Fisher's Original Bad Boy: The 50A Power Amplifier

dcgillespie

Fisher SA-100 Clone
Subscriber
When the 50A showed up on the 1952 audio scene, everybody took note. In one fell swoop, Avery both broke long established rules, and could arguably be credited with starting the first High Fidelity power war. Frank McIntosh had recently introduced the world to his innovative 50 watt Unity Coupled amplifier design, when along comes Avery with his 40 watt 50A, leaving everybody else trying to catch up.

But Frank's design used a pentode output stage -- which was well known to not be the way to achieve audio nervana with the high fidelity crowds of 1952. Against that, Avery's design used a triode output stage that was producing 4 times the power that everybody else was getting out of their triode amplifiers. With triodes and high power, Avery had the wind at his back, whereas Frank's pentode design faced a head wind.

And what's not to like? Big beefly transformers, lots of tubes, and a chassis layout with a terminal board and military grade wiring harness that just screams "can't touch this"! The kind of stuff that Tim Taylor would drool over. The thing is simply beautiful to look at when operating, even if it's not producing a sound. When it is producing sound, it produces that immediate if almost patented midrange clarity and presence that the model is famous for.

Built like a tank, the design of the 50A was certainly innovative for its day: Dual heavy duty rectifier tubes for a 40 watt amplifier with a real choke input power supply gives the 50A a level of B+ regulation like no other in its power class: The B+ supply drops a mere 20 volts from zero to full sustained power output -- which it can do all day long if required. Compare that to a popular single rectifier tube 120 watt amplifier of today. It uses the popular Williamson front end with a single preamp stage tacked on in front of it, and a driver stage whose tubes glow blue, and can easily drive a speaker all by itself. Think of the snob appeal: I mean after all, does your amplifier have driver tubes that glow blue? :) Global NFB that includes all but the preamp stage rounded out the design. It was truly a unique effort that includes 6 glorious stages of vacuum tube amplification (three of which are push-pull), the likes of which was never seen before -- or since. Compare that to those today of the less is better crowd, who would (if they could) have their CD player directly drive a SE DHT output stage!

Avery's secret to getting all the extra power output was that the beefy driver stage allowed the ability to drive the output stage into Class AB2, which has the effect of strapping afterburners on to a conventional triode output stage. When Class AB2 commences, power output rises significantly beyond the normal power levels achievable. Now just as Frank faced a head wind due to his use of pentodes, Avery also would have faced a stiff headwind were the fact of his amplifier operating in Class AB2 widely known, as that mode of operation has never been equated with high fidelity performance (i.e., low distortion) either -- unless it is designed very carefully. The good news is that it's precisely because Avery did design it very carefully with the heavy duty power supply and driver stage, that the amplifier does in fact perform very admirably for high fidelity service. And, because Avery was ever the marketer, he chose not to advertise the fact that his amplifier used a mode of operation generally reserved for a steelyard PA system but rather, heralded the 50A as the first home high power high fidelity all triode amplifier, since he knew that the tecknoids of the day would all be drawn in by the well known siren call of an all triode amplifier, and likely had little knowledge of Class AB2 operation, or the design concerns surrounding it anyway. Smart move.

My current client has sent me two of these beasts for electrical restoration and performance assessment -- one being an early A, and the other a somewhat later AZ. In assessing the results of my archeological dig into both units, it appears that my work represents the third layer of recapping and service these two amplifiers have seen -- though its clear they never knew each other in their former life. In any event, two layers of some rather junky work had to be removed from both units before any serious restoration work could begin. As well, the AZ had clearly been either the recipient of some mad professor's efforts to address some of the design's deficiencies, or the result of junior trying to make a guitar amplifier out of it after dad updated his system. I mean, this thing had a completely different OPT installed on it at one time (looks like Acro), and has various holes drilled all over the chassis to accommodate (apparently) the various different circuits that were tried out in it. What was left of the original beautiful Fisher wiring was sad.

The well known voltage surge of these amplifiers at turn-on is now taken care of with the use of modern 5V4GA rectifier tubes (the original 5V4G tube had very poor dependability), and can caps from Hayseed Hamfest specified at 600 vdc for the first can, and 550 vdc for the rest. The extra safety margin on the can caps is really icing on the cake, as between the 5V4GA rectifier tubes and a CL-80 current limiter, there is no voltage over-surge at turn-on.

In terms of the actual project, I'm rounding third for home, with a lot of mop-up details to attend to, as now, both amplifiers have finally been stirred from their extended multi-decade hibernation (both still had all original can caps installed dating from mid '51 to mid '54, and anywhere in between), and had a visit to the listening room for the first time today. Along the way, I've developed a lot of info that the other 50A fanatics out there might find useful, and can offer some options for improved performance along the way as well. For now however, a few pics are offered. The rest will be added to the thread as the details become finalized.

Dave

As received, with neither apparently having received any love in a very long time.......
SAM_2307.JPG

Underside of the AZ. Apparently, after all the Frankenstein experiments were done, it was hap-hazzardly "restored" back to the original circuit at some point (although not correctly), with of course much the beautiful Fisher wiring basically trashed.
SAM_2300.JPG

Underside of the A, in progress.
SAM_2321.JPG

In the listening room operating together for the first time.......
SAM_2332.JPG

My 202-T is enjoying some time with Fisher equipment older than it is........
SAM_2333.JPG
 
Wow, beautiful amps and nice work! I didn't know much of anything about the 50A, interesting to read about them. Does the 50A share much with the later 200A?
 
