1. Time for some upgrades in server hardware and software to enhance security and take AK to the next level. Please contribute what you can to sales@audiokarma.org at PayPal.com - Thanks from the AK Team
    Dismiss Notice

Fisher's First High Power Stereo Amplifier: The SA-300

Discussion in 'Fisher' started by dcgillespie, Jun 27, 2018.

  1. dcgillespie

    dcgillespie Fisher SA-100 Clone Subscriber

    Messages:
    9,707
    Location:
    Ball Ground, GA
    I may get a whole lot of blow-back on this one, but here goes: I've never been particularly fond of the Fisher SA-300. Besides the SA-1000, it was the last of the high performance stand alone stereo basic amplifiers Fisher produced, serving as big brother to the SA-100 during its five year run. But whereas the measures taken to achieve high performance in the SA-100 and SA-1000 all work together in harmony to achieve that end goal, those applied to the SA-300 more resemble management of medication side effects, where one medication is taken to address the side effects of another medication, etc. In other words, the measures applied are collectively working as much to contain the side effects as they are to actually improve the performance of the amplifier.

    The finished product all works well enough as designed: The amplifier will meet its distortion specification (.5% THD at 1 kHz within 1 db of rated output) and power rating (35 watts RMS per channel) well enough when only one channel is driven (typical of how amplifiers were rated back in the day), and even meet its noteworthy hum and noise specification (100 db below rated output) in Channel B, and come darn close to it in Channel A, where EMF radiation presents challenges in meeting the specification for that channel. But the high tube killing quiescent current draw (78 mA / tube) and resulting heat produced (continuous operation at 110% of rated plate dissipation/tube) in achieving these specifications are simply unprecedented. Add to this the heat dissipated by the heavy duty power supply employed to support this beast and you've got all the makings of a true space heater. With the cost of tubes and AC power (both to operate the amplifier and cool the room where the amplifier resides) ever on the rise, the SA-300 is not an amplifier to turn on and forget about in today's audio environment. Quite simply, for all the design work that went into it, the SA-300 comes off as incredibly inefficient and extraordinarily complex, producing performance results no better than other competing designs that are simpler and more efficient. Put in more classic terms, if a camel is a horse designed by committee, then the SA-300 is Fisher's camel. But how did it get that way?

    No doubt, as with any product development, there was likely a tug of war going on between marketing and the engineering department, with the bean counters acting as judge and referee. And as usual, marketing probably ended up joining forces with the bean counters to rule the day, leaving the poor engineers to figure out how to pull it all off. The engineers were not completely silent however, as the SA-300 ended up being engineered like no other Fisher product before it, or after. The conversation must have gone down something like this:

    Marketing: We want a new 60-70 watt basic stereo amplifier developed around the EL34 tube, because it's the hot European tube (pun intended) that everybody is designing with right now.
    Engineering: Well to operate that tube efficiently in that power range, it needs to operate in Ultra-Linear or tapped screen operation mode.
    Bean Counter: Oh no. We're not going down that road again. We did that with the 70 series amplifiers. We don't want to pay any more licensing or transformer development costs for something that is not an original Fisher concept to begin with.
    Marketing: Fine. But this new amplifier must use the EL34 one way or another to be competitive.
    Engineering: But we've already used that tube in the 55-A series, and can't claim it's first use anyway. If we can't go the UL route, then the 6L6 would be a far better way to go.
    Marketing: No, no. We've already used various versions of that tube before in the 50-A and 70, 80, and 100 amplifiers -- they're just considered too out of date and passe' now compared to more modern tubes. It must be the EL34.
    Engineering: OK -- but given the restrictions, it's going to take a lot of engineering hi-jinks to make a competitive product out of a non-UL EL34 amplifier specification wise.
    Marketing: We nearly live and die by specifications -- we've got to have respectable specifications!
    Bean Counter: Hi-jinks equals high cost. Will it cost more than if we had to go down the UL route?
    Engineering: You tell me. You're the Bean Counter -- but probably not. Likely just more R&D time up front, and then the build will take a little longer on the line.
    Bean Counter: That's it? I like it! -- A fixed cost that ultimately goes away, with just a little more build time.
    Marketing: OK, it's settled then. EL34s with engineering hi-jinks. You guys figure it out. Who's up for lunch?

    And so the SA-300 was born. Obviously none of us were there to know how the conversation actually went, but given the square peg into a round hole outcome that the design represents, the above offering just might be closer than you think.

