The Power Output Stage Pt. 1
The output stage of the ST-70 harkens back to Hafler's earliest days at Dynaco. His first output transformer -- the A-430, as used in the inaugural MK II amplifier -- was specifically designed for use with EL34 tubes. With a 4300 ohm primary, and screen taps placed at 33% of the primary winding, this load and screen tap position was experimentally determined to produce the greatest linearity from these tubes, coupled with a good degree of load regulation. The A-470 output transformer, as used in the ST-70, has the same exact specifications, at a somewhat lower power rating. Unlike the original A-430 transformer, the A-470 also includes a 4 ohm output tap as well -- this to accommodate the 4 ohm speakers (notably AR) that were starting to appear at the time.
So the ST-70 operates the output tubes with the same loading conditions as Hafler's original amplifier does, but operates them at a lower B+ level (~410 vdc vrs ~450 vdc) to achieve a lower power output level. Which brings us to the first myth about the ST-70's output stage: Lowering the quiescent current level (i.e, Biaset voltage) to extend tube life. From a standpoint of absolutes, lowering current levels will extend tube life. But in doing so in the ST-70, what is the practical return?
A common battle cry for this practice comes from those who categorically say something like: "Because tubes were so cheap back then, manufacturers ran the tubes right to the ragged edge for the performance ratings wars, and didn't care about tube life". No doubt in some cases that's true. But that line is said so often, that the statement has nearly become fact for all cases, regardless of what the individual facts actually are. So what about in the ST-70?
The 15.6 ohm cathode resistor causes the output stage to operate at 100 ma total cathode current in each channel, when the Biaset voltage is set for 1.56 volts as instructed. With a matched tube set, this is 50 ma per tube, with about 45 ma of it conducted by the plate. With 410 volts across the plate and cathode then, this equates to 18.45 watts, which as a percentage of the very conservative Design Center 25 watt plate dissipation rating for these tubes, equates to operating them at only 73.8% of their rated dissipation level. Even when operating the amplifier from a 122 volt line and still using the recommended Biaset level, the dissipation only climbs to 19.4 watts, for 77.6% of rated capability. These are both very conservative levels to operate these tubes at -- particularly when you consider it is being judged against the Design Center Rating value. The more common Design Maximum Rating System that was in place at the end of the tube production era rates the EL34 as a 30 watt tube, meaning the tubes would only be operating at 61.5% of capability (64.7% at high line voltage) under that rating system. With such conservative operation of tubes that are designed to conduct a continuous 165 ma each (Design Maximum rating) -- both from a dissipation and current flow standpoint -- backing the quiescent current level down by 10 ma per tube to save on tube life will gain you precious little of what you are trying to achieve, but it will give you a whole bunch more distortion!
If a THD test set is connected to the ST-70 and each channel is individually driven with a 1 kHz signal to 1 db below the rated power of 35 watts (about 27 watts), it will be found that the lowest distortion operation point is achieved with a quiescent Biaset setting of -- you guessed it -- 1.56 vdc, or veeery close to that setting. If you go below OR above this setting, distortion will climb, and do so rather rapidly.
For example, with a Biaset voltage of 1.40 vdc (45 ma total current flow per tube), 1 kHz THD climbs nearly 55%. If you reduce it further to 1.25 vdc (40 ma per tube) it climbs another 170%, for a total of 225%!
So, reducing the current draw of each tube by 6% of its rated cathode current, when it is already operating at just 30% of rated cathode current and 62% of rated plate dissipation, all to gain what, 225% more distortion? That is a very poor compromise to make, versus the very well thought out operating conditions that Hafler set up for the tubes to operate at.
But what about using different tubes? The 6L6 in many of its various forms, manufacturers, and sources, is a popular alternate for the ST-70. The pin out is the same, but requires slight alterations to the bias supply to bias the tubes at a reasonable value.
Assuming that appropriate bias supply changes were made then, using the popular 6N3C-E Russian derivatives will both reduce power, and increase distortion, due to the less than ideal loading conditions offered for these tubes. Maximum power drops to 24 watts RMS per channel at 1 kHz, with both channels driven, while THD rose to .5%. This represents a 36% loss of power, coupled with a 61% increase in distortion as compared to sockets stuffed with EL34s under the same conditions.
6L6 class tubes also had a minimum distortion operating point of around 50 ma per tube as well. But at this level, these tubes are operating at nearly 90% of rated plate dissipation capacity, and much more of rated cathode current capacity as well. Additionally, high power 20 kHz sine waveforms take on somewhat of a triangle waveform at this frequency, with THD running nearly 6% at 24 watts output -- and this at the optimum 50 ma current setting. At 40 ma, distortion at this frequency is nearly 10%.
There are all manner of accolades available in abundance about the sound produced from using 6L6 class tubes, and reduced quiescent current operation as well. I would never suggest what someone does or does not enjoy in their sonic quests, but I can say that given the performance intentions that Hafler established for the ST-70, using 6L6 class tubes, and/or reduced quiescent current levels in place the the designated tubes/setting, will cause it to fall far short of that mark.
The only pic to show is that of 20 kHz sine waves at the onset of clipping being squeezed into somewhat of a triangle formation at elevated power levels when using 6N3C-E tubes. Distortion of this waveform is ~5.5%. This is invariably due to the lower than optimum loading conditions being offered to 6L6 class tubes, which are very much happier when employing a 6600 ohm load, when operating in UL mode.
In part 2, alternate modes of operating the output stage will be examined.
Dave