Un peu d'information technique sur l'ampli... C'est seulement en anglais...
All-discrete fully-complementary topology
Complete symmetry from input to output.
Cancels even order harmonic distortion (even though β22 has negligible distortion of any kind to begin with).
Symmetric start-up and shut-down behavior: When powered by the σ22 tracking power supply, no "thump" noise is heard when the power is turned on or off.
No integrated circuits used, allows for complete design control over all circuit parameters for optimum performance.
JFET + BJT + MOSFET
Three types of transistors used to best exploit their strengths.
JFET input cascoded complementary differential stage has high input impedance and low noise characteristic.
The input N-channel differential pair is connected directly to the P-channel differential pair via a bias-adjustment trimpot. The two pairs acts as current sources to each other, eliminating the need for additional constant current sources. This topology also has the benefit of increased slew rate.
The BJT voltage amplification stage (VAS) is also in a cascoded complementary differential configuration. Each side is driven by their counterpart in the input stage.
The output stage is a complementary push-pull follower design featuring high current MOSFETs. These are the same devices already well proven in the M³ amplifier, and are used here in cascoded form for even better performance.
The rugged and stable MOSFETs have no "secondary breakdown" region typical of BJTs, and have a negative temperature coefficient. This allows the amp to be designed without a safe operating area (SOA) V-I limiting circuit nor does it require any thermal compensation to control "thermal runaway" conditions. This also reduces complexity, eliminates any amplitude compression effects associated with such circuitry, and provides maximum sonic transparency.
MOSFETs are majority-carrier devices and are faster than BJTs of comparable power ratings, because they do not exhibit the BJT's "hole storage and recombination delay" characteristics.
MOSFETs are voltage-to-current amplification devices (contrast to the BJT, which are current-to-current). The gate impedance of a MOSFET is very high and thus simplifies the needed drive circuit. They do have a capacitive gate characteristic but cascoding reduces that considerably, and high quiescent current in the VAS stage provides an abundant charge for high speed drive of the MOSFETs.
Pure class A operation
Every stage operates in class A, even at extremely high output levels into low impedance headphone loads.
No crossover distortion because no transistor turns off throughout the voltage swing.
In class A, all transistors flow constant current regardless of output power. This reduces harmonic and intermodulation distortions.
Every stage is dynamically cascoded.
Instead of statically-biasing the cascode transistors with resistors (that are referenced to signal ground), β22 uses CRDs (current regulator diodes) as current sources to feed low-noise zeners diodes, which are then used as voltage references relative to the outputs of each amplification transistor. This biases the cascodes in a dynamic manner -- it tracks the signal swings to make the voltage across each of the amplification transistors constant.
Dynamic cascoding dramatically reduces the input capacitance of each amplification transistor.
Dynamic cascoding also dramatically improves the linearity of the amplification transistor due to the constant voltage. Nonlinear device transfer behavior due to voltage swing is eliminated.
Combining dynamic cascoding and class A operation makes the amplification transistors supremely linear even prior to negative feedback correction. This is because the amplication transistors "see" constant voltage and constant current even as the signal level is swinging. See an article by Nelson Pass which describes the cascode technique employed in the β22.
Cascoding also divides the voltage "seen" by each transistor, neatly reducing power disspation on each. In the VAS stage, this made possible the use of the high-performance BC550C/BC560C TO-92 transistors without requiring heatsinking. In the output stage, this makes thermal management simpler with the onboard heatsink option. Compare this to the common technique of paralleling multiple output transistors to divide power dissipation, which causes the total input capacitance to be the sum of the capacitance of every device.
Moderate global negative feedback
Since the amplifier circuit is highly linear to begin with, careful application of local feedback in the input and VAS stages sets the open loop gain of the amplifier at a low 56dB.
This allows only a moderate amount of global negative feedback to be applied for further reduction of distortion, lowering of output impedance and extending the bandwidth.
Global negative feedback is not used as a band-aid to bad circuit behavior.
High speed, wide bandwidth
Due to the use of cascoding in conjunction with wide bandwidth devices, the β22 has very low propagation delay, high speed and extended frequency response.
Judicious use of compensation capacitors in the VAS stage and in the feedback loop, and careful tuning of the MOSFET gate resistances, an optimal balance between wide bandwidth and phase margin is achieved for stability.
Rather than relying on the transistor's input capacitance (which is non-linear) to control bandwidth, cascoding drastically reduces the input capacitance. The compensation capacitors were then added to accomplish the task in a far more linear fashion.
The result is frequency response up to 2.5MHz, high slew rate approaching 200V/µS, and square wave response without overshoot or ringing.
There is no phase shift anywhere in the audio band. Even at 100KHz the phase shift remains negligible.
The high speed characteristics, along with a wide overload margin at the input stage, makes the β22 free of transient intermodulation distortion (TIM).
There are no signal-degrading coupling capacitors at the input, output or negative feedback loop.