Al -- It's a McIntosh cabinet. Matt does superb work. Having one built was about the only way I was ever going to find one for this unit in my lifetime..........

Jail -- The 200 was the last iteration of the original 50A design. While the two designs were near identical in many ways, there were certainly some differences too, in that:

1. The preamp stage was eliminated.

2. The 12AU7 AF Amplifier/Phase Inverter tube was changed to a 12AX7 -- with the added gain this provided necessary to compensate for the loss of the preamp stage.

3. The bias regulator 12AU7 tube section was electrically stiffened by connecting the other half of this tube (originally used as the preamp stage) in parallel with it, which made for an effective improvement in bias regulation during Class AB2 operation.

4. The unit now used GZ34 rectifier tubes. The 5V4GA had already been released as an excellent replacement for the problematic 5V4G, but since the GZ34 was still superior yet and available (it had been used in some versions of the 55A as well), Fisher never again returned to using the newer 5V4 after having so many problems with the original G version of the tube.

5. Output tubes were now EL34s (also used in some versions of the 55A), but this was maining a marketing choice, as the EL34 was not really a good choice for Class AB2 operation. Fisher therefore took measures (in the 55A as well) to significantly limit how far into Class AB2 the amplifier could be driven with these tubes -- all in the name of protecting them from melt down. Using the EL34 however was apparently important to Fisher, and indicates he had a read on the fickle portion of the audiophile crowd of that day as well.....

6. The biggest difference with the 200 is that the output stage no longer operates in triode mode, but in pentode mode. The inherent power increase this mode of operation provides means little class AB2 operation is really required to produce a slight power increase over that of the 55 watt 55A series. The 200 is rated for 60 watts.

Vendo -- The output tubes used/shown are yet another variation of the 6L6GC family, the 6BG6GA. This tube has 6L6GC guts in it, but brings the plate connection out to a plate cap terminal. Hence, the tube must be used with an adapter that provides for the plate cap connection.

Dave
 
Very cool. Good thing the AZ's original transformer wasn't lost so it can at least be put back right.

original tubes in this were what, 6L6GC then? Or were these a 1614?
 
OPPORTUNITIES

For all of its high flying design/build qualities, opportunities exist to put a polish on the 50A that brings it to new levels of performance and capability in the modern audio age. None are particularly complicated, yet all are effective in result. One is brought over from previous development work on a 55A, one is brand new, and one -- its need depending on whether you execute one of the modifications -- corrects a rather bone headed build mistake on Fisher's part. But first, some mop-up comments regarding the power supply and related topics.

1. For typical AC line voltages of today of around 121-122 vac, a single CL80 should be installed between the AC receptacle and fuse post to minimize the turn on current surge from all the cold heaters. So installed and with today's line voltages as indicated, the heater voltage at the output tube socket terminals is about 6.45 vac which is excellent for performance and tube life.

2. Regardless of what rectifier tube type is inked onto the chassis, the optimum rectifier tube to use when using any of the 6L6 family of tubes is the 5V4GA. This tube will allow maximum usable B+ voltage without over-dissipating the 30 watt versions of the 6L6 family of tubes, provides excellent regulation, eliminates the turn on surge, and is extremely dependable.

3. If GZ34 rectifier tubes are used, then the amplifier is capable of supporting and maximizing performance from output tubes as large as KT-120 tubes.

4. When replacing the selenium bias rectifier with silicon, do not add any extra resistance for compensation.

5. With the scenario described in #1&2 above, B+ at the output of the choke is typically 440-450 vdc when the output tubes are drawing the optimum quiescent level of current. With the 6L6 family of tubes, this will be a cathode current 120 mA. Screen grid current under this condition is < 2 mA per tube, and plate dissipation is just under 26.0 watts per tube. 30 watt versions of the tube take this in stride, and show no color in the plate, although the output tubes do need to be well matched. Power output will reliably be 45-50 watts RMS at low and midband frequencies. Power output does drop off slightly at high frequencies due to the winding capacitance within the output transformer.

INITIAL MODIFICATIONS

1. Ground Loop Elimination: If you do nothing to these amplifiers, do this: At V1, pin #6 is grounded to the same ground tag (located at the V1socket front mounting rivet), that the cathode resistor from pin #3, and the ground terminal of the input jack are grounded to. The problem with this is that pin #6 of this tube is the plate element of the section acting as the bias voltage regulator. Because the regulator is passing about 5 mA even under quiescent conditions, and is fed by a half-wave rectifier, it injects the noise of this half-wave current flow into the audio ground of the input tube -- the most sensitive ground point in the amplifier. When the level control is advanced, the noise is minimized. But when the level control is turned full down or nearly so, the buzzing sound of half-wave rectification comes through loud and clear, and particularly so when high sensitivity speakers are used. The answer is to remove the ground connection to pin #6 from that ground point, and then ground pin #6 to one of the ground tabs over at C-15 (bias filter cap). This corrects a well defined ground loop built into the unit from the factory. When corrected, the reduction in noise is immediate and quite obvious, making for a very black background. If you perform the Mosfet modification coming up, the ground loop issue will be corrected automatically during the course of executing that modification.


2. Mosfet Bias Regulator: This was developed during my work with the 55A, and since that unit uses the same bias regulator circuit as the 50A/AZ (the 55A is the successor of the 50AZ), that makes it a drop in fit for the 50A as well. In the 50A, the updated regulator allows the driver stage to develop less output for a given level of Class AB2 drive by working to reduce bias creep to an absolute minimum. This allows the driver stage to operate with less output and distortion, which is the stage where the most distortion is generated within the design. But there's an additional benefit as well.

