Does biasing the output tubes for (relatively) no
cross-over distortion at idle or at low volumes mean
that you won't get cross-over distortion when you start
to crank the amp? I'm thinking you you'll start to see
that cross-over distortion at the higher volumes.

Different answers probably exist for the following
conditions: cold and cathode-bias, hot and
cathode-bias, cold and grid bias, cold and grid bias.
Comments?

(For those relatively new to the idea of cross-over
distortion, it sounds bad, otherwise, I wouldn't worry
about it. Being a "Guitarista," I only want to eliminate
sucky distortion, not the good distortion.)

For the "auto-bias" current regulated cathode biasing
scheme proposed by Hrvoje, I can say from experience
that the tendency for cross-over distortion increases
with the output power. For a given bias setting, a small
output signal would give no indication of the cross-over
notch. With that same bias setting, but driving near to
full output, the cross-over notch would show up. I had
to increase the bias current for no cross-over distortion
at full-power.

Maybe this all boils down to the following: do class AB
circuits by definition exhibit the cross-over notch at
some point before the onset of clipping? (A corollary
would be: "If you don't want the cross-over notch, bias
at full power for no cross-over notch. That means class
A operation, and for that you have to lower the plate
volts to keep plate dissipation low enough to not fry
your output tubes." Is that also correct?)

Thanks,

There will be no waveform to look at under no signal
condition, Carl. 
But you know that.
At low output levels the amp can be biased any normal
way and there probably will not be enough signal
voltage to drive it to anything other then class A.
I think the "O'scope biasing" guys idea is to adjust the
bias voltage at full power with the tops just starting to
show a flat spot and move the bias voltage to show no
Xover.
If the driver circuit is well balanced and(based on the
bias voltage applied to the power tube's grids)... until
there is enough AC drive to force one tube off on the
negative going swing, there should be no cross over
distortion.
Since bias is more of a DC term it would be easy to set
the amp overbiased and have all kinds of Xover
distortion even at low to moderate drive levels.
I used to work on Class C RF single ended transmitter
amps and they would draw very very little current at
idle. 
But when you drive them hard enough, they'd put out
thousands of watts of power, plus the duty cycle is very
low so you can abuse the livin' crap out of them! 

q{Maybe this all boils down to the following: do class
AB circuits by definition exhibit the cross-over notch at
some point before the onset of clipping? (A corollary
would be: "If you don't want the cross-over notch, bias
at full power for no cross-over notch. That means class
A operation, and for that you have to lower the plate
volts to keep plate dissipation low enough to not fry
your output tubes." Is that also correct?)}

Boy, I think that's close to it.
But IMO, not always. Some amps with enough drive
behind it can drive into class AB2 also.
I think the whole object of a push pull amp, beside
more deliverable power, is to get one tube to carry the
load while the other one is driven off. If the Xover
current is a smoother line because both tubes are "on"
(one is going off while the other is coming on)in that
overlap region, then it should have very little distortion
because of it.
If they both are being driven hard (+ an-) and one is on
hard while the other hasn't had a chance to come back
on... there's the xover distortion.
I think you can still have class AB1 at full power and
not have xover distortion if the grids can't draw any
current.

But, what is "full power"?
Clean with no flat tops at all?
5% distortion...15%... All that makes a difference in
this when refering to that notch you see.
But really, lets all take a little break on the "do
everything possible to get rid of the xover distortion"
hysteria.
Some of the best rippin' GUITAR amps I've heard
have plenty of Xover distortion at high output levels and
very ugly looking waveforms!! 
Maybe that's what you are alluding to also.

An interesting circuit would bias the amp at low levels in
normal class A and as the drive levels went up higher
and higher the bias voltage would change too.
Does it need too? Probably not.

Bruce

The increase in crossover distortion at large drive levels
is actually caused by the fact that the normally
high-impedance grid suddenly starts acting like a
forward-biased diode, and becomes a low-impedance
path to the cathode. 

Since most guitar amps use a relatively high impedance
AC-coupled drive circuit (from the typical phase
inverter circuits), the circuit acts like a standard diode
clamper circuit, clamping the peak of the grid signal to a
point slightly above the cathode voltage, while pushing
the rest of the waveform downward, or more negative.
This has the effect of pushing the amp further towards
class B operation, with the resultant increase in
crossover distortion. The phase inverters and driver
stages cannot source enough current to drive the output
stage into class AB2 because of the high impedance
and relatively long time constant AC coupling to the
grids of the output tubes. Therefore, you cannot get
away from the increase in crossover distortion as the
signal is pushed into clipping in a standard guitar amp
circuit.