    Now by itself, the EL34 is a wonderful tube. In true Class A triode mode, it barely produces any distortion at all. In UL mode, it is very efficient, producing both high power output and low distortion in very economical circuits. And in pentode mode, it can even keep up with the power output capabilities of the 6550. But good as the EL34 is, it's characteristics make for a double edged sword when it comes to design considerations.

    As a high Gm tube, the EL34 is a quite sensitive which is nice because it eases drive requirements. But in addition to high control grid sensitivity, that also makes it sensitive to any changes in operating supply voltages presented to its screen grid as well. Further, because it is a true pentode design without aligned grids, the change in screen grid current from quiescent to full power conditions is quite large. Collectively then, that's the rub. The EL34 itself is capable of operating with low distortion with the operating parameters Fisher used. But operating them that way in practical real world installations can get them into trouble real quick. Here's why:

    In medium voltage pentode circuits (as Fisher used here), the screen grids are basically powered from the same voltage source as the plates are, this to achieve maximum power output. To deal with the large change in screen current, any dropping resistance used for filtering purposes in the screen circuit must be kept small. As a result, for minimum hum, maintaining a tight DC balance is critical. But the primary concern, is that with an operating voltage so close to and powered from the plate supply voltage, any sag in the plate supply voltage produced by a practical power supply design will translate directly to a sagging screen grid voltage as well -- which in the EL34, causes distortion to increase quite significantly -- and is exactly what Fisher engineers were having to deal with given the design restrictions they had. Fisher being Fisher however, they were bent on producing good performance results anyway, and so the engineering hi-jinks began. So just what are they?

    That is discussed in the next installment.

    Dave
     
    Last edited: Aug 10, 2018

     

    Please register to disable this ad.

  2. dcgillespie

    dcgillespie Fisher SA-100 Clone Subscriber

    Messages:
    9,707
    Location:
    Ball Ground, GA
    Engineering Hijinks in the SA-300:

    1. Heavy Negative Feedback: The panacea for correcting all amplifier ills is NFB. It can lower distortion, lower noise, increase frequency response, flatten frequency response, and leap tall buildings in a single bound. The better classic designs of the day typically employed about 20 db of global NFB (producing a 10X distortion reduction), which is invariably enough to bring distortion and response well into the High Fidelity arena for most quality designs (of any type), and all without creating unstable operation in the process, which is a negative byproduct of using too much, or too much uncontrolled NFB.

    Fisher did its development work well however, as the NFB applied in the SA-300 is quite stable -- and that's a good thing too, because there's a whole bunch of it -- 25.5 db in the main global loop. This may not sound like a whole lot more than 20 db, but it represents a reduction level of almost 19X, or basically double the normal amount of feedback employed. Some amplifiers have trouble remaining stable with just 20 db of feedback, so the SA-300 remaining stable with nearly double that amount is impressive -- and also explains why there are so many HF step networks (i.e: medication) used in the design as well. They are needed to keep that much feedback stable at supersonic frequencies (meds for the meds), while that much feedback is needed to keep the distortion at bay created by the original issue of operating the output stage from an economical power supply.

    2. Partial Direct Coupling: There are 6 nodes in the circuit where LF phase shift can take place within the NFB loop, which can make for serious LF stability problems in an amplifier employing large amounts of feedback if precautions are not taken. Fisher took them. By partially direct coupling the driver stage to the output stage, it effectively eliminates one of these nodes so that only 5 remain relative to the (sub-sonic) frequencies of concern. Of those, two have exceedingly long time constants to effectively eliminate them as well. Two more are then tailored as necessary so that in conjunction with the inductance of the OPT (the last node), good LF stability is produced in the final product under all loading conditions. Partial direct coupling then is an aid (a med) to achieving good LF stability when heavy amounts of NFB are used. To maximize it effectiveness in this design, a Long Tail Diff Amp phase inverter is also employed (med for a med), which presents an equal low impedance drive from it's outputs. It is telling that this is the only piece of modern Fisher equipment to use this type of phase inverter design.

