BIAS STABILIZATION, 2014.
This page explains cathode bias and fixed bias for all output
tubes and application of
Dynamic Bias Stabilization for PP amps with cathode bias to give
much better class AB operation.
Biasing of tube amps is a critical design issue and is not much
understood by anyone without technical training.
Before I explain Dynamic Bias Stabilization, DBS, and its role in
my 300W monobloc amps, readers should
be aware of terminology used in relation to the use of KT88 and
6550 or similar output tubes.
The mention of a whole lot of differently named Vdc voltages such
as Ea, Eg2, B+ Eg1 are confusing to many but
I must write with the many abbreviated terms to be concise for
those prepared to learn the language of tube electronics so
they may more easily build an amp and understand what they are
Practice makes perfect, after I began 1993 I found the book with
consistent terminology was the
Radiotron Designer's Handbook, 4th Ed, 1955, so I have tried to
keep terms used in 1955 here as this website.
"Bias" is a term used to describe the relation between the DC
voltage and DC current operation of an output tube
in relation to the applied Eg1. Bias is usually stated for the
idle condition with no signal present. The amp operation
is optimal when the bias condition at idle is maintained
throughout the range of signal operation from idle to beyond
Unless stated otherwise, signal voltages are 400Hz to 1kHz sine
waves at constant amplitude and THD < 1%.
Briefly, "the amp bias" means the relationship between Eg1 and
Iadc. But bias can also be "screen bias" which
nominates the Eg2 which is usually between slightly less than Ea
or much less than Ea, and which ideally should
be well regulated. In Triode of Ultralinear amps the screen bias
is very close to Ea because the screen is connected
to anode for triode, or to a tap along OPT primary winding which
has very low winding resistance. But for pure beam
tetrode or cathode-feedback (CFB) operation the Eg2 is a fixed Vdc
Setting the bias, or adjusting bias or "auto biasing" or "cathode
biasing" describes the method used to stabilize DC
working conditions of a tube at the idle condition without any
signal present. "Biasing" may not remain constant during
signal operations especially at elevated levels and in class AB,
and usually much unwanted signal distortion occurs
when biasing changes during such operation. There are two common
methods for biasing,
1. cathode bias, aka auto bias.
2. fixed grid bias.
Fig 2, Bias methods....
Fig 2 shows Cathode Bias aka Auto Bias on left, and Fixed Bias on
The KT88 are in triode in both schematics and have identical idle
conditions of Ea = +379V, Eg2 = 379V,
Ia+Ig2 = 50mAdc, Eg1 = -41Vdc. The biasing level is chosen so an
output tube produces a safe and constant
level of heat from the anode and screen which allows optimal class
A or AB operation with minimal crossover distortion.
With screen connected to anode we may include screen Idc as part
Anode heating is called Pda, power dissipated at anode, in Watts.
For simplicity, Pda here means anode and screen heating.
So, Pda = Ea x ( Ia + Ig2 ) = Ea x Ikdc, where Ikdc is the cathode
dc current. For Fig 2, KT88 Pda at idle = 390V x 0.05A = 19.5W.
This is a comfortable level of heat well below the maximum Pda
rating = 42W. For class AB, Idle Pda should be < 50% max
Pda and for pure class A can be up to 67% of max Pda. No octal
tubes should ever be idled at the maximum Pda.
In the Ea and Ia idle condition for Fig2, KT88 triode has Ra =
1k6, µ = 8, gm = 5mA/V. This is very different from KT88
data figures measured with much higher Ia and gm with lower Ea but
if you carefully examine the KT88 triode Ra curves
you will see what I mean.
Cathode bias - Advantages:-
Cathode bias aka auto-bias offers excellent and simple idle
current and grid bias voltage regulation with use of
Rk & Ck parallel network between each cathode and 0V. There is
no need for an owner to make any periodic adjustments.
