If you're familiar with tube ampliers, you know that the

two major methods of power stage operation are single

ended (SE) and push pull (PP). As with so many things in
life, most people are highly opinionated when it comes
to these two choices. While any confrontations between
the two camps are less likely to end with blood spilled
on the oor than say, vinyl versus digital or tube versus
solid state (proponents of SE and PP are, after all, still
"brothers in tubes"), their clashes can nevertheless be-
come a bit heated at times. The real danger here exists to
that poor, well-meaning soul who enters into the discus-
sion and tries to play the peacemaker by extolling the
virtues of each method and proclaiming them equals,
which is the "raw nerve" for both camps. When this hap-
pens, you'll usually see Camp SE and Camp PP rise up
in unison, beat him like a rented mule and toss his limp,
lifeless body aside as fodder for the vultures, and then
resume hostilities as usual betwixt themselves.
Yes, even among the brethren of The Brotherhood Of The
Firebottle, there are many schisms. Paper in oil caps ver-
sus plastic lm caps. Vintage carbon composition resis-
tors versus fancy metal foils and metal lms. Vintage
amp designs versus modern amp designs. Grid bias ver-
sus cathode bias. And then there are various, very ada-
mant "splinter groups." Here, you'll nd the "Power
Mongers" branch of the Brotherhood. The Power Mon-
gers drive their gargantuan speakers with behemoth de-
vices that give off an infrared heat signature detectable
by satellites, and more closely resemble 800 megawatt
nuclear reactors than tube amps. No "girlie man" amps
for them! Sadly, many Power Mongers wind up in jail,
after nally getting busted for the racketeering and
smuggling operations they had operated in order to pay
their electric bills.
Then, there are the Old Timers, those easy going, laid
back fellows who love to enjoy their music with the
"warm and fuzzy" sounds of the bygone Golden Era of
Hi-Fi. Old Timer is not an age, it's a mindset. They are
some pretty young Old Timers out there, who have usu-
ally been accepted into the Brotherhood under the wing
of an older Old Timer. Every bit as dedicated as the other
splinter groups, Old Timers would rather have their Eico
HF-87 than a free lifetime membership in the Ferrari of
the Month Club and a date with Norah Jones. Life is
good.
Finally, there is the most hardcore sect of the entire
Brotherhood, The Triode Junkies, which is further divided
into the Transmitter Triode Vikings, the DHT Dandys, the
Sweep Tube Ninjas, and the fractional-watt Spudmeis-
ters.
Transmitter Triode Vikings are those aggressive individu-
als who want to crank it to SPLs that give you a nose-
bleed, but want to do so with class and nesse. Their
way is an iron st in a velvet glove. They differ from the
Power Mongers in that they posess a seething dislike for
push pull. As a matter of fact, some Transmitter Triode
Vikings began life as Power Mongers, but were lured by
the seductive sounds of SET, and became Triode Junkies.
They do however still retain some of their old Power
Monger ways, and 2 watts just won't cut it for them.
Their quest for single ended nesse along with enough
power to arc weld often drives them to the brink of
madness. They do not at all mind the fact that there is
nothing more than a thin ceramic cap standing between
them and 1500 volts, neither do they mind a level of
radiated heat that would leave even Power Mongers
reaching for an asbestos suit.
The DHT Dandys are a crew of rened tastes, who enjoy
the ner things in life. They mate their Avantgardes with
their $35,000 monoblocks via speaker cables that cost
more than the yearly Gross Domestic Product of most
developing nations. If invited to their home for a listen to
their system, you'll nd most of them to be very hospita-
ble and civil gentleman, with all the prim and proper
ways of a British butler. Alas, there is the occasional
DHT Dandy who can be a bit snobbish about the fact
they spent more on their stereo than you did your house,
but most are true gentlemen who simply love music.
However, the situation can deteriorate very quickly if
you mention that your favorite amplier does not con-
tain a DHT power tube, leading you to be escorted from
the premises by some very scary looking bodyguards
with large bulges under the torsos of their suits. So
please, keep your opinions to yourself, be quick to com-
pliment their system, and even if you do not like the cav-
iar and trufes they offer you, pretend to like them any-
way.
Sweep Tube Ninjas are "the thinking man's triode ad-
dicts". Cunning and crafty, with a million tricks up their
sleeve, they are true to their namesake. Being the think-
ers and doers thay they are, they can recite the pinout of
every television vertical amplier tube in existence from
memory, and delight in building a chintzy looking amp
for less than $250 (including $6 tubes) that will amaze
and confound the DHT Dandys by it hanging with their
$35,000 monoblocks.
The Spudmeisters are a relatively young group that is
small and unique, just like their ampliers. Their com-
mon bond is the spud (one tuber) amp. One resistor, one
tube, and one output transformer in the signal path per
channel, nothing else. Any more than that, and a Spud-
Single Ended vs. Push Pull: The Fight of the Century
by Eddie Vaughn
meister will fall to the oor convulsing, ears covered, at
the thought of putting anything extra in the signal path to
contaminate the signal! Coupling capacitors are the ul-
timate statement of blasphemy to the Spudmeister.
Though they only have one scant watt on tap, they don't
seem to mind. It's all about quality, not quantity.
Various individuals within the rank and le of The Triode
Junkies belong to a dark, erce, and shadowy sub-sect,
the Triode Zombies, who are the mindlessly dedicated
Dark Overlords of the Three Element Tube. People who
listen to pentode or ultralinear amps are no friends of
theirs and are not welcome in their home. Enough said.
Getting serious for a moment, there are two broad cate-
gories that all these sects fall into, whether you're an
Old Timer or a Spudmeister. Those two categories are
Single Ended and Push Pull. Much has been written on
how each works, but very little on their direct compari-
son. So, in this article I'll attempt to bring out the virtues
and drawbacks of both single ended and push pull op-
eration, lay them out side-by-side, and try to avoid rais-
ing the ire of both Camp Push Pull and Camp Single
Ended along the way. Enjoy. :^)


How SE Works
Single ended means just that. In layman's terms, a single
power tube does all the work. Actually, SE ampliers
may have more than one power tube per channel (paral-
lel single ended, or "PSE"), but each of the parallel
power tubes function exactly alike, and handle the en-
tire AC music signal sinewave form together in unison.
In other words, they together function as a single tube.
Because it must handle the entire waveform, the SE out-
put stage runs in Class A operating class. This means it is
biased so that it conducts plate current throughout the
entire 360 of the AC signal cycle, and maintains the
tube in a highly linear region of it's operation at all
times.
Fig.1 Single Ended Output Stage
Fig. 1 above is a typical single ended output stage. This
1940's schematic diagram depicts a directly heated tri-
ode, it's output transformer, and the loudspeaker. You'll
notice that the single output device handles the entire
waveform. As a side note here, some people are shocked
(no pun intended) to nd out that the music signal in
their system is nothing more than an AC voltage, just like
your home's AC wall voltage! Depending on where you
live, your AC wall voltage has a frequency of either 50 or
60Hz, a single musical tone. The only difference in it
and the music signal is that the music signal of course
consists of many simultaneous frequencies.

Back to our SE output stage. The reason it must be biased
so hot is because of the effect that the AC music signal's
negative half-cycle has on the tube itself. Any tube stage
must have it's operating point, or "bias" set by making
it's control grid some negative voltage value versus it's
cathode voltage. This sets the idle point of the tube, or
else it would run wide open and quickly burn up. The
AC signal's negative half-cycle has the effect of adding
to the negative bias voltage present, which reduces cur-
rent ow just as if you had increased the negative bias
voltage. You can think of this as waves in water. Every
wave has a crest and a trough. We can liken the crest to
our positive AC half-cycle, and the trough as our nega-
tive AC half-cycle. The trough is a depression, and when
it is added to another depression, the sum result is an
even deeper depression! In Class A operation, the tube is
biased hot enough that it does not stop conducting cur-
rent at the peak of the negative half-cycle. If it did, it
would not be able to amplify that portion of the signal. It
must not only be biased hot enough to merely conduct
at the peak of the negative half-cycle, it must still be
conducting enough current at this point to keep the tube
out of the highly non-linear (high distortion) region near
it's conduction cutoff (zero plate current). To do so is
very inefcient, and Class A's lower distortion and sweet
sound come at a high cost of low power, high heat, and
faster tube wear.
