The Lorelei, Beautiful Sounding EL95 / 6021 PSE Amplifier
Small sized, small budged, EL95 based PSE amplifier with 6021 sub- miniature
tubes in the driver stage.
The EL95 is a miniature 7 pin tube designed for use in car radio sets. It has a higher
distortion (dtot: 12%, penthode class A) than the EL84 and a lower power output, but holy
distortion (!!!) it has a truly beatiful sound.
The EL95 has a charming middle and high frequency tone, but I found a single EL95 in triode
mode has a poor bass when no feedback is used and too little power.
In the PSE configuration the whole frequency range sounds beautifully clean, detailed with a tight and
powerfull bass. Every tone nuance, every color hue is there at the right place on the stage and there is depth!
When I had the PSE prototype ready, connected to the backloaded horns and with Lester Bowie's
Brass Fantasy – Serious Fun turning on the CD tray... Oh my! The brass sounded so vividly
three-dimensional that I was overjoyed.
This EL95 PSE is doubtlessly on a par with my bigger amplifiers.
In the output stage two EL95 connected in triode mode are working in parallel.
Each EL95 sinks about 21 mA addind up to 42 mA per channel. A suitable OPT can be that of an EL84 SE.
The ones used here have taps at 4K, 5K2 and 7K, I picked the 4K tap by rule of thumb.
The core should be designed for 45 to 50 mA, not bigger.
A medium mu 6021 sub-miniature double triode is used in the driver stage.
Sub-miniature tubes are part of the last generation in tube development. They are rugged and often
designed to MIL specs. to provide dependable operation under severe operating conditions.
Note that the two halves of the 6021 are not connected as an SRPP circuit, as it may appear at first glance;
The lower triode is in common cathode mode working against the upper half acting as active load and wired
the same way. SRPP is a push-pull circuit while this one works in single ended manner, just like the
common cathode driver working against a passive load. The wiring difference lies in where the coupling
capacitor is connected to, either at the plate of the lower triode (SE) or at the cathode of the upper
triode (SRPP). The way the coupling capacitor and the power stage input's grid resistor interact with the
driver is different in each case. Both SRPP and this have excellent linearity.
Check out SRPP Decoded
by John Broskie for further study.
The high tension Maida regulator supplies the driver stages. With 208V the drivers together sink about 4 mA,
2 mA each channel. Each triode in the 6021 'sees' half the supply voltage, around 104V.
Because the 6021 allows 200V heater-cathode voltage difference there is no need to elevate the
heaters, what simplifies things a little.
The output voltage can be fine adjusted varying the 1K8 adjust resistor.
A lower resistance lowers the output voltage.
Each output stage has its own LC filter. The chokes are Hammond 154M with 2H @ 100mA.
Since each channel sinks only about 42mA the effective inductance may be somewhere between 3H and 4H
improving ripple rejection a bit further. The amplifier is very very quiet, no hiss on the tweeters
nor midrange and a very faint and distant hum on the bass to be heard only with the ear next to the
A couple of resistors on the 6,3V winding were required to set the heater suppy to 6,3V.
All heaters are connected together in parallel and sink around 1,4A.
6.3Vx1.4A + 220Vx84mA = 27.3VA. Just under 30VA, the toroidal's nominal rating.
The driver stage consumption is negligible.
The unconventional mounting of the power transformer was chosen to provide good
ventilation (the transformer does get hot) and to minimise stray magnetic fields
coupling into the output transformers (also toroidals leak).
Towards the front, after the power transformer, are the Shotky rectifier and
the smoothing caps, about 33uF in total.
The Hammond power supply filter chokes are fixed to the wooden sides of the chassis.
Between them are the output stage reservoir capacitors.
The high tension Maida regulator sits between the two OPTs. It is fixed by means of
thick solid wires soldered to a one-sided copper clad PCB board beneath it. This in
turn is fixed with double sided tape to the
serves as the top cover for the amplifier's chassis.
At the front, the amplifier sub assembly. All components are mounted on a one-sided copper
clad PCB that is fixed with screws to the chassis. This one-sided copper clad PCB only serves
as a support and is connected to ground at a single point where the ground connection of the
left and right drivers and output stages come together as well (main star ground).
Close up take shows heater wiring and component layout. Main star ground and heater
connecting points are made using self adhesive PCB stripes. The star ground is not
yet finished. In the picture the driver and power stage grounds are joined but left and
right grounds are still separated. When completed, left and right and PCB copper are
connected together. This becomes the ground connection point towards the power supply.
The input resistors will get their own self adhesive stripes to make an input connecting
point. The volume control will later be connected to it.
The picture above shows a detail of the tube sockets mounting. For the 6021 tubes I used
TO-type 10 pin IC sockets (two pins were removed). The 6021 tubes snuggle nicely in and sit
tight, as if the sockets were made for them.
Above a detail of the high tension Maida regulator fixed to the one-sided copper clad PCB
underneath. It supplies both driver stages. Right in the middle the LM317 and to the right
of it the darlington pair made up of two MJE13007.
No heat sink is needed since the voltage differential between in and out and the current
drain are moderate. A higher input voltage or a higher current drain may require
Between the two 1uF metal film capacitors the regulator's fine voltage adjustment resistor
'hangs' between two mates. Mounted this way it can easily be adjusted on test.
I aimed for 2mA plate current through the 6021 for each driver stage.
The early mono prototype sourced from the Ipanema DAC. Love at first tone!
June. 3. 2012, Carlos Durandal