Owners do not need to remember to do bias checks and there is no
need to to use a screw driver and voltmeter after
removing amp covers to make voltage measurements while exposed to
Cathode bias is most ideal for all pure class A amplifiers such as
all SE amps. If there are multiple output tubes there
must be separate Rk & Ck for each tube and not one Rk & Ck
network for commoned cathodes. The separate Rk & Ck
ensure all tubes have equal Ia and Ek. With separate Rk & Ck a
single tube of say 4 output tubes can be replaced
without needing to worry about it being closely matched to others,
ie, with exactly the same electronic parameters.
So if the replacement is a different brand and year of production,
it will work fine. If output tube parameters of gm, µ,
Ra change over years, cathode bias tends to keep Ia and Ek the
same and maintain the same tube operation and
extend their life. Pda of output tubes is kept fairly constant
despite large mains voltage changes between say 220Vac
and 250Vac, because at DC the effective Ra of tubes is much
increased with presence of Rk.
Cathode bias - Disadvantages:-
All tube amplifiers whether SE or PP will be used with speakers
which have higher or lower impedance than
recommended by the amp manufacturer. Owners routinely ignore user
manuals. The SE class A amps with cathode
bias usually or fixed bias can tolerate a range of RL between half
and double the ideal load for maximum Po and which
should be labelled on output. The SE amp may have only one
labelled outlet as 6r0 for 25W and if 3r0 or 12r0 are
used the Po may be 12.5Watts in each case. The SE amp cannot
produce more Po if RL is lowered, and the practical load
limit is 12.5W for all loads between 3r0 and 12r0, but it will be
all class A and usually sound well if that is enough power
without any clipping. SE amps usually have output tubes idling at
67% of max Pda and neither fixed or cathode bias can
improve bias stability when RL is low, nor allow higher Po when
load is low.
In PP amps the idle Pda is often less than 50% of Pda maximum and
the maximum class A Po might be 12.5W with 12r0.
When 6ro is used, Po is 25W class AB and with 3r0 is is perhaps
40W class AB. With cathode bias, the Ek remains stable
for all loads above 12r0, but for any load below 12r0 the Ek will
to more positive and cathode biasing is the least suitable,
with high 40W into 3r0 "not worth having" so in fact the PP amp
with cathode bias offers little more than the SE.
Many old PP amps from 1950s did use cathode biasing such as
Quad-II and Leak. The makers realized that owners
seldom used more than 1/2 a W average to each of two speakers
which had sensitivity of 93dB/W/M, so class AB was fine
because even with 4r0 speakers connected to amps labelled for
16r0, the outcome was listenable. But many ppl now
want louder levels of music with high bass content and with
speakers rated for 87db/W/M, so 4 times the Po is needed,
so instead of 2 x EL84 or 6V6 per channel, you now need 2 x KT88
or 6550. Instead of having a pair of KT66 or 6L6 in
triode for 12W pure class A, people now will prefer a 30W
capability for all loads between 3r0 and 9r0 from 2 x KT88 or
Fixed bias works better for where low RL loads are
Fixed bias - Advantages:-
In general fixed bias benefits operation of PP amps which are to
be used with a wide range of load values usually all
below the RLa-a which gives pure class A Po. In Fig 2, the fixed
bias example requires 17r0 at "8r0" outlet of OPT to
get the same 12.5W of pure class A Po as for cathode bias. When
the RL is reduced 4r0 about 20W is available with
fixed bias and there is no bias drift of Ek.
Nearly all guitar amps are PP with fixed bias but have only one
adjustable -Eg1 which is applied to say 2 or 4 x 6L6GC
in the amp output stage for 50Watts or 100W class A Po. The reason
for fixed bias is always the better high Po
performance and crossover distortion caused by cathode bias drift
when all output tubes are driven way beyond class
AB clipping and into class C operation.
Fixed bias - Disadvantages:-
Owners need to check the bias regularly and keep re-adjusting as
years go by, which they forget to do, or they adjust
everything wrongly because it involves technical skill, knowledge
of exactly what they are doing, use of Vdc voltmeter,
and a screwdriver.