How PP Works
In push pull operation, two tubes work together as in
PSE, but they do so on alternate cycles of the input
waveform, instead of the same cycle. Each tube of the
push pull pair bears half of the work, so to speak, in-
stead of SE where any and all power tubes bear all the
work. To accomplish push pull operation, the tubes of
the push pull pair must be fed signals that are in oppo-
site phase to each other, but are otherwise identical in
all respects. The "phase inverter" does this. There are
several different tubed phase inverter geometries, and
each has advantages and disadvantages. Each of them
works by splitting the signal into a copy of the original
that is in-phase, and a copy of the original that is 180
out of phase. The two phase inverter outputs are simul-
taneously fed to their respective power tubes. This
Single Ended vs. Push Pull: The Fight of the Century 2
means the output signals of the power tubes are of
course out of phase with one another. They are summed
together in the OPT as the signal that drives the speak-
ers. Fig. 2 below illustrates this.
Fig.2 Push Pull Output Stage
Note that the input signal going to each tube of the push
pull pair is a duplicate of the other, just 180 out of
phase, or "inverted phase." You can see how that when
one tube is at it's peak conduction when being fed with
the peak of the positive half-cycle, his brother across the
aisle is chilling out and resting as he is simultaneously
fed with the peak of the negative half-cycle. Then, the
cycle switches. Without going into a long technical ex-
planation of precisely what happens in a PP output
stage, sufce it to say that it does just that, one tube
pushes while the other pulls, and back and forth. It can
be compared to how SE operates with a very simple
analogy.
Think of PP as two men in a small boat, with one rowing
on each side. One rows while the other removes his
paddle from the water and moves it forward to row
again. This motion is repeated back and forth in alternat-
ing cycles. Though one is rowing on the starboard side
and one on the port side, the sum of their rowing actions
is a straight motion. Each expends only half the energy
required to propel the boat at a given speed. You can
liken SE to only one man in the same boat. He alone not
only has to muster the same rowing energy that two men
did together, but he must use a lot of his energy in turn-
ing his torso from side to side as he rows on each side of
the boat in order to make it go straight, wasted energy
that does not prot in making the boat move. It is waste-
ful and inefcient compared to two men rowing to-
gether, and the one man is simply not capable of ex-
pending the same level of energy as the two men com-
bined are capable of.
Efciency and Operating Class
Here we see the major reason for using PP, it's high ef-
ciency versus SE. Actually, if the PP topology is ran at the
same high current Class A operating point as a SE amp
("high" or "hard" Class A operation), it would produce
the same power as a parallel single ended amplier. No
power is gained over PSE by using the PP topology if it is
ran in high Class A, because high Class A is so inefcient
(typically around 10% efciency). To capitalize on the
power potential of the PP effect, hi- gear is generally
operated in Class A at a lower current, higher voltage
operating point than SE, where the tubes barely conduct
through the full 360 of the AC input cycle. The need to
conduct substantial plate current for the entire 360 is
negated by the PP effect, as each tube of the PP pair
must handle only half the waveform (180) anyway, not
the entire 360 waveform as in SE operation. This "low
Class A" operation is somewhat more efcient and
wastes less power as heat, while still maintaining low
distortion. Figure 3 demonstrates the opposing phase
input signals, and that the tubes conduct throughout the
entire 360 of the waveforms.
Fig 3 Power tube conduction of a typical Class A PP
output stage
But if even more power is needed, PP is ran in Class AB
operating class to further increase the efciency. Class
AB biases the tubes so that they conduct for less than
360 of the AC music signal cycle, but still more than
the necessary 180, in order to keep them out of the
very high distortion region near their cutoff point. Be-
cause it idles at generally no more than 70% of maxi-
mum dissipation, and conducts full current only on loud
peaks at high volume, the tube life is usually longer than
with Class A operation. Figure 4 is a representation of
the input signals to each power tube of a push pull am-
plier in Class AB operation.
Single Ended vs. Push Pull: The Fight of the Century 3
Fig 4 Power tube conduction of a typical Class AB PP
output stage
Note that in the Class A push-pull amplier in Figure 3,
both tubes conducted throughout the entire 360 of the
AC sine wave cycle. In Figure 4's Class AB operation
amplier, the tubes are biased "colder" and conduct for
less than 360 but still substantially more than 180. The
black shaded area of the input signal sine wave repre-
sents where the tubes cut off on the peak of the negative
half-cycle. Because the tubes are wasting less potential
power as heat in Class AB than in Class A, the power
output is increased substantially. However, due to the
lower plate current, they are operating in a less linear
region as the postive half-cycle nears the zero (center)
line than with Class A. This increases distortion and gives
a different sound. This is the conundrum of operating
class. When you try for more power, you get increased
distortion. When you try to lower the distortion, you lose
power.
There are yet other operating classes, but they cannot be
used for audio due to their high distortion. Class B op-
eration conducts for exactly half the AC signal cycle
(180), and has extremely high distortion near the center
zero line of the AC sine wave. Class C actually conducts
for less than 180. Class B and Class C are used only in
RF applications, where the high distortion does not mat-
ter. Class AB is so named because it is intermediate be-
tween Class A and Class B in function, and behaves in a
manner somewhat similar to (but not identical to) Class
A at low output. A common urban myth is that "Class AB
amps are Class A at low output". This is about as far from
truth as it gets. A Class AB amplier is just as Class AB at
idle as it is at full power, and vice versa. As stated, they
behave in a manner similar to Class A at low output, but
they are still running at a Class AB operating point.
Biasing Methods Used in SE and PP
We mentioned bias in our basic primer on SE amplier
theory. Bias is simply making the tube's control grid
(where the signal is applied) more negative than the
cathode (where the electrons are emitted), in order to set
the operating point of the tube. Without it, the plate cur-
rent ow would be uninhibited and the the tube's inter-
nals would melt in mere moments if the circuit's supply
voltage was very high at all. Not a happy thing, needless
to say. Figure 5 illustrates the two major methods of bias-
ing used in tube ampliers, grid bias and cathode bias.
Fig 5 Methods of biasing a tube amplier stage

GRID BIAS Grid bias (also called xed or adjustable
bias) is achieved by applying the desired negative volt-
age to the control grid, from a special bias power supply.
The cathode emits electrons, which have a negative
charge. If the grid is "more negative" than the cathode, it
will repel some of the electrons (like charges repel, re-
member high school science class?), keeping them from
leaving the cathode. The higher the negative control grid
voltage, the fewer electrons that will be allowed through
to the positively charged plate, regardless of the plate's
strong, high voltage electrostatic attraction. In Figure 4,
we've applied 20 volts negative to the control grid, and
the cathode is grounded. The cathode is at ground po-
tential (zero volts), and the grid is 20 volts more negative
than the cathode's zero volts, so we have 20 volts of
bias.
Grid bias is easily adjusted by varying the applied nega-
tive voltage, hence it is often called "adjustable bias." So
how can it be adjustable and xed at the same time?
This sounds like a contradiction, but it's really not. It's
also called xed bias because it's xed to a certain volt-
age value and does not change regardless of the tube's
operation or anything else in the amplier. It does not
mean that it can't be varied if one so desires.
Single Ended vs. Push Pull: The Fight of the Century 4
CATHODE BIAS Cathode bias (also called "self-bias")
is a very different method from grid bias, but it accom-
plishes the same thing. Well, sort of... Grid bias makes
the control grid more negative than the cathode. Cath-
ode bias makes the cathode more positive than the con-
trol grid, which is exactly the same thing. The grid is still
more negative than the cathode, it's just that we
achieved it in another way. Remember, in a tube circuit,
voltages are relative to one another rather than to earth
ground. You can see in Figure 5 that the cathode is 20
volts positive with respect to the grid (which has zero
volts), so we end up with 20 volts bias.
Cathode bias also differs in that it doesn't use a separate
power supply to apply a positive voltage to the cathode,
the way that grid bias uses a power supply to give it a
negative voltage. The cathode is connected to ground by
a resistor of appropriate value, through which it draws
it's current. According to Ohm's Law, a positive bias
voltage will be created when the tube's current is drawn
through the cathode resistor. There you have it, our posi-
tive bias voltage is created out of thin air, just like magic,
eh? Even neater, the positive bias voltage will regulate
itself, unlike grid bias which requires adjustment. If you
put in a new tube that draws a little higher than normal
current, the increased current draw through the resistor
will create a higher bias voltage, shutting down the
tube's current draw a bit and forcing it to "chill out."
Likewise, a tube with lower current draw will create less
bias voltage and allow the gate to open a bit more in
compensation.
COMPARISON OF BIAS METHODS We can see that
cathode biasing's self-regulating property tends to pre-
vent or at least alleviate "bias runaway," the condition in
which the tube loses bias, conducts wide open, and
melts. The more current the tube tries to draw, the higher
the bias voltage created to stop it. On the other hand,
grid bias can fail and cause the unthinkable to happen.
Grid bias failure is caused by a malfunction of the nega-
tive bias power supply or the associated componentry
that feeds the bias voltage to the tube, or loss of proper
contact between the tube socket and the tube's control
grid pin due to corrosion, loose contacts, etc.
