"Crossover Notch" Biasing - Why It Should Be Avoided

General

The "crossover notch" method of biasing is touted by some as the only "correct" method of biasing a guitar amplifier.  This couldn't be further from the truth, and this method should be avoided, as it has numerous flaws that make it not only unreliable, but in some cases, even hazardous to the operation of the amplifier.  At best, particularly with lower plate voltages, it biases the amplifier to a very cold point of operation, just above class B.  At worst, particularly at high plate voltages, it can bias the tubes too hot, because you cannot remove all the crossover distortion in some amplifiers without exceeding the plate dissipation, either at idle or at full power.

What is the "crossover notch" method, and how do you bias an amplifier using this method?

The method, as proposed by several tube amp "gurus", requires the use of a signal generator, an oscilloscope, and a dummy load for the amplifier.  The bias adjustment is first set to the most negative voltage on the grids of the output tubes, and a 1kHz or 2kHz sine wave (depending on which "guru" you follow) is applied to the amplifier input.  The volume level of the amp is then increased until the output waveform at the dummy load ,as seen on the scope, is at a point "just prior to clipping".  The bias control is then adjusted until the crossover notch  "just disappears", supposedly setting the tubes to the only "correct" bias point.

A few reasons to avoid this method of biasing:

Biasing by the crossover distortion method is extremely inaccurate and non-repeatable because the point at which crossover distortion appears can very hard to detect and is subject to changes with load impedance, amount of negative feedback, and, in particular, with grid drive if the phase inverter is AC-coupled to the output tube grids (as it is in almost all guitar amps).  Following are additional reasons why this method should be avoided:
    • When driven into the positive grid region at clipping, the output tube grid acts as a forward biased diode and clamps the positive peaks of the grid waveform to a point slightly above the cathode voltage. As the input signal level is increased, the clamping action forces the average value of the grid waveform downward, effectively increasing the average negative grid bias.  This results in more crossover distortion, even if the amp is biased higher into class AB.  Because of this clamping effect, the amount of crossover distortion that you are trying to "bias out" will change depending upon how far into clipping you set the grid drive.  If you keep trying to eliminate the "notch", you will bias the amplifier too hot, and your tubes will be destroyed.
      • This method gives no indication of the actual bias current or plate dissipation in the tubes, so you have no idea whether or not your amp is biased into a safe region of operation.

 

         
      • If you have two class AB1 amplifiers, one with a plate voltage of 350V, the other with a plate voltage of 550V, and set them both using this method, the amplifier with the 350V plate voltage will be biased too cold, while the amp with the 550V plate voltage may be biased too hot.

 

         
      • If you try to bias a push-pull class A amplifier using this method, you will end up biased right to the cold side of class AB operation, and you will no longer have a class A amplifier!

 

       
    • If you have a single-ended amplifier, there is no crossover notch to be seen, so this method is useless. In some cases, particularly with large amounts of negative feedback, the notch on a class AB amplifier cannot be easily seen, either, except at very cold bias settings.

An example of the pitfalls of this method:

In order to illustrate the problems you can run into using this method of biasing, following are two simulations of a "typical" 50W class AB1 2-EL34 output stage, one running at 350V, the other at 550V.  The bias on each was adjusted using the "crossover notch" method, as shown in the accompanying output plots.   The bias voltage was adjusted until the crossover notch "just disappeared", after first setting the input drive level to the point where the output was just entering clipping, as is claimed to be the "correct" procedure, by those who promote this method.

The 550V amplifier:

This first image shows the output of the amplifier when the bias is set to -61V, which corresponds to a plate current of 15.4mA, and a plate dissipation of 8.5W.   A large amount of crossover distortion is present, as indicated by the "kink" at the crossover point (intersection of the 0V axis).

xover 550 61

 

 As can be seen in the following output waveforms, the crossover notch is still visible at a grid voltage of -59V, which corresponds to a plate current of 19.4mA, and a plate dissipation of 10.7W.  The schematic for this amplifier, showing all voltages and currents, can be seen here.

xover 550 59

If the bias voltage is decreased to -57V, the crossover notch is still just barely visible, corresponding to a plate current of 25.6mA, and a static plate dissipation of 14.1W.
The schematic for this amplifier, showing all voltages and currents, can be seen here.

xover 550 57

Finally, if the bias is reduced to -55V, the crossover notch has "just disappeared", as is claimed to be the "optimum" bias point by those who recommend this method.  This corresponds to a plate current of 32.5mA, and a static plate dissipation of 17.9W. The schematic for this amplifier, showing all voltages and currents, can be seen here.
 
