Amplifiers Homepage

Stephen W. Moore

EL84 Push Pull

Transformer Coupled EL84 Push Pull


This amplifier design uses a Lundahl LL1540 input transformer for the phase splitter and Edcor power (XPWR057) and output (CXPP25-6-7.6K) transformers. Shown below in prototype monoblock implementation.



The amplifier uses an LL1540 input transformer to perform the phase splitting to a 12A*7 driver valve. The LL1540 primaries can be wired in series, resulting in a 1:1 ratio with the secondaries, or in parallel, resulting in a 1:2 ratio with the secondaries.  Wiring the input windings in parallel will double gain but halve the input impedance, so make sure the source component can handle driving low impedances. The LL1540 is wired according to the Lundahl data sheet, with each secondary loaded with 12K resistors. I also added the 22K and 1nF resonance suppressor according to the data sheet.


I tested the amplifier using the LL1540 with the primaries in series, for higher input impedance and 1:1 ratio.


For testing, the source is connected to the LL1540 through a coupling capacitor. I did this because the LL1540 is intolerant to DC offset. Some source components (i.e. preamps) have some DC offset. I will remove the coupling capacitor after confirming the preamps in question do not have any DC offset.


Appropriate driver valves, depending on amplification requirements, are the 12AT7 and 12AX7. Therefore, the amplifier’s total gain is determined by the LL1540 configuration (1X or 2X gain) and the driver valve’s gain. I tested the amplifier with the 12AX7 driver.


The driver valves have a shared cathode resistor and a trimpot for fine adjustment with a total 111KΩ plate resistance. This results in 1.2mA at 300V B+. The plate potential is 150V, exactly half of B+. 220Ω grid stopper resistors prevent parasitic oscillations, along with 0.1uF capacitors soldered directly across the filament pins (4 and 5).


The driver valves are coupled to the output valves through 0.47uF capacitors. I provided plenty of space on the board for a variety of capacitor selections. I have tested Solen (black), Jensen (blue), Dayton (black), and Mundorf (white). The driver grids are pulled to ground through 470KΩ resistors. 220Ω grid stoppers prevent parasitic oscillations, along with 0.1uF capacitors soldered directly across the filament pins (4 and 5). The EL84s each operate at about 40mA for 12W plate dissipation by using a shared bypassed cathode resistor. The bypass capacitor can accommodate both 5mm and 7.5mm pin spacing to test different bypass capacitor types. I stock both Elna Cerafine and Panasonic FM capacitors to try. 2Ω resistors serve as convenient bias current measurement points and a trimpot is used to make fine adjustments. Solder pads are configurable for Triode, UL, and Pentode operation.

noise (the MOSFET takes care of 60Hz/120Hz ripple). The next revision will use a larger toroidal choke. The final output is filtered by 2 x 22uF Solen film capacitors, with everything using star grounding.


I measured 3.2mV of noise on the B+ when under load.


The DC filament supply uses the 6.3V and 5.0V windings of the power transformer in series, for a total of 11.3VAC through the full-wave diode bridge. The diodes have 0.1uF snubbers and the output is filtered by two 0.47uF high-frequency capacitors. Two 0.56Ω resistors in series provide current limiting to reduce DC voltage and reduce the peak currents in the power transformer. This is filtered by 2 x 10000uF 25V electrolytic capacitors separated by a 18mH power toroidal choke. The final output is filtered by 0.47uF to catch any high-frequency components. I measured 16mV of ripple under load.


The next version will increase the capacitance significantly, add a third electrolytic capacitor, and use two sets of common-mode chokes. This should reduce filament DC ripple to below 4mV under load.


The total amplifier hum, measured across the EL84 plates, was 12mV. The hum was not measurable using my oscilloscope at the speaker terminals, however, the amplifier hum is audible using high efficiency speakers (i.e. Fostex FE207E). The power supply upgrades outlined above should reduce the hum significantly.



I found solid-state diode commutation noise objectionable. The high frequency spikes from the commutation backfeeds into the power transformer and inductively couples to the other windings. Therefore, the high-current filament diodes commutate loudly, and the resulting high-frequency hash inductively couples with the B+ windings.


A quick fix was using 0.1uF 50V capacitors across all four filament diodes. A more sophisticated solution would be using 100Ω resistors in series with each capacitor.


I also installed 1250V capacitors across the two B+ diodes, along with a 0.47uF capacitor bypass to return.

Schematics — B+ Power Supply



Schematics — Filament Power Supply



Schematics — Amplifier



Schematics — Improved Filament Power Supply



Optional Modifications


Some optional (experimental) features were integrated into the design. The coupling capacitor pads have variable spacing to accommodate large or small capacitors. The EL84 cathode resistor bypass capacitor can accommodate several sizes of electorlytics. Screw terminals allow the placement of an external (i.e. Hammond 159T) choke on the B+ power supply.


The most interesting feature is using the UL transformer taps as feedback to the driver plates. Populating a resistor pad on the circuit board engages the feedback, which reduces gain, distortion, and output impedance. However, while the feedback reduces overall distortion, the removed distortion reappears (albeit attenuated) at higher frequencies, which is arguably more objectionable than the original distortion.



The EL84 filaments operate at 6.3V, so they are connected in series to be compatible with the 12A*7 filament. The filament circuit is grounded between the EL84s to control noise (see schematic below).


The filaments operate with a DC power supply.


Final output is through an Edcor 7.6K : 6Ω output transformer.


The B+ and filament power supplies use diode rectification. I had a lot of problems with commutation noise, so crude snubbers were installed (see photo) using 0.1uF and 0.47uF capacitors. More sophisticated snubbers would use resistors in series with the capacitors.


The B+ power supply uses a 60Ω current limiting  resistor to decrease the peak current through the transformer. This is connected to a Solen film 2.2uF capacitor followed by a 450V 470uF electrolytic capacitor. A MOSFET “capacitor multiplier” follows with 68uF storage fed through a 50KΩ resistor. I added a 1MΩ resistor to ground to create voltage divider with the 50KΩ resistor, making the MOSFET drop a few more volts and have better ripple rejection.


After the MOSFET capacitor multiplier, there is another 450V 470uF electrolytic and a small 18mH inductor designed to control high frequency


Pass Zen v.4

Tube/MOSFET Hybrid

Fostex FE207E

EL84 Push Pull