You're probably asking yourself, "Why on Earth would
anyone want to use grid bias, if it requires a separate
power supply and can fail, causing a catastrophic, ery
thermonuclear holocaust?" The answer is vefold.
(1) Grid bias produces more power. All your higher
powered push pull ampliers use grid bias (see be-
low), although it can be used for Class A operation
as well, and often is. Cathode bias is less efcient,
and wastes some power as heat dissipated in the
cathode resistor.
(2) With grid bias, the actual plate voltage of a tube is
the measured voltage between the plate and ground.
With cathode bias, it's the voltage difference be-
tween the plate and cathode. Since cathode bias
applies a positive voltage to the cathode, the actual
plate voltage is the measured positive plate voltage
minus the positive cathode bias voltage. This means
that in many cases, grid bias allows for a lower volt-
age main power supply than what is needed with
cathode bias.
(3) Grid bias must be used for Class AB operation. Un-
like Class A operation, which draws the same cur-
rent regardless of signal level, Class AB operation
proportionally draws much more current as the sig-
nal level increases. The self-regulating property of
cathode bias ghts against this increase in current.
Grid bias works completely independent of every-
thing else in the amp. It's called "xed" voltage for
this reason, remember? This turns out for the best
anyway, as the whole reason for using Class AB is
for higher power, and as we just read grid bias pro-
duces more power than cathode bias anyway.
(4) Grid bias produces lower distortion than cathode
bias.
(5) Grid bias allows for ease of experimentation with
different plate currents. You can turn the plate cur-
rent up or down and nd the sweet spot that sounds
best to you, usually with the ease of turning a trim
pot.
Well, now your question has probably been turned
around in the opposite direction, and you're asking
yourself, "Why on Earth would anyone want to use cath-
ode bias if it wastes power, creates more heat, lacks ad-
justability, has higher distortion, and sometimes requires
a higher voltage power supply? Who cares if it's self ad-
justing? That's too much compromise for gaining nothing
but self-adjustment." The reason is simple. The sound.
Cathode bias sounds very different from grid bias. Grid
bias fans claim their method sounds clean, neutral, un-
colored, and rm, while cathode bias sounds soft,
mushy, and hopelessly colored. Fans of cathode bias
claim their method sounds sweet, warm, musical, and
natural, while grid bias sounds harsh, sterile, and arti-
cial.
As we saw, grid bias is always used in Class AB PP amps
(by necessity), and in a large percentage of Class A op-
eration PP amps as well. It is even used in some SE
amps, but only a few. Most SE amps use cathode bias,
which is an essential ingredient of the SET sound. Most
people who prefer the PP pentode sound also prefer grid
bias, as it tightens and cleans the sound. For what it's
worth, the author greatly prefers cathode bias in his hi-
gear (both SE and PP) and grid bias in his geetar amps.
:^)
Single Ended vs. Push Pull: The Fight of the Century 5
PP OPT Phase Cancellation
Besides higher output power, PP exercises another huge
advantage over SE, which is phase cancellation of cer-
tain artifacts. Here's an analogy: If you toss two rocks
into a pond at exactly the right time and distance apart
so that the crest of one wavefront exactly met the trough
of the other, they would cancel each other out. The net
result of a crest and a trough together is zero. In other
words, the two waves met 180 out of phase with each
other, and cancelled each other out. Because of the op-
posing directions of current ow in the PP OPT, phase
cancellation of any "foreign bodies" such as power sup-
ply hum and noise occurs. The hum frequency from the
power supply enters into the PP OPT's primary winding
center taps, ows in opposing directions toward each
tube of the push pull pair, and is cancelled by phase
opposition. Figure 2 demonstrates this current ow from
the center tap outwards. In a SE amplier, it all goes
straight through, and is transferred to the speaker. PP
ampliers therefore need much less power supply ltra-
tion to prevent hum that their SE counterparts do. An-
other benet of PP is that second order distortion pro-
duced in the PP outputs stage is cancelled out in the
OPT, which brings us to our next topic:
Harmonic Distortion Series of SE and PP
SE and PP ampliers naturally produce different har-
monic distortion series. Different distortion series sound
quite different to the ear. In general, the higher the order,
the more offensive to the ear, and in correspondingly
smaller amounts with progressively higher orders. Also,
odd order harmonics are more offensive up to a point in
the series, where it all sounds ghastly horrible in very
tiny amounts, whether odd or even order. For example,
as much as 3% second harmonic is unnoticed by most
people, while .3% (ten times less) fth order is very no-
ticeable to most people. Second order is manifest as be-
ing exactly one octave above the fundamental tone, and
therefore it is not dissonant. In small amounts, it is per-
ceived by most ears as adding "sweetess" or "liquidity" to
the sound. Third order is a musical fourth above the fun-
damental tone, which is not terribly harmonious also but
not terribly dissonant if held to low enough levels.
Fourth order is two octaves above the fundamental, and
is not extremely dissonant but blurs focus. Fifth order
and up just sounds very dissonant and nasty, period.
A SET (single ended triode) amplier's main distortion
product is second harmonic, which is relatively benign
to the ear. The single ended driver stages of a PP amp
produce predominantly second order distortion, but the
PP output stage itself produces mostly high/odd order
distortion. Remember, the second harmonic created in
the output stage is cancelled, but any distortion (whether
second harmonic or not) produced in earlier stages is
passed on through.
The reason why PP and SE output stages produce differ-
ent distortion series is because they clip differently. This
"clipping" occurs when the tube reaches the maximum
output attainable at the operating point chosen by the
circuit designer. The tops of the sinewave form cannot go
any higher (higher amplitude), because the tube is at it's
maximum, it cannot amplify any further. Still, on the
oscilloscope screen the waveform amplitude tries to
grow taller in response the increased input signal. Since
it cannot, a at line appears across the tops of the sine
waveform at the "maxed out" point, giving them the ap-
pearance of having been clipped off with scissors, hence
the name clipping. A PP amplier clips symmetrically,
which means the opposing sides of the sinewave clip
alike and at the same time. This produces high/odd order
distortion products, which are more offensive to the ear
than second order. A SE amplier typically clips asym-
metrically (more on one side of the sinewave), which
produces predominantly second order as it's chief har-
monic distortion artifact. These different distortion series
are one of the predominant reasons as to why PP and SE
sound so very different.
Damping Factors of SE and PP
Damping factor is a function of an amplier's output
impedance and how it relates to controlling the loud-
speaker, especially at low frequencies. It is expressed as
a ratio, calculated by dividing the load impedance by
the amplier's output impedance (Z). For example, if an
amplier has an output Z of 2 ohms and the load is 8
ohms, that gives a DF of 4. This method can be used for
estimating the DF, but is not highly accurate. In real life,
a speaker's impedance curve is not at, therefore the
actual DF is more or less at certain frequencies. In other
words, DF tracks the speaker's impedance curve. When
calculating the DF, one must not only gure the output Z
of the amplier's output stage, but must also consider
the resistances of the output transformer's secondary
winding and the speaker cables as well.
So what does DF do and how does it apply in my sys-
tem? In general, the higher the DF, the stronger the am-
plier's "grip" on the woofer. A low DF makes for tubby,
mushy bass with a slow, loose, boomy decay of notes. A
high DF makes for better control over the woofer, and
yields tighter, drier, faster bass without over-ring and
sloppy note decay. A typical SE amplier using directly
heated triodes may have a DF of perhaps 4 or 5 at best.
A SE amplier using triode-strapped tetrodes or pentodes
will have an even lower DF, depending on the tube type
used. Some have a DF under 2. A vintage PP amplier
from the '60s may have a DF of perhaps 10 or 12. A
Single Ended vs. Push Pull: The Fight of the Century 6
typical solid state amplier may have a DF of anywhere
from around 100 to over 1000.
MU, TRANSCONDUCTANCE, AND PLATE RESISTANCE
So what does a tube's electrical characteristics have to
do with DF? We'll get to that. We saw in the section
"HOW SE WORKS" how the plate current and plate
voltage of a tube swing up and down, back and forth in
a reciprocating fasion in the output stage. The tubes
themselves do not amplify, they manipulate the power
supply by this voltage/current swing to produce a higher
amplitude copy of their input signal across their load. A
better term for amplier would be "modulated power
supply" and a better term for an amplier stage would be
"power supply modulator". If there is no load present
across the plate of the tube, there will be no amplica-
tion. Again, it's not the tube itself that amplies! Back to
the voltage/current swing I just mentioned. The plate
voltage/current swing occurs in an orderly manner in
response to the AC input signal. As we had read in the
"HOW SE WORKS" section, on the negative half-cycle of
the AC input signal the plate voltage swings high and the
plate current low. On the positive half-cycle, the plate
voltage swings low and the plate current high. The width
of this voltage swing determines the degree of amplica-
tion, and the symmetry of the swing determines the or-
ders of harmonic distortion present. It also creates a dif-
ference in the DF of SE and PP ampliers, which we'll
see demonstrated shortly.