The only way around this is to use a driver stage that
has a low output impedance and can source the large
grid currents that are drawn by the output tubes when
the grid goes positive, into class AB2 operation. A
cathode follower with the cathode resistor sourced
from a negative supply, DC-coupled to the output tube
grids (with the bias set by the DC voltage on the grid of
the cathode follower), is one way of providing this
non-bias-shifting drive. Another is to use a step-down
transformer to provide a low-impedance drive source
capable of providing the grid drive power necessary for
class AB2 operation.

This is one main reason why the "oscilloscope
crossover notch" bias method is such a bad idea. If you
crank up the amp until you hit clipping, you have
already clamped the signal peak and started pushing the
waveform downward. If you then adjust the bias until
the crossover distortion goes away, you will find it
comes right back if you increase the gain to clipping
again. If you then adjust the bias to get rid of that
crossover distortion, you will see it happen again at
another level. This makes it a very non-repeatable bias
method, not to mention the fact that you have no idea
what your idle or max signal plate dissipation is.

As for the "autobias" circuit, the important thing to note
is that the cathode waveform of a class AB output stage
is an asymmetrically-clipped, rectified sine wave. As
you increase the drive signal, the average DC value of
the cathode will change, resulting in a bias shift during
transient signals (run a PSpice simulation using a burst
waveform, or use a test signal generator capable of
providing a gated sine wave burst to see this effect.
Alternately, observing the steady-state DC voltage on
the cathode with no signal and then with full signal will
show an average DC level shift. Simply bypassing the
cathode with a large cap will not fix the problem
because of the change in DC level due to the different
average level of the rectified signal). Since your
autobias is servoing the drive current to inversely
compensate the cathode voltage change, you will get
bias shifts and transient distortion.

I have had success with an autobias circuit design of my
own that uses a tracking bipolar clipper circuit to
provide a cathode feedback signal that is independent
of the rectification effects and transient signals. This
control signal then adjusts a shunt regulating transistor
to control the negative grid bias voltage to keep the
cathode current constant, with a suitably long time
constant to the loop. This produces a biasing voltage
(and therefore current) that remains constant and
independent of the type of tube used, but doesn't create
bias shifts due to rectification effects at the cathode. 

Randall,

Thanks for the detailed reply.

From what I know and from your post, there are two
mechanisms for distortion in power-tubes:

1) The input signal tries to go larger than the bias
voltage. Example, an input signal trying to be 20Vpp
when your bias voltage on the output tubes is -9Vdc.
As you point out, the bias will start to push down-ward
(in this case, by about a volt), some top-edge clipping
of the input signal will occur, and the amp is effectively
colder biased. The input signal's Vp (trying to be
10Vdc in this instance) goes past the negative bias
level, so the more the output stage is put in class B
operations. Hence the cross-over distortion.
2) The B+ rail being small enough and the gain of the
power tubes being large enough such that input signal
causes the plate voltage hits the rail before the input
signal's Vp reaches Vbias. To raise the gain, more bias
current is needed - and a lower the B+ to avoid
overheating plate or screen. (Hey, this sounds alot like
class A biasing!)

So, I can think of three ways to address the cross-over
distortion. All involve limiting the input signal to the
output stage.

A) Design a fast-acting limiting circuit that keeps Vp
below Vbias (a zener diode or something?), or
B) Get the previous tube stage to clip before the Vp of
power tube input signal reaches Vbias, or
C) Reduce the input signal, increase the output tube
gain, and decrease the output tube B+ such that the
output tubes "rail" significantly before the input Vp
reaches Vbias. So in operation, cross-over distortion
really can't be reached.

Any of the above limits the power available from amp.
The first two because more headroom is still available
from the output tubes before the input starts to clip, and
last because the output rail has been lowered
considerably.

I missed one way of avoiding the cross-over distortion
(you wrote about this one): drive the power tube grids
with a low-impedance source, and let them go into
classAB2 operation. I've heard that this doesn't sound
that great, however. Have you been able to listen to
class AB2 for guitar operations, if so what do you think
of its sound?

For now, I think I'll stick with class A operation on the
output of the combo I'm building. But I guess I'm not
really choosing because it minimalizes the cross-over
distortion, more because I like it's overall sound. As
Bruce notes, some great sounding amps show
cross-over distortion when cranking ...

