Epilog
What a wonderful journey this has been! Class AB2 designs are rare indeed, but to be able to work with one under the Fisher Flag has been a real treat. It's been fascinating to unravel all the design hoops that Fisher had to jump through to produce a practical Class AB2 design for commercial hifi use.
In a way though, it was already a relic before the first one rolled off the production line: Class AB2 designs were primarily a tool for high noise level industrial applications where high power and high efficiency were required, but high fidelity definitely was not. Think large industrial plants, stadium, or ship-at-sea applications. Simple in concept, but not so simple to pull off in reality, they are really an amplifier driving an amplifier when the afterburners of turbo-Class AB2 operation kicks in. They are heavy and associated with high distortion levels -- but ingenuity at its best none the less.
Because it is difficult to produce a high quality AB2 amplifier, they were more than rare for home use -- with the Fisher 50-A platform being the ONLY AB2 design I am aware of for home high fidelity use. But at the time it was on the drawing board, it was the only practical way to get any significant amount of power out of triodes, which at the time, was still the best way to achieve a low output impedance for driving a loudspeaker. The idea of 20+ db of NFB that would allow pentodes to achieve such low impedance levels was in its infancy, with the early designs employing that much FB generally being more unstable than stable. UL was brand spankin' new, still unproven -- and a patented concept. Against that backdrop, Fisher went with tried and true technology of the 40's: Triode operation through out, and Class AB2 operation to do an end run around everybody else power wise. They dressed it up with a good power supply, phase inverter, OPT, and a better driver stage/transformer than any industrial designed ever hoped to have, to coax some good performance from it for its day. It was a gutsy move to say the least at a time when AB2 operation was widely considered as producing low brow performance. By the end of it's run however, the seeds of UL and higher power pentodes planted at the time of the 50-A's introduction were in full bloom: UL operation was thoroughly vetted as capable of producing triode performance with pentode power, and double the power output with greater quality could be achieved from a single pair of tubes -- without the heavier weight and complications of Class AB2 designs. Fisher made a good run with it, but by the 2nd half of the 50's, the 50-A's platform had run its course. Today however, these amplifiers continue to enjoy a huge following -- but beyond simply being a Fisher, why?
These amplifiers are not the kind you sit down and discuss performance specifications with over a tall cold one. If that were the criteria, there are many other designs that can easily beat it in any performance category you could imagine. In fact, I dare say most owners of these amplifiers could care less about anything but the most general of specifications like, how much does it weigh, and will it operate safely on today's line voltages! As far as any other specifications, shut up and listen to the music! When you listen to these amplifiers, its hard to argue with that logic. There are reasons for that.
From a design standpoint, the biggest sonic contributor to this series of amplifiers is that they are based on an open loop design that targets good response, low output impedance, and minimum distortion without the need for large amounts of corrective overall (global) NFB. To that point:
1. Plate load resistors are keep low in value to maximize good HF response.
2. Maximum use of local NFB is employed via unbypassed cathode resistors.
3. Direct coupling is used where practical.
4. A triode output stage is employed.
5. A (relatively) high quality driver and high quality OPT are used.
6. A stiff choke input power supply is employed.
To this recipe, Fisher added a meager amount of global NFB -- just 8 db in this example! (with more frequency poles than you can keep track of in the design, it really can't employ much more and still remain so stable)
This approach produces a high degree of stability (square waves don't ring, and pulses settle very quickly), a very flat frequency response (within .1 db across the audio spectrum), reasonable distortion (<2% at 1 kHz at 50 watts), and reasonable output impedance (~1.5 ohms) -- but also, it produces a very large, open, and and full sounding sound stage. Realizing that ANY electronically reproduced sound will sound smaller than real life, then any amplifier that causes the sound to appear larger than others do will present a more realistic and favorable listening experience for many people.
