DYNAMIC BIAS STABILIZATION FOR 300W AMP.

Before moving to my bias stabilizer schematic, some basic biasing ideas need to be explained, since there are so many questions raised whenever anyone seriously disusses tube amps.

Biasing of tube amps is an important  issue in the mind of any amplifier designer.
What is meant by the term biasing?
Bias is a word meaning preference for one direction of thought when many choices are optional.
All vacuum tubes need to have a certain value of current flow through them even with no signal present
and the amount of current in the idle condition is known as the anode bias current, and the applied grid voltage is known as the
grid bias voltage. Where there is a resistance load in the case of a small signal amp, the anode voltage is set for an optimal voltage working point by adjusting the dc voltage between cathode and grid to get a wanted amount of anode current
and hence set the anode voltage. With power tubes connected to a load with a transformer,  the same cathode to grid
dc voltage adjustment needs to be made to set the wanted anode current.

So for example, in a 6550 output tube there are *two* idle bias conditions to worry about.
We need to set the tube up for the wanted anode current at idle by applying the correct value of control grid voltage
and screen grid voltage. Both these grids have control over anode current flow, ie, they have transconductance, gm,
which is measured in mA/V of anode current change with grid or screen voltage change.
In a 6550, grid gm = approx 10mA/V, and screen gm = approximately 2mA/V
The latter figure is one rarely quoted for screen data.

For pure class A operation the idle anode bias current is usually chosen so anode current x anode to cathode voltage, Ia x Ea,
= 2/3 of the maximum rated anode dissipation of the tube. This product in watts is known as the anode heat disspation, Pda,
or 'power dissipated at anode'. Ea is the dc voltage  between anode and cathode, and not the B+, which is the dc V between the supply and 0V, and includes Ea, and Ek, the cathode dcV between cathode and 0V, and any voltage drop across
load or lower winding resistance.
So in the case of a 6550, Pda from the data sheets is 42 watts, and we might set the tube up for Pda = 28 watts.
We may also have an anode B+ = +370V. The anode current will therefore = 28 / 400 = 76 mA.
Suppose we set the screen voltage, Eg2, for +330V for the best linearity in a class A circuit, ie, just a little below Ea,
then it is unknown what value of grid voltage needs to be applied to get the wanted current.
Where the cathode is directly grounded and thus at 0V potential, the grid dcV can be adjusted until it produces the wanted Ia and the voltage applied in the above 6550 will be between about -25V and -40V.
This situation is called fixed bias; the bias grid voltage is variable until the adjustment has been made manually
and then when adjusted it remains fixed until the tube ages and re-adjustment is needed.
The "grid bias voltage" is applied usually from a resistance of about 100k ohms from a bias supply network shown in many
of my schematics. Signal voltage is applied to the grid via a preceding voltage amplifier and coupled with a capacitor, usually about 0.47uF, so the dcV at the grid remains always at its bias voltage regardless of signal voltage +/- changes
and dc volt bias on the the input side of the grid coupling capacitor.

But we can also connect the grid to 0V via a 100k resistance so that very low frequencies the voltage at the grid is always at 0V potential. The grid *must* be negative with respect to the cathode to get a wanted bias current flow. So with the grid at 0V,
the cathode voltage must some how be raised to between +25 and +40V to get the cathode to grid voltage
wanted to produce the right idle current. We will place a resistance between the 6550 cathode and 0V and allow whatever
anode current which flows also in this 'cathode resistance', Rk, to develop a positive voltage. This positive voltage is known as the cathode bias voltage, or Ek.
You must remember that not only anode current also flows from cathode to 0V; there is also the screen current, approximately
5% of anode current, so about 4mA in this case, so a total of 76 + 4 = 80mA of cathode current flows at idle..
Setting the wanted bias current is done by adjusting the value of Rk until Ia is correct.
So if Ek = +30V, and Ia = 80mA, then by using Ohms law, Rk = 30 / 0.08 = 375 ohms.
This method of biasing is called cathode bias, or sometimes called 'auto' bias, because the Rk
acts as a negative current feedback element.

With a transformer coupled 6550, at dc the tube has virtually no anode load and acts like a 'cathode follower' where the
cathode output resistance = 1 / gm of grid1. Since gm = approx 10mA/V, Rout at cathode = 1 / 0.01A = 100ohms.
Suppose a a slight increase in Ik occurs. Ek will rise a little, making the grid seem more negative and the very
rise of Ek tends to reduce Ia, so the Ek is allowed to regulate the Ia and does so remarkably well when you have
the Rk = several times the cathode output impedance.
The advantage of cathode bias is that it self regulates the bias current and any unmatched tube can be plugged in and the
bias current will automatically adjust itself close to the wanted value.
Many class A amps use cathode biasing.