Eliminating Class AB2 bias creep not only allows the driver stage to be more effective, but it also means that the output stage can idle at a notably lower quiescent current as well. Bias creep in a Class AB2 output stage is always in the direction of pushing the stage towards cut-off during AB2 operation, which increases distortion. To compensate for this, the quiescent current must be set higher so that when the creep commences, it creeps towards the lowest distortion operating point. But if the creep can be minimized, then no compensation is required. The quiescent current can be set lower, and the lowest distortion operating point is maintained at all power levels.

Now granted, a lower quiescent current means slightly more bias voltage for the driver to overcome. But the increase in bias voltage due to a lower quiescent current represents only a fraction of the bias voltage saved due to elimination of the stock design's bias creep. So even with the reduced quiescent current afforded by the new regulator, there is still a net reduction in required drive voltage from the driver stage. Between the reduced driver stage output, and holding the output stage operating point much more consistently at the lowest distortion operating point, the new mosfet regulator arrangement - all else being equal -- drops all distortion by about 33%.

It was mentioned that in the stock 50A with 6L6 family tubes, the lowest distortion quiescent current operating point will be found to be at 120 mA. However, with the Mosfet bias regulator installed, a new, and lower distortion operating point will then be found at just 95 mA of quiescent current, dropping the plate dissipation in the output tubes significantly, and making just as significant an impact on improving tube life as well.

So the new regulator lowers distortion, reduces output stage quiescent current, improves tube life, and promotes cooler overall operation. And, because the regulator's dynamic action doesn't even exist electronically below about 20 watts RMS of power output, it is also completely neutral with respect to any audible characteristics. There are simply no down sides to the mosfet bias regulator modification -- only benefits.


3. Negative Feedback and Stability: This is an aspect of the design that Fisher clearly struggled with, with four different versions of NFB circuits used through out the production life of the 50A/AZ. It is also what caused the Naval Research Laboratory in April of 1953 to write the following about the 50A when they tested several different popular premium power amplifiers of the day:

"The Fisher Model 50-A Amplifier

This amplifier (Figure 34) was somewhat of a disappointment inasmuch as about the
only manufacturer's advertised specification which was met by the amplifier tested was its
power rating (Figure 21). It will deliver the full 40 watts before any sharp increase in
distortion. Frequency response and transient response were not good at either the high or
the low frequencies."


Indeed! And there were/are in fact such problems with the design -- which will be addressed and resolved, starting next time.

Dave


Ground loop hum: The short lead grounding pin #6 from the stock bias regulator to the same point where the input jack, volume control, and cathode resistor of V1-A is grounded to is a text book recipe for generating ground loop hum. It does. Correct by removing this connection and grounding pin #6 over at a ground tab on the Bias Can Cap. Or much better yet:
SAM_2334.JPG

The new Mosfet Bias Regulator installed. The installation inherently corrects the ground loop hum noted, and provides all the benefits noted in the discussion above.
SAM_2335.JPG
 
Last edited:
NEGATIVE FEEDBACK AND LOW AND HIGH FREQUENCY TRANSIENT RESPONSE

Applying NFB over five stages of audio amplification that include two with full or partial transformer coupling, and three with full or partial R/C coupling, can certainly present design challenges at both ends of the audio spectrum if good frequency and transient response, and a reasonable level of stability are to be maintained under various conditions of loading. Based on the number of circuit versions known to exist (four), Fisher was clearly struggling in tying to achieve a good balance between these performance criteria, as it was changes in the NFB and stability related components (which primarily determine the performance achieved in these areas) that made up the principle changes between the various versions.

Now regardless of version, Fisher was always able to maintain a basic level of stability in the 50A, as none of the versions produce outright oscillation under typical loading conditions. Achieving good stability however is always at odds with achieving good frequency and transient response, with measures that improve the performance in one category resulting in a loss of performance in the other, and visa-versa. From the very beginning then, this frames the juggling act that Fisher was struggling with in the various versions, as will be seen.

The Naval Research Laboratory's tests appear to document the beginning of this struggle, as based on the schematic they provided for the 50A, the version they tested was in fact the original design employing 6S4 driver tubes. Since Fisher published extensive performance specifications for this version from the get-go, it made for fair game in verifying the performance claims made. Tests by the NRL's facilities would be about as fair and unbiased as it gets, as there would be no advertising connections to flavor any of the findings.

Since the principle changes between each version primarily involved changes to the NFB and stability circuits, it was a rather easy matter to test the relevant performance changes that each configuration produced with the units I have here. What follows then is the results produced from those tests, which ultimately give merit to the NRL's comments. The results and the NRL's comments however, are only relevant as judged against Fisher's published specifications. Therefore, best place to start then would be to take a look at just what those specifications are.

That presents some problems however, in that graphs and text provided in their advertising literature, doesn't always agree with each other, with the text always supporting a higher level of performance specification. I've provided a copy of these specifications as downloaded from the Yahoo Fisher Group, although it won't provide much more than giving you a general indication as to what the document looks like, as it certainly won't be legible here, since it's barely legible on the Yahoo site to begin with:

_50a2.jpg

In trying to make sense of the mumbo-jumbo, we can generally take away that the 50A is rated as a 40 watt amplifier, with the following additional information:

1. THD at 40 watts RMS is shown graphically to hover in the .50-.75% range 20Hz to 20 kHz (never falling below .50%) , except for rising slightly over 1% at 20 Hz. THD is stated in the text to be < .30% at 40 watts.

