6AN8 Pentode Section, Voltage Amplifier


Heathkit UA1 Schematic: click here for full image


Heathkit SA2 Schematic: click here for full image


The circuit I built from the Heathkit idea actually works, and well. It is, and was used, as evidenced in the Heathkit UA-1 (a 12W amp dating from early 1958), the subsequent (mid 1961), and only slightly changed UA-2, and the Heathkit SA-2, and Dynaco Classic MkII and Dynaco Classic MkIII.

A key design feature is the use of the 6AN8 pentode-triode compactron. It's an interesting valve. It has pentode section sufficiently suitable to fill the input voltage amplifier spot where say an EF86 might typically be seen. Operation at low anode and screen voltage is eminently practical, as will be discussed at length below. Other aspects of performance in this critical role, such as susceptibility to instability, noise, hum and microphonics all appear minimal, and readily acceptable. Then there's a reasonably beefy triode, eminently suitable for subsequent concertina phase splitter duty; ra is 4,500 Ohms, just above 20, and typical operation at Ia to 15mA is suggested in the data.

A second key design feature is the use of direct coupling between these first and second sections;

"It is usual to direct couple to the anode of the input stage, and let that determine the DC conditions of the concertina, resulting in the saving of a coupling capacitor and a low frequency time constant." "Valve Amplifiers", 2nd Ed., M. Jones, p282.

The operating point of the 6AN8 pentode section, as a consequence of this coupling strategy, is a compromise, since the same B+ supply is used for both stages (rather than a separate and higher B+ for the concertina, which would be a viable but more parts-costly alternative).

Detail from Dynaco Classic MkII Schematic: click here for full image


Detail from Dynaco Classic MkII Schematic: click here for full image


6AN8 pinout


The ability to use the 6AN8 pentode section linearly at anode and screen voltages much lower than recommended spec. is of direct application to being able to direct couple to the concertina phase splitter following. A concertina phase splitter desirably has about of the available B+ across each of the (equal) cathode and anode resistors, the remaining across the valve itself, such as to facilitate maximum voltage swing.

The pentode to triode direct coupling results in a compromise such that whatever B+ is available gets split (less than desirably) approximately three equal ways; 1/3 across the anode resistor, 1/3 across the valve anode-cathode, 1/3 across the cathode resistor. The available B+ was unlikely to be more than say about +300V; this would put 1/3 of that, around 100V, on the phase splitter triode's cathode, and roughly the same, give or take some bias volts, on the grid; hence the anode voltage of the preceding pentode stage (if direct coupled) would be this (relatively low) value. Hence it was "interesting" to see Heathkit, Dynaco and Univox circuits using the 6AN8 pentode section in this way.

So, on the one hand the 6AN8 pentode operating point needs to be relatively high, to give best linearity, and low to give an appropriate voltage at the cathode of the triode following (since with direct coupling between the two the operating point of the pentode sets the operating point of the subsequent triode, effectively. On the one hand the manufacturer recommendation of 125V at pentode anode and screen would be good, and on the other, one quarter the B+, (say 75V for a +300V B+) - the typical design centre for a concertina phase splitter - plus a few volts triode grid bias - would be best for the phase splitter. It comes as little surprise then to see the actual voltage at the 6AN8 pentode anode in the Heathkit designs being around halfway between these two considerations at 80~100V.

Detail of 6AN8 stage in Heathkit UA1 amplifier.  Ignore the switched feedback arrangement shown, for the purposes of this article.

Heathkit UA-1

Inspection of a few other similar application examples showed usage at lower voltages still - right down to about 40V on the anode, 30V on the screen in one case! I was interested in looking further at this. How did the 6AN8 pentode section perform at anode voltages substantially lower than the suggested operating conditions given in the valve data? Surely it was poor design practice to use the valve "so far from home"?

The immediate problem was finding curves to inspect; I easily found published curves from a couple of sources, but both were for a screen grid voltage of 150V; that was the problem!