    3. Partially Cathode Biased Output Stage: While the partial direct coupling into the output stage enhances LF stability, it also allows any change in power supply operating voltages to directly impact the operating point of the output stage by way of it's control grid. The worst case is when high power output is demanded from the amplifier. When that happens, B+ voltage sags to all circuits of the amplifier, and of course at the driver tube plates as well. Due to the partial direct coupling, this then allows the output tube grids to sink further negative, causing the output tubes to become seriously over biased under conditions of high power output. Already being impacted by sagging screen grid voltages during this time, exposing such sensitive tubes to changes in control grid bias as well amounts to bullying and piling on. High distortion now rises even higher. But by adding a precise amount of cathode bias to the output stage, it works to appropriately counter the effects of the partial direct coupling (more meds for meds) and thereby helps to stabilize the operating point. But this remedy introduces its own issues in the form of significantly elevating 20 kHz THD. However, that was the end of the medication rope, as there was no answer for that.

    4. High Output Stage Quiescent Current: To further assist in dealing with the shifts in grid bias produced during periods of high power output, the output stage is biased for maximum possible quiescent current. Therefore, when it shifts downward from that point during periods of high power output, it too helps to keep the tubes from otherwise becoming severely over biased during periods of high power demand (yet more meds for a med).

    The rest of the amplifier's design is rather conventional -- although the heavy duty power supply is nice (but required with the extraordinary quiescent current demands), and paralleling the sections of the input 12AX7 is also a unique measure used to help increase the open loop gain (OLG) of the design, so that a reasonable input sensitivity (0.90 vac for maximum output) could be maintained after all that global NFB was applied.

    That Fisher got the design to operate as well as they did is a testament to their engineering capability. However, in my opinion, that capability doesn't come without its sonic issues. Amplifiers with high levels of NFB tend to sound dry compared to other designs that use less feedback due to the increased damping the elevated feedback produces. But in the SA-300, there are also time related effects as the time constants of the partial cathode bias and partial direct coupling networks react to dynamic voltage changes in the B+ delivery system with all of its various time constants as well. In the lab, this can clearly be seen on a scope even at low power levels, when a square wave is (for example) instantly doubled or tripled in p-p value. When that happens, the leading edge of the square wave can be seen to "correct itself" in response to the new input level -- and independent of the rest of the wave -- immediately after the increase is applied. This is clearly a form of transient distortion, and a direct result of the measures used to stabilize the high level of NFB employed -- with that of course tracing back to the issues surrounding the choice of output tubes, and topology of the power supply.

    Fisher's engineers knew all of this. The proof of that pudding lies in the complexity of the design resulting from their work that was required to pull it all off -- which is unprecedented in Fisher's history for such a basic audio device. To their eternal credit, the results are also about the best that can be expected when so clearly trying to push a square peg into a round hole.

    There are some wonderful fixes for all of the issues noted -- fixes that leave the unit as still very much being a Fisher SA-300, with the same major components, tube complement, circuit topology, and at least the same original performance levels, but with output tubes that only require drawing about half of the quiescent current level from that of the stock design, and dissipating about half the heat as well. Distortion is lowered, stability improved, tube life extended, performance improved in many ways, and complexity reduced. Yes, EFB is involved, but it's only part of the answer. The rest of it lies in a comprehensive look at the design as a whole, so fixes no longer needed are removed, and fixes that remain become fully effective. But presenting that can wait for another day.

    The piece here is meant to establish just how unique the SA-300 is in the Fisher lineup -- by ultimately showing just how un-Fisher like the SA-300 really is. It's almost like a completely different design team developed it. On the outside, you'd never know the difference, but under the hood, it is like no other Fisher product ever. If interest is there, then this offering can serve as a launching pad for a future modification presentation. For now however, it's only meant to enlighten how it operates to those who own them or wish to. My personal preferences aside, it is not meant to demean the product in any way. One thing is for sure and for certain however: The SA-300 stands as proof positive as to just how much power those in marketing yield -- and we all know who was in charge of marketing at Fisher!

    Keep those A/C units charged!

    Dave
     
    Last edited: Jun 29, 2018
    Archguy, tcdriver and Tim D like this.
  3. thornev

    thornev Well-Known Member

    Messages:
    515
    Location:
    Mid Hudson Valley, NY
    I can confirm, at least in an abstract and general sort of way, that the engineer versus marketing conversation is very real. The engineers want the expensive functionality to satisfy the saviest of audiophiles; Marketing wants to make the sale which usually means at the lowest possible cost to the consumer. It's an age-old story. Technologists versus Executives. My approach as a technician (for 35 years at a major computer company) has always been KISS - Keep It Simple Stupid. I usually lose as usability doesn't seem to be important to the expert technologists. They want to market to those who will appreciate all the bells and whistles they want in the product. Marketing says it's too much for Service and the average Joe and Jane to digest. To wit... the success of Apple who locks down their OS and yet the success of the open platform of Android. Just look at the the customers of those products and we'll see that both factions exist. So.. To Whom It May Concern... Thorne
     
