All about Tubes, Tube Circuits, Tube Gear

Thursday, June 14, 2012

The Aikido Comedy (pt 6): Thinking/Testing Failures

Here's my suggestion if you want to test my point:

First leave the circuit as is, but:

(1) lower the B+ for safety reasons, then

(2) set the 1st tube idling with plate at half the B+ (balancing the load resistance)

(3) inject a 120 Hz exaggerated PS hum signal, say 10 volts,

(4) measure the voltage of this PS hum signal at the plate
(this will give an idea of the AC impedance of the tube as viewed by the PS,
if it differs from that of a simple resistor (it won't significantly differ, but go ahead).

You should easily be able to pull this up on your scope.

(5) Now add a 6-10 Hz input signal on the grid.

My prediction is that you will see the PS hum signal being AM modulated by the input signal:

Click the image to open in full size.

This is enough to test the first assertion,
namely that PS noise is modulated by the input signal.

What happens when you raise the frequency of the input signal above the frequency of the 'carrier' (PS 120 Hz)?
Essentially the "Ring Modulator" sound effect.
Its non-musical and also an unwanted intrusion into the input signal (the music).

Although the 'real life' situation involves much smaller voltages (signal and noise), the type of artifacts produced are the same.

If you have the equipment, display an FFT in the frequency domain of this output.

Of course one is free to calculate what you think is PSRR any way you like.

That is precisely how absurd claims about impossibly naive techniques are perpetuated. 

Recall again Broskie's own explanation:

How it works

This circuit eliminates power-supply noise from the output,
by injecting the same amount of PS noise at the top and bottom of the two-tube cathode follower circuit.
One source of PS noise is coming from the 1st stage,
the other from his added resistor-divider.

The top tube in the 2nd stage is fed the PS noise from the 1st stage (along with the signal).

The bottom tube in the 2nd stage is fed the independent copy.

The bottom tube however is not a cathode-follower:
Its signal is inverted.
The two triodes in the 2nd stage are equally matched,
(same cathode resistors, same DC current, same gain),
so the two noise signals are supposed to cancel.

The way it works is that the input stage (the first two triodes)
defines a voltage divider of 50%, so that 50% of the PS noise
is presented to the CF's grid; at the same time the 100k resistors
also define a voltage divider of 50%, so the bottom triode's grid
also sees 50% of the PS noise.

Since both of these signals are equal in amplitude

and phase, they cancel each other out, as each triode
sees an identical increase in plate current
(imagine two equally strong men in a tug of war contest).
Broskie could have explained the opposing tube configurations better (see my note above),
but the point ends the same: opposite signals (phases) cancel.

Note his perception of the operation of the 1st stage however.
Clearly Broskie himself perceives the SRPP as a simple voltage divider in regard to the AC noise.
He is concerned to avoid bypass caps, for just this reason,
because they would imbalance the AC resistance,
"add distortion", and complicate the noise-biasing.

CONTRA DF96's assertion that
"Nazaroo doesn't know the difference between static and dynamic resistance",
we find that it is Broskie who treats the SRPP as a simple DC circuit,
even eliminating capacitors to "make it so".

------- SIDEBAR ---- Flashback to 2002 ---

The above is actually an SRPP stage, one Broskie(?) has analyzed before,
in Tube Cad Journal (2002).

There he describes it as "controversial", because there were
several varied analyses of this circuit available.
Part of the reason that controversy brewed,
was because Broskie himself was so unfamiliar with Mu-Stages,
that he confused the two, and blurred their distinction.
In a very convoluted series of manipulations,
he ends up with
Alan Kimmel's Mu Stage from 1993,
but he deliberately fails to acknowledge this,
and to cover his tracks,
he misnames the circuit an "SRPP Amplifier with an Internal Coupling Cap"
and he tries to re-name the circuit an "SRSE",
and even tries to describe it as a kind of Loftin-White circuit.

After stirring up a lot of fuss and muddying the waters,
in regard to the operation of both the SRPP and the MU-stage,
leaving the reader with the impression that Mu-Followers
have poor CSS characteristics (another gross error by Broskie),
we have come full circle.

Now Broskie makes some more modifcations,
resulting in a "Mu-follower + Cathode-follower".
Finally, he finishes by eliminating the Mu-follower,
and restoring the SRPP as a 1st stage + Cathode-follower,
complete with a capacitor 'voltage-divider' to inject PS noise,
back into the 2nd stage, just as in the later "Aikido" design.
Presumably he discovered that two caps don't make a reliable voltage divider,
especially if they are frequency-sensitive,
alter the phase of the noise-signal, and short it to ground
instead of presenting it to the 2nd stage lower input!

All of this was just a long-way-round way of
crapping all over the Mu-follower/Mu-Stage,
(actually a far superior circuit to his own 'brilliant' concoction),
to setup the selling of the soon-to-be "Aikido" preamp.


Clearly Broskie believes that the PS noise coming from stage 1
is constant in amplitude, to match the signal he taps
from the B+ through his resistor-divider.

Further, he perceives the the PS noise in the 1st stage
as coming from the B+ through the 1st stage as a voltage divider.


If the output connection is taken from the
cathode follower's cathode, then the balance will be broken.
The same holds true if the cathode follower's cathode resistor
is removed. (Besides, this resistor actually makes for a better
sounding cathode follower, as it linearizes the cathode follower
at the expense of a higher output impedance.)

Here again Broskie is concerned with matching the gain
of the two triodes in the 2nd stage.
Why? because only then will the two signals match in amplitude and cancel at the output.
This again presumes that PS hum from stage 1 is a constant amplitude and shape.

But is any of Broskie's analysis valid?

(1) First in regard to the source of noise in the 1st stage,
its amplitude will be affected by several factors, including
the real sources from which it comes.

