A neutralized preamplifier & driver

[LW1DSE]

I want to write about the design of my preamplifier still under development.

The trouble:

It is well known that triodes are made of small pieces of metal in proximity, and in this way, they have some capacitances
between them. I will no longer explain amplification theory, there are lots of text about this item. I will concentrate in
the effect of this capacitances, and amplification.

When a signal is applied to the grid of a triode, it is amplified, and appears at the plate as a current variation, and in
the load resistor it is converted into voltage variations again. But this is true that this voltage variations returns to the
grid via plate to grid capacitances (not only internal to the tube, but also in the socket and wiring circuit capacitances.)

As this coupling is capacitive, the phase of the signal returned to the grid is rotated 90 electrical degrees. This causes
negative feedback, and increases the grid´s conductance, and then high frequency lost of gain and phase rotation at the
plate´s output.

This feedback (Plate to grid) capacitance can be referenced to the grid to cathode circuit y means of the well known "Miller
Effect", and added to the grid-cathode capacitance.

The solutions:

Many solutions had been proposed:

1) The use of a screen grid, converting the triode in a tetrode, and next in pentode, but both of them becomes noisier
than triodes because of the Partition Noise of electrons in the grid-screen space.

2) To tune this capacitance with an inductance, a resource used in high frequency amplifiers, but as tuning is
essentially narrow band, it can’t be used in audio applications. Plus, that inductors needed becomes large and bulky, and
their high internal capacitances become worse the trouble.

3) To neutralize the amplifier. This consists of inject a same voltage amplitude at the grid of the circuit that is
available from the plate, but with the phase inverted.

The last of this has been extensively used in HF, VHF and UHF amplifiers, even when pentodes are used. This not only
stabilizes the amplifier, it also becomes less noisy and easy to adjust in wide bandwidth circuits as in tube TV tuners.
Usually, when a tuned circuit is included in the amplifier, it is easy to obtain an out of phase signal and then return it to
the grid circuit via a variable capacitor, then a bridge circuit is formed, cancelling the effect of the plate-grid
capacitance. This capacitor is made “trimmable” to get the optimal value.

My circuit of a exciter for PP EL34’s UL wired, makes use of this technique. Observe pictures of my circuits, in which can
be seen its complexity. It starts with a von Skoyok phase splitter(V1 : 12AU7, V2 : 12AX7), and then “cascoded” to V3
(12AX7), finally the output is via a White Cathode Follower (V4 and V5 12AU7’s) in which the signal is bootstrapped to the
plate circuit increasing output voltage and linearity, as load impedance becomes very high. (Image #1)

But as impedance becomes higher, stray capacitances (Including Miller ones) become very important, as a high frequency pole
appears at the point in which capacitive impedance (Frequency dependant) is made equal to resistance in the circuit (Above 6
megaohms in my calculus).

Observe Image #1 capacitors C9 and C10, they are the neutralizing capacitances. They can be adjusted to carry an equal amplitude signal from the opposite amplifier to act cancelling the node capacitances.

Image #2 illustrates what happens in my circuit when C9 and C10 are less than 9 pF (Picofarads).
Image #3 shows the same circuit with 10 pF. Note overshoot.

But observe when C9 and C10 are made 9.4 pF exactly(Pic. #4). The bandwidth is extended more than a decade by cancelling the parasitic capacitances, and the of Miller Effect.

Please note, that capacitances per se don't disappear by this way, only I can suppress their effect in the circuit

[LW1DSE]

As I haven't in my lab a sweep audio generator, I imagined that the improvement in bandwidth must be viewed placing a square wave at the input of the amplifier of such amplitude that not cause amplifier limiting, neither by grid current nor plate
saturation, then maintaining the amp in the linear region. The square wave has lots of high frequency components in their
rise and falling tracts (This is demonstrated by means of the Fourier’s Theorem). So, a lack of high frequency response can
be seen as a rounded step, and an excess as a peak in the rising and falling edges.

Also, I tried it in the real world, not only simulations, and results are surprising.

Only one channel viewed, 50VPP (limiting output voltage is 80V, so no grid current, no plate saturation, only amplification)
1 KHz (The test output from oscilo to calibrate probes.).

1) The amplifier undercompensated. Note the cursors approaching the slew time.

2) The amplifier just compensated.

3) A bit of overcompensation (Note peaking).

4) More overcompensation (Peaking and ringing).

5) Still more overcompensation (The amp breaks into oscillation).

(Once again: sorry because photos are out of focus, when I took them I didn't understand the photo machine)

[LW1DSE]

Here, the effect of the trimmer capacitor showing at the same time, both out of phase preamplifier's output.

Note the improvements in the rise and fall times, a clear symptom of high frequency better response.

[LW1DSE]

Here is the physical layout. Observe the 3-30pF mica compensators playing the role of neutralizing capacitors.

[LW1DSE]

Unfortunately, the page doesn't allow to upload higher resolution images for the simulation.
Please, tell me how to improve them.

[Shannon Parks]

Unfortunately, the page doesn't allow to upload higher resolution images for the simulation.
Please, tell me how to improve them.

If you click the hard to see magnifying glass icon in the upper left with the "+" once, and then once again, you'll get the original resolution. Looks good to me!

Shannon

[LW1DSE]

Thank you, boy‼