Let's see how this plate voltage/current swing in re-
sponse to the control grid signal voltage works to dene
damping factor. To do that, we rst need to look at the
three main electrical characteristics that make a given
tube type what it is, which are mu, transconductance,
and plate resistance. They are inter-related and insepara-
bly linked. If two of these are known, the unknown third
one can easily be calculated.
Mu is the ratio of the change in plate voltage to a change
in control grid voltage in the opposite direction, and is
usually referred to as the amplication factor of the tube.
It's symbol is .
Transconductance (represented by it's symbol Gm) is
dened as the ratio of the change in plate current to a
change in control grid voltage. The term "transconduc-
tance" is derived from "transfer conductance" and is
measured in siemens (S), where one siemens equals one
amp of plate current change per one volt of control grid
voltage change. It is also commonly expressed as mho,
which is equivalent to a micro-siemens (1/1000 sie-
mens). Mho is ohm spelled backwards, which is a very
tting term because conductance and resistance are op-
posites. A transconductance amplier is an amplier in
which a change in input voltage causes a linear change
in output current. Tube ampliers fall into this category.
The other main type of amplier is known as a transim-
pedance amplier, where a change in input current
causes a linear change in output voltage, exactly the
opposite of a transconductance amplier.
Plate resistance is dened as the change in plate voltage
divided by the change in plate current it produces as a
result. Plate resistance is expressed in ohms, just like any
other resistance. Plate resistance is symbolized by Rp.
PLATE CURRENT AND DAMPING FACTOR We can
see how these three parameters are related and interac-
tive. The property of their interrelation we are most con-
cerned with here is how plate current affects plate resis-
tance. Plate resistance drops with an increase in plate
current, provided the other variables remain constant. To
see how this relates to damping factor, let's build a theo-
retical SE amplier and a theoretical PP amplier, for the
purpose of observing how each topology's inherit way of
operating affects it's output impedance. Let's use the
same triode power tubes in both amps, and run both in
hard Class A operation at the same voltage and idle cur-
rent. Both ampliers use grid bias, and both their OPTs
have the same secondary winding resistance.
Let's power up our SE amplier rst, and give it a signal.
Remember, we just read where plate resistance drops
with an increase in plate current, provided the other
variables remain constant. On the positive half-cycle of
the AC input signal, the plate current swings high, and
therefore the plate resistance is lower on the positive
half-cycle than the negative half-cycle, where the plate
current drops. By this, we can see that the plate resis-
tance goes up and down in sync with the polarity swings
of the AC input signal. The higher the plate resistance,
the higher the output Z, and the lower the DF. Here's
where we start getting to the core reasons of why most
PP ampliers have rmer bass than SE amps. The DF
ip-ops up and down with each 360 cycle of the AC
input signal. This tends to give SE ampliers an odd bass
performance characteristic. The DF is different on the
"out" movement of the speaker cone than it is on the "in"
movement!
Let's try this same experiment on our PP amplier. Each
tube of the PP pair sees a signal that is in opposite phase
to the other. In other words, when one signal is at the
peak of it's positive half cycle, the other is at the peak of
it's negative half-cycle. As a result, one power tube's
plate current is rising while the other's is dropping, so
that the two always average out to the same median
combined plate current. To better illustrate this, we'll use
integers and say that while at idle, both tubes are at
zero. At the peaks of the signal cycle, one tube will be at
say, +10, while the other is at -10, which averages to
zero. Halfway through the cycle, one tube will be at +5
and the other at -5, which still averages out to zero. As a
result, while one tube's Rp is dropping as current in-
creases, the other's Rp increases as it's current drops.
Single Ended vs. Push Pull: The Fight of the Century 7
Their Rp averages out to a median gure, and therefore
the output Z also averages out to a median gure. This
means (you guessed it!) the DF is steady as well, very
much unlike our SE amplier where the DF changed
with each half-cycle of the input signal. This is what
gives PP an inherit advantage in the bass department
over SE, all other things remaining the same.
However, we used triodes in our test PP amplier, and
ran them in hard Class A. Triodes (especially low mu
triodes) have far lower Rp than tetrodes and pentodes,
and therefore will naturally yield a higher DF. But most
PP amps use tetrodes or pentodes instead of triodes,
which means very high output Z, a very poor DF, and
almost non-existent bass. Why then do most PP amps
have at least twice the DF of SE amps? The next section
explains it all.
Negative Feedback
To further widen the sonic gap between SE and PP, con-
sider that most PP amps use negative feedback (NFB),
while very few SE amps use NFB nowadays. "Okay,
that's nice" you say, "now, what is NFB?"
Negative feedback is a sample of the signal taken from
the output of an amplier stage and re-injected into the
front of that same stage (local NFB), or into an earlier
stage (global NFB), where it is 180 out of phase with
the signal present at the re-injection point. In other
words, "feeding back." Pretty crafty, eh? The circuitry
contained between the two points is called the "feed-
back loop." Usually, the global feedback loop in a PP hi-
amplier will run from the speaker terminal back to
the amp's input stage, thereby encompassing the entire
amplier circuit. Figure 6 points out the global feedback
loop of a typical push pull amplier schematic in bright
red.
Fig 6 Global negative feedback loop
Your next question is probably, "Okay, now I know what
it is. What does it do?" Good question. Well, it does a lot
of good things.
(1) NFB reduces the distortion and noise in the stages
contained inside the feedback loop. It cancels distortions
in the same way our PP OPT cancelled power supply
hum and second order harmonics earlier.
(2) NFB lowers the output impedance. This tightens and
extends the bass.
(3) NFB attens and extends the overall frequency re-
sponse.
Most people at this point would be thinking, "Cool!
Bring on the negative feedback!" Not so fast. NFB also
does some bad things.
(1) NFB kills the space and air between the instruments,
parts, and voices, rendering music dull and uninterest-
ing.
(2) NFB reduces the stability of amplier circuits
(3) NFB reduces gain.
Oh well...... Nothing's perfect, including NFB. Especially
NFB. This rst brings us to the early days of telephone
service, where long distance telephone transmission
lines were powered by tube ampliers. Due to parasitic
losses in the long, cross-country runs of wire, repeater
ampliers had to be used at intervals. The cumulative
distortion of several of these repeater ampliers was hor-
rid, and telephone communications were difcult to
understand as a result. NFB was invented at Bell Labs to
lower this distortion, and it indeed worked very well.
Without NFB, early long distance telephone communi-
cations across entire nations and continents would have
been impossible.
Fast forward to the Golden Era of Hi-Fi, which was the
original heyday of tube audio back in the 1950s and
'60s, before mainstream solid state audio gear even ex-
isted. Just as it is in today's "big box store" world of mass
consumer audio electronics, specs on paper meant eve-
rything back then. The lowest distortion at the highest
power was what sold biggest. Lots of NFB was applied to
everything, often even in multiple feedback loops. This
was done because NFB lowers distortion, and to the un-
informed masses the lowest distortion means the best
sound. "Lower distortion has to sound better, doesn't it? I
mean, distortion is a deviation from the original music
signal, so lowest distortion means the most "perfect"
sound, right?" No, it denitely doesn't. We saw this in
the last section, "HARMONIC DISTORTION SERIES OF
PP AND SE." The amp with the lowest distortion is not
necessarily the best sounding. It more strongly depends
on which orders of harmonic distortion are present,
more so than how much. The human brain often dis-
agrees with specs on paper as to what sounds best! Be-
cause NFB is taken from the very rear of the amp and
injected back into the front, there is a frequency-
Single Ended vs. Push Pull: The Fight of the Century 8
dependent time smearing effect that occurs, which the
ear is extremely sensitive to. It blurs the focus and ho-
mogenizes the sound. The excessive amounts of NFB
that was used to get the best distortion specs on paper
was what made many of these ampliers of yore sound
lifeless and uninteresting to listen to, and is a major rea-
son why today's big box store electronics sound so bad
as well. Both may have ridiculously low distortion, yet
there are ampliers with literally a hundred times more
distortion that sound better. Human psychoacoustics do
not obey the laws of science very often!
So why do most PP ampliers use NFB and most SE am-
pliers don't? The main reasons are distortion and output
impedance. Power triodes and triode-wired power
tetrodes/pentodes were not used very much during the
Golden Era of Hi-Fi, as a matter of fact almost never.
