2 x 50 watts Tube Power Amplifier
Important Note: The ACCP 2 output transformers used in this amplifier
are no longer available for individual DIY projects.
(Info from the Welter Company 12/2001).
However, concept and schematic may still be of theoretical interest
The amplifier comprises four stages, input stage, phase splitter, driver and output stage (see schematic). Input tube is a UHF twin triode ECC 81 (12AT7), operating in single-ended differential mode. System I is a cathode follower that feeds the signal into the cathode of system II. Hence the input has very high impedance, determined only by the volume control (or grid resistor). Input capacity also is very low, for there is only static capacitance but no dynamic capacitance (Miller capacitance) as usual in common cathode triode circuitry. Besides cable capacitance, the preamplifier output only “sees” the static capacitance of grid/plate CGA and grid/cathode CGC respectively. Using triodes in conventional common cathode circuitry you may have an input capacitance of over 100pF according to equation 1:
Cdyn = CGA x (1+gain) + CGC (1) (EQ 1)
System II is a differential amplifier working with the input signal from the cathode follower at its cathode and a local negative feed back (NFB) at its grid which is defined by two resistors of 270k and 68k. By shuffling these resistor values the local NFB and thus the input gain can be adjusted. Main purposes of the local NFB are enhancing bandwidth (250kHz at -3db point in my application) and reducing adverse effects of aging or tolerances in tube data on tonality and stereo image.
The cathode idle currents of the ECC 81 automatically set to a ratio of 4mA to 1mA due to plate of system I being directly connected to the +rail ("through grip" effect). This is sonically advantageous. A cathode follower's distortion is directly related to it's current swing because internal plate resistance and transconductance of the tube change with current. If the current of system I is substantially higher compared to system II, current changes with signal and hence changes in internal plate resistance and transconductance are low. For similar reasons the cathodes are coupled by a current source instead of a conventional resistor.
Current sources look like resistors of almost infinite resistance and still allow to keep currents high. The cascode of the BF245 and BC548 build a " rock solid" current source with an internal resistance of several GigaOhms. The BC548 can be considered a current source for the BF245, the emitter resistor allows current adjustment according to the formula:
ICS = (VD - 0.6V) / RE (EQ 2)
ICS = Current defined by the current source
VD = voltage across the four silicon diodes
RE = resistor value at the emitter.
Another advantage of the current source is an absolutely constant load at the +rail. The total current (IAI + IAII) across the two systems stays absolutely the same, no matter whether heavy bass or diminutive transients are run through the input tube. Replacing the ordinary tail resistor by a current source gave a remarkable improve in stability of tonality, stereo image and DDR (Downward Dynamic Resolution, as Allen Wright calls it in his very remarkable tube preamp cook book). The serial regulator HIP 6300 reduces hum to almost zero. Serial regulators normally should be banned from audio circuitry, but in this case in worked fine.
Fig 1: Front view of the power amplifier.
Please note: This picture shows an older version of the amp than the schematic. This version used EF 86 in the phase splitter stages similar to the QUAD II design and 6550A output tubes.
Phase splitter stage
The ECC83 (12AX7) converts the unbalanced signal into balanced signals. It works as a long tail phase splitter and has the advantage of high gain, good frequency response and large voltage swings. Negatives are a poor balance and the necessity to tweak the plate resistors for balanced voltage, if you do not use a current source!
For optimal symmetry the cathodes must be coupled by a stable current source instead of a tail resistor. Then the current of the first system can only rise to the extent the current of the second system is lowered and vice versa. If the current in system I increases by 0.01µA, the current in system II must decrease by 0.01µA. So the phase split signals are forced to be perfect mirror images, depending only on optimum similarity of the anode resistors. And again, you have the advantage of an absolutely constant load at the +rail. Differential amplifiers were always considered less precise than single ended amps. From my experience differential amps with conventional tail resistors are loose and they indeed loose signal detail because the circuit parameters can shift with voltage swing. A good current source locks the total current down and differential amps become as precise as single ended amps.
The anode resistors of the 12AT7 are relatively high in value for good linearity at large signals. This is because the KT88 beam power tetrodes are run in an anode/cathode coupled circuit similar to the QUAD II amps by Peter Walker (the McIntosh circuits also look similar in the first, but they work differently). The voltage swings of the phase splitter must cover the grid voltages necessary to drive KT88 plus the voltages that are fed back into the KT88 by the cathode coils of the output transformers.
270 kOhm at the anodes of the phase splitter and low idle currents of 1.1 mA give a relatively high output impedance of the 12AT7. Thus with respect to the low grid resistors of 68 kOhms at the KT88, cathode followers were added in galvanic coupling for low impedance. The ECC82 (12AU7) was chosen, because it has acceptable transconductance for low impedance and allows 180 volts between cathodes and heater. Again current sources were chosen as load resistors at the cathodes for the reasons discussed previously. They allow the ECC82 to run at idle high currents, but current swing is only depending on the tetrodes grid resistors of 68kOhms. With ordinary tail resistors of 30kOhms and idle currents of 5mA the current swing would be between 3mA and 11mA at signals of 80 volts peak to peak. With a current source the current swing is limited between 6mA and 8mA.
The MOSFETs IRF 830 can handle up to 500 volts and work very reliable. To dissipate their wattage, they are mounted on small heat sinks. The gate resistors at the MOSFETs are essential to avoid RF oscillation. For the same reason the ECC82 and KT88 have grid stopper resistors. The currents of the cathode followers are adusted to 5mA by the source resistor of 2.4kOhm.
As already mentioned the KT88 tetrodes are run in an anode/cathode coupled circuitry (ACPP in the schematic stands for anode-cathode-push-pull). This circuitry goes back to Peter Walker and is also used in the bigger Jadis tube amps as far as I know. The trick is a local feed back from the plates into the cathode/grid loop. The voltage controlled feedback grant the tetrodes characteristics similar to triodes with low impedance and low distortion while maintaining their ability to high output power. With 430 volts at the plates the output power is 50 watts. Contrary to ultra-linear stages the anode/cathode coupling allows free choice of optimum voltages at the screen grids. Stabilizing the screen grid voltage at 370 volts gave further improvements in sonic performance in my application. I tried the IRF 830 as regulator but I could not keep it from oscillating. So I used a BUT13 darlington transistor.
The plates of the KT88 dissipate around 30 watts. According to the data sheets the tetrodes should have a minimum distance of 10 cm (4 inches) from each other to avoid picking up radiated heat from the neighbor tube (a frequently ignored recommondation !).
The amplifier has a negative feed back (NFB) loop from the secondary coil of the output transformer into the phase split stage. In my experience NFB can be somehow compared with pharmaceutics. They have effects and they have adverse effects. To achieve best benefit, you carefully have to adjust the dose and in some cases the best dose may be zero.
However, the stability of the stereo image is very much dependent on an equal gain in both channels. Even the smallest gain fluctuation blurs the image especially when it comes to the reproduction of depth. That NFB is the best tool for stabilizing is often overlooked, despite the fact that everyone agrees on the importance of gain stability. A tube amplifier is more prone to gain fluctuations than a solid state amplifier. Small movements of the electrodes due to heating change the properties of a tube at least to some degree.
In my application a good compromise between woofer control, stability of the stereo image and vitality was achieved with NFB between 10db and 12db. The capacitor was tweaked to 240pF by adjusting to best square wave response at the output.
Schematic of the power source unit.