However, with cathode bias there is a need for a cathode bypass resistor, Ck. Without Ck, with signal operation as distinct from dc operation, the effective anode dynamic resistance, Ra', increases dramatically. Ra' is increased because the tube
is subject to the local series current feedback from Rk.
In the case of a 6550 connected as a beam tetrode, Ra = 18,000 ohms approximately,
and the amplification factor, µ, = 200 approximately, and the effective Ra' with an unbypassed Rk
= Ra + ( µ x Rk ) = 18,000 + ( 200 x 375 ) = 93,000 ohms, a rise of about 5 times,
and the aim is to build amplifiers with low rather than high output resistance so unbypassed cathode resistors are
never used in tube amp output stages. Even with triode or UL connected the use of a single unbypassed Rk
to each output tube raises the Ra' of the tube too much.
I bypass Rk usually with 1,000 uF which has a reactance, ZC, = 16 ohms at 10Hz, and only 0.16 ohms at 1 kHz,
so effectively with Ck the ac signal from the cathode is almost completely shunted to 0V.

During pure class A operation up to clipping in a pure class A amp the Ek of each cathode hardly changes, and cathode biasing is simple and works well.

Difficulties arise with cathode bias when you try to use it in a class AB push pull amp where the Pda may be
less than 2/3 maximum Pda, and the signal current change at the cathode is much more when the tube turns on
than when it turns off when passing ac signals. 

Below I explain further what happens in class AB cathode biased amps and what can be done to improve the
circuit working, and improve the music if such an amp is turned up loud.

As far as I know, trying to dynamically stabilize or clamp the cathode voltage of class AB amplifiers
using cathode biasing has never been *officially* invented before 2006, and when I tried to
describe a basic version of it to a news group ( rec.audio.tubes ) a couple of years ago in 2004, 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.

So allow me to introduce a new idea for bias control in push pull amplifiers.

First, let's have a look at the basic sketch i submitted to the news group 2 years ago :-

Schematic basic dynamic bias stabilization.

The above schematic shows a typical ultralinear 35w+ amplifier output stage which has been used in countless
amplifiers since about 1955, except for one major addition which wasn't possible in 1955.
There are two power transistors in the cathode circuits.
( Now just be calm, its ok, they are not bandits waiting to rob the amp of its marvellous sound quality!)

Let us examine the schematic for a minute and imagine that the transistors and the 10 ohm R are not in the schematic,
and all we have is standard a cathode bias circuit, and unlike my more complex 300w design shown further down this page, there is no cathode feedback windings or multiple tubes to befuddle people when looking into the schematic.

Consider this output circuit working with a sine wave at 1 kHz.
Consider that under normal class A conditions that the signal currents at the cathode of each output tube have a maximum undistorted +ve and -ve going value equal to approximately the idle Ik during each sine wave.
So if Ia = 50mA per tube, then we will have about +55mA, and -47mA current swings.
The slight difference in Ia swings is due to the even order distortion and mainly 2H currents that flow in each output tube
during normal class A operation, where each output tube acts as though it was a single ended class A tube.
The RL a-a shown = 6.6k, so while in class A each tube's class A load = 3.3k, and with load current
equal to 52mA peak, maximum class A from each output tube = peak Ia squared x RL / 2 = 4.5 watts.
So we will only get 9 watts in class A from this amp but class AB power maximum can be calculated at 47 watts.

The voltage across the 1,000 uF cathode bypass caps and 750 ohm cathode R will rise slightly at 9 watts perhaps by around 5% since there is only slightly more current trying to charge the Ck than is drained out of it during current cycles of each sine wave.

While the amp works in class A there is no major bias stability problem since the bias changes only slightly
between the grids at 0V and cathode voltages at say +38V.
A slight change of +2 V will not displace the bias working point of the tubes and won't increase thd very much at all.

But now let us consider class AB working. Above the 9 watt level produced by the above output stage the tube that has its
anode current reduced by a -ve grid input signal reaches a current change limit, and turns off completely, so no more
current reduction is possible, and since it then is effectively not connected to the load and not giving and V x I change to the load its contribution to power production ceases and its as if the tube was removed from the chassis. But while one tube is cut off, the other is being turned harder by a +ve grid signal and it  is the only tube connected to the load through the output transformer. The turn ratio between tubes and output load winding then becomes 1/2 what it is with class A operation and the load seen by the single tube = 1/4 of the RLa-a value, or 1,650 ohms in this case.
Load line analysis will show us that when grid voltage = 0V is at its maximum at clipping, anode voltage
has pulled down to just +80V and there is 390V across the 1.65k so the RL peak flow = 390 / 1,650 =  236mA,
which is nearly 5 times the idle current or maximum peak class A current swing.
In this case at clipping the there will be a much bigger positive peak current flowing in the tube than the the negative current swing when the tubes are turned off.
So during each wave cycle at clipping, each cathode bypass cap will be charged up by the difference in +/- cathode current to about +70V just like the caps working off a tube rectifier, because more current runs into the cap than drains out during
each high level wave cycle.
This is the result of large even order distortion currents in each output tube in class AB PP amps.
This phenomena causes the tubes to then act as if the bias conditions had been changed so that applied grid bias voltage
had been increased from -30V to -70V, which is a case of over biased tubes, and serious crossover distortion is generated which cannot be much reduced by NFB, because while the cathode voltage is at 70V, both the tubes are in cut off where the signal is at the zero crossing point and so for part of the wave cycle there is no voltage gain , so large distortions occur.
Fixed bias is mostly used to avoid the problem of cathode voltage upward drift due to the heavy
2H tube current flows in class AB.