Triodes produce comparatively less power than tetrodes/
pentodes, and remember, specs on paper were every-
thing so the most power won the day (or at least the
sale!) Practically 99.9% of ampliers used beam power
tetrodes or power pentodes. As we saw in the last sec-
tion, both of these tube types have very high plate resis-
tance and therefore high output impedance. Neither is
as linear as triodes; both types have much higher distor-
tion. A zero feedback amplier built using these types
has high distortion, with much of it being the high/odd
order type, plus the very high output impedance means
the bass will be weak and abby. It sounds bright, harsh,
and thin, and literally unlistenable. So, NFB was a nec-
essary evil with ampliers using these tubes, both to
lower distortion and yield an acceptable damping factor.
To make matters worse, many of these PP ampliers had
very complex, multi-stage signal paths to improve the
specs on paper. This created more distortion, which of
course required more NFB to correct! NFB reduces gain,
so higher gain driver stages were needed to recoup the
lost gain. Higher gain also means more distortion is cre-
ated than in a lower gain scenario, which means (you
guessed it) more NFB is needed to lower the distortion!
It's like a dog chasing his tail, a vicious cycle to which
there is no end, and nothing good can ever come from
it. Plus, the immense complexity of such amplifers also
degraded the sound, due to the sheer volume of compo-
nents, solder joints, and jumper wires the signal had to
travel through. Your music got ran through a huge net-
work of sonically degrading components, and it sounded
like it.
What's even sadder is that most of today's commercially
produced PP ampliers are still nothing more than
tweaked-over copies of these same vintage circuits! Fact
is, many (most?) of today's companies producing these
PP ampliers are still caught in this same "specs on pa-
per" rut and can't seem to get out of it. I have this fancy
saying about this, "Test equipment ain't got ears". The
oscilloscope is not gonna be listening to it. The har-
monic distortion analyzer is not gonna be listening to it.
So who is? Humans! We have ears and brains, not
screens and displays. As I had said earlier, psychoacous-
tics usually doesn't obey numbers written on a piece of
paper.
Now that we've beaten the high feedback PP horse to
death, let's move back to triodes and single ended op-
eration. Triodes sound less bright, thin, and harsh when
not using NFB, and have lower output impedance and
lower distortion. A SE amplier with a single, highly lin-
ear driver stage and a low plate resistance triode power
tube can usually get by without NFB, which means the
average SET amp is usually more open, airy, and purer
sounding than an otherwise identical SEP (single ended
pentode) amplier using NFB to correct it's problems of
high distortion and output impedance.

In Summary
We've seen that single ended and push pull ampliers
both have advantages and disadvantages. It's difcult to
make blanket statements about each, as there are many
variations of circuitry and quality in each broad cate-
gory, so keep in mind these statements are generaliza-
tions. There are exceptions to the rule
ADVANTAGES OF SE OPERATION
Less signal deterioration because of fewer signal
path components
Better detail and coherency, especially at low vol-
umes
More open and airy sound
DISADVANTAGES OF SE OPERATION
Poor efciency, with low power and high heat pro-
duction
Almost no rejection of power supply hum and noise,
which necessitates more power supply ltering
No cancellation of second order harmonic distor-
tion (some do not consider this a disadvantage)
(Usually) shorter power tube life
Has speaker compatibility issues, is less tolerant of
wide impedance curves and complex crossover
networks
Poor damping factor
Single Ended vs. Push Pull: The Fight of the Century 9
ADVANTAGES OF PP OPERATION
More power
(Usually) more authoritative and powerful sounding
Better bass performance
Fewer speaker compatibility issues
Lower distortion
DISADVANTAGES OF PP OPERATION
Homogenized, uninteresting sound
More signal deterioration due to greater signal path
complexity
More power supply intermodulations between am-
plier stages, and other power supply anomalies
Poor low level detail and dynamics
Again, do note that these are generalizations. There are
some horrible sounding SE ampliers out there, and
some very open and musical sounding PP ampliers. It
all depends on the design. If we apply SET design con-
cepts to PP, using engineering practices aimed at achiev-
ing the best sound rather than the best specs on paper,
the results can be astonishingly good. A PP amplier
using low Rp triode power tubes such as the 2A3 or
300B can get by with little or no NFB. For one, the tri-
odes are an advantage in themselves, in both linearity
and output impedance. Two, the front end voltage am-
plifer and phase inverter stages can be designed with
much lower gain, if no NFB is present to reduce the
available gain. Lower gain means lower distortion and
lower output Z, so NFB is not necessary to attain these
desirable properties. You end up with a much more lin-
ear amplier, that has a simpler circuit with lower parts
count and therefore less signal deterioration, and more
air, breath, and vibrancy. See how much better every-
thing gets when you set out to design it right from the
beginning, instead of trying to x it with NFB?
You're probably thinking, "There must be a catch." Well,
there is. The power from such an amplier is very low
compared to high feedback PP amps using big pentodes.
But, which would you rather have? 15 watts that sound
scary good and leave you wanting more of the same, or
70 watts that bore you to death and leave you fatigued?
Furthermore, in light of the huge number of highly ef-
cient, great sounding speakers on the market today, am-
plier wattage is a moot point. The pitfall to be avoided
is not so much SE or PP as it is high power. By using
very efcient speakers with minimalist or no crossovers,
we can sidestep the whole power issue, and end up with
much better sound, usually at lower cost as well.
High efciency and low power is the way to go in this
author's opinion. I grew up in a house full of vintage
tube gear, and have owned both high and low powered
tube and solid state gear. I've heard a lot of gear in my
lifetime, and quite frankly, most of it isn't worth remem-
bering. Most of what is put out well less than 10 watts.
Nothing I've ever heard can compare to a good low
powered tube amplier with highly efcient speakers. SE
or PP? Name your poison. Both have their place, and
both can be fabulous if correctly designed and imple-
mented. We can sum it all up in these words of wisdom,
from one of the greatest tube amp transformer designers
the world has ever known, Dutch electrical engineer
Menno van der Veen:
"In reality, measurements are ne and can help you. But
we have the nest measurement devices available on the
two sides of our head: our EARS, and are not they fan-
tastic? Well, use them and rely on them."
Eddie Vaughn
Copyright 2006 by Eddie Vaughn
This article may not be reproduced in whole or in part,
or used for commercial or prot purposes except with
the express written consent of the author.
Single Ended vs. Push Pull: The Fight of the Century 10
A Burning Question
Here's a question I've been asked by several folks, in-
cluding a gentleman who has a hearing condition that
makes tizzy, grainy systems and loud listening levels
painful. "Does the Carina SE EL84 amplier have good
detail and dynamics at low listening volumes, or must it
be cranked up so that you can hear everything clearly?" I
explained that it sounds very detailed and articulate at
sub-watt output levels, and as a matter of fact it's inside
the rst watt where it truly excels. I immediately fol-
lowed my statement with a question, "What amplier
are you using now?" It turns out that this particular gen-
tleman has a push pull amp. Ah-ha, just as I'd sus-
pected! While his PP amplier is different (I'd say better)
from most in that it operates it's four 6BM8s in strapped-
triode and uses very minimal negative feedback, it still
suffers from one of the drawbacks inherit to PP. He says
that while it sounds positively wonderful at higher levels,
he nds that it lacks detail and dynamics when played at
very low volume. His main reservation about buying a
Carina was the fear that it will exhibit this same trait.
It won't. Denitely not.
This is a conundrum that confounds the "spec fanatics."
Despite having much poorer distortion specs on paper
compared to a push pull amplier using feedback, a
zero feedback SET usually produces better detail, espe-
cially at low volumes. Some of this lies within the am-
plier itself, due to the fact that single ended amps usu-
ally have a lower signal path component count and
hence less signal degradation, and less power supply
intermodulations and such to due the fact there are usu-
ally less power supply stages. But to nd the REAL rea-
son why, you must go beyond the amplier section and
power supply, and look at.......the output transformer.
The SET's output transformer (OPT) also confounds the
spec fanatics, because at rst glance to the untrained eye
it shouldn't work well at all, because of it's air gap. The
air gap is a "band-aid x" that allows it to function with
some degree of practicality. Because it reduces the in-
ductive coupling efciency, one would therefore think it
would also only accurately convey the "macro" parts of
the sound, and lose a lot of the ne details.
Again, it won't. Denitely not.
Neither the low volume ne-detail resolution of SET
ampliers, nor the seemingly inherit inability of PP am-
pliers to properly render ne details and microdynam-
ics at low volume are any kinds of secrets. Anyone who
has owned both PP and SET ampliers has observed it.
Yet all too often, it gets blamed on the amp being PP.