Once you have 4 or more output tubes per channel the number of
adjust pots and time needed to get bias correctly
adjusted becomes a challenge for non-technical owners. Bias
adjustments are interactive, and when you plug in 4 new
tubes and turn on for first time the bias currents of each may
So you must first turn all bias pots low, and the pot position for
low bias current may not be known. So an owner needs
to measure current quickly and turn bias down quickly before any
tube overheats during the bias setting procedure.
With bias on each tube turned low, each is then raised slowly. One
first to be raised to 50mA will then go lower as
other tubes are raised because the anode B+ is being pulled low
under increasing total Ia.
So one may have to repeat 8 adjustments several times before all
bias is set at the maker's recommendations and all
have equal Ia.
Tubes will change with age, and after say 5,000 hours of use they
may develop faults with air molecules leaking past
the glass to pin junctions. Often as a tube ages, the silver
gettering turns brownish-grey, and grid draws small grid current
at idle, so hence the need for grid bias resistors for fixed bias
amps to be to 100k or less. Expensive NOS tubes which
have sat on a shelf unused since 1960 will have a short life
because even without use some air leaks into the tube,
or soon will when tube has had a month in an amp after doing
nothing for so long.
I have known owners to regret not buying an extra couple of spare
tubes to cover early losses. In all ageing tubes
gettering becomes unable to absorb all the gas. The tube seldom
slowly reduces electron emission from cathode, but
tends to emit more and increase Ia for a given EG1 bias voltage.
With fixed bias, there is no cathode Rk to regulate
tube current and tubes can overheat and enter thermal run-away
operation with spectacular and expensively smoky
My previous attitudes and solutions for biasing:-
SE amps. In all the SE amps I have built I have used cathode
biasing with separate Rk & Ck networks for each output tube.
Sometimes I have used a combination of cathode bias plus a fixed
Eg1 because the Rk value could be lower value than
if the full Eg1 needed was all across Rk to get Ek = Eg1.
PP amps. In many PP amps I have used fixed bias, and in 100W ULAB
amp with 6 x EL34 I have used 6 adjust pots.
In later versions I have used 3 pots for fixed bias balancing with
LED indication which tells an owner if any one output
tube is not working.
All amps have active protection against bias failure.
In 2003, I invented the idea of Dynamic Bias Stabilization for
biasing all PP amps so that:-
1. Cathode bias with separate Rk & Ck could be used to
eliminate all adjustments,
2. Operation with very low RL and at high Po up to and beyond
clipping could occur without any significant positive
Fig 3 schematic was drawn in 2014 and is virtually the same as I
originally posted up in 2003, but it shows ALL the working
voltages correctly and for use with Sovtek KT88 which are very
similar to Sovtek or EH 6550. Use of 6CA7 ( EL34 )
were also tried in the same circuit and Ek at idle = +35Vdc, so
Ikdc = 40mAdc approximately, and R14 and R15 could be
increased to 8r2. 10r will a little too high.
The data for some common NPN TO220
package power bjts is :-
|Ic max, Amps
|Hfe bias 1A
The cathode Iac and Idc currents must flow through R14&R15. At
idle with Ikdc = 50mAdc, the Vdc +0.33Vdc and in class A
operation the Iac flow in R14&R15 is +/- 50mA approximately,
so the V change is from +0.33V to +0.66Vpk to 0.0V approximately.
In class AB the peak positive Iac at cathode is up to +350mA and
Vac at R14&R15 could be +2.4Vpk. But well before this Vpk is
reached Q1 & Q2 are turned on via R18&R19, and Q1 & Q2
conduct peak current through R20&R21 33r so the peak cathode
current is bypassed through Q1&Q2 and C5&C6 do not charge
up very much. C5&C6 do act to prevent high distortion voltages
which would appear between cathode and 0V.