2. IMD is stated and graphically shown to be .8% at 40 watts equivalent RMS power output.

3. Frequency response is stated and graphically shown to be within 1 db to 100 kHz.

4. Power response is graphically shown to be flat from 15 Hz to 60 kHz at 40 watts, yet stated to be flat within this range within 1 db of 40 watts (or stated more accurately, "while producing at least 31 watts").

5. Sensitivity is graphically shown to require 1.13 vac to produce 40 watts output, but the text states a sensitivity of less than one volt input for full output".

6. Internal output impedance is stated to be .53Ω relative to the 16Ω tap.

7. Hum and noise is stated to be > 92 db below full output.

8. Other general specifications state that the design uses 20 db of NFB, and providing for completely stable feedback.

Between the graphs and text, there is a notable specification discrepancy for numbers 1, 4, and 5, with good agreement for numbers 3 and 4. Numbers 6-8 were drawn from and supported by text only.

Relative to the NRL's comments and my own measurements then, depending on which version is being discussed, all 8 performance criteria come into significant question over the life of the 50A, save for number 4, with all versions meeting the text specification.

So with that, let's get started.


Circuit Differences Between the 6S4 and 6CL6 Versions

The versions here were not so early as to employ 6S4 driver tubes, but one (the 50A -- Serial # 1782F) was clearly the version released immediately after those tubes were changed to the 6CL6. The change in tubes would have caused no significant change in performance, as the driver stage operates in a cathode follower configuration, with both tube types when so configured requiring basically identical bias voltage and drawing identical quiescent current. Because of the configuration and great similarity in tube characteristics then (when the 6CL6 is strapped as a triode), the two different tubes will produce basically identical performance in the circuit. But the 6S4 units also employed a different and notably smaller driver transformer as well. This too would not produce any significant change in performance, because both transformers employ a 1:1 turns ratio. But a change in transformers could change both the sub and supersonic characteristics of the amplifier before the NFB loop is closed.

On the supersonic end, a change in transformers would be minimized by the significant capacitive coupling used around both sides of the push-pull driver transformer, regardless of which transformer is installed. On the subsonic end however, a change was made to the time constant of the R/C network coupling the pre-driver and driver stages, almost certainly to accommodate the greater inductance that the larger driver transformer of the 6CL6 versions would have over the smaller one of the 6S4 versions. The increased inductance would make for improved LF response under a Class AB2 load, and so a change in the R/C time constant was made (basically cutting it in half) to account for the improved LF characteristics of the new transformer. Within the components used for the new R/C network, the change in driver tubes also required a change in the value of grid return resistance used for the driver stage as well -- this to account for the lower Maximum Grid #1 DC Resistance specification of the 6CL6 versus that of the 6S4. The driver stage coupling cap value was changed then not only to shorten the overall R/C time constant to account for the larger transformer, but also to account for the new lower grid return resistance employed to accommodate the 6CL6 tubes.

The change to 6CL6 tubes itself was almost surely initiated at the time by the diminished forecast for the future of the 6S4 tube. An A version of this tube came out in the mid 50's that included a controlled heater warmup characteristic to improve its performance when used in series connected heater string designs. But overall, the fate of power triodes had pretty much been cast at the time, and it was doomed to be discontinued as quickly as possible. Therefore, the change was made to a power pentode connected to operate as a triode, which filled the bill quite nicely.

Ultimately, regardless of version, LF stability is basically acceptable with all versions of the 50 series, with the latter two versions displaying the better performance in this regard, and the final version best overall.

With the difference in driver tubes and transformers largely accounted for then relative to this discussion, that leaves just the changes in the NFB and HF stability circuits to evaluate between the various versions, which without a doubt is where the lions share of differences -- both circuit and performance wise -- reside. And so we begin.


6S4 Version

Besides the basic NFB and stability circuits, this version also uniquely employs a 2200 pF resonant cap strapped across the full primary of the OPT, which was dropped in all later versions of the 50A. Adding such a cap produces a rather significant bump in supersonic response (as would be expected), producing (in this case) a +2.7db rise at 40 kHz (ref: 1 kHz = 0 db). Adding a cap in this location can be done for a variety of reasons, be it to smooth out the response of the OPT, prevent parasitic oscillation due to unequal characteristics between to two halves of the primary winding, or even to intentionally provide a HF "kick" to counteract HF circuit losses. Providing a boost in this manner can help to ensure that feedback at supersonic frequencies remains negative instead of positive, which helps to ensure adequate HF stability. Since the same OPT was used for all versions, then the probable thought for including the cap was to provide a HF kick, as the inherent HF response of the transformer is quite good, and there are no tendencies shown towards producing any parasitic oscillations under any condition of use.

The down side of such a cap however, is that it can significantly load down the output stage at high frequencies. While that can be good (as stated) for stability purposes, it can also be quite detrimental to any ability of producing low THD in the upper ranges of the audio spectrum. Consider that in the original release of this amplifier as published by Sams, THD at 20 kHz at just 12.5 watts output already reaches 5.0%! When the cap is removed, THD at 20 kHz drops like a rock. As a result, most manufacturers of high quality audio equipment worked to avoid using such caps whenever possible. At 1 kHz, THD at 40 watts came in at .30%, which is very close to Fisher's stated performance level at this frequency and power level.