While this does have somewhat to do with it, it has a lot
more to do with the output transformer than the ampli-
er itself, as this article on PP OPTs and transformer
lamination materials will attempt to explain. We'll ad-
dress the output transformer differences between PP and
SE here, then we'll work towards the front end of the PP
amp in consecutive articles. (WARNING: COPIOUS
AMOUNTS OF WIGGY-HEADED, MIND NUMBING
TECHNO-BABBLE AHEAD)
PP Core Saturation, or "Bass Iindigestion"
No transformer lamination material can achieve a theo-
retically innite magnetic ux density. All "saturate" at a
certain ux density. Flux density is achieved by applying
a magnetomotive force, or mmf. The mmf in an OPT is
EMF (electromotive force), more commonly known to us
as voltage. Actually, mmf is the magnetic equivalent of
electrical voltage! Here, it's the AC music signal from
the power tubes that supplies the EMF, or voltage. The
saturation point of a given OPT is mostly determined by
the core dimensions and the lamination material used.
The transformer's magnetic ux density increases pro-
portionally with mmf, which increases proportionally
with the applied EMF. Once the core reaches it's satura-
tion ux density, any further increase in mmf does not
(in fact, cannot) result in any further increase in mag-
netic ux density. An overloaded, saturated output trans-
former distorts the waveform in much the same way a
saturated tube clips and distorts, creating harmonics in
the secondary winding's output signal (NOT good!).
So as we can see, there is a limit to how much AC signal
voltage a given OPT (whether PP or SE) can tolerate be-
Single Ended vs. Push Pull (Continued):
The Deep, Dark Secrets of Output Transformers
by Eddie Vaughn
fore core saturation rears it's head and everything falls
off a cliff sonically. Actually, it's how much total energy
is contained in each half-cycle, not just the voltage. A
sinusoidal waveform of a given EMF (voltage/amplitude)
increases in total energy as the frequency decreases.
Why and how? The energy contained in a sinewave is
proportional to the total area accumulated in the curve
of that sinewave. Since each half-cycle of say, a 40 Hz
waveform accumulates more area between it's ends (the
zero line of the oscilloscope graph) than an 80 Hz wave-
form of equal amplitude (voltage) does, the ux density
will attain higher values due to the higher mmf energy
contained in the 40 Hz waveform. Therefore, the signal
amplitude required to saturate a transformer decreases in
a linear manner with a drop in frequency, and rises with
an increase in frequency. In other words, if a transformer
saturates at 20V RMS at 80 Hz, it'll saturate at around
only 10V RMS at 40 Hz, and 40V RMS at 160 Hz.
Some transformer builders embellish their low frequency
response specs by measuring them at very low signal
levels, often only a few hundred milliwatts. In actual
listening at substantial volumes, such an OPT will satu-
rate rather easily at bass frequencies and distort. Quite
often, an OPT nominally rated for 100 watts continuous
RMS will distort at low bass frequencies at well under 25
watts. This problem also exists in power transformers. If
a power transformer designed for 60 Hz line voltage is
operated at 50 Hz instead, the ux density will reach a
higher peak level on each AC half-cycle due to the
greater energy contained in the 50 Hz sinewave, thus
causing core saturation at a lower secondary current
draw even if the voltage of each sinewave is the same.
On the ip side of the frequency band coin, it's nearly
impossible to saturate any normal OPT with frequencies
from the upper midrange audio band and up due to the
low energy contained in the waveform. As a matter of
fact, most OPTs have sufcient mutual inductive cou-
pling to work without a core from about 2 kHz on up!
The sound wouldn't be as coherent, but it would work
without a core nevertheless.
SE Core Saturation and Air-Gapping
Now that we understand core saturation, let's hit on
some basic PP OPT theory. In a PP amplier's OPT, the
power supply is connected to the center tap of the pri-
mary winding and a tube is connected to each end of
the primary. This allows the tubes to conduct on alter-
nate cycles of the AC input waveform, much like two
persons in a boat with one rowing on each side. Since
the DC bias currents to each tube ow in opposing di-
rections in each half of the primary winding, they cancel
each other out. Remember integers from high school
math? Positive 10 plus negative 10 equals zero! How-
ever, in a SE amplier's OPT, the primary has only two
connections instead of the PP OPT's three. One goes to
the power supply, the other to the plate of the power
tube (or tubes if PSE). Therefore, the DC bias current
ows in one direction only, and the end result is that
there is none of the cancellation effect we saw with the
PP OPT.
Those who know something of the fundamentals of
transformer operation might be thinking, "But aren't we
talking about AC voltage here, not DC current? Trans-
formers turn AC voltage to magnetic energy and then
back to AC, don't they? They don't pass DC from pri-
mary to secondary." This is true, but in the case of the SE
OPT, a magnetizing voltage is created by the DC bias
current owing through the primary winding's resistance
according to Ohm's Law. To your SE OPT, this DC volt-
age looks like one half-cycle of 0 Hz AC, if such a thing
could actually exist. Low frequency AC was bad enough,
but 0 Hz (innitely low frequency) is really bad news for
our OPT!
This voltage creates higher mmf, added on top of the
mmf resulting from the AC signal.The resulting addi-
tional "bias" or "offset" will push the alternating ux
waveform closer to saturation on one half-cycle than the
other, thereby lowering the amount of AC signal a given
OPT can handle before saturation. This offset DC bias
effect is more easily understood by looking at in electri-
cal, rather than magnetic terms. Figure 1 is an oscillo-
scope screen with a symmetrical AC sine wave, which
for our demonstration purposes here depicts the mag-
netic ux created in a PP OPT.
Fig 1 Sine Wave with no DC offset
Single Ended vs. Push Pull (Continued): The Deep, Dark Secrets of Output Transformers 2
We can see that both the positive and negative half-
cycles are of equal amplitude, just as the magnetization
in a PP OPT is equal in both directions. But what hap-
pens in a SE OPT, where offset DC is present? Figure 2
depicts our sine wave with an offset DC bias added.
Fig 2 Sine Wave with positive DC offset
Notice that it is of greater amplitude on the positive half-
cycle, which means that magnetic "headroom" before
saturation is decreased in that direction in our SE OPT.
Offset DC bias is bad news, to say the least. If you could
operate your OPTs at extremely low cryogenic tempera-
tures, the windings would become superconductive and
you wouldn't have to worry as much about this condi-
tion. But, I for one prefer my listening room at around
72 F. :-) The inescapable reality is that the higher the
DC current is in our SE OPT, the closer and closer the
core is pushed to it's saturation point ux density.
Therefore you must do something, anything to raise the
saturation point. Solution: A SE OPT has a shim in it's
core between the stacks of E and I shaped laminations,
to hold an air gap between them and reduce the induc-
tance by breaking the magnetic "circuit". This helps to
stave off early core saturation resulting from the offset
DC. But, it also lowers the efciency, limits bandwidth,
and increases distortion and leakage inductance (more
on that in a minute). When it comes to output trans-
formers, you can't have your cake and eat it too!
We could just forego the air gap and make a bigger
transformer, which spreads the mmf out over a larger
core area, so that any given cross-sectional area of the
core has a lower ux density than with a smaller core.
But, this exacts a price in other areas as well. It increases
stray magnetic elds in the core (elds not involved in
the actual mutual inductive coupling), known as "leak-
age inductance." This leakage inductance "appears" to be
in series with both the primary and secondary windings,
so it manifests as a series impedance, reducing the sig-
nal level delivered to the speaker. Inductive reactance
increases with frequency, so the effect is increased with
rising frequency. It therefore rolls off the highs, just as in
a speaker crossover.
Eddy currents are another problem. They are random,
circular currents (just like eddies in a body of water,
hence the name) induced in the lamination materials by
the magnetic ux, that cause magnetic "ripples" in the
ux eld. Eddy currents are a major reason why trans-
formers (and also electric motors) are built with stacked
laminations and not solid blocks of metal for cores. The
laminations are heat-treated to form a non-conductive
oxide lm on their faces, which electrically isolates each
lamination from the next and prevents the conduction of
the eddy currents throughout the core. This also helps to
break up the eddy currents by limiting their physical
size. The thinner each lamination is, the smaller the area
that an eddy current is restricted to. The effect of lamina-
tion thickness on eddy currents is demonstrated in Figure
3.
Fig 3 Eddy currents and lamination thickness
The eddy current effect increases with rising frequency,
so the higher the desired maximum operating frequency,
the thinner the laminations must be in order to minimize
eddy currents and preserve high frequency coherency
and bandwidth. A smaller core mass, thinner lamina-
tions, and high quality lamination materials all work to
help reduce leakage inductance and eddy currents, and
preserve high frequency coherency.