The actual Vac at cathodes is the same as appears at R14&R15,
and is a square wave of +/- 0.7pk with top having a slightly
rounded top. This cathode waveform has extremely low contribution
to THD, and the total operation at clipping with very low
RL is equal to fixed bias operation. The 33r value is not critical
and can be between 27r and 47r so that maximum peak
current when shunting 1,000uF with Q1&Q2 is limited to less
than the bjt maximum Ic rating, often about 5Amps. The 33r and
bjts will become warm during continuous sine wave clipping. I
found they did not need heat sinking even with a 2r0 output load.
If Ik peak = 400mA, average cathode Ik = 126mA.
I have included a basic protection board ( bottom left ) which
turns off the large mains PT if excessive Idc should ever flow at
idle in V1 or V2. The same board could be used with many output
But in September 2014, I found this schematic better than the 2003
Fig 4 schematic shows R14 and R15 without any Idc flow which
allows the R value to be chosen for
better Ek regulation during class AB operation without any worry
about the effects of idle Ikdc.
I found R14&R15 needed to be higher ohms than for Fig 2,
because the Idc current flow does not flow in R14&R15,
and at idle the Vdc across each is 0.0Vdc. For Q1&Q2 to turn
on the base voltage must be above +0.55Vpk.
The minimum Vac pk at R14&R15 for beginning of Q1&Q2
conduction must be as a result of peak Ik = 2 x Ikdc,
100mA pk in this case. I found 15r quite suitable and Ek rise was
less than +10% with very low RLa-a at clipping.
The operation becomes equal to fixed bias operation.
With normal low average Po being all in pure class A, the Q1 and
Q2 never do anything and there is no effect on
signal path. But where a sudden music transient occurs, all high
peak currents are instantly bypassed by Q1&Q2 and
Ek remains constant. During high Po in class AB, it is difficult
to measure any additional THD above what is measured
with fixed bias.
The DBS action is reduced at HF by the effect of C7&C8 20nF
which create pole so that DBS is reduced at F above
10kHz where little musical energy exists.
Some of you may try to add DBS to an amp where idle Ia is higher
or lower than what I have here.
The values of R14+R15 and R18+R19 can be reduced and in my 300W
amps equivalents to R14+R15 are 2.35r, and
I have R18+R19 at 200r, with diodes limiting applied Vpk to bases
of gutsy 2SD424 bjts.
With just 2 output tubes you could use 22r for R14&R15, and
then have an additional R from bjt bases to 0V to determine
beginning of bjt base current turn on.
It is possible that if bjts turn on with Ik peaks less than 2 x
Idle Idc, then Ek will decrease after a few first Watts, then
slowly increase with class AB, or stay lower which means that
tubes are pushed into heavier class A operation and this
extends the amount of class A Po available before class AB
threshold is reached, but I favor keeping Ek constant and allowing
a maximum of +10% at sine wave clipping.
Some may like this DBS idea, and they are not entirely baffled.
The bjts act as silent slaves to allow the tubes to work better.
Without DBS as I have it, operation with a sine wave is at
clipping with low RL is entirely depressing to behold!
The protection circuit.
Fig 4 offers best protection against any one of more output tubes
conducting too much Idc at idle. The best way to prevent
overheated tubes from causing damage to themselves of other
amplifier parts is to turn off the amplifier.
In Fig 4 the idle Pda = Ea x Iadc = 382V x 0.05A = 19.1Watts. Ek
Suppose idle Idc increased from 50mA to 70mA, then Ek = +57Vdc. Ea
will then be 423V - 57V = 366V, so Pda = 366V x 0.07A = 25.6W.
If Ek at one KT88 were to rise to +57Vdc at idle it indicates
something is very wrong with the KT88, because the increase is
normal Ek +25%.
If there is a difference of say 20mA of Ikdc between both KT88, it
means there is a net 20mAdc flow across the whole of the OPT
primary which would cause the ungapped OPT core to become
saturated by Idc, therefore causing serious THD & IMD and
high peak signal currents.
The amp should be shut down before anything gets too hot, and KT88
replaced. If a known good KT88 is fitted, and problem
continues, the grid coupling cap may be faulty.
