Biasing questions...

Ok... now that this amp is doing what I expect, I'm trying to draw some conclusions about its biasing... I'm reading the Aiken paper on biasing, trying to become more learn-ed.

The amp is an AB push-pull 50W amp using EL34s, I think this is what you'd call fixed-bias, in that the cathodes sit at ground and the grids are fed through a bias network from a -34V source (adjustable via the bias trimpot.

I measured the idle current using one of Aiken's methods, by measuring the open resistance of each tap of the output transformer, then measuring the voltage of each leg relative to the CT, and finding the current that way. I measured 108.1R, 3.76V ~= 35mA on one side, and 72.6R, 2.55V ~= 35mA on the other, and those values seem more than reasonable, so I'm confident in that measurement.

Now, Aiken's page says "The plate load impedance is extremely important, as it sets the maximum average plate dissipation under signal conditions, and thus, the maximum idle current that can be used." I don't have any good way to know the load impedance, as the documentation I have does not mention the ratio of the Output trafo at all, so I had to resort to measurement.

I fed the secondary a 1KHz, 1VRMS sine wave, and observed a 21.47VRMS sine wave on the primary, so 21.47:1 would be the turns ratio, making the impedance reflection ratio 461:1. As the load speaker is 8R, that makes the reflected load 3688R. (hopefully I'm still sane at this point).

His paper then goes on to show a bunch of graphs taken from a piece of software that does calculations based on the plate load, plate voltage, and idle current, but he doesn't actually give the method to calculate such, and just uses that as proof that 70% of the max plate dissipation is a good bias point.

ANYWAY, so I figure that 35mA at 420V is 14.7W, which is only 59% of the maximum 25W dissipation for the EL34, and that 70% would be 17.5W, which at 420V is 41.67mA. Is that a reasonable value, and should I tweak it up?

Thanks folks!

> I'm reading the Aiken paper on biasing,

Be nice to give a link for that. Not all of us have read it.

> I measured the idle current using one of Aiken's methods, by measuring the open resistance of each tap of the output transformer, then measuring the voltage of each leg relative to the CT

That's dangerous. Put a 10Ω or 1Ω resistor in each cathode. It won't upset the bias, and with 200mV DVMs selling for $10, accuracy is high.

The cathode current is plate+screen current, so you have to make an allowance. But at idle, the ratio of plate/screen current is pretty constant.

> just uses that as proof that 70% of the max plate dissipation is a good bias point.

It isn't that simple. In fact there is NO good answer. In the general case it depends how hard you feed and drive the tube. And which tube it is: linear 6F6 gives different transfer curve than bent 6L6. The EL34's bentness depends a lot on load and swing. Remember that an "8Ω" speaker is NOT 8Ω but 6Ω to 50Ω.

Remember that the gain of a tube is a function of current. More current, more gain.

If bias voltage is small, idle current is large, tubes run hot. And if you load the tubes so one side goes to cut-off at peak swing, then gain reduces near maximum power so THD is higher than it needs to be.

If bias voltage is large, tubes idle near cut-off, then gain is low for small signals and larger for large signals, THD is high for small and medium signal levels. At an extreme, you get "crossover distortion" just like an underbiased transistor power amp. The 24W book-values for 7189 run awful close to crossover distortion. And the large bias requires large grid swing to get to full power, possibly increasing driver distortion. However the tubes run cool.

You get the maximum power from a push-pull pentode by rasing the plate voltage to the max, then reducing the load impedance until you hit the plate dissipation rating at full power. If you go this way, very high plate voltage, you often do have to idle at very low current to keep from melting the plates at idle.

Most less-insane tube amps really work class AAB: class A for all powers up to almost max, then just cutting-off one tube on peaks of full power. If we pretend that Gm varies as √Ip, then the optimum bias current per tube is 1/4 of the peak plate current. Peak plate current is 1.4 times the total average DC plate current at full power with small bias current. So if full power current is about 200mA when you trim the idle current down to 10mA, linearity should be best with 70mA idle current in each tube. From there: one tube swings 70mA to 280mA, its gain doubles, while the other tube swings from 70mA to <10mA, its gain vanishes. So gain is the same for small signals on both tubes and for large signal peaks carried by one tube.

But real tubes deviate widely from Gm~~√Ip. Cancelling the 3rd harmonic like this usually (for tubes, always) brings up the 5th harmonic. And in this case 70mA puts you way over the dissipation rating. And this ignores the shape of the THD plot between small-signal and full-power.

Indeed, 70% Pd is not a bad bias point for the less extreme amplifiers.

35mA, 41.67mA..... uh, this makes NO difference. I would worry more about 10% balance between tubes (both idle current and the voltage needed to get there) than any 20% difference of opinion about "ideal" bias current.

Try it. Remove global feedback, use a nominal load, and reduce idle current to 10mA. Scope the output. Coming up from zero signal, you won't get much output, then it will leap to the several-watt level with an "S"-shape around the zero-crossings. This sounds bad.

Bring it up to 70mA per tube (for a short test). Now the zero-crossings are smooth, and drive is easy. What may not be obvious (and might not be true for EL34) is that the full-power peaks are smoothly rounded-down. That may even be a useful effect for guitar, though the short tube life at 30W dissipation may not be good for long or low-pay gigs.

But I bet you can't see any difference 30mA-50mA, and the audible differences may be very subtle. Slight shifts in the ratio of 3rd to 5th harmonic at different levels. (This may be masked by driver distortion, which is often not much less than output distortion but varies in different ways.)

Here are some tips from the trenches from a guy (me) who used to do nothing but fix guitar amps all day. The following applies to push-pull output sections with fixed bias.

I strongly recommend inserting 1-ohm flameproof resistors in series between the cathodes and ground. The resistors needn't be high-wattage, but they really ought to be flameproof. The voltage dropped across each resistor corresponds directly to the cathode current of the tube. This is the best and safest way to measure it.

Aiming for 50% to 70% of design-center max plate dissipation, at idle, will get you in the ballpark. But you have to go deeper if you really want to find the optimum bias point. Here's the method I used. I would connect a distortion analyzer across my dummy load and look at both the output waveform AND the distortion product (from the analyzer output) on a dual-trace scope. The scope was triggered off the output waveform.

I'd crank the bias to maximum, inject a 1kHz sine and adjust the amplifier output to 1 watt. Superimposing the distortion product on the output waveform, I could see distinct "spikes" falling at the zero crossings of the output waveform; I could also see visible crossover distortion in the output waveform itself. I'd reduce the bias slowly until the crossover distortion disappeared. But at that point, there'd still be spikes visible in the distortion product, even though the output waveform looked fine "to the naked eye." So I'd continue reducing the bias JUST until the spikes disappeared. Then I'd check my plate dissipation and make sure I was within reasonable limits.

I'd like to go into an explanation of WHY I used this method, and not one of the more common methods of bias-setting, but my dinner is ready and I don't want to eat cold food tonight, so I'm signing off for now...

If you don't mind drilling a hole in the back of the amp, sub in a cermet pot you can put a knob on. Move the wiring over. Now you can adjust the bias by ear, which is how it should be done. Just make sure you can not get a bias point that causes meltdown. I played with this mod for many years and really liked it.

Different rooms sounded better with diferent bias settings. So at each room, I would tweak the bias to get the amp to sound right. Also, if you switch genres often, you can dial in whats best for what you are playing, which guitar, etc.