Single Ended vs. Push Pull (Continued): The Deep, Dark Secrets of Output Transformers 3
Another unwanted occurrence in inductive devices is
"stray" or "parasitic" capacitance. This is capacitance
from winding to winding or winding to core. The larger
the core, the longer the distance around the circumfer-
ence of the bobbin on which the wire is wound. The
larger the bobbin circumference, the longer each turn of
wire is, and the more winding area that exists to develop
capacitance. This capacitance varies with signal voltage
and frequency, and forms inductive-capacitive (L-C)
resonances with the OPT's inductance. Needless to say,
this wreaks havoc on our frequency response atness.
Not a happy thing. But, certain special winding topolo-
gies are used to reduce stray capacitance, and preserve
the frequency response. Properly designed and wound, a
good SE OPT can have a substantially large core mass
before high frequency response suffers substantially.
The Hidden Glory of the SE OPT
Leakage inductance and stray capacitance are not the
exclusive problems of SE OPTs however, they plague PP
OPTs as well, and also inductors, power transformers,
and autoformers. So if both PP and SE OPTs suffer from
these same anomalies, then PP is superior because it
doesn't need an air gapped core and is less susceptible
to saturation, right?
Wrong. Although the SE OPT's offset DC is a drawback,
it's also a blessing in that it keeps the core magnetized.
The offset DC creates a static magnetic eld that holds
the transformer in a highly linear region of it's magneti-
zation curve, even at very low signal levels. It avoids
ever seeing a zero ux condition, whereas PP does not.
In a PP OPT there is no offset DC, so with no signal pre-
sent there is no mmf, and no magnetic ux.
The PP OPT depends solely on the AC signal
to provide the magnetizing or "excitation"
current, while a SE OPT stays magnetized
100% of the time. It's very important to un-
derstand this, because it's a critical reason
why SE ampliers are more coherent and
detailed at low volumes than PP! There is no
hysteresis or core loss (more on that in the
next section). Granted, this comes at a price,
as the air gap and necessitated larger core
mass do lower the efciency and create
more distortion and frequency response
problems. But, it does knock hysteresis in the
head.
I've read claims that the air gap is actually
benecial to the sound, and is what causes
this phenomenon. This is 100% untrue! The
air gap is detrimental in every way, and is a "necessary
evil". it would be innitely preferable to not have an air
gap in a SE OPT if that were possible. In spite of escap-
ing the need for an air gap, the PP OPT is faced with
another, inescapable handicap in that the two halves of
the AC cycle are of opposing polarity. This means the
magnetic ux is in a different direction on each half-
cycle. The PP OPT faces a grievous hurdle here, at the
sinewave's zero signal crossover point where the effects
of hysteresis take their toll on it.
Hysteresis, the Hysteria Inside Your PP OPT
Hysteresis is an unfortunate phenomena, and also an
unavoidable one. Hysteresis is due to the existence of
retentivity in the magnetic domains of the lamination
material, the very thing that makes certain metals per-
manently magnetic in the rst place. In simplest terms,
retentivity can be thought of as residual magnetism, al-
though it will be addressed more in depth later. Mag-
netic domains are regions of magnetic alignment in fer-
romagnetic materials, resulting from quantum mechani-
cal interactions at the atomic level. A high degree of
magnetization naturally occurs in ferromagnetic lamina-
tion materials within individual domains, but in the ab-
sence of an external mmf these domains are randomly
oriented. An externally applied unidirectional mmf
aligns the domains with the direction of the magnetiza-
tion eld it creates. The end result is a large net magnetic
ux, a multiplication or amplication of the applied
mmf.
Figure 4 depicts a hysteresis loop and it's parts. Examin-
ing it will help you to better understand the rest of this
section, and how it applies to our OPT.
Fig 4 A typical hysteresis loop, or "B-H loop"
Single Ended vs. Push Pull (Continued): The Deep, Dark Secrets of Output Transformers 4
When the applied mmf in a transformer ceases (the zero
voltage line of the AC sinewave), the magnetic ux
should theoretically return to zero. However, because of
retentivity (the tendency to retain residual magnetism)
the ux does not fully relax back to zero when the mmf
is removed, due to the fact that some of the domains will
still retain their unidirectionally oriented alignment. Just
as we read at the beginning that no transformer lamina-
tion material can achieve a theoretically innite mag-
netic ux density, no ferromagnetic lamination material
completely loses it's domain alignment when the mmf is
removed, and returns to zero ux density. All exhibit a
certain degree of retentivity. This retentivity makes possi-
ble the creation of permanent magnets, but in our PP
OPT this is highly undesirable! The point at which the
mmf ceases (the zero line on the oscilloscope graph) and
ux due to only residual magnetism remains is called
the point of retentivity. The level of residual ux at the
point of retentivity is called the remanence. It requires
considerable energy to overcome the remanence and
drive the core to the point of true zero ux, called the
point of coercivity. In our PP OPT the mmf expended to
take the core from the point of retentivity to the point of
coercivity is called the coercive force or coercivity. The
coercive force can be considered wasted energy, and
this inefciency is called core loss. This is one reason
why PP is less articulate, dynamic, and detailed at low
levels than SE.
After full negative-direction magnetization has occurred,
the cycle reverses again and begins it's positive magneti-
zation swing. However, the point of coercivity is differ-
ent on the return cycle, because it is skewed by what is
now a different retentivity polarity. Because the magnetic
polarity of the point of retentivity is opposite on each
half-cycle, so is the point of coercivity. Because of this,
the magnetization curve is not a uniform, linear line in
both directions. This impossibility to precisely track a
linear remagnetization curve when the alternating mmf
is applied is the property called hysteresis. Hysteresis
causes the magnetization curve to trace out an S shaped
loop, which is where the hysteresis loop in Figure 4 gets
it's name. The left-to-right width of the hysteresis loop at
the center point or waist is the degree of hysteresis pre-
sent. It is also sometimes called the B-H loop, where B is
the ux density and H is the magnetomotive force, or
mmf. All ferromagnetic materials exhibit a certain de-
gree of hysteresis, as the relationship between B and H is
never perfectly linear. The hysteresis loop can be simply
thought of as a representation of the magnetic ux den-
sity (B) as a function of the mmf (H). The more magneti-
cally linear the material is, the narrower the waist of the
hysteresis loop.
Lamination Materials and Their Properties
OPTs demand lamination materials with high magnetic
permeability (the ease with which it is magnetized) and
low retentivity and coercivity. The higher the magnetic
permeability, the lower the mmf that is required to attain
a given ux density, and oftentimes a smaller core mass
can be used as a result, along with it's benets of lower
leakage inductance and stray winding capacitance. If
saturation becomes an issue, core size can be increased
accordingly, and special winding techniques used to
help overcome the increased stray capacitance that usu-
ally results from the use of larger core mass. The lower
the retentivity of the lamination material, then usually
less coercive force that is required to overcome it and
remagnetize the steel in the opposite direction. So, the
higher the magnetic permeability and the lower the re-
tentivity, the better that details and microdynamics will
be rendered at lower signal levels.
Such high quality lamination materials are very expen-
sive. GOSS (grain oriented silicon steel) is one of the
most commonly used lamination materials for trans-
formers and electric motors, because of it's desirable
magnetic properties. Grain orientation lowers core
losses and extends the saturation point, and about 4 to
4! percent silicon lowers the remanence (retentivity).
GOSS laminations are annealed at about 1500 F (820
C) before use. Annealing relaxes the internal stresses
caused by rolling out the sheet and then stamping the E
and I shapes from it, and forms a non-conductive oxide
lm to reduce eddy currents. M6 is the grade of GOSS
most often used in high quality power transformers and
PP OPTs.
M19 is a non-oriented steel commonly used in lter in-
ductors, low-end power transformers for tube amps, and
in some SE OPTs. It has lower magnetic permeability
than M6, and requires greater core mass per volt-amp of
output in power transformers. It requires more energy for
magnetization reversal, and has a wider, less linear hys-
teresis loop. Because of this, it is less efcient in power
transformers than M6. Due to the fact that more of the
AC sinewave's energy is wasted as heat in reversing the
ux direction than with M6, power transformers built
from M19 exhibit higher heat rise due to the higher core
losses. Because of it's inferior properties, it should never
be used in PP output transformers.
Some very high-end PP OPTs use exotic materials such
as M4 GOSS, which has superior properties to M6 but is
very expensive. Amorphous iron is another exotic mate-
rial used in very high-end OPTs. It has an amorphous
molecular structure similar to that found in glass. It has
Single Ended vs. Push Pull (Continued): The Deep, Dark Secrets of Output Transformers 5
low retentivity/coercivity and high permeability, and is
therefore extremely efcient, with a very narrow hystere-
sis loop. However, it saturates at a lower ux density
than many materials, especially at low frequencies.
