawing
error in the l/80 article which showed an
LM340 regulator on the negative sche-
matic, where an LM320T-15 should
have been. (Present kits include LM340T
and LM320T regulators. -ED./
We swapped the ST-150 output tran-
sistors with the regulator transistors
(TR,) because the TCG180 and 181’s
supplied by Old Colony are 2OOW,
lOOV, 30A devices, and the amplifier
output transistors will see larger power
products t h a n t h e r e g u l a t o r p a s s
elements. This change just gives a little
46 THE AUDIO AMATEUR Z/1981
* .225V sinewave) plus an equivalent of B and AB amplifiers. This gives reduced
channel B’s distortion when operated at crossover distortion and the quasi-Class
that level into 8ti. In this case a .28V A operation at lower levels which other-
peak distortion spike would indicate a wise we could only achieve in a bipolar
1% peak distortion (not uncommon in circuit with dynamic biasing.
amplifier’s with .1 ‘$J average distortion). The amplifier is sonically a significant
improvement over the original, par-
PERFORMANCE ticularly in the high end where the Cita-
Figures 15-24 document the performance tion 12’s veiled characteristic is replaced
of the prototype amplifier, which was by a detailed, somewhat sweet sound.
built without selected components. Of The imaging and midrange definition
particular note are the distortion and are also much improved; but the bass
square wave comparisons against the response (one of the Citation 12’s
original, showing a significant improve- strongest points) remains much the
ment in distortion characteristic and same-ideal for planar loudspeakers like
transient response. Figure 24 shows the MG II’s, less so for acoustic suspension
current waveform through half the out- woofers.
put stage. Note that the current through In conclusion, I hope many of you
the output device idles down instead of will try this one and enjoy yourselves.
shutting off abruptly as in bipolar Class No amplifier sounds as good as the one
margin to the design and was not really tions get more pronounced at higher
necessary. Watch the lead length on the power levels, since the output im-
transistors themselves if you swap, pedance of the unregulated supply is
however. The ST-150 heatsink is so fairly high, and also strongly frequency
thick that the leads barely make it into dependent.
the sockets, and some brands of tran- The fact that we can control sine wave
sistor make marginal contact. If in steady-state distortion with regulation
doubt, use the old devices. almost certainly means that operating-
Pat’s listening comments are more point-dependent distortions which are
than supported by my own experience; I the result of supply variations, such as
think the combination of changes dynamic compression and some types of
covered in his article turned a rather TIM, are being controlled as well. We
lackluster, low/medium power amp into have solid, measurable evidence that
an open, dynamic, exciting regulation of the supply can make a
high/medium (140W/chan) amp which significant contribution to overall sonic
is sonically the equal of anything I have performance.
heard. The photos show the care and
craftsmanship Pat has applied to the TEST RESULTS
changes, and demonstrate that the We performed tests into an 8fI nonin-
ST-150 was indeed an ideal candidate ductive load, both channels driven, after
both physically and electrically for the a 1 hour warmup period at 25W and
addition of full regulation. 1kHz signal. Unregulated readings are
taken from Dyna supplied charts.
RESULTS AND SIGNIFICANCE
The test results show something even
more interesting than low distortion or IM vs. POWER
high power. The distortion of the
regulated amplifier is nearly constant at POWER REG. AMP UNREG.
all power levels, and across the entire (W) IM % AMP’ IM %
audio bandwidth. Let me stress the im- .1 .Ol ,011
portance of this finding. 1 .Ol ,018
The normal pattern of transistor 10 ,014 ,022
amplifier designs, and of the unmodified 50 ,016 ,040
ST-150, is to have the THD measure- 75 ,017 .050
ments fluctuate by 50% of their 1 kHz 100 ,018 n/a
values, or more, at various points in the 140 ,020 n/a
frequency spectrum. Also, the varia-
THD vs. FREQUENCY, SEVERAL POWERS
Regulated Amp Unregulated Amp
Power .75W 7.5w 75w 140w .75w 7.5w 75w
Freq.
20Hz ,030 ,030 ,030 ,033 ,035 ,035 ,045
1OOHz ,030 ,030 ,030 ,033 ,035 ,035 ,045
1OOOHz .030 ,030 ,030 ,033 ,030 ,030 ,040
1OkHz ,030 .030 ,030 ,033 .031 ,031 ,035
20kHz ,030 ,030 ,030 ,032 .035 .035 ,040
 you built yourself”, and (
commercial amplifier will give ydu a:
much satisfaction or frustration. A:
careful as the Editor and I have been,
Ir;
there is probably an error or two; 1
recommend that you cross-check thr
parts list. schematic, and pc boarc
layout.
My amplifier has performed well fo1
six months, as of this writing, and I en-
countered no exotic problems during its
construction and testing. Nevertheless,
inevitably some of you will need help,
which you may obtain by calling or
writing me at Threshold Corporation,
1832 Tribute Road, Suite E, Sacramen-