And we dare suggest more hum is coming from the grid inputs
and possibly the heaters (depending on the tube) than from the B+
directly across the tubes acting as a voltage divider.

This part of his "theory" is already very suspicious,
and is disproved by experience with hum problems
re: input and heaters.

But so what?
Even if Broskie is wrong about the source of PS hum in stage 1,
surely his method will cancel it out regardless.
To an extent this must be true,
provided it remains constant in amplitude,
and he has adjusted his hum-injection right.
This is where again however, Broskie's methodology fails:

(2) Why should the PS hum magically be exactly 50%
of total PS hum found in the common supply?

There is no logical reason for this, especially when we can
pinpoint several different sources combining to create it.

This is nonsense, and we'd be far better off with a simple
adjustment resistor in his voltage-divider,
to nail the right amount of hum in any version actually built,
by LISTENING to it.

(3) Provided the actual wave-shape of the hum coming from stage 1
is the same as the hum directly tapped,
We should be able to eliminate 90% of it, or at least minimize it.

The reason for this is not in Broskie's analysis,
but in the simple fact that:
IF the PS hum is coming from tube heaters or grid-pickup,
it is likely to be constant, whatever value it is.
(of course this doesn't preclude improving the stage 1 hum,
by for instance moving heater-lines or adding shielding,
in which case we would also have to ADJUST AGAIN the
injected PS noise amplitude.

 here's the simplest experiment possible.

Replace Broskie's voltage-divider with an adjustable one that has a wide swing (like 30% to 70%).
i.e., keep the 100k grid-resistor, and add a 100k linear pot bypassed by another 100k resistor. so you can swing the injection-signal amplitude significantly.

Now listen, and set to minimum hum if possible.
Disconnect and take a reading from the pot, and see if 50% of
the B+ hum was the minimum balance-point.

This isn't rocket-science.

Next try something even more daring:
use a strong sine-wave input signal on stage 1.
Take a copy of this, and put it too on the grid of the lower tube of stage 2.
Adjust that to completely if possible cancel out the sine-signal.

NOW measure the A.C. hum/noise and see if it has increased.

 Lets talk about why the Broskie noise-injection would work, and when it won't.

List of things necessary for it to work as described:

(1) PS hum/noise coming out of stage 1 must be 50% of raw PS hum/noise.

(2) gain of both triodes in stage 2 must be closely matched.

(3) voltage-divider for injection must be accurately 50%/50%, i.e., matched 1% resistors.

(4) Both stage 1 and voltage-divider must be feeding off same PS / point.

(5) PS hum/noise at either sample point must not be influenced by input signal in first stage.

(6) PS hum/noise at either sample point must not be influenced by input signal in second stage top triode.

(7) Phase of complex PS hum/noise signal must not shift in any significant frequency band.

List of some things that could go wrong:

(1) PS hum/noise varies in output of stage 1.

(2) PS hum/noise amplitude does not match that of voltage divider.

(3) PS hum/noise varies in power-supply itself.

(4) Gain varies between tube-triodes in either stage, either out of box or over time.

(5) Stage 1 generates no significant PS hum/noise, in which case Stage 2 simply injects noise!

(6) quiescent operating points of tubes involved shifts based on signal amplitude (non-linear Rp etc.).

(7) Hum/noise injected into circuit from other causes shifts phase of either PS hum/noise signal.

Things going wrong caused by DIY circuit-builder:

(1) Failing to set tube-bias correctly,

(2) Failing to match tube-gains.

(3) Running each stage off different PS.

(4) Running voltage-divider off different PS or section.

(5) Eliminating noise from stage 1 by rectifying heaters or shielding grids.

(6) Eliminating PS hum/noise in one part of PS but not others.

(7) Failing to adjust the circuit for minimum hum/noise.


"Why you shouldn't just test the output?"  

Its obvious.
The whole point as stated right from the start,
was that the circuit does not operate as claimed.

It has been conceded repeatedly that the circuit is indeed low noise,
when powered by a well-designed power supply and good tubes.

That the circuit can cancel some hum is acknowledged.
Testing the output however tells us nothing about what is going on
inside the stages, which is what the thread is about.

Its like this:

(1) Stage 1 adds some hum/noise from various sources in a real-world build.

(2) Stage 2 also adds some hum/noise from various sources, including amplifying hum/noise from previous stages.

(3) Broskie's circuit indeed eliminates most of the small amount of hum/noise present, if adjusted properly.

(4) This leaves sweet F.A. at the output to measure and analyze,
and tells us nothing about how the circuit(s) really operate,
where the noise is coming from, or how much noise was generated and cancelled.

(5) Its the worst method for trying to determine what is happening in the circuits.

Here's an analogy.

(1) I take a can of white paint.

(2) Jerry Lewis adds an unknown color to it.

(3) Broskie shuts the lights off.

(4)  try to figure out the color that was added.

(5) I analyze the paint-shelf instead and notice the green can has been used.


This is an Information Theory type problem.

When you shuffle a deck of cards properly,
there is no way to establish exactly how the cards were shuffled,
by examining the deck after shuffling.
There are a near-infinite number of ways it could have happened.

Put even straighter,
There is no real way to tell what you wrote in the sand,
after you erase it.

Certain processes are 'lossy';
they lose information permanently.
For instance, once a picture is reduced or compressed using a 'lossy' process,
there is no way to recover the original exactly.

Trying to guess what the distortion process was after the circuit has removed 90% of it,
is like trying to guess what the dog ate after he poops it out.
You are far better just watching what he ate.
The output is all but useless for detailed analysis
of a complex but lossy process with internal interactions,
like a multi-stage amplifier system with various kinds of
noise sources and cancelling feedback loops.

No comments:

Post a Comment