Nickel is added to iron to improve the permeability and
lower the core losses. It's drawback is a far lower satura-
tion point than GOSS. M6 typically saturates at around
20 kilogauss/2 Tesla, while iron with only 50% nickel
content saturates at about 15 kilogauss/1.5 Tesla. Be-
cause of it's extremely low core loss, nickel iron is one
of the nest materials available for transformers. 80%
nickel iron was originally christened Permalloy back in
1914, when Gustav Elmen of Bell Laboratories discov-
ered it's superb properties, but today these alloys may
contain different nickel contents. The premium trans-
former grade Permalloy consists of 79% nickel, 17%
iron, and 4% molybdenum. It has the highest permeabil-
ity and lowest core loss of any common transformer core
material, which makes it a very sensitive and low distor-
tion lamination material. It's low saturation ux density
of 8 kilogauss/.8 Tesla and extremely high cost limits it to
certain specialty applications, such as recording studio
microphone transformers, step-up transformers for low
output, moving coil phono cartridges, and other critical
small signal applications. It is also used as record head
cores in the nest tape recorders.
79% - 80% nickel Permalloy is also known by the
names Supermalloy, Mu-Metal, Ultraperm, and it's AISI/
SAE classication, M-1040. 36% nickel Permalloy is
commonly known as Radiometal, and 48% to 50%
nickel is called SuperRadiometal. Permeability falls as
the nickel content falls, and the saturation ux density
point rises. 80% nickel Permalloy has a saturation ux
density of only .75 Tesla, 79% nickel saturates at .8
Tesla, 50% nickel saturates at 1.5 Tesla, and 48% nickel
SuperRadiometal has a saturation ux density of 1.6
Tesla. So, the transformer engineer must carefully con-
sider his application and choose the appropriate nickel
content to avoid core saturation. Nickel iron laminations
must be correctly annealed to ensure the best magnetic
properties, and care must be taken to not expose them
to high impact after annealing. Hard, high-G impacts
(such as dropping a completed transformer on a hard
oor) can change the magnetic properties of nickel iron,
especially high-nickel content alloys.
High-cobalt steels such are sometimes used in certain
specialty applications. They can attain very high ux
densities before saturating, higher than even M4 or M6,
but have higher retentivity and therefore higher core
losses. Were it not for their extremely high cost, cobalt
alloy steels would be very useful for power supply lter
inductors. Size for size, a cobalt steel inductor is capa-
ble of handling much higher DC currents before satura-
tion than the commonly used M19 or M6 materials.
Because SE OPTs largely avoid the hysteresis effect from
maintaining magnetization in a linear region, their con-
struction permits lamination materials that could never
be used in PP OPTs, such as M19. However, most high
quality SE OPTs are made of M6 nonetheless. Because
SE OPTs suffer less from core losses than PP OPTs, co-
balt steels are sometimes used in certain SE OPT appli-
cations.
Torodial Output Transformers
Although toroidal power transformers are very common,
toroidal audio transformers are not seen much, due to
their inherit problem of high intolerance to offset DC.
They will saturate very quickly because their ux density
level typically is already higher than that of an E-I lami-
nation transformer, and there is no air gap in their core.
However, a few companies have "worked out the bugs"
somewhat with winding topologies and core geometries
that they are very secretive about. Despite their lower
tolerance to offset DC bias, toroidal OPTs enjoy several
specic advantages over E-I OPTs.
Toroidal transformer cores are made by winding a con-
tinuous strip of the core material into a donut shape,
called a toroid. Because of the way the core is made,
100% of the grain is oriented in the same magnetic di-
rection. This eliminates one of the major disadvantages
of E-I lamination transformers, in which only about 60%
of the grain is oriented in the same magnetic direction.
Although the grain of a stamped-out lamination is ori-
ented in one direction only, the magnetic ux in an E-I
transformer runs in all directions, and is asymmetrical in
different directions as well, as shown in Figure 5.
Fig 5 Grain orientation direction versus magnetic ux
direction in an E-I core
Single Ended vs. Push Pull (Continued): The Deep, Dark Secrets of Output Transformers 6
The E-I core's magnetic ux is omnidirectional, but the
lamination steel's grain is unidirectional, so much of the
core is "out of sync" with the ux direction. In a toroidal
transformer, the ux is symmetrical about the core, in
one direction only. In other words, any given cross-
section of an E-I transformer sees a different ux direc-
tion and density, where any given cross-section of a tor-
oidal transformer sees the same ux direction and den-
sity. This gives toroidals very low leakage inductance.
The windings are then evenly wound directly on the
symmetrical, gapless, round toroidal core, without a
bobbin. Because the windings are symmetrical and
even, the inter-winding capacitance can be very high
unless special winding techniques are employed to re-
duce it. However, because the toroidal's windings and
ux eld are symmetrical, several advantages are real-
ized. The ux density is much higher than in an E-I
lamination transformer, and every turn of wire sees the
same ux density. As a result, the level of mmf necessary
to establish and maintain the minimum ux density re-
quired for coherent, full bandwidth mutual inductive
coupling is very low. This minimal mmf, called the "qui-
escient" or "excitation" power, is typically over a dozen
times lower for toroidals than E-I transformers. Not only
are toroidal OPTs very efcient (typically over 95%) ,
they are ideal for push pull applications where little or
no offset DC is present to cause core saturation. The tor-
oidal OPT's low hysteresis and excitation mmf makes for
much better detail and dynamic response at low vol-
umes.
C-Core Output Transformers
C-cores are made in a manner similar to toroid cores, by
winding a strip of material, usually M6, into a shape as
shown below in Figure 5. It is bonded together and an-
nealed just like a toroid core, but is then cut into two
identical C shaped halves. The cut faces are precision
ground exactly square and parallel, as it is of the utmost
importance that they t together perfectly, or a large loss
of efciency will result. The wound bobbins are placed
over the legs of one core half, and the other core half
inserted into the bobbin until their faces seat together.
For SE operation, an air gap shim is added before as-
sembly. Then, the core is banded tightly around it's cir-
cumference with a steel strap to ensure tight contact of
the core halves.
Fig 6 C-core transformer construction
Like toroidal OPTs, C-core OPTs are not highly common
in today's tube audio world. However, like toroids, they
do exhibit certain advantages over the much more
common E-I lamination transformer. They are more ef-
cient than E-I transformers, though not as efcient as
toroidal designs. The bobbins can be wound on conven-
tional equipment, unlike toroidal transformers. Because
the primary and secondary are wound on separate bob-
bins and reside on different legs of the C-core, they are
physically separated from one another and their inter-
winding capacitance is practically nonexistant. The main
disadvantage of the C-core topology is that it's leakage
inductance is quite high.
To Parafeed or Not To Parafeed, That is the
Question
Parafeed (parallel feed) is worth mentioning here in clos-
ing, while we're on the topic of OPTs. In a parafeed to-
pology, the power tube is loaded by an inductor, resistor,
or active constant current source, and a DC blocking
capacitor couples it to the OPT. It has became very
trendy due to several reasons. It has far better immunity
Single Ended vs. Push Pull (Continued): The Deep, Dark Secrets of Output Transformers 7
to power supply ripple and noise (higher PSRR, or Power
Supply Rejection Ratio), because the power supply is not
routed through the OPT as in conventional SE output
stages. The output stage's AC ground is the OPT primary
directly to ground versus back through the power supply,
which nets a lower impedance. There is no offset DC
present in the OPT, because it is blocked by the
"parafeed capacitor". In the absence of DC bias current,
the OPT can be made much smaller, and without the air
gap required by conventional SE OPTs. Due to the small
core mass, leakage inductance and stray capacitance are
very low. Manufacturers of parafeed ampliers tout the
parafeed topology as the be all, end all of output topolo-
gies, with the best specs and best sound. However, they
do not tell you the whole story.....
It's worth noting here that although they are single
ended, parafeed OPTs suffer from hysteresis in the same
manner as PP OPTs do. This is of course due to the fact
that no offset DC is present to hold a linear static mag-
netization. But, a high permeability, low core loss mate-
rial such as Permalloy can be used to great effect in the
parafeed OPT to minimize hysteresis effects. However,
there is no way to avoid the other problems inherit to the
parafeed topology cause by the DC blocking capacitor,
such as frequency-dependent phase shifts, a charge/
discharge curve that is non-linear with frequency, and
losses caused by the capacitance being in series with the
OPT's inductance. Although parafeed can sound in-
credible if done correctly, nothing is perfect. Especially
transformers ...
Eddie Vaughn
Single Ended vs. Push Pull (Continued): The Deep, Dark Secrets of Output Transformers 8