In any event, the OPT resonant cap is part of the original design, so it was included along with the relevant NFB and HF stability components when the 6S4 version was mocked up and tested. The resulting amplifier is characterized by having absolute stability, although LF stability is a little slow to settle when loaded, and slows further when unloaded. The amplifier displayed a very slow rise time with 10 kHz square waves either loaded, or unloaded. It operates with a measured 20.4 db of NFB in the global loop (@ 1 kHz), and requires 1.30 vac to develop full power output. Frequency response in the supersonic region is quite poor, showing itself (Ref: 1 kHz) as:

+1.25 db @ 20 kHz.
+1.20 db @ 40 kHz.
- 6.50 db @ 60 Hz.
- 13.6 db @ 80 kHz.
- 19.3 db db @ 100 kHz.

Measured internal output impedance is nearly double the claimed value clocking in at just over 1.0Ω, THD at 20 kHz at 40 watts is off the chart since sine waves at that frequency look more triangular than sinusoidal, and hum and noise struggles to be even 80 db below 40 watts output.

The comments from the Naval Research Laboratory then hold true for the 6S4 version of the 50A as depicted by Sams. Whether there were other versions of the 6S4 design or not is unknown. Minimum THD in this design was produced with an output stage quiescent current of 120 mA, which holds true for all versions of the 50A in stock form. A scope shot of a 10 kHz square wave is provided as produced by this version when loaded with a 16Ω resistive load. The response produced at both the 8Ω and 16Ω taps is identical.

Below: The poor supersonic response and slow rise time of the 6S4 version can be seen in how the amplifier reproduces a 10 kHz square wave:
SAM_2348.JPG

For reference, here is the 6S4 version of the 50A, as presented by Sams:
Fisher 50A 6S4 ROTATED.jpg

End Part I
 

Attachments

  • SAM_2329.JPG
    SAM_2329.JPG
    74 KB · Views: 94
  • Fisher 50A First 6CL6 NFB:HF Stability.jpg
    Fisher 50A First 6CL6 NFB:HF Stability.jpg
    22.1 KB · Views: 78
Last edited:
Part II

First 6CL6 Version

This version of the amplifier switched from the originally problematic 5V4G rectifier tube, to the very dependable but quick warming 5AW4. As mentioned earlier, Fisher took a hard look at the driver transformer, and started using a new (much bigger) one with this version. The new driver transformer is at least twice as big as the original piece. Within the NFB and HF stability components, changes were made that produced a dramatic improvement in frequency response in the supersonic region (ref: 1 kHz):

+.25 db @ 20 kHz.
+.20 db @ 60 kHz.
+.60 db @ 100 kHz.

What a turn around! The NFB, sensitivity, and hum & noise levels remained the same as did the internal output impedance with that of the 6S4 version. But 20 kHz THD levels dropped to about 4.0% at 40 watts output, and the rise time of 10 kHz square wave forms improved notably as well. But along with the improvements came a commensurate reduction in stability. The amplifier was no longer absolutely stable, although could still tolerate cap only loads of as much as .05 uF, which is generally quite acceptable for most speaker loads. But a new stability problem developed as well: While the HF transient response of the amplifier had been greatly improved, the HF transient stability had now notably deteriorated: On 10 kHz square wave displays, the wave forms now showed noticeable ringing:
View attachment 1107924

The improvements discussed as well as the ringing noted above all resulted from very simple changes to the NFB and HF Stability circuits over that of the 6S4 design, as shown here:
50A 1st ver 6CL6 NFB:HF Stability CORRECTED.jpg

NOTE: Please disregard the depiction of these simple changes as shown as an attachment at the bottom of my previous post, as it contains an error. The depiction shown here is the correct one.

Overall, the first 6CL6 version is a very much improved amplifier relative to the original (and only) published specifications for it. It still did not meet published specifications for distortion, sensitivity, internal output impedance, or hum & noise. But it was moving in the right direction. Fisher wasn't done with it however, as more changes were to be made in future versions as Fisher sought to round out the performance of the amplifier. We'll cover the last two versions, as well as how the two amplifiers here ended up -- next time.

Dave
 
Last edited:
Second 6CL6 Version

With the advent of the second 6CL6 version, the original straight 50A seems to have apparently all but been replaced by the 50AZ. The "Z-Matic" feature (variable damping control), was all the rage throughout the mid 50s with many manufacturers. With speaker design and understanding improving rapidly at the time however, the benefits of variable amplifier damping were all but gone by the end of the 50s. A few amplifiers of the very early 60s retained elements of this feature, but only to service those speaker systems from the prior decade that could benefit from it. Today, this feature is a complete anachronism, so the tests done on the NFB and HF stability circuits of the second 6CL6 version are all based on the Z-Matic feature being turned off, which effectively turns the unit into the second 6CL6 version of the 50A. The changes made that were discussed earlier regarding the LF response of the amplifier and introduced with the first 6CL6 version, remain unchanged throughout all 6CL6 versions of the 50 series amplifiers, and in fact follow the design into the succeeding 55 series of amplifiers.

So what changed with the second 6CL6 version? The biggest is that Fisher reduced the amount of NFB employed by 5 db, from 20.4 db -- as used in the original 6S4 and first 6CL6 versions -- down to now 15.4 db, with changes to the components controlling HF stability as well. This represents a major reduction in the amount of NFB used, and certainly changes the performance capability of the amplifier as well. The maximum power and power response of course did not change, but consider all that did as a result of the new simple changes applied to this version:

1. Sensitivity is increased so that now just .50 vac drives the amplifier to full power output.

2. General THD and IMD figures all increased by about 100%.

3. Internal output impedance also increased by at least 100%.

4. Hum and noise performance deteriorated due to the increased sensitivity.

5. The supersonic frequency response curve changed once again as follows (ref: 1 kHz):

+.40 db @ 20 kHz.
+2.0 db @ 60 kHz.
-2.5 db @ 100 kHz.

