General

This paper details a comparison of the JJ/Tesla and the new Electro-Harmonix EL84 power pentodes.  The EH EL84 tubes were graciously provided as free samples from Lord Valve at NBS Electronics in Denver, Colorado.  I had to buy the JJ/Teslas myself. :)
 
 

The two contenders face off on a field of white levant tolex....

A fixed-bias, push-pull class AB1 output stage driven by a "long-tailed pair" phase inverter was used for the tests.  The phase inverter was designed around a 12AX7 stage to provide the proper drive levels for the EL84 tubes, allowing them to be driven to fully clipped levels before the phase inverter itself exhibited noticeable clipping.  Provisions were made for a global negative feedback loop that could be disconnected to test the performance as both a feedback and non-feedback amplifier.  The input signal was applied directly to the phase inverter input, no preamps or tone controls were in the path.  The tubes were run at a plate voltage of 368V at idle, sagging to 342V at full clip (using a 5AR4 tube rectifier).  Three screen resistor values were used: 100 ohm, 750 ohm, and 2.5K (more on this in the output power section).

All measurements were taken using a PC-based network analyzer system, composed of a 96kHz 16-bit signal generator, and a 48kHz 16-bit 16,384-point FFT/network analyzer.  Listening tests were done using an Aiken Invader 18W head into a 4x12 cabinet loaded with Celestion Greenbacks.  FFT spectral data was composed of an average of 25 runs per plot.  A pair of JJ/Tesla EL84 tubes were chosen to carefully match the bias current of the EH tubes, allowing for "on the fly" swapping without rebiasing.  The bias was set to 28mA cathode current (combined plate and screen current) per tube.

The output stage described above is relatively hard on the JJ/Tesla EL84, particularly with lower values of screen grid resistors, which is due to the high plate and screen grid voltages.   A 100 ohm screen grid resistor will cause the screen element to glow very brightly, and it takes a 2.5K resistor to get the dissipation down to a mild glow.  The EH EL84 reportedly has an improved screen element, but it also glowed quite brightly at 100 ohms and 750 ohms.  At 2.5K,. however, it did not glow at all, while the JJ still exhibited some glow, which indicates it is indeed a more robust screen element.  As a side note, the EH exhibited some glow at 1K and 2K as well, so the "safest" value is 2.5K, unless the screen voltage is decreased.  It is useful to note that the max rated screen voltage for the EL84 is 300V, so this test is exceeding the manufacturer's recommendations unless the higher value resistors are used to limit the screen dissipation.

The 100 ohm screen resistor was deemed unsafe at any speed, so no further tests were conducted on it.

The JJ/Tesla pair put out 50.6V p-p at the onset of clipping into a 16 ohm resistive load with the 750 ohm screen resistors, indicating an output power of 20W.  When the screen resistor was raised to 2.5K, the output voltage dropped to 48V into 16 ohms, for a power of 18W.

The EH pair put out 50V p-p at the onset of clipping into the 16 ohm resistive load with the 750 ohm screen resistors, indicating a slightly lower output power of 19.5W.  The output with the 2.5K screen resistors was also lower, at 46.6V into 16 ohms, or 17W.

The first tests were frequency response sweeps into a reactive load (a Marshall Power Brake set to the "load" position), with the global negative feedback loop disconnected.  The output was first set for 1W at 110Hz into a 16 ohm resistive load using the JJ/Tesla tubes, then a reactive load was connected to the amplifier for the test. The input level was left at this same setting when switching to the EH tubes, to allow a direct comparison of relative output power and spectral response.  Below is a graph of the frequency response of both tubes overlayed, as well as a plot of the relative difference in response:

This first graph above shows the JJ/Tesla puts out a slightly higher power than the EH, but the overall frequency response is similar.  Note the reactive load variation results in a peak deviation in frequency response of 12dB from the resonant peak at 100 Hz to the "valley" at 400 Hz.
 
 

The second graph shows the EH is not as responsive to the reactive load impedance variations, which indicates it has a lower plate impedance or higher damping factor when loaded with the same reflected plate load.  I believe this is responsible for most of the perceived tonal variations between the two tubes (more on this later).
 