 

xover 550 55

The 350V amplifier:

 As can be seen in the following output waveforms, the crossover notch is quite visible at a grid voltage of -39V, which corresponds to a plate current of 7.95mA, and a plate dissipation of 2.8W.  The schematic for this amplifier, showing all voltages and currents, can be seen here.
xover 350 39

 

If the bias voltage is decreased to -37V, the crossover notch is still barely visible, corresponding to a plate current of 11.9mA, and a static plate dissipation of  4.2W.
The schematic for this amplifier, showing all voltages and currents, can be seen here.

xover 350 37





Finally, if the bias is reduced to -35V, the crossover notch has "just disappeared", as is claimed to be the "optimum" bias point by those who recommend this method.  This corresponds to a plate current of 17.4mA, and a static plate dissipation of 6.1W.  The schematic for this amplifier, showing all voltages and currents, can be seen here.


xover 350 35



 
 
 
 
 
 

Now, as can be seen by the above output waveforms, the 550V amp is biased to a static plate dissipation of 17.9W, which is 72% of the max plate dissipation of the tube (25W for an EL34).   The 350V amplifier, however, ends up biased to 6.1W, which is 24% of the max dissipation of the tube.  Clearly, the 350V amp is biased far too cold (it can be safely biased up to around 50mA or more at this plate voltage), while the 550V amp is possibly a bit too hot, because of the increase in average dissipation that occurs in a class AB amplifier under full-signal conditions.

The effect of global negative feedback:

A far more insidious problem creeps in if the amplifier employs global negative feedback.  Feedback acts to reduce the distortion generated in the stages encompassed by the feedback loop.  Since the crossover distortion is generated in the output stage, and the feedback loop "wraps around" this stage, it will act to reduce the amount of crossover distortion in the output, by "feeding back" a correction signal to the input which predistorts the input signal in such a way as to reduce the crossover distortion at the output.

If we take our example of the 550V amplifier output stage and remove the global negative feedback as shown here (reducing the input drive voltage accordingly by the amount of the feedback factor, in order to set the output back to the point "just before clipping" as recommended), the results become even more interesting.  Following is the output of the 550V amplifier with no negative feedback at the -57V bias point. Compare this with the -57V bias point shown above in the amplifier with feedback, and you will notice that there is more crossover distortion, essentially the same amount as is seen in the -55V bias point.  This clearly indicates the effect of global negative feedback on the reduction of crossover distortion.  If this amplifier were biased using the crossover notch method, it would end up biased hotter than the feedback amplifier, because the crossover notch would "just disappear" at a lower bias voltage, corresponding to a higher plate current and a higher plate dissipation.
 
 

xover 550 57 no nfb

Conclusions:

Taken to the extreme,  if a tech were to bias a 250V 2-EL34 amplifier, which is a push-pull class A amplifier, using this method,  he would end up with a very cold class AB amplifier!

It is clear that this method gives no indication of the actual plate current and plate dissipation, so the tech hasn't got a clue whether or not the amplifier is operating safely.  If he then chooses to measure the plate current and calculate plate dissipation, he faces a dilemma - does he leave it at the spot where the crossover notch "just goes away", which is the supposed "correct" spot, or does he increase the bias current further because it is "safe" to do so?  If he adjusts it to a higher point, what is the purpose of looking at the crossover notch in the first place?  How about the other extreme - what if the amp is biased up to where the crossover notch "just disappears", and he measures the plate current, calculates plate dissipation, and finds it exceeds the level for safe operation of the tube?  Does he then adjust the bias colder, where the crossover notch reappears, even though this violates the recommended procedure?

The change in bias point for the same plate voltage and OT primary impedance just by removing the global negative feedback also is an indication of the flaws in this method.  If the plate voltage is high enough, and there is no global negative feedback, the bias current would be increased to unsafe dissipation levels.  If the plate voltage is low enough, the amp ends up biased very coldly, and may not have the best tone.

Because of the above illustrated flaws, among others, this method of biasing should be avoided.


Copyright ©  2000  Randall Aiken.  May not be reproduced in any form without written approval from Aiken Amplification.

Revised 02/19/14