The 6AN8A pentode data curves I was able to locate were Va versus Ia curves with the screen voltage held at 150V, way above that which it was proposed to employ in this circuit.

Published data curves for the 6AN8 pentode.  The 150V screen voltage at which these are done was of little application here.  Click here for full image.


The application examples I had were for screen voltages considerably different to this; 40~70V or so only. Attempts to plot a load-line on these Vs = 150V curves, linking points (Va=307V, Ia=0) and (Va=0, Ia = 307V/(220K = 1.4mA) produced meaningless results, coming nowhere near the Vg = -1.3 curve that applied in practice.

Eventually I thought I'd have a go at plotting some myself using an AVO 160 borrowed from good friend Hugh Dean. It wasn't too hard a process, and I soon came up with anode voltage versus anode voltage curve diagrams for four or five grid voltage settings, with screen voltage held fixed at 60V.

At this point I had no idea how relevant or accurate any of this data I'd gathered was. The AVO 160 is certainly "uncalibrated" and I have no standard against which to establish its performance or accuracy. I also don't have a standard "6AN8", not even a "production" one from the same manufacturer as any versions of the published data curves. At least I had a "new" valve to use for the measurements! So it came as little surprise, when I used the test set-up to mimic the published curves (simply by setting the screen voltage at 150V and repeating the measurement exercise) that there was considerable "error".

AVO 160 valve tester; a wonderful instrument to work with but in this case calibration errors became apparent . . .

Comparison of the Sylvania published data versus the characteristics I measured on the AVO160 for a NOS CEI branded 6AN8, for Vg = -0.5V, -1.0V, and -1.5V, revealed (ahem!) a substantial offset (putting it nicely) between measured and spec'd curves, I began to think of what sources of "error" I could influence.

The only "adjustment" on the AVO 160 is the setting for mains voltage input to the instrument, and this setting is quite important. The applicable available mains voltage settings are 235Vac and 240Vac for my house-mains usually lie somewhere between these two values, nominally closer to 240V than 235V. Changing the instrument from 235 to 240V was interesting, "elevating" the Vg curves I measured. Still the offset error remained substantial, and on closer inspection there was a slope error as well. Ho-hum. Well at least the relative positions between curves produced by different grid voltages looked valid, and there was an obvious general correlation between what I measured and what was published. I'd say the 60V screen voltage curves I subsequently made might be considered "indicative" and "informative" rather than quantitatively useful; (a nice way of saying they look pretty but probably aren't much use for design calc's!).

On this rather slip-shod "guestimate" basis I made the following curves for Vs = 60V. I think the diagram as presented here is probably close-ish to about right, er, somewhat . . .imprecisely . . .

My home grown 6AN8 pentode curves for 60V as the screen voltage.

Drawing the load-line is easy; the B+ measured in practice for the stage, at +307V. This is the extreme of operation where the valve is completely cut off; no anode current flows (so Ia=0) and the B+ appears in full at the anode (i.e. Va=B+). No problemo!

The other end of the load-line is the point where the valve is assumed totally conductive, anode to cathode, such that it appears as a short circuit.  The current assumed to flow in this condition is  just B+/Ra, (Ra being the load in the anode circuit of the valve), that is, the Ia for Va=0 is B+/Ra = 307/220K = 1.4mA)

On the diagram I had also plotted curves for what I knew the bias Vg was; -1.3V or so. The actual operating point falls on this curve, as it intersects the approximate load-line I drew across the diagram. It looks about right; as a cross check the operating point should occur at about Ia = (307V-80V)/220K = 1mA

I also plotted a curve for Vg = -1.1V. This value is 0.2V away from the actual bias point of Vg = -1.3V. The curve for Vg = -1.5V was there already, so then it was possible to compare two curves, each 0.2V away from the bias point. The point of doing so was to try to get some idea of the likely linearity of response to (small) input signals (in this case of 0.4Vpp amplitude).

