| I am one of many who built the PASS A40 power amplifier. Operating Class A at
40W/channel into 8 ohm, it is a fine amplifier. However, I wondered if similar performance
could be achieved with a MOSFET output stage. I was finally motivated to tackle such a
design by the failure of a second Darlington output device in my A40. Several
enhancements to the original design which I thought would be worth adding included
response down to DC, and achieving a suitably high-input impedance without the use of
"bootstrapping." I had previously designed 100W Class A monoblocks which met
these criteria by use of JFETs and MOSFETs. I constructed four of these amps, and they
proved over the course of two years to be reliable performers despite woefully inadequate
heatsinking.
The output stage was composed of 12 250W bipolar devices in common base driven by 12
medium-power MOSFETs in common source. This cascoded output displayed absolutely no
tendency toward thermal runaway even though bipolar case temperatures were as high as
75øC.
Nelson was kind enough to test the monoblocks for me, which tested well on the bench,
but he thought they were not particularly "dynamic" in listening tests. He
thought this was similar to the situation he had experienced with the Threshold amplifiers
using cascoded output stages. For this reason, I elected not to cascode the output stage
of the A40 successor. While some doubts about the sonic qualities of cascoded output
stages may be valid, no one has raised an issue regarding cascoded front ends. Nelson
employs similar cascode methods in the front end of his own amps.
As I set about the task of designing a suitable front end, I became more and more
determined not to add another gain stage to boost the original 7mA to the higher level
needed to drive the capacitance of the output MOSFETs at high frequencies. I found the
answer to achieving the needed drive current in only two stages in the form of
International Rectifier's low-power MOSFETs in the IRFD series. By using these devices in
the second (driver) stage, I obtained the higher slew rate. Because the Darlington output
stage of the A40 is really two stages, I had produced an A40 equivalent with only three
stages instead of the A40's four.
The output stage used two N- and two P-channel power MOSFETs, also from the IRF line.
Measured slew rate was over 100V/uS, high frequency distortion was less than 0.1% at rated
output, and a single capacitor of a few picofarads in so-called "phantom zero"
configuration was sufficient to compensate the closed loop amplifier.
As I had done with the earlier monoblocks, I powered the front ends with a separate
regulated power supply whose voltages were about 5V higher than that of the unregulated
{but well filtered) main supply. This regulated supply was derived from the main supply
via a conventional half-wave voltage doubler.
The amplifier's sound was remarkable, with a smooth high end and a very dynamic low
end. Its most vivid characteristic, however, was its low-level detail including ambience
cues, which gave a sense of "being there" better than any other amp I had heard.
I knew Nelson was committed to updating the A40, so I immediately wrote to him
suggesting that the design be offered in satisfaction of his update obligation. He liked
many features of the design, but had some ideas of his own for improvement.
His first suggestion was to use a "folded" cascode. This would reduce the
front end, in effect, to a single compound stage and the total amp to a two-stage affair.
This had obvious appeal, but a true folded cascode would restrict the available drive
current to that of the input stage. The answer was then to add a follower stage which
could supply the required 20mA. In the end, we used a modified folded cascode which could
produce the needed drive current by itself. We deleted the follower.
Current source loads replaced the resistive loads of the differential amp for a time,
but Nelson was not satisfied with their performance. Upon returning to resistive loads, he
used fixed and variable resistors in a scheme which allowed precise adjustment of the
driver bias current. Although Nelson is no advocate of dual differential designs, he left
that part of the original proposal undisturbed.
The final topology, presented here, is capable of superior performance. As I look at
it, I can see enough of my original design to feel comfortable that I have contributed
significantly to the project. It also has Nelson's indelible stamp on it. For his many
fans (of which I am one), that is welcome news indeed. |
