[ NON-NFB Amplifier ]
In an ordinary negative feedback amplifier, when there is a reaction from the load, it is fed back to the initial stage of the amp via a negative feedback circuit. Therefore, inverse reaction develops inside the amplifier to cancel out the reaction. It is difficult for this inverse reaction to emerge as a measurement value, but actually it causes inter-modulation distortion, and this a major factor contributing to sound quality degradation. Therefore, ONKYO employs a non-NFB amp to completely eliminate this factor.

Simulation of sound quality degradation due to inverse reaction

The following experiment was conducted to investigate the effects of reaction from the speaker. As indicated in the diagram below, the method involved comparison of power amp internal response when a disturbance signal is input from the output terminal side. In an NFB amp, a large internal response waveform is measured, and furthermore there is a cancellation effect due to feedback, and phase inverts. In the experiment, a simple square waveform is used, but in an actual complex audio signal, inter-modulation distortion is generated.
Even in a negative feedback amplifier, this effect is apparent to some degree (1 over the current amplification rate). However, this is at a level which can be ignored, compared with a negative feedback amp. Furthermore, the phase becomes the same, so there is almost no generation of inter-modulation distortion.

Circuit diagram and measurement data for inverse-drive test

Problem points with Non-NFB amplifiers

With an ordinary NFB amp, distortion and output impedance are reduced by controlling the amount of feedback. Naturally, in a non-NFB amp this cannot be done, so it is crucial to lower the distortion rate and output impedance of the power amp section itself. At ONKYO, instead of using a conventional power amp section with multi-level emitter/follower connection, we use an inverted Darlington circuit with multi-level connection to an inversion amp with emitter ground. The basic advantages of an inverted Darlington circuit, over an ordinary Darlington circuit, are as follows:

Good non-linear distortion between the base voltage and collector current.
Little distortion due to local feedback, and low output impedance.

The following explains these advantages.

One major factor contributing to transistor amp distortion is non-linear distortion of the base voltage - collector current (Vbe-Ic) characteristic -- a basic characteristic of the transistor itself. As indicated in the diagram, with an ordinary 3-level Darlington circuit, the distortion increases because each of the 3-level transistors contributes its part to non-linear distortion.
In contrast, with an inverted Darlington system, only the initial level Vbe emerges in output, so non-linear distortion is improved on all levels. Furthermore, the initial level has A-grade operation (with extremely little distortion), so the system has even less distortion than a conventional Darlington circuit (combined with the above-mentioned non-linear distortion).
Diagram comparing output level circuit systems

Three-Level Darlington Connection

Double-Inverted Darlington Connection

The mainstream for today's transistor amps is a Darlington circuit, with multi-level connection of emitter followers, to each of which 100% local feedback is applied. In contrast, with an integrated Darlington circuit, 100% local NFB is applied from the output level to the previous level via a 2-level connection of emitter-ground inversion amps. Each level has its own gain, and 100% of the earned gain is fed back to the previous level. Therefore the system features low distortion. At ONKYO, our idea is not regard this as an inversion amp, and to add a current booster to the emitter follower. Thus we add two levels of current boosters, thereby using a 3-level circuit with higher performance to reduce the distortion rate of the power amp section itself.
With an ordinary Darlington circuit, the output impedance will not drop lower than the resistance applied to the emitter. However, with an inverted Darlington, this can be lowered further by applying local feedback. Furthermore, with an emitter follower, the output impedance varies due to the collector current, and since there is non-linearity, this becomes a problem with a non-NFB amp. However, with an inverted Darlington system, this type of non-linearity is not present.

Diagram comparing output circuits

Problems with inverted Darlington circuits

The following are the reasons why these outstanding circuits are not used more commonly:

It is extremely difficult to maintain thermal stability for bias current.
Each stage has its own gain, so it is easy for a slight oscillation to arise in the phase margin, and thus sophisticated mounting technology is required.
With a normal Darlington system, the variation direction of the base voltage (Vbe) due to temperature is always the same, so total temperature compensation can be achieved at once by attaching a temperature compensation transistor to a heat sink equipped with an output level transistor. However, with an inverted Darlington, there is inverted amplification, so the direction of variation is reversed between the previous and later levels, and compensation cannot be achieved with the same method. For this reason, temperature compensation is conducted on two levels (as shown in the diagram) with special temperature compensation for the first driver, and separate temperature compensation for subsequent levels. With an inverted Darlington in particular, the most critical point is stabilization of the first level, and this almost completely determines the bias current on the output level. Therefore the bond for first level temperature compensation transistor was strengthened by mounting, together with the 1st driver, to an aluminum heat radiator which has a small thermal time constant.
Diagram of temperature compensation for power amp section
The diagram shows analysis, with a simulator, of a 2-level Darlington and 2-level inverted Darlington circuit. A particularly striking difference can be seen in the output impedance and its phase characteristic. In an ordinary Darlington, the phase (in the high range in particular) fluctuates in both the + and - direction due to bypass current, so there is only a small amount of total phase shift near the bypass current which is actually used. With an inverted Darlington, however, there is only a shift in the + direction, so the phase margin is small and there is greater tendency to oscillate. The diagram shows a simulation of the 3-level inverted system used in this case, but the amount of phase shift increases, and furthermore there is a greater tendency to oscillate. As can be seen from the diagram, there is an output impedance peak at 20-30MHz, and it is crucial to control this phase shift skillfully.
At ONKYO, we use an impedance for phase correction at the base of the output level transistor. However, with an ordinary air-core coil, Q is high, and there are problems with diving and high-frequency fusion, so we use ferrite beads in the jumper wire. These ferrite beads have extremely high magnetic permeability at low frequencies, high loss in the 20-30MHz band whose phase is to be shifted, and low Q. The jumper wire is simply threaded with ferrite beads, so there is total shorting in the audible range, and there is no degradation of sound quality.