This all sounds dreadful, but in practice when you only want 1 watt average from any amp and the reserve of 47 watts
is only used for occasional transients that don't last longer than a split second, the bias bias voltage hardly moves, since it takes
a long time to charge up the 1,000 uF cathode caps, and no music is a constant amplitude single tone signal.

But where you want 6 watts average from an amp like this then the bias voltage is constantly varying, wandering up and down
because of the tendency to move into class AB from class A.

Fixed bias avoids the varying bias problem with high output power, but bias needs adjustment
and tubes will age, so bias currents will vary, although not because of signals applied.

But what if there was a way to avoid the problems of tube ageing and having to adjust the fixed bias currents
with adjust pots to set the grid bias voltage, and yet still be able to use cathode RC biasing?
Well there is a way to avoid both, and enjoy the natural excellent quiescent bias stability of ordinary cathode bias
while the amp works in class A, and also control cathode voltages lurching around on musical signals
caused by class AB action.

So far we have examined the schematic without the R9, R10, R11, R12, and two transistors and R13, R14. 

Now let us add all those in and see what happens...

Consider R9, R10, 10 ohm resistors. They will have 0.5 Vdc across them at idle since that is caused by the 50mA
bias current dc flow from cathode to 0V.
But during heavy class AB working, the tube load becomes 1.65k, and the Vswing at the anode
is say -390V peak, so peak current at the cathode = 50mA plus peak load current of 236mA = 286mA,
and this would generate a feedback voltage at the 10 ohms = 2.86V peak.
The 1,000uF has a very low impedance at audio signal frequencies, so the cathode is subject to such current feedback voltages except at very low frequencies below the audio band.
The current feedback voltage will have a negligible effect on the tube gain because 10 ohms is such a small value of cathode resistance.

Now consider the action of the added BD239B transistors. (We can use almost any normal garden variety TO220 low voltage power transistor; I used a pair of BD339B laying around.)
The emitter is taken to 0V, the base is fed through 1k, R11/12, from the top of the 10 ohms.
The 1k limits the base input current to the transistor by forming a voltage divider with the base input resistance
which is quite non linear, ie, Rbin is high when no current flows, but low when a lot of current flows from collector to emitter.
The 0.02 uF, C7/8,  prevents the circuit from working when the signal rises above audio frequencies,
where we don't need or want the circuit to work.
The collector is taken via a 47 ohm R to the cathode.
No current ever flows in the 47 ohms unless the transistor is turned on.
The transistor is fully turned on when base voltage is at about +0.8V,  and fully turned off when base is at 0.5V.
When the +ve moving voltages in the 10 ohms rises in excess of approximately 0.5V,
the transistor just begins to turn on, and cathode current peak surges that would normally charge up the 1,000 uF cap
are bypassed though the transistor and 47 ohms to 0V, so the voltage at the Ck remains at near the idle voltage.
The use of the 47 ohms is about right for an average amp and the value can be anywhere
from say 10 ohms to maybe 100 ohms; it isn't very critical; there must be a collector resistor to limit the collector
maximum current to prevent the transistor from fusing.
47 ohms also seemed to work to stop the cathode voltage from *reducing* below the idle level when large sine wave
signals were used because too much ac signal current shunting occurs between idle and maximum class AB action.
 
Transistors are good, but we do have to molly coddle them and prevent their all so easy failure when currents could exceed their SOA, safe operational area, so the 47 ohms will prevent excessive collector current.
When such precautions are taken, transistors can last almost indefinitely, and in this case to act as slaves to make tubes work better, and without causing any extra horrid distortions during the tube operations; in fact the transistors act to reduce distortions that otherwise will occur.

If the cathode voltages were allowed to rise from say 30V to 70V with a sine wave at clipping then the thd of an AB cathode bias amp ( with NFB ) can increase to say 5% easily at clipping, which is bad when compared to a fixed bias amp with the same tubes, where the same load and similar power might produce only 0.3% thd.

