All about Tubes, Tube Circuits, Tube Gear

Thursday, June 14, 2012

Ultralinear Guitar Amp (5): Output Loadlines

Okay, so now its time to take a good look at the output section expectations:

The Load Line is an interesting problem, and opens a whole pile of questions.

I'll be talking about this for a few posts.
In the diagram above, we have the usual 6L6 curve family, where we are supposed to draw a load line.

Trouble is, the output stage doesn't have a resistive load, so the normal facts don't apply.

With an ordinary resistor load, the tube acts like a variable voltage divider,
and naturally, we can draw a load line by taking the extreme points:

(1) Tube off - complete current shut down (resistance = infinity): now the full power-supply voltage appears across the tube (sort of). Since there is no current, there is no voltage drop across the plate resistor, and the full B+ is at the plate (to ground). This is where the tube goes into "Class AB/B" and turns off for part of a cycle with a big signal swing. The danger at this point is arcing inside the tube, or surface traces/flashes across the pins/socket, if the voltage is too high. or the sockets are dirty or the air is too humid.

(2) Tube full on - minimal resistance, maximum current (resistance = remaining load): now there is no significant voltage across the tube, and most of the voltage appears across the resistor load. For a tube with a plate-load, the plate reads zero volts relative to ground, and the voltage drop across the resistor is the full voltage (minus any voltage drop across a cathode resistor in series with both the tube and load). The danger here is overloading thin wires (usually the grids) with too much current, red heat, meltdown, and tube death. Another potential danger is tube wear: cathode-stripping from over-current and over-heating. Nonetheless even when fully conducting, there isn't really a full short-circuit, since the resistive load (and any cathode resistor) is still in series with the B+

These two extremes serve to provide end-points for a load line, drawn as above (from highest voltage across tube /no current over to lowest voltage / max current). The tube voltage and current is assumed to follow Ohm's Law (a straight line on the chart between the points).

Normally, we would pick an appropriate HV (B+) which fixes one end of the load line,
and then pick an appropriate load impedance (Z-primary) that nails the other end down.
However, we already have a power transformer (fixing choices of B+),
and an output transformer (fixing maximum current and the angle of the load line).

The only question remaining, is where the tube will 'idle' (i..e, its quiescent current), that is, what resistance will it present, and what current will be flowing through it when there is no other signal on the control grid (g1) other than the DC Bias voltage. This idle point is what WE get to set, by picking an appropriate Bias voltage.

We do it by picking an appropriate mid-point on our line, and a modest idle current, then reading (or estimating) what the BIAS voltage will have to be to hold the tube there.

For Class A (full operation) usually we want a good (equal) swing in both directions that stays in safe range and keeps the heat dissipation below the maximum ratings.

The same basic procedure is used for pentode, Ultralinear or Triode modes.
But with pentodes, we want the load angle to avoid dipping into the 'elbow'
and with triode modes we want to follow more conservative ratings and watch screen current, but its the same idea.

In our amplifier, we can't actually pick our own load line, because we already have the transformer, which is fixed at a supposedly 5k ohm impedance.
We can still pick an 'idle' point however, just as with a car we can't change the horsepower or torque, but we can adjust the 'idle RPM' to save gas but keep from stalling.

Looking at the 'load line' here again,
you'll notice I have highlighted the bottom right corner.

Here is where, if the grid was driven far enough, we would cross
the 500 volt barrier, and find the full voltage bearing down on the tube.

Now a 5881 is only rated at 400-450 volts,
and even our tube of choice, the big 6L6GC, is rated at 500 max.

We dropped about 50 volts across the 1.3k cathode resistor
we plan to use for self-bias. This significantly lowers the voltage across the tube,
as long as it is conducting current
(Class A/AB).
This seems to offer enough protection for a 6L6GC (540 - 50 = 490 volts avail.) ...

...and here comes the catch: as long as it is conducting!.

However, as soon as the tubes start shutting down, the voltages again climb too high for safety.

We need to look at why the 6L6GC was rated at 500 volts in the first place:
People were getting shocked trying to change tubes with a top plate-cap carrying 500-800 volts,
and makers (tube makers too) wanted to reduce risk and liability,
so tubes without plate-caps and designed to run on lower voltages were promoted.

Moving the plate-cap to the base however, made higher voltages unreliable,
because of arcing and tracing at the socket
(and arcing inside the tube with smaller gaps between elements inside the tubes!).

We can overcome this problem by (you guessed it!)
reverting back to top-cap plate connections, namely using a
6BG6A instead of a 6L6 as our tube of choice.

Click the image to open in full size.

As it turns out, the 6BG6A is a Higher Voltage version of a 6L6,
like the 807 and 1625 tubes (only more modern).
And its a hell of a lot cheaper than good 6L6GC tubes too!

I happened to pick up a box of them for a few dollars apiece.

With proper HV wire and caps, and a safety-screen to keep idiot-fingers
out of harm's way, this seems like the ultimate solution.

Even if we could squeak a 6L6GC in our circuit using self-bias,
at lower risk, the 5881 would be right out, at a max of 400 volts.

We will test the amp with the 6BG6s, then look into making some options
for retrofitting 6L6GCs or 5881s in a pinch.

The 807 and 1625 would do equally well electrically,
however, we'd also have to invest in special 5-pin or 7-pin sockets (expensive),
which are also unsuitable for tube-substitution or swap-outs. anyway, to recap (no pun intended),

We're going for 6BG6As instead of 6L6GCs.
The basic curves stay the same, but the maximum voltage ratings are increased, allowing "class B" operation,
or excursion into current cutoff as the driving signal crosses over between output tubes (or pairs!).

But we still have some problems with the 'load line':

Although we treat the reactive transformer and speaker-load as if it were a resistive load for purposes of drawing the load line,
the premise is basically false, and the real tube behavior will be nowhere near what the load-line suggests:

(1) Suppose we pick an idle-point of 50 mA, which would require a bias of 50 volts, and according to the load-line would leave 300 volts on the plate (dropping a vertical down to the x-axis).

(2) Of course we all should know that the transformer is not a resistor, but only a 'nominal A.C. impedance' of 5k. Its actual (measured) resistance is only 50 ohms per side (on mine anyway).
This means that with 50 mA coursing through it on idle with no AC signal, its only dropping about 2.5 volts, leaving 538.5 volts from ground on the plate! Of course, our self-bias cathode resistor takes up about 67 volts (at 1350 ohms), leaving 471 volts from cathode to plate. Thats safe (while idling) for a 6L6, but not a 5881.

(3) All this suggests that any AC signal at the plate is really riding on a 471 volt D.C. offset voltage (relative to the cathode), and we should really shove our load line over to the right about 171 volts! (or move the scale 171 volts to the left).

(4) A 50v +-, or 100v Peak to peak input voltage, on one tube grid would swing the voltage at the plate about 360 volts, or from 290 to 650 volts + relative to the cathode. Only the 6BG6A will be able to handle that kind of spread.

No comments:

Post a Comment