EL84/6BQ5 Output Stage

My intention with the output stage of this project amplifier was to use a completely conventional output stage; with my odd and somewhat suspect output transformers to worry about I wanted to be able to eliminate any other output stage variables I could. This wasn't the time to experiment with something wonderous and new in terms of stage topology or valve types! So I had a fair idea that EL84's or something similar would be the go, striking an appropriate balance between the available chassis sockets, B+ and heater supplies available, and physical size, and the power I thought I might drag out of my odd output transformers.

The stereo push-pull EL84 based output stage in my project amp.

EL84 valve data was readily available at Nostalgia and Triode Electronics. The EL84 (a.k.a. 6BQ5) is a relatively common medium power output pentode, with many many application examples. In Class AB1 a  push-pull pair is capable of in excess of 10Wrms into an appropriate transformer coupled speaker load. Anode power dissipation is 12W. Design anode voltage maximum is 300V although it is not at all uncommon to see it used well over this rating, to 350V or so is not unusual, and it appears to survive such abuse well enough! These Pdiss and Va ratings suggest a quiescent bias anode-cathode current to around 40mA or so.

At a Va of 300V, (the highest for which manufacturer ratings are given, since that's the design max for the valve), an Raa of the appropriate output transformer of 8 kOhms is given, making the primary impedance of the transformer I proposed to use substantially too low. I appreciated that that would result in a lower output power realisable in my amp than would otherwise be possible.

I had decided that I would at least try to maximise the albeit low power that at least I could achieve. The speakers I use are poor quality and low efficiency, and the power available from this amp would be low in any event.

I felt I could not afford the "luxury" of sacrificing more, and looked wistfully but merely passingly at the "triode mode" data also available for the EL84. I would have to pass up the better sonics but lower power of the triode mode connection.

DEtail of PP EL84 output stage from Leak model "Stereo 20"; click here to enlarge the image.

Stereo 20

I had to search a bit to find an example or two of the output stage I had in mind; EL84 Class AB1 push-pull in ultralinear mode, with cathode (self) bias. Many examples existed for triode mode or pentode mode, and using grid bias.  But the web provides such an accessable and immense resource for such a purpose. The output stage of the Leak Stereo 20 and Leak TL12Plus3 at Kiewa were interesting, and included most of the circuit elements and component values (near enough) that I would eventually use.

Then I came across website references to the Heathkit UA-1 at several websites; Heath Hi-Fi (or Heath Hi-Fi), SDS Labs (Sheldon Stokes), and "Tom80".  The Heathkit SA2 also used much the same circuit, again seen at a number of websites including Heath Hi-Fi (or Heath Hi-Fi), and "Tom80".

Output stage detail from Leak model "TL12 Plus 3"; click here to enlarge the image.


Then followed an extremely useful and informative note on the Heathkit UA1from Sheldon Stokes;

Prolific and talented engineer, Sheldon Stokes

"The amp's -3dB points are at 8Hz and 65kHz at full power. Plus the square-wave response at 20kHz (5 Watts) is very clean; it looks like there isn't a transformer there at all. After all the most critical parts of all tube amp's are the output transformers."[Notwithstanding Sheldon's comments, I got pretty much the same frequency response measurements with my el-cheepo transformers, which confused me considerably, but left me satisfied enough!]

Sheldon continues; "The amp was originally designed with two feedback schemes. The first is the traditional voltage feedback scheme, where some of the output signal is taken from the 16 Ohm tap on the output transformer, and fed back into the cathode of the input pentode. The amp sounds amazing in this mode.

The second scheme uses some current feedback as well as voltage feedback. With this second scheme, the current from the speaker is sensed via a small value resistor and fed back into the cathode of the input stage. The voltage feedback from the output of the amp is reduced by about half to keep the total feedback about the same. This may have worked well with old speakers, but doesn't sound very good with my Mission bookshelf speakers. And it sounds just awful with the Quad ESL's. You can flip a switch on the front of the amp to select either the great sounding traditional all voltage feedback, or the bad sounding current/voltage feedback scheme. When I rebuilt these amps, I removed that feature completely from the amp, and used the voltage feedback scheme."

I constructed my derivative version without this second type of current/voltage feedback scheme, and on the strength of Sheldon's adverse comment decided simply to leave it that way, with a conventional feedback scheme only!

"The UA-1 and UA-2 amps are essentially identical. The only differences between the UA-1 and UA-2 are that Heath increased the value of the cathode resistor on the output tubes because they were running them WAY too hot, and burning them out quickly. The UA-1 had a 130 Ohm resistor in the cathode at first. They changed that pretty quickly though, and the schematic here reflects that change. They also reduced the value of the screen grid resistor on the input stage to 680 kOhms. I have tried both the 1 MOhm and 680 kOhm resistor and there is no difference that I can hear or measure."

I decided to leave the bias I had on the EL84's, which was roughly the same as having one 130 Ohm per pair, and suffer the consequence of reduced valve life. The 1 MOhm to the screen on the 6AN8A pentode was easy enough to change, and since at least it didn't appear to have a detrimental effect I decided to go with the Heath mod and change down to the 680 kOhm value. Presumably this would pull the screen voltage up somewhat, presumably to improve the pentode's linearity, I guess.

EL84 output stage cathode bias arrangement detail; click here to enlarge the image.


My nifty little 120 Ohm preset's. I wonder where I "liberated" these from! Damn handy when I needed them though! 120 Ohm WW
Detail of the bias arrangements I used for each of the EL84's

Bias Detail

I then looked at one or two ways of doing the cathode biasing; that shown above was interesting although in the end I didn't use it.

Rummaging through my parts collection I discovered four (one for each output valve for the two channels) nifty little 120 Ohm wirewound presets.  Bingo!  On the strength of this discovery I decided to use discrete (unlinked) bias arrangements for each valve, selecting a series resistor that I hoped would put these pots at about 60 Ohms, the middle of their range, at a nominal cathode current of 35~40mA.  With a bit of fiddling the resistor value turned out to be 270 Ohms.  I added a 1 Ohm resistor in series for test purposes (the voltage across it in mV is the current through it in mA), to make set-up easy.  I found it relatively easy to set the bias of each valve for the same cathode current by these means, although I guess I'll have to recheck the bias setting regularly as the valves change over time.