So the supersonic response now displays a notable rising characteristic peaking at 60 kHz before beginning its decline. But the amplifier still has acceptable stability, remaining conditionally stable as before by tolerating up to .05 uF as a cap only load on the output, and it retains the improved HF transient response from the first 6CL6 version as well. However, the changes in performance that the new NFB and Stability circuits in this version produced made a complete scrambled egg mess of any relationship the amplifier now had with the original published specifications for it, and likely gave rise to the beginning of Fisher's coverall statement of (something like): "Due to the efforts of Fisher Laboratories to continually improve its products, Fisher reserves the right to change their specifications at any time without notice...........".

So besides the addition of the Z-Matic feature to the second 6CL6 version, what was behind all the other performance changes noted? The answer likely goes back to the ringing 10 kHz square waveform of the first 6CL6 version. In my experience, it can add a harsh element to the sound, represents reduced transient stability, and certainly does not market well - something we know Avery liked to do and was very good at. So the big improvement of the second 6CL6 version? The 10 kHz square waves no longer ring, as shown here:
SAM_2327.JPG
The NFB and HF stability circuit changes of the second 6CL6 version are shown in a partial schematic at the end of this post.


Third 6CL6 (and final) Version

In the final version of the 50A(Z), Fisher mixed up the NFB circuit yet one more time, still trying to achieve that elusive goal of no ringing on square waves, absolute stability, and reasonable distortion levels in the upper portion of the audio spectrum -- all at the same time. The original 6S4 version didn't accomplish this, as the square waves didn't ring and the stability was absolute, but the THD at 20 kHz was astronomical. The first 6CL6 version didn't, as 20 kHz distortion was acceptable, but square wave ringing was significant and stability was conditional. The second 6CL6 version didn't, as while the 20 kHz THD was still good and the square waves didn't ring anymore, the stability was still conditional. So how did they do with the third 6CL6/final final version of the amplifier? The results are still mixed, as in the design was now absolutely stable, square waves don't ring, and 20 kHz THD was still quite reasonable, but other things previously resolved, now suffered once again to achieve this. More specifically, as with the original 6S4 design, supersonic frequency response was seriously compromised once more, as measurements of the final design reflects (Ref: 1 kHz):

1. -.80 db @ 20 kHz.
2. -6.0 db @ 60 kHz.
3. -13.0 db @ 100 kHz.

The negative impact this had on the HF transient response is notable, with 10 kHz square waves certainly looking more polite, but displaying only somewhat better transient response than where it started from in the original 6S4 design:
SAM_2328.JPG

Finally, sensitivity remained basically unchanged from the second version, but because this third version reduces the NFB level even further, distortion, hum and noise levels, and internal output impedance all rose even further above the levels displayed in the second version. Agreement with most published specifications was now basically limited to rated power output and power bandwidth -- all as originally stated by the Naval Research Laboratory in their summary of their testing results -- with most other performance parameters now deviating considerably from the published specifications.

The changes to the NFB and HF stability circuits for both the second and third 6CL6 versions are shown here:
50A- 3ed, 4th version NFB, HF Stability.jpg


Summary of Versions

In some ways, by the time the final version of the Model 50 amplifier was released, it had gone through enough changes in performance (both good and bad) to basically make a complete circle and come back upon its own performance levels established (actually produced) with/by the original 6S4 design. Along the way, other performance criteria either improved or deteriorated as the case may be, depending on which version is being discussed. None of them ever met all of the lofty performance standards that Fisher originally publicized for the model. Of them all however, the second 6CL6 version (third version overall with introduction of AZ model) likely represents the best mix of performance criteria, while the first and third 6CL6 versions can be converted to the second one easily enough if so desired. When looking for two units to achieve stereo operation, just be aware of at least the significant sensitivity levels that exist between the various versions produced.

As for the two Model 50 amplifiers here, after their basic electrical restorations, the pool of various versions offered by Fisher as presented here were analyzed, with the decision ultimately made to take a fresh look at trying to achieve the best overall balance of performance from the 50A(Z) amplifiers. That effort will be documented in the final chapter of this project, next time.

Dave
 
Last edited:
Hello everyone,

I would like to give my impressions of the 50a/59az that Dave has restored and fully modified for me.
First, I would like to thank Dave for the great job he did and for his expertise. Workmanship that is first rate.
I've not heard these amps before in their original state as they needed restoration when i got them so i cant give a before and after comments.
Now to how these bad boys sing. To my ears, and compared to some of the amps i have (Jeff Korneff 45, audio note p2se, rh84, icepower class d, lance Cochrane and jwn 6l6 amps) the fishers presents a more natural sound character. It is very easy to listen to but that doesn't mean its not dynamic. It is effortless as hell. It shows its dynamic prowess when its needed. Its Midrange is to die for. It sounds oh so real with a very good recording. It betters my single ended jeff korneff 45 and all the amps listed above. Highs that are oh so sweet and never rolled off. Oh, the bass. My oh my my, it is big, tight, and deep. Bass drums sounds like real bass drums. No kidding. It gets close to the real thing. Oh one more thing, its micro details is outstanding, im hearing details especially in the bass and high frequencies that i hear are already there but this time they are more present or fleshed out. These are the most natural sounding amps I've ever heard. It draws you emotionally to the music and this is the most important for me. The best sounding amps ive ever heard.