Without my regulation the cathode bias amp will experience slightly moving cathode bias voltages well before clipping with music signals since drum beats and transients cause momentary changes to cathode voltages, Ek, and hence the changes cause the bias between Ek and Eg1 to change, and the working point, so hence the distortion due to poor bias conditions,
and momentary dc current imbalances in the halves of the OPT primary,
thus causing unwanted patches of intermodulation distortions.

With my regulation, the Ek tends to remain very steady and the circuit can be trimmed to give only say 10% Ek rise with a sine wave ( worst signal condition ) where otherwise a rise of +100% may occur.
During a heavy drumbeat or sudden transient, the excess cathode current is shunted only as long as the transient occurs,
so the bias remains stable.

So my method renders the operation of a cathode bias amp to that of a fixed bias type.

The bias stabilizer should produce a slight increase in Ek without any fall in EK with increasing power between idle and clipping, with any load down to 1/3 of that required for full class A at clipping.
Experimentors with the circuit should try 10ohms for the current sensor resistances but may need to add an R from
transistor bases to increase the turn on threshold voltage at the 10 ohms at which current shunting begins.
The collector resistance values can also be trimmed for optimal operation.
The two transistors don't need to be accurately matched.

What if a tube decides to conduct more dc than usual due to ageing or a fault?

Say the bias idle current increased from say 50mA to 65mA.
The dc V across the 10 ohms would rise from 0.5V to 0.65V, and the transistor would
just be thinking about turning on.
meanwhile the Ek would have risen from say +38V to +49V,
and if we had an active protection circuit to detect such a dramatic Ek rise, it would
trip and turn off the B+, thus saving the user from a lounge room full of smoke.
So dc faults in tubes will not interfere with active protection measures.

And for those wondering just what the load would be for pure class A operation of the schematic above,
it is easily worked out.
In class A for any single UL connected or pentode or bean tube,
the peak voltage load swing positively and negatively = 0.9 x Ea.
In this case load V swing = 0.9 x 470V = 423Vpk = 299Vrms.
The peak current swing is simply +/- the idle bias current = 50mA = 35mA in this case.
RL = V / I = 299 / 0.035 = 8.5k ohms.

Since there are two output tubes in class A acting effectively in series on the load from anode to anode, the class A load
for both tubes is 2 x RL for one tube = 17k ohms in this case. But the output voltage anode to anode will be twice what it is for one tube because each anode's voltage is oppositely phased. Va-a = 598Vrms
Maximum power output in pure class A = Va-a squared / RL = 598 x 598 / 17,000 = 21 watts.
The input power to the 2 tubes is 2 x Ea x Ia  = 2 x 462 x 0.05 = 46.2 watts.
The efficiency in % while in class A = 100 x output power / input power = 100 x 21 / 46.2  =  45%.

Notice that for the above output stage the load that gives only class A power requires a load of 3 times the 6.6k
arbitarily chosen for class AB. Many amps have been built to this recipe to enable then to produce double the
maximum class A power with such mild bias conditions where Pda is only 23 watts per tube at idle.

Now let us move on to the 300W amps with a dozen 6550 output tubes.........

sheet 5
Schematic of bias stabilizer, 300w amp
 

The schematic has the current sensing R = 1.67 ohms consisting of 6 x 10 ohm 5 watt resistors in parallel and
connected between the ends of the cathode feedback winding and the commonly connected cathodes on each side of the push pull circuit.

The voltages in the 1.67 ohms are applied to the bases of the power transistors Q1 and Q2, which
in this case are 10 amp capable TO3 or flat pack types on a small Al angle heatsink under the amp chassis.
I used a pair of 2SD424.

R3, R4, C1, C2 act as base current limiters and a low pass filter. The collector current limiting R5 to R16 can bypass excess increase in signal currents without ever having the the peak bypass signal current rise to the rated collector current limit of about 8 amps.

All the IN4004 keep each cathode isolated from each other, but allow the common bypass of ac currents from the cathodes to the OPT cathode winding, so the cathode caps do not charge up under low load class AB conditions.
Some tubes will experience more bypass currents than others because EK varies between unmatched tubes, and some begin to bypass before others. But this is of little concern and I found that all Ek were fairly similar at idle, and all tubes conducted similarly, and under tests all 12 Ek voltages moved only slightly up and to about the same voltage. The excessive ac signal currents were being discharged from each cathode through each
of the 100 ohm resistors. 

So these 300W amps do not have to have 12 bias adjustments which would be a cause for anxiety for some if not many amplifier owners who will never become technically proficient no matter how much teaching and training they undergo.
Some people just hate biasing rituals, they are only eager to listen to music without having
to worry if their amps are biased correctly, or if they are likely to produce more smoke than Mozart.

Back to index page