Global NFB improves measured performance significantly, but unchecked, can also make for a more clinical or sterile sound that can be un-engaging. This design however is anything but that. It is in fact extremely musical and listenable for long periods of time without fatigue. Fisher thread the needle very well with this design -- even if they were forced to. The specifications are good enough to ensure a reasonable degree of accuracy, while not sucking the life out of the presentation.
So what do the modifications do to the sound? Well, nothing -- until you light the fuse. Low and medium power presentations are unchanged from before: Big open presentation with plenty of detail, but without any sense of restraint. However, high power levels are now produced with an improved level of effortlessness. Is that because of the knowledge that the driver stage can now drive the output tubes right out of their sockets before giving up the ghost? Maybe -- but anytime effective driver impedance can be lowered during class AB2 operation, performance improves: power goes up, and distortion is reduced, which is exactly what happened.
In this case, the amplifier can now deliver greater than 55 watts RMS at ~ .70% THD at 1 kHz, which is a significant improvement over the stock performance (1.8% at 50 watts) -- and this is actually produced with a slight decrease in NFB. Since the lowest distortion operating point of the output stage is just 120 mA, it can now operate at that level without fear of outstripping the driver stage capabilities (lower quiescent current = higher grid bias voltage for the driver stage to overcome). The stock quiescent current level of 180 mA (requiring less grid bias voltage) was required to keep the drive requirements of the 6550 output stage within the limits of the driver stage.
But in lowering the quiescent current level of the output stage, the effective gain of the stage is also reduced, meaning open loop gain is reduced, meaning that the global NFB level is also reduced when the loop is closed. The amplifier can still be operated at a quiescent level of 180 mA of course. But because of the improved linearity of the pre-driver stage, and the lower effective drive impedance presented to the output stage during Class AB2 conditions, distortion actually increases now above a 120 mA setting, due to moving the output stage away from its low distortion operating point. In this case, the driver stage impedance itself wasn't lowered, but because the bias regulator -- whose impedance WAS lowered -- is in series with the driver stage as far as the output stage is concerned -- the effect is just the same: Effective driver impedance is lower, leading to improved high power performance.
Pics include:
1. A 1 kHz sine wave at ~ 47 watts RMS with the stock design, and output stage quiescent current set at 180 mA. Maximum power output is 50 watts RMS. But at 47 watts, you can see crossover distortion beginning to set in from the driver section already crying uncle before rated power output is reached. This is with the unit operating from a 121 vac line.
2. The same waveform under the same conditions at 56 watts RMS with the new bias regulator and pre-driver cathode resistor in place, and the output stage set for 120 mA. Even though the driver is delivering more power into the output stage under these conditions to produce greater power output, there is no sign of driver distress at the onset of output stage clipping. Before, power output was limited by the power the driver could deliver into the output stage. Now, the output stage is the limiting factor, which makes for improved Class AB2 performance.
3. The lower quiescent current of 120 mA does upset accurate power indications, but heck, even that could be dealt with if an addition bias current of 60 ma from a separate internal source is applied to the meter to fool it into submission. But that's another discussion for another time.
4. In good company with a modified (but naked) Fisher 30C driving this bad boy. It's been running flawlessly for about 8 hours in this pic.
Turning the unit on now no longer leaves me cringing until the circuit starts to draw current to draw down the previously very high turn on voltage surge. In fact, start up is boringly unspectacular, even with the new lower quiescent current level: Turn it on, 32 seconds later, the relay kicks in, the rectifiers heat, and the output tube B+ ramps up to 445 vdc -- all quite noiselessly and uneventfully. The same 3A fast blow fuse holds up well at the new higher power output levels that the unit is now capable of, while also powering the delay relay and timer as well. If the power burps, the rectifiers immediately cool, and then automatically reengage after 32 seconds.
So, this project comes to an end. I should have a rough schematic of the modifications installed posted by tomorrow morning. If one of you computer capable folks want to work it into the original schematic and re-post it, so much the better.
It's been quite a ride for me -- and what a bonus to listen to it as well!!
Dave