 
For this test, the feedback loop was connected and the same response plots were made.
 
 

The above graph shows the effect of the global negative feedback loop is to reduce the gain and also reduce the variations caused by the reactive load. The new peak deviation at the resonant frequency of 100Hz is only 4.6dB, compared to the 12dB difference obtained without global negative feedback.  This is a result of the increased damping factor and reduced output impedance caused by the application of negative feedback.
 
 

The second graph shows the response after negative feedback to be quite different than before.  The response of both tubes is much more similar, except the EH is a bit brighter on the high end, and a bit lower on the low end, but only a few tenths of a dB in either case.

Test number 4:  Spectral response plots for undistorted outputs, no negative feedback
 

For this test, a 110Hz high-spectral-purity tone was applied to the phase inverter input, with the output level again set for 1W into a 16 ohm resistive load, and a reactive load connected for the test.  The output spectrum of the amplifier was then plotted for each tube type.  The first set of plots was done with the global feedback loop disconnected, and the second set of plots shows the response with global negative feedback.

The above plot shows the EH to have a 4dB higher second harmonic, while the JJ/Tesla has a 16dB higher fourth harmonic.

The above graph shows the difference between the spectrums of the two tubes over a 20Hz to 20kHz range.

For this test, the feedback loop was connected, and the plots were repeated.

The above plot shows a much lower overall harmonic content, and very little difference between the JJ/Tesla and EH harmonic spectrums.  This is a result of the reduction in harmonic distortion gained by the addition of global negative feedback.

The above plot shows the spectral differences between the two tubes with global negative feedback.  Note the much lower difference in spectral content.  Global negative feedback tends to compensate for the variations in tubes, making the resulting output relatively independent of tube parameters.

For this test, a 110Hz high-spectral-purity tone was again applied to the phase inverter input, but this time the output level was cranked up to well into hard clipping.  The output spectrum of the amplifier was then plotted

Note the fourth harmonic of the EH is about 4dB higher than the JJ/Tesla.  The JJ/Tesla shows a 4dB higher second harmonic, and the third harmonic is about the same for the two tubes.

The above plot shows the difference between the two responses.
 

The global negative feedback loop was connected and the test repeated.

The above plot shows a very similar harmonic content, with the exception of a 4dB higher second harmonic on the JJ/Tesla pair.
 
 

The above difference plot shows a much smaller variation in harmonic content than the no-feedback case.

Listening tests

Both sets of tubes were tested in an Aiken Invader amplifier (18W, fixed-bias, class AB1, using global negative feedback) driving a 4x12 cabinet loaded with Celestion Greenbacks.   To my ears, the JJ/Tesla tubes had a subjectively "better" sound than the EH tubes, both clean and distorted.  The JJ/Tesla's sounded a bit "fuller" than the EH tubes.  The EH tubes were a bit brighter sounding than the JJ/Tesla's. Overall, the two sets were very close in tone, but there is a bit more complexity to the tone of the JJ/Tesla that gave it the edge over the EH tubes, in my opinion.

Both the listening tests and the measured data indicate the two tubes are remarkably similar in performance and tone, given the differences in construction and appearance.  The JJ/Tesla tubes put out a bit more power into the same load impedance, and in my opinion, have a slightly better tone.  However, the increase in screen dissipation of the EH tubes makes them a nice choice for amplifiers that tend to go through tubes more often.  The EH would be highly recommended for amplifiers that run the tubes (especially the screens) at high voltages.

And now the caveats:  Since I only had one set of EH EL84 tubes to compare to the JJ/Tesla tubes, this test is not in any way statistically accurate.  In order to properly test the tubes and draw accurate conclusions, a larger sample size is required.  This test made no allowances for the variation that can occur from tube to tube or in tubes from different production batches.   In addition, the listening tests were purely subjective, based on a sample size of one, using my ears.  This must not be taken as factual information, as it was just my personal preference.

Again, thanks to Lord Valve at NBS Electronics for providing the samples of the new EH EL84 tubes for testing.

Copyright © 2002,2003  Randall Aiken.  May not be reproduced in any form without written approval from Aiken Amplification.

Revised 02/16/03