The output DC offset is adjustable with a trimpot, which makes it easy to zero the DC offset during initial setup.
The use of well-matched JFETs at the input stage, along with their close proximity to each other, provides minimal DC drift over time. This eliminates the need for a DC servo mechanism to control the offset (which would otherwise add complexity and amplitude/phase response irregularities in the subsonic frequency region).
This is a true DC amplifier. Care must be taken to ensure that the input source does not have DC offset at its output.
High PSRR design
The power rails are connected directly to the output stage MOSFETs, but everything else are isolated via a capacitance multiplier on each rail.
The capacitance multipliers use the same high-gain BC550C/BC560C transistors as the rest of the amplifier. This provides a "virtual" capacitance in the order of 50000µF per rail, smoothing out supply rail noise and ripples.
The liberal use of constant current sources throughout the amplifier further removes the effect of supply rail fluctuations, providing very high PSRR (power supply rejection ratio).
This makes the amplifier virtually immune to power supply noise despite the moderate amount of negative feedback.
Combine the ultra-clean, tracking output of the σ22 PSU with the high PSRR design, the result is uncompromising performance.
High power output
β22's default power supply rail voltage has been raised to 2.5 times that of the M³ amplifier. In conjunction with beefed up heatsinking, the output power is increased by about 9 times.
The increased output voltage swing capability of over 40Vp-p comfortably drives the AKG K1000 headphones (which are notoriously inefficient) to very high levels,
The output power is 18Wrms into 8Ω, making the β22 an excellent high-quality small power amplifier for speakers. In a moderately-sized room and medium efficiency speakers, good listening levels could be achieved.
When operating in fully balanced configuration (BTL - bridge tied load), the β22's output power is increased to beyond 50Wrms into 8Ω.
See the Board & heatsinks section about heatsinking considerations and the Power supply section about power supply requirements for speaker driving duty.
Adjustable bias and output DC offset
Precise bias adjustment of the input complementary differential stage and the MOSFET output stage are done via separate multi-turn trimpots.
The complemetary VAS stage bias is scaled automatically according to the input stage bias.
Output DC offset is adjusted with a third multi-turn trimpot.
Passive ground, active ground or fully balanced configurations
The β22 amplifier can be built in one of the following configurations:
Conventional 2-channel "passive ground" amplifier (2 β22 boards required)
This is a minimal configuration for those who have limited chassis space and want to build a β22 at the lowest possible cost. As a headphone amplifier, two β22 boards with 1.5" onboard heatsinks will fit nicely in a Hammond 1455T220x enclosure, with the σ22 PSU and power transformer located in another Hammond 1455T220x.
2-channel passive ground is also the recommended configuration if the β22 is to be built as a dedicated speaker power amp. The heatsinking should be scaled up accordingly, and two σ22 PSUs should be used (one per β22 board).
3-channel "active ground" amplifier (3 β22 boards required)
This is the recommended configuration for standard 3-wire headphones, and offers improved performance by having an active ground channel amplifier for the headphone's shared "ground return" wire. The ground channel amplifier sources or sinks the return current from the transducers, which would otherwise have been dumped into signal ground or power supply ground. This shifts responsibility for the high current reactive load of the headphones from signal ground to the tightly regulated power supply rails, thus removing the primary source of signal ground contamination. The headphone transducer "sees" symmetrical output buffers with equal impedance and transfer characteristics on both sides, rather than an amplifier on one side and a capacitor bank of the power supply ground on the other.
The 3-channel configuration could also be used for driving speakers (the ground channel output should be connected to both the left and right speakers' negative terminals). The ground channel will bear the return current of both stereo channels and must therefore have larger heatsinking. A total of three σ22 PSUs is recommended for this configuration (one per β22 board).
4-channel differential "fully balanced" amplifier (4 β22 boards required)
This configuration is recommended if your signal source (such as a CD player, preamplifier, DAC, computer sound card, etc.) has XLR balanced outputs. If can be used to drive headphones or speakers, but the headphone plug and wiring modifications must be made to convert it into a 4-wire device (with separate and isolated negative wires for each channel).
In this configuration, each stereo channel has two amplifiers, a positive amplifier and a negative amplifier to handle the two differential signals. The balanced scheme is preserved from the source all the way to the headphone trasducers. Like the 3-channel system, this configuration does not contaminate the signal ground because both terminals of each headphone transducer is actively driven and the return current is not dumped into signal ground.
This configuratin also nearly doubles the output voltage swing as seen by the headphone or speaker load, resulting in great output power increase, and doubles the effective slew rate of the amplifier. With adequate heatsinking and power supply current, this configuration will deliver in excess of 50Wrms of output power into 8Ω.
Headphones will need to be re-terminated (usually to dual 3-pin male XLR connectors) and may also need to be re-cabled (to separate the left and right ground returns) in order to be used with a fully-balanced amplifier.
As a speaker power amp, this configuration should employ four σ22 PSU boards (one for each β22 board).
See the Wiring & ground section for more details.