The protection circuit turns off the amp if Ek1 or Ek2 or both
reach +57Vdc or higher.
The other reason Ek must not be allowed to rise too high is that
the V rating for Ck 1,000uF caps is usually 63Vdc, and if
Ek stays above +63Vdc for minutes, the cap may leak current, short
circuit, or overheat and explode.
Ek1 and Ek2 are applied to two R&C filter networks
R3&R4&C1, and R6&R7&C2. The Vdc at tops of C1 and
C2 have low
Vac content and slow down rise signal from Ek. Both Vdc at
C1&C2 are applied to diodes to R5 and base of Qx which is a
cathode follower which is high base Z in, and low Z emitter output
which feeds current through 5.6V zener diode to R2 and
R1 and scr1 gate. The Vdc at Qx base is determined by whichever Ek
is highest. One could have cathodes and 100 diodes,
and the Qx base is still only driven by the highest Vdc before the
Scr1 is sensitive gate C106D scr needing +0.67Vdc at 0.03mAdc to
latch on. For this to happen, Ek must be +57Vdc.
During normal operation, Qx has no collector to emitter current
because no current flows in 5.6V zener diode.
When scr1 latches on, it turns on the coil of Relay 1 in PSU. The
relay opens the Neutral line to mains input to large PT1
and the whole amp is turned off. The green "on" LED turns off, red
"fault" LED turns on. The +12Vdc rail for relay 1 is supplied
via a 7VA auxiliary mains transformer with 12Vac secondary. It is
turned on with PT1 but its Neutral line is before Relay 1.
Scr1 stops conducting only when the amp is turned off at mains
switch and reset by turning back on after 2 seconds.
In my 300Watt monobloc amps I have applied the principles of DBS.
The amps have CFB windings which make DBS
application slightly more complicated. I found I only needed 2 x
2SD424 to look after cathode current management of 12 x 6550.
In my 300W amps I also have a regulated Eg2 at a lower Vdc than
Ea. If Eg2 were ever to rise to same value of Ea, the
protection circuit would turn off the amp. However this is very
unlikely because the Eg2 SS regulator has a current limiting
collector R and zener diodes tend to fail by reduction of their
Now let us consider a 300Watt+ monobloc amp.
With 12 x KT88/6550 output tubes, fixed bias would mean having 12
adjust pots and with 2 channels there would be 24 pots,
and all are a little interactive, and bias setting would be an
The use of cathode bias plus fixed bias worked just fine if I used
1. Anode B+ = +512Vdc. Anode idle current = 40mAdc,
2. Screen B+ = Regulated +375Vdc to +387Vdc, depending on real
I found "75V" zener diodes idled at 78Vdc with only 4mAdc. Ig2 at
12 x 6550 = 48mAdc and maximum at clipping with low RL = 150mA
3. Cathode resistors = 500r, using 3 x 1k5 x 5W in parallel per
4. Cathode bypass caps = 1,000uF x 63V per output tube.
5. Fixed grid bias = -17.6Vdc applied to all 6550 grids, and
derived from rectified 12.6Vac heater windings.
If I used my method of Dynamic Bias Stabilization in conjunction
with the 12 x R&C cathode bias networks, the rise of Ek at
each of 12 cathodes was from +23.7Vdc to +27Vdc at clipping when
using 1/2 the recommended RL at Po = 390W.
As far as I know, my idea of dynamic bias stabilization or
"clamping" of Ek in class AB amplifiers using cathode biasing
has never been officially invented before 2004 when I tried to
describe a basic version of it to a news group, rec.audio.tubes.
I don't think anyone who saw my invention could understand how it
worked, even after posting the basic schematic
to the news group alt.binaries.schematics.electronic.
One lad tried to tell me my idea could not work at all, so I asked
him to test it, and of course, he would do no such thing,
so typical of men on the Internet. You are at....
amp dynamic bias stabilization
amp input/driver and output stages
power vs load graphs
300W amp images,
tubes with blue glow, and more views of amps
Back to Power
Back to Index Page