The following list covers the TEM 3200's key points and explanations by its designer:

1./ 100% symmetrical amplifier and 100% freedom of using any active part in the circuit
Based on my patented topology, I can build an all-transistor amp (all BJTs or all MOSFETs or a mix thereof) as well as an all-tube amp. I prefer to use active devices in places where they operate the most linear. For voltage amplification, I rate tubes highest because they have a lot of headroom (if the operating points and bias are correct) and a long straight usable working range. For power and/or current amplification, I rate discrete power transistors best. With them there is no need for an output transformer and it's easier to achieve very low output impedance to match complex speaker loads. For a hybrid amplifier, MOSFETs combine well with tubes because they need no base current.
2./ Use of only one extremely powerful MOSFET of the most modern type for each output stage
The output device is the major source of distortion but there's a simple trick - use a really beefy device and it will behave like a low-level signal device used at a mere fraction of its potential for higher linearity and highly increased reliability.
The use of the SOT-227 package (connection of source, drain and gate done by M4 screws instead of welding!) is very expensive but has the major advantage of an isolated large heatsink combined with very low thermal resistance. Thermal isolation layers necessary for ordinary TO-220 or TO-247 packages add a large thermal resistance to the system so that the theoretical consumption of the transistor is pure theory. Therefore in common amplifiers, many paralleled transistors are necessary to spread the heat and
lower the total thermal resistance. Not so in the TEM 3200. The big and heavy MOSFET is widely overspec'd. Although the heatsink on the back of the amp is generously dimensioned, the overall size of the amp is pretty compact given its output power. An additional advantage is that no matching is required and the performance of the amp is very predictable even under serial production.


3./ Use of high transconductance tubes for the voltage amplifier and driver stage
For the input stage, one of the lowest-noise tubes is used, the EC86. All tubes in the TEM 3200 have a frame grid and hence a very high transconductance of >15mA/V. Using these types of tubes is tricky because the linear operating range is critical and they oscillate very easily. The were originally designed for very high frequencies up to 800MHz. But properly biased and used, they excel because of high gain combined with low-resistance circuitry which ensures very wide bandwidth. The latter is ultra important for dealing with the extremely large parasitic capacitances of the power MOSFET.


4./ Local feedback loop in the output stage
For further linearization of the output stage, a local feedback loop including the power MOSFET and 2nd and 3rd tube stage lowers the non-linearity of the power stage to very low levels even without a global feedback loop.


5./ Never switched output devices
The output MOSFETS never switch off. There always flows a well-defined drain-source current. As a consequence, there are no crossover or switching distortions and the amplifier behaves sonically like a class A amplifier - but with far reduced power consumption.


6./ DC coupling throughout
The beefy power MOSFET has an incredible transconductance of about 40A/V. That means that a one-volt change of gate-source voltage causes a change of 40 Amps in the output circuit. Put differently, just 1 Millivolt (!) causes a very audible change of 40mA in the load (speaker). Music of course isn't just sinusoidal oscillations. Music consists of a string of pulses. Only the sum of all pulses is DC-free but large bass-induced pulses are not. That means if such a heavy-duty device like the TEM 3200 MOSFET is AC-coupled with a capacitor and a necessary high-pass resistor, the time constant of this RC network must be designed to match the lower corner frequency. But - this time defines the time length wherein the RC high-pass acts as an analog memory for every DC fraction of the signal. So it is very likely that for a large bass pulse of say 10V, the DC fraction is at least 100mV. This 100mV will cause a dramatic shift of the working range of the MOSFET - 4 Amps more quiescent current!!! This shift will last as long as the lower corner frequency defines it. For 20Hz, this means 50ms. That's simply not tolerable so DC coupling without any time constants is a must.


However, DC coupling from input to output including 6 thermally drifting tubes and 2 MOSFETs with 40A/V transconductance is extremely difficult. Bias precision in the microvolt range becomes hyper critical. My topology insures near perfection. After warming up, the remaining output offset voltage is lower than 1mV - and there is no low-resistance output divider! All operating parameters are automatically regulated. The quiescent currents in the two floating output circles are set with only one regulating device. The two currents are always set to one and the same current, with offsets in the microAmp range. That is the very heart of my patented system and very stable and reliable. All TEM 3200s in the market have been working with zero failure - and some have been working for a full 2 years. One Swiss customer believes in never turning off his amps. I don't believe or encourage that (tube aging) but his amps have been constantly powered up since January 2007. That means more than 9000 working hours without any signs of wear or degeneration. I like this customer because he conducts a great stress test experiment for me...


7./ Foolproof protection against overload and thermal runaway
The amp has very fast yet intelligent electronic fusing against shorted output leads. That ensures very high pulse currents even at 2-ohm speaker loads but limits the current flow in the power MOSFETs in case of a short to harmless figures. Keeping in mind the 132000uF storage capacitance of the power supply, this is very important. Even though the amp contains a lot of protection circuits, I never recommend to experiment in that way (shorting the outputs) because I designed the amp for music listening, not Tesla spark experiments... The thermal condition of the amp is constantly monitored for perfect stability under all conditions. For the current monitoring only, I use two very costly induction-free 100-watt Caddock power resistors per half wave.


8./ Ultra wide bandwidth and very low open-loop distortion
Every stage of the amp is designed for best possible linearity without any feedback. Naturally, there are no switching diodes or transistors. The bandwidht is very high and the amp has a very good phase reserve of nearly 90 degrees (closed-loop). Due to >100MHz open loop bandwidth, the global feedback loop works nearly perfectly.


9./ Self-cancelling magnetic fields
Inside the amp is a real-world 4-wire connection between the power supply and power stage (2 wires per half). The heavy currents which might flow to the speaker and back will not cause magnetic field disturbances because the vector sign causes self cancellation of the magnetic induction. This is impossible with common complementary circuits and +/- and ground power supplies (in fact a 3-wire connection). And an additional surprise is this: The connection between the voltage amp and driver stage is the most simple - one piece of copper wire. But along this wire does not travel any current from the power stage to the voltage amplification stage. The reason is pretty simple: The electrons coming from the two floating power supplies "go back home" and not to the ground-referenced power supply of the voltage and driver stages.


10./ Symmetrical slew rate
Keeping in mind again that music consists of pulses, it is very important that the leading edge of these pulses be amplified with the same speed as the trailing edge. The TEM 3200 has the same slew rates for both rising and falling slopes due to perfect symmetry. Almost all common complementary amps suffer a slower falling edge because the pnp or p-channel devices are slower than their npn or n-channel counterparts in the other half of the circuit.