Hmmm. Maybe most of why I like class A is due to
that minimal cross-over distortion, though. I'm
scratching my head, but continuing to build anyway! I'll
only really know one thing when I'm done: all this
collection of stuff I threw into the amp came together as
"magic," or it didn't. Sigh.

Thanks much,

Carl

Q{I missed one way of avoiding the cross-over
distortion (you wrote about this one): drive the power
tube grids with a low-impedance source, and let them
go into classAB2 operation. I've heard that this doesn't
sound that great, however. Have you been able to listen
to class AB2 for guitar operations, if so what do you
think of its sound?}

Carl, let me tell you from first-hand experience... Class
AB2 sounds DAMNED fine in a guitar amp. 

I may even go so far as to say it sounds BETTER than
AB1. The reason I say this is that when you design a
driver for a class AB2 output stage, you're already
thinking of a low impedance circuit. A low impedance
driver has many benefits... 

One of which, of course, is that it may push the grids
into conduction without the penalty of grid blocking and
subsequent crossover distortion.

Another is that the low driving impedance serves to
provide excellent protection from thermal runaway, and
permits the use of tubes that would be too gassy for a
higher impedance grid circuit. The bias remains more
stable, and the tubes are happier.

Yet another is the fact that a lower impedance driver
circuit will improve HF response. The capacitance of
the grid circuit does not come into play nearly as much
as in the case of a high impedance driver. If you think
triode mode is too dark, try pushing it with a low Z
driver... You may be surprised.

Due to the design goal of pushing into grid current,
either one of two driver topologies may be implimented.
One is a direct coupled cathode follower, which is the
only method I have utilized up to this point. This avoids
the cost of a chunk of iron, which would be the other
method. With a DC'd CF, you eliminate the coupling
cap between the cathode of the driver and the grid of
the output tube. By doing so, recovery from overdriving
the grid circuit is (for all intents and purposes)
instantaneous. Grid blocking of all sort goes out the
window. This makes a damned big difference, let me
tell you. The amp sounds FAST.

The Ampeg SVT used a direct connected CF (a
12BH7) as a driver for the 6550's. We all know how
that amp sounds!

The other method, that of using iron, can be done in a
coupled of different ways. First is to use the obvious,
most easy method: an interstage transformer. Simply
create a small power amp as a driver stage, and take
the secondary of the IT and connect it to your grids.
Bias it up and you're done.

Slightly more involved would be to use a center-tapped
choke, which you can use in place of the grid leak (aka
grid bias) resistors. This provides a low DCR path for
the grid circuit. If you cap-couple the choke to the
plates of the driver stage, you should use a driver
tube(s) with a low plate impedance and a low plate
resistor. I wouldn't overlook the lower-power power
tubes, triode connected, like the 6BQ5 or the 6V6. The
lower your driver's plate circuit impedance, the better
the LF frequency response of the choke will be.

I'm in the process of talking to Mike LaFerve of
Magnequest to source some suitable push pull
interstage transformers for a class AB2 or class B
power amp. I can keep anyone informed of my findings
if they'd like.

So anyway, Carl, don't be afraid of grid current, in any
form. If you're prepared for it adequately, it will sound
every bit as good as (if not better than) a design
without.

~KG~

Quite a few high power guitar amps are designed to run
in class AB2. Most of the higher output fender amps
are. Even the Prosonic, has two of it's three power
modes specified as AB2. A low impedance driver and
a current-capable power supply are basically all that's
required.

Typical blackface fenders used 1meg grid leak resistors
and the 82k/100k plate resistors on the 12AT7 phase
inverter. Later silverface versions, where higher clean
power was the objective, set up the PI stage using
330k grid leaks and pair of 47k-1w plate load resistors
for better drive capability, at the expense of slightly
lower voltage gain of the BF circuit. Just "jumping in" on
a SF amp and "blackfacing" it, particularly the PI stage,
may not be the best thing to do, especially if clean
power and output tube reliability are issues

doc

Some of the best rippin' GUITAR amps I've heard
have plenty of Xover distortion at high outputlevels and
very ugly looking waveforms!! 

Thank you for your "experienced ears" observations,
Bruce! Cross-over distortion may not be the ugly thing
I've thought it to be. Maybe just too much of it, as in an
amp that's biased too cold. I'll have to do my own
listening test with various bias levels and cranking levels,
compare that with what the scope says, and see if I can
come to any conclusions, like, "Cross-over distortion
starts to get awful sounding when …"

- Carl