t o , C A 95815-(916) 927-5061.
Although I am pleased to help, those of
you who call should be reconciled to the
possibility of only a short conversation.
Good luck, and have fun. 0 high quality ‘in sound reproduction
Audio Aids
Continued from page 50
were used throughout, not so much for
accuracy as for lower noise and increas-
ed stability.
I used LM353 bi-FET op amps in
place of the 5558’s.
The project finally completed, I hook-
ed it up to my scope to view it. Since I
don’t have a distortion analyzer, I can
only report the results of listening tests
made when I attached it to my system.
PE reports that the unweighted noise
level of the equalizer in the output was
66dB below 1V and when the measure-
ment bandwidth was restricted
250-20,OOOHz. To exclude hum and
random noise, the measurement was
74dB below 1V.
With the modifications incorporated
into the project, this 74dB can be easily
extended through the entire audio spec-
trum. Listening tests confirm this when
the equalizer is switched in and out of
the system and there is absolutely no
audible difference in noise level
throughout the spectrum. (This is with
all controls flat as I assume the PE unit
was when tested, although the article
does not state that this was done.) Even
with the lower bands turned up there is
no appreciable hum.
Although I have never heard an un-
modified unit, I am convinced these
Audio Amateur Publications
are pleased to announce their appointment
as agent in the United States for
The distinguished French publication
dedicated to a new approach to high
quality audio: both construction and
sonic arts.
L’Audiophile is a 140 page bi-
monthly, 7x9% “, beautifully printed
and illustrated magazine which ex-
plores the engineering reasons for
equipment. Published now for nearly
four years by Editions Frequences
and edited by Gerard Chretien and
Jean Hiraga, the articles critically ex-
amine all sorts of equipment and
study such matters as component ef-
fects on sound and the question of
the relevance of measurements ver-
sus subjective evaluations. M. Hiraga
brings a strong Japanese interest to
the magazine and often discusses un-
usual Japanaese products.
The magazine also devotes a large
section to Arts Sonores (sonic arts).
Articles deal with the acousti& of
Bayreuth, the characteristics of vari-
ous musical instruments, old instru-
ments and their reproduction sonic-
ally. The Arts Sonores section is
edited by Jean-Marie Piel.
Typical articles:
Techniques Sonores; output
Transformerless amplifiers; The Sound of
Turntables, a t e n t a t i v e e v a l u a t i o n ; enough for it to be a useful inpui for
Thoughts on turntables; The Koetsu car
-------------------------_--_________
Send subscriptions to:
Audio Amateur Circulation/L’Audiophile
Post Office Box 576, Peterborough NH 03458
0 Enter my subscription to L’Audiophile
for the next six issues
0 Back Issues: (circle issues desired) -
Issues 3.4,5.6 for 1978 @ 36 each ppd.
Issues 7,8,9, 10. 11, 12 @ 97 each ppd.
Issues 13. 14, 15, 16 @ S9each ppd.
(For a translated set of each issue’s contents pages,
addressed #l 0 envelope to the above address).
Please allow at least eight weeks for delivery of first copy on regular subscription. Orders
cannot be accepted from areas other than North America. Ramittancm in US S only.
Til
tridge; Defining and measuring the prin-
cipal characteristics of the high frequency
speaker; Do Asians and Europeans hear
differently? Arts Sonores; Sound
Engineers’ View: Pierre Lavoix of Erato;
Listening to the Onken-Mahul system; The
Saxhorns, a study of the tuba’s cousins;
The Concertgebouw of Amsterdam, study
of a hall; Outstanding new discs; Impulse
testing linear and non-linear systems.
The viewpoints expressed by L’Aud-
iophile’s editors are personal, opin-
ionated, and are a refreshing depar-
ture from the views generally current
in the USA. The magazine does not
take advertising and unhesitatingly
takes positions about relative merits
of equipment which includes US, UK
Japanese and F rench gear.
Although L’hdiophile is published
in French, the text is relatively easy to
translate for anyone with a year or
two of high school language study.
The written word is augmented by
copious illustrations and diagrams,
whose designations are in almost all
cases identical with those used in
English publications. With a simple
French/English dictionary and a year
or two of either Latin, French or
Spanish, the dedicated audiophile
can translate L’Audiophile well
fresh ideas about sound.
taa Ysl
USA
S 48.00
Total
Total
Total
send a stamped.

modifications were more than worth the Remittance details:

extra money necessary to purchase the I enclose S for the above in: q check, q money order
upgraded parts needed to do the job. In q MasterChargeNisaiIl (S 10.00 minimum)
listening tests the equalizer is clear, card number:
responsive, and versatile enough to
tailor many rooms to your audio needs.
It is definitely one of my favorite pieces Expire I Interbank I
of equipment.
A LAN M. VALENTE Signature
Loudonville, NY 122 11

THE AUDIO AMATEUR 2/1981 47