    Last edited: Jun 28, 2018
  4. Dave451

    Dave451 AK Subscriber Subscriber

    Messages:
    1,133
    Location:
    Philly area
    Great piece, Dave! I recently did some work on a SA-300 and fast impressions were 1) crowded!--lots of stuff in a very compact chassis 2) heavy!--big iron all around 3) Strong!--loved the dual-rectifier, heavy duty power supply (see heavy!) 4) complicated!--(see above) 5) sounds great!--even with low-end output tubes, I thought the sound was robust, clean and detailed (OK, maybe a little less warm).

    This info gives me a much deeper appreciation for the SA-300 and it goes into the file with the rest of my notes (on the top of the pile!). Really appreciate these in-depth discussions of classic gear, Dave.
     
  5. heyraz

    heyraz AK Subscriber Subscriber

    Messages:
    1,906
    Location:
    Long Branch, NJ
    Dave,
    So glad you are finally taking a look at my favorite sounding amplifier!
    No test equipment used....the music just "sounds right" with my system. There's just something about that amp I like....

    It's a "heater" all right. I have a spare SA300 that I may perform your mods on if I have the time.
    I'm looking forward to more on this thread.
    Rich

    PS: I had my Central A/C replaced last month, it won't be uncomfortable listening to music.
     
  6. Tim D

    Tim D AK Subscriber Subscriber

    Messages:
    1,346
    Location:
    Melbourne, FL

     

    Please register to disable this ad.

  7. larryderouin

    larryderouin Turn it UP, POP? PLLUUEEEZZZZZEE Subscriber

    Messages:
    21,834
    Location:
    Glen Burnie Md.
  8. golana

    golana analog Subscriber

    Messages:
    1,836
    Location:
    Akron, Ohio
    Great read Dave. If. You want a unit to play with, please take mine.:)
     
    trainbuftony likes this.
  9. larryderouin

    larryderouin Turn it UP, POP? PLLUUEEEZZZZZEE Subscriber

    Messages:
    21,834
    Location:
    Glen Burnie Md.
    OH boy, a potential Guinea pig for Surgery!
     
  10. tcdriver

    tcdriver AK Subscriber Subscriber

    Messages:
    5,382
    Location:
    Valley of Heart's Delight
    Thank you for the information you are providing on the Fisher SA 300. :thumbsup:
     
  11. dcgillespie

    dcgillespie Fisher SA-100 Clone Subscriber

    Messages:
    9,707
    Location:
    Ball Ground, GA
    And So We Begin........

    But first, some initial thoughts and admissions:

    One of the things I find myself doing a really lousy job of anymore is documenting the "before" appearance of a project to provide a context of the before and after look of a piece presented for modification. That's almost entirely due to the fact that I most often don't start these projects with the thought in mind of it being a piece for presentation on AK. Virtually all of the threads I've contributed over the years start out entirely as something I want to do for myself, on my own piece of equipment. As such, I typically don't document such work or think to do so anyway. Later, if the success is such that I feel it might be worthwhile to contribute to the group, then I start thinking about before and after thoughts -- when it's too late. Such is the case at hand. This project actually started a year or more ago, when I started taking stock of all my equipment pieces, trying to make some sense of it all, and then determining what other equipment I might like to pair it up with to form a complete system -- a very long and drawn out effort that is ultimately supposed to end up with me paring down a whole bunch 'o stuff to that which I really want to keep. The effort with the SA-300 was a side project that needed to be done, so as to determine what its status would be in terms of the larger picture -- in other words, to see if it could make the cut.

    There are a number of things that determine if a piece from my collection reaches the "inner circle" of my equipment. Sound quality is of course a big one, but design quality, manufacturer, and industry impact are also hugely important, as is the circuit design itself. Therefore, many of my pieces reach the mountain top simply because they are (for example) the best UL EL84 design, best triode KT88 design, or best pentode 6L6 design, etc., etc., be it commercially made, or home grown. Rare and obscure equipment then is not nearly so important to me then as is the importance of what a piece represented to the golden age of high fidelity, in terms of design, and historical impact.