The load-line crosses the Vg = - 1.1V curve well past the knee of that curve. This makes it a bit meaningless to try and calculate actual percentage distortion this way. It suffices to say that for 0.4Vpp signals the distortion must be pretty horrific!

So why does the implementation actually sound even half-way decent? Well, I surmise that the gain structure has a lot to do with it, as does the asymmentry of the distortion producing mechanism seen here. The open loop gain of the overall amp, from input to loudspeaker, I measured at roughly +42dB (x126). The closed loop gain I ended up with was +25.7dB (x19.2). (As an aside, I was surprised at how close this was to the "op-amp" formula involving the feedback resistor (22 kOhm) and the 1 kOhm from 6AN8 pentode cathode to ground; (22K + 1K)/1K = 23 (+27.2dB). This formula assumes an infinitely high open loop gain for the amplifier, something far, far from the case in a valve amplifier of this nature. I recall seeing a modified formula for use in this situation, but it is understandably more complex, involving additional terms).

Anyway, the difference between those open and closed loop gain figures is the amount of loop negative feedback; 42 - 25.7 = 16.3dB worth. Relatively and dangerously "generally" speaking, for valve amplifiers, my understanding is that this is quite a high figure. High levels of loop negative feedback in such an application of this are certainly questionable for firstly stability reasons, and secondly sonics; the resultant distortive harmonic artifacts, while very much lower in level, are undesirably pushed to higher orders.

The other point is the blatant (read "gross") asymmetry in the nature of the distortion produced - it's this asymmetry that gives rise to even rather than odd order harmonic content of the distortion so generated; a euphonic rather than discordant characteristic. For this very same reason one may well favour the concertina phase splitter over some of the alternate, (dual valve and more symmetrical and hence odd-order harmonic dominant in nature), phase splitter topology alternatives. It may just be that with relatively high loop-negative-feedback to reduce the distortion level overall, and then a dominant asymmetrical distortive characteristic, (both at the 6AN8 pentode voltage amplifier stage AND at the subsequent phase splitter), that an overall sonic signature is conveyed that is linear but "warm".

In the Univox U1221 and U1226 line driver schematic details at right, other typical usage and circuit values can be seen. Relatively even lower circuit voltages around the pentode section are observed; +57V and +78V on the anode and +30V, +39V at the screen. So it was interesting to produce and look at curves for an even lower screen voltage; 40V. (I haven't reproduced them here as my measurement inaccuracies make them otherwise all but valueless).  In these examples the same 6AN8 pentode circuit as considered above is being used as a line driver for a reverb spring. Guitar amp or not, in this application linearity would be a design emphasis rather than deliberate distortion - that might well be engendered elsewhere in a guitar amp(!), but there's no particular reason to deliberately design it into this part. So one might still surmise an expectation of reasonable linearity in practice in this application, 40V screen grid voltage or not! These applications at least showed that even lower Vs operation WAS indeed workable!

Detail of Univox U1221 Schematic: click here for full image


Deatail of Univox U1226 Schematic: click here for full image


Amplifier schematic - click here for full size image


As can be seen in the schematic at left, I first started with the Heathkit schematic component values in this part of the circuit. These included a 1 kOhm cathode resistor (just the sum of the two resistors, 510 Ohm and 470 Ohm, seen in the UA-1 circuit), a 15 kOhm in the grid-stopper position (without knowing why this particular value?, a 220 kOhm as anode load (off approx. the same +310Vdc B+), and 1 MOhm, 0.1F network for the screen.

1 kOhm turned out to be quite a convenient value for the cathode resistor; the DC voltage across it in Volts was direct readable as the cathode current (the sum of anode and screen currents remember) in mA. I subsequently changed the 1 MOhm on the screen to 680 kOhm, following the Heathkit change seen in the UA-2, and observed the screen voltage to rise somewhat. I later changed the anode resistor(s) to 180 kOhms, to raise and better position the DC voltages around the subsequent (direct coupled) triode (phase splitter) section.