Picture to follow

Regards to all.
 
Final Modifications for the 50 Series Amplifiers


NFB and HF Stability Circuits

So with the performance information digested from all the various versions, it was decided to take a fresh look at the NFB and stability circuits of the 50 Series amplifiers. Throughout the series, LF stability is generally good enough, so it is the HF end of the spectrum where the greatest opportunity lies. As a reminder, it is specifically the 50A, or in AZ versions, effectively the 50A (Z-matic feature removed) that is being addressed.

Regarding the Z-Matic feature, It is highly recommended that this feature and its components be completely removed from the amplifier. These versions provide a switch and receptacle on the side of the amplifier, with the switch determining which output impedance the amplifier is set for, and the receptacle being where the Z-Matic control is attached to the amplifier -- this to allow for its location being remote from the amplifier itself. Understand that for any Z-Matic version of the amplifier, use of the Z-Matic feature is optional, since the the control when turned to the off position, primarily takes the Z-matic feature out of the circuit -- for the most part (more on that in a moment). But also understand that just because it is not desired to use the Z-Matic feature, doesn't mean that the control doesn't need to be plugged in. Quite the contrary: Unless the Z-Matic circuit and components are properly removed from the amplifier, the Z-Matic control must remain plugged into the amplifier for it to perform properly whether the Z-Matic function is turned on or off! Today, many of the as found Z-Matic amplifiers no longer have the Z-Matic control, and since its function is an anachronism anyway, it all but mandates that the circuit be properly removed.

There is an additional reason for this recommendation as well. With the Z-Matic control connected but set to off, and the amplifier set for 16Ω operation, the speaker -- by way of the Z-Matic control switch and wiring -- is then connected directly to the OPT Com and 16Ω terminals as it should be. But in the same scenario except set for 8Ω operation, such is not the case. Now, the speaker connects to the 8Ω OPT tap as it should, but connects through a .50Ω resistor to the OPT Com terminal, due to a design flaw. This causes a loss of damping, and a loss of power output as well. Since most users today will operate the amplifier set for 8Ω operation, this then becomes an additional reason to completely remove the Z-Matic circuit and its components. Unless you are actively wanting to intentionally keep a unit as original as possible for historical purposes then, there is simply no good reason to keep this feature installed in the amplifier, and plenty of good reasons to remove it. The NFB and HF stability modifications to be presented are based on the Z-Matic circuit and its components being completely removed (except for the side panel slide switch), with the switch then being configured to simply switch the speaker "hot" output terminal between the 8Ω and 16Ω terminals of the OPT. The Common speaker output terminal then remains permanently connected to the OPT Com terminal at all times, and of course ground as well.

From the discussion of the various versions then, the HF performance areas of interest include:

1. Supersonic Frequency Response: This needs to extend high enough with minimal attenuation to provide for good transient response within the audio spectrum, but not so high as to produce instability problems when complex loads are connected to the amplifier.

2. Supersonic Transient Response: This is a product of the rise time as depicted with (typically) 10 kHz square waves. Nice perfectly vertical sides to such a square wave may look nice, but are a recipe for HF instability. On the other hand, lazy square waves that take their time to reach the wave peak make for poor HF transient reproduction, as often demonstrated with the delicate tapping of a cymbal for example. As with Supersonic Frequency Response, a proper balance is needed for best performance.

3. Supersonic Transient Stability: Again turning to 10 kHz square waves (themselves being transient in nature), once the peak of the wave front is initially reached, there should not be a trail of oscillations across the wave top after it, which would demonstrate poor damping of the circuit to such transients, and therefore poor transient stability.

4. 20 kHz Total Harmonic Distortion: Efforts to tame HF transient stability and produce complex load stability work against maintaining low THD in the upper reaches of the audio spectrum. Once again, a balanced approach is required to produce both good HF stability, and acceptable HF distortion.

5. Stable Operation Into Complex Loads: All of the factors affecting the previous four areas directly impact this area as well, which completes the circle right back to the defining comment of #1 above.

These are the issues that Fisher was struggling with throughout all four known versions of the 50 series of amplifiers. In all fairness, back in the day, there was no definition of what a good balance between these performance parameters represented, with the marketing arm of most companies actually doing serious damage to their company's product, by demanding such outrageously high frequency response numbers. That led to so many the problems previously noted, and the various versions to fix the issues of a previous version. I guess Windows wasn't the first to play that game!

In looking at the overall issue, one of the things Fisher did was to realize that the design could not produce the lofty distortion goals it set for itself along with good stability -- not both at once anyway. As a result, they made the hard decision to reduce the amount of NFB employed in the final two versions, which slammed the door on ever achieving the published distortion results. But it was the right move, as reducing the NFB then produced a workable level of distortion and stability, which together resulted in a much improved amplifier. What was still on the table then was developing any sort of definition as to what a good balance of these performance attributes really represents -- which defines the never ending argument between marketing, and the long hairs. Fisher ultimately reduced the NFB from 20 db, all the way down to 11.75 db -- which is basically the level I continued to work with. But with no marketing pressures on the design anymore (other than maybe what the DeHavilland copy produced), it meant a fresh start could be had in attacking the balancing act. So going back to the five identified HF performance areas of interest, the goals were now to:

1. Achieve a supersonic frequency response that allows for good load stability, and good transient response. Frequency response should be flat to 20 kHz, down no more that 1db at 40 kHz,, with a -3 db point around 100 kHz. Output should be stable into any value of capacitance only loading. Rise time on 10 kHz square waves no more than 4 uS.
2. Achieve target supersonic transient response with good transient stability. 10 kHz square waves should display fast damping and no trailing oscillations, producing a flat wave top.
3. Maintain a reasonable level of full power 20 kHz THD, commensurate with the performance levels achieve in the other three areas. A full power 20 kHz goal of no more than 3% THD would be reasonable for this design as a fully stabilized NFB amplifier.
4. Achieve absolute load stability, which brings us right back to #1 again.

