Regrouping (continued from 11/28)
Where are the computer amplifier’s parts values? Two events have forced me to do a major rethink. The first was two readers complaining that 32 watts was excessive for the application: too much wattage for background music from loudspeakers close enough to touch. I can readily see their point—if all 32 watts were engaged at once, but my view was that the average wattage into the loudspeakers would be closer to one third of a watt and the rest would serve as backup for crescendos. The second event was more troubling: I could not get the distortion down with the limited amount of open-loop gain and, thus, limited feedback available. The tubes are not particularly linear to start with and the high-voltage and the relatively low idle current conspire to produce excessive distortion.

I used every trick I could think of and I even tried a cross-coupled driver stage topology; alas, the distortion remained too high for my taste and I didn’t want to up the number of input and driver tubes.

Click on image to see a more workable schematic

Solving the first “problem,” too much output wattage, by reducing the number of output tubes will only make the second problem worse, if the same 32-ohm load is specified. On the other hand, solving the second problem, too much distortion, by decreasing the B+ voltage and upping the idle current will both reduce the output wattage and lower the distortion. The combined plate curves makes the point:

Notice how much more linear the bottom set of curves is. By increasing the idle current (80mA vs 40mA) we have moved the amplifier’s output stage further into class-A mode. The amplifier is still a class-AB amplifier, but a much-more-heavily-biased-into-class-A amplifier. Additionally, the output stage becomes safer, as the 6AS7 is known to arc at high voltages, halving the B+ voltage probably quarters the likelihood of a mishap.

All of this sounds great, but what did we lose in the bargain? The potential output wattage has been reduced to a sixth of its old value (5 watts, instead of 30 watts) and the input and driver stages lose potential B+ voltage, as their power supply is parasitically derived from the output stage’s power supply. The danger with losing input and driver stage B+ voltage lies in potentially adding distortion from the frontend to the backend. In other words, we might have linearized the output stage at the expense of losing linearity in the input and driver stages because of the new lower voltages. But we may still come out ahead, as the new configuration requires only a third of the driver swing as before, so the effective feedback ratio of the driver stage has also been increased.

How could we cheat the system and retain the high output wattage and reduce the distortion? The only answer would be to do what the solid-state boys do: pack as much gain into the amplifier as we can, so that more feedback can be applied. This would require more input and driver tubes. it would also require a great more caution and tweaking than any computer amplifier deserves.

A little more testing and the values will finally arrive.