Output Stage:

This was the object focused on the previous revision of this report.

Number of Tubes and Bias Point

Initial investigations were carried out with 9 tubes per channel, with the power supply and system scaled to allow expansion to 12 or so tubes per channel. What I determined was that you could get a fairly high power output, at reasonably low distortion, but the overall amplifier didn't sound very good. A smaller number of tubes produced lower power output, but could be biased over a different operating point that allowed the sound to be improved greatly. The current implementation uses 6 6080 tubes (12 sections) in parallel in inverted mode for the output. At the chosen bias point, this allows slightly over 2 watts to be delivered and it sounds good. I'm still working to make it sound as delicate as the driver stage was. The chosen bias points are: -23 volts on the cathodes, about -70 volts bias (-93 volts considering where the cathode sits), 0 volts on the "plates" directly coupled to the speaker. The plate load is a 45 ohm 100 watt resistor to a 49 volt supply. Interestingly enough, the measured output resistance of the 6 paralleled tubes is actually lower than the output resistance of the 9 tube version, due primarily to the changed bias point. The current + supply did not have enough range of adjustment to allow 9 tube operation at the same bias point as the 6 tube point (or rather, I would need to lower the plate load resistor so much that the AC power lost in the plate resistor started to become significant; so no benefit in power output was achieved. The amp currently has 1 GE, 2 Svetlana, 3 RCA, 6 Sylvania devices in the power section. These were chosen for match, not brand.

Heat

I found the plate load and bank of 6080s was getting too hot for comfort, so I ended up moving the plate load resistors to their own sub chassis. This distributed the heat better without otherwise affecting anything. Thus, in the picture on the main report page,the "open space" behind the high voltage PS now is occupied by the load resistors. Also, the middle "row" of 6080s have been removed.

Power Sequencing

I took some care to work this through. Initially a set of relays shorts the outputs to ground. When power is applied, the bias voltage stabilizes relatively quickly (within seconds). The +49 volts comes on, but is loaded into the 45 ohm resistors, since the output is shorted to ground. The -23 volt supply is regulated, and the regulator is set so that it takes about 30 seconds for the -23 volts to come up. This provides time for the 6080s to warm up without drawing any current, which could otherwise potentially harm the cathodes. If the bias supply becomes bad, the -23 volts is never applied, as the reference for the -23 volt regulator is the bias supply. The relay "unshorts" the output about 30 seconds after the -23 volts and +49 volts are present; providing an overall "startup" time of 60 seconds or so. If either the -23 volts or the +49 volts is not present, the relay will not unshort. In this way, the tubes as well as the speakers are protected against failure conditions. If one or 2 or 3 of the output tubes fail, there is still sufficient current draw to prevent the DC sent to the speaker from harming it. If any of the tubes short, the -23 volt supply will drop sufficiently to cause the relay to de-energize, shorting the output to ground (and blowing the fuse, since we postulated the cause as a shorted output tube, which would short the -23 to ground.)

There is a 2 color 3 leaded LED present to give an indication of the power sequencing: when first turned on the RED light indicates power is present. After the -23 becomes available, the GREEN light is turned on, providing a "yellow" composite color. After the relays energize, the RED is turned OFF, providing a green indication.

Performance Tests

See the "What's it sound like" section for sonic impressions. The distortion measures as almost all second order, until you get right to the clip point. It sounds much better than the 4+% second order distortion would seem to indicate. The distortion is almost independent of frequency at all power levels. For instance, at 1 kHz, the distortion at 1 watt measured 4.4% second and 0.12% third harmonic. At 20 Hz, these numbers were 4.4% and 0.12%, and at 20 kHz they were 4.2% and 0.09%. The second sonic audition preferred this distortion to a rebiasing of the input stage that allowed the composite to have about 2% distortion (but higher third order relatively). See the second sonic audition results.