Other than working with the NFB and HF Stability networks, the design of the basic amplifier circuit within the FB loop in the units here is exactly the same as Fisher worked with. But there was an additional tool at my disposal that Fisher didn't have: The modified bias regulator: It represents such an improvement over the original design, that open loop distortion is nearly cut in half. By allowing little if any bias shift when Class AB2 operation commences, the output stage operating point is maintained, and required driver power is reduced, which collectively makes for a significant drop in open loop distortion. Ultimately, this means that less NFB is required to achieve a given level of distortion when the loop is closed. Even with this advantage however, achieving all of the original published specifications for the 50A are still technically out of reach. But with the new bias regulator and the reduced FB level, the design can get closer than ever, and achieve a greatly improved balance of performance characteristics in the process.

Based on these thoughts then, new NFB and HF stability circuits were developed. They are not radical re-inventions of the wheel. In fact, they are much in keeping with the types of simple changes Fisher made from one version to the next. The modified circuits then could easily be considered as just another version. In keeping with that thought then, just as Fisher's simple changes to these circuits had a pronounced effect on performance, so do the changes made and installed in the amplifiers here, which will be seen when the final performance results are posted.

Continued in following post.
 
New Input Stage Design

In the original design of the 50A, V1 serves the dual function of acting as a single stage input preamp, and output stage bias regulator. In the modified amplifier however, the section serving as the bias regulator is no longer used. This presents the opportunity to redesign the input preamp to achieve improved performance.

The original preamp is a simple single stage design, that operates outside of the global FB loop, with only local current FB to help stabilize its performance parameters. As such, it provides a gain of ~ 8.0, but has two significant drawbacks:

1. The design is such that the HF frequency response of the amplifier is very dependent on the setting of the level control. For example, at a 50% setting, response at 20 kHz is down nearly 3 db when the input is driven from a low impedance source (i.e.≤ 1000Ω), and worse as the drive impedance is elevated.

2. The design causes the overall amplifier from input jack to output terminals to invert the phase of the signal passing through it.

By using both sections of V1 as the input preamp however, a number of benefits are achieved:

A. Amplifier performance is greatly stabilized, and sensitivity of tube choice for V1 is all but eliminated. The sensitivity was due to that fact that with the original design, the preamp section of the tube had little corrective ability built into its design, while changing the bias regulator section impacts the bias setting of the output stage. Therefore, changing V1 potentially changed two performance areas of the design. In the modified amplifier, bias regulation is now accomplished by the new Mosfet bias regulator modification, while the new input preamp is now a two stage design, with enough corrective ability built into its design to stabilize performance against changes in tubes at the V1 location.

B. The new input preamp has the same effective gain as that of the original preamp, but is non-inverting by design, meaning that the entire amplifier is now non-inverting as well.

C. The new input preamp is insensitive to the setting of the level control. Response of the amplifier remains unchanged at any setting even with an elevated drive impedance.

D. The new input preamp produces lower distortion, helping to minimize overall distortion produced by the amplifier in light of the reduced global FB it operates with.

So the new preamp stage then has a number of advantages over the original design, with no known disadvantages. No changes to the B+ supply are required with its installation, as the new dual stage design draws little more current than the old single stage design did. The same tube is used as well. When this modification is combined then with the modifications for the bias regulator, NFB and HF stability circuits, and ground configuration, it makes for a greatly improved amplifier, that can be measured and heard.

Modified Performance Summary

1. Absolutely black background due to elimination of ground loop. Hum and noise now approach 95 db below maximum power output.

2. Inherently lower distortion due to greater efficiencies achieved in the driver and output stages. At 40 watts RMS, 1 kHz THD is typically .5%, in spite of using less than one half of the original specification NFB. (at 20Hz, THD at this power is just 1.15%)

3. Excellent frequency and HF transient response, with balanced performance between these areas. Frequency response is now:
-.5 db @ 40 kHz.
-1.0 db @ 50 kHz
-2.0 db @ 70 kHz
-3.5 db @ 100 kHz.
Rise time on 10 kHz square wave is now just 3.5 µS, compared to nearly 5 µS for the original and final versions of the amplifier.

4. Superb HF transient stability with an absolutely flat top on square waves, and rapid damping after the rise of the leading edge.

5. Improved 20 kHz distortion characteristics. THD at 40 watts is now just 2.9%.

6. Absolute stability with any cap only load on either output tap.

7. Amplifier now maintains absolute phase.

8. High frequency response is insensitive the the setting of the level control.

9. Amplifier is no longer sensitive to tube changes at V1.

10. With reduced quiescent current required to achieve optimum performance (now 95 mA), rectifier and output tube life is enhanced, as is cooler overall operation.


In the final post, a schematic of the modifications and a scope shot of the final HF transient performance achieved will be presented.

Dave
 
Back
Top Bottom