    Given these parameters then, the SA-300 should be a shoe-in to my inner circle, given the uniquity of its design, which gives it a big advantage. After all, there were not many commercially offered pentode based EL34 designs ever produced, no doubt due to the problematic design considerations given earlier when operating the tube this way. The EL34 is a tube that simply begs to be operated in UL mode, with high power output, low distortion, and high efficiency easily being achieved all at the same time, and all in an economical package to boot when operated that way. Witness the Dynaco ST-70, and it's countless "me-too" UL imitators from the likes of Eico, Heath, Pilot, and many others. On the other hand, only precious few manufacturers went down the pentode EL34 road (HH Scott, Altec 345A, and Fisher again with the X-1000), all of which were certainly low production units (and had the same design problems, too). So if only by design, the SA-300 starts off strong as needing to be a part of my inner circle, but my logical side says only if at least the gross inefficiency, space heating, and tube killing aspects of its design can be appropriately dealt with.

    The stock SA-300 is an excellent sounding piece, and truly a credit to Fisher's engineers. But good as it is, it is very hard for me to enjoy music from a piece of equipment when I know the toll it takes on itself by simply operating it. Inefficiency alone I can deal with: A good Class A amplifier is grossly inefficient and produces significant heat for its power output capabilities. But designed properly, it won't kill the tubes, and will operate with a long and happy life. But that's not the case here. I know that most SA-300s have lived to tell the tale. But with a design that operates the output tubes beyond their specified limits, I shutter to think of the temps reached by a stock SA-300 when buttoned up inside the back of a Fisher President console (without its cage even!), and operating in the summertime in a non-air conditioned home in 1962! Today's line voltages only add insult to this injury. This type of operation is simply unacceptable for me today, and marks the beginning point for any modifications performed. That is, if the heat generated and tube life produced by operation of the stock design cannot be impacted in a positive way -- and significantly so -- while maintaining at least the same sonic and measured performance capabilities that Fisher engineers achieved with the original design -- then any modification of the SA-300 (for my purposes) is a waste of time -- and, keeps it out of my inner circle.

    The typical approach of cooling the output tubes down by simply adjusting the bias for cooler operation is not an option with this design, as such a move causes distortion to rise dramatically (and quickly), and power output to fall as well. And yet, to cool the output tubes down, that's ultimately what must happen, but without increasing distortion in the process. I know, you may only use a few watts in actual practice. So do I. But that's like saying that since I only drive the speed limit, how my car performs beyond that point is of no concern. That kind of thinking is unacceptable to me for a product of this caliber, compromising the intended integrity of the product, and the efforts by the engineers to create it. So it is against this backdrop and expectation then that a modification of the SA-300 was approached, making it a tall order to accomplish indeed.

    To kick things off then, any performance enhancements start from a base line of stock performance capability. My stock unit, biased correctly and operating from 120.0 vac, driven through the unfiltered inputs, returned the following results, as an average of both channels, measured with both channels driven:

    1. Power Output and Distortion (at the onset of clipping), measured into 16Ω:

    @20 Hz: 30.30 watts RMS per channel, with 2.4% THD.
    @1 kHz: 37.5 watts RMS per channel, with 0.60% THD.
    @20 KHZ: 33.1 watts RMS per channel, with 3.8% THD.


    2. Frequency Response (1 kHz ref):

    @20 Hz = -0.20 db.
    @20 KHZ = -0.50 db.
    @40 kHz = -1.0 db.


    3. Stability: Absolute. No amount of capacitance only loading will cause the amplifiers to oscillate.

    4. Sensitivity: 0.90 vac rms for maximum power output.

    5. Output Tube Quiescent Operating Conditions (each tube):

    Plate Voltage = 400 vdc.
    Plate Current = 0.07 A.
    Plate Dissipation = 28 watts (112% of rating).

    6. Hum and Noise: 98 db below 35 watts.


    For 1962, this was pretty respectable performance. It's not Mac performance levels mind you, but you'll still have plenty of fun at the party. It is in fact very Dynaco ST-70ish, which by that time -- having been released earlier -- may likely have been their performance target goal. If so, it is particularly commendable given the design considerations they were dealing with. That the SA-300 uses more tubes, has ~ 70% greater heat output, and basically double the circuit complexity to accomplish it however, is quite telling.

    With a base line of information established then, discussion of the modification itself can now begin. Next time.

    Dave
     
    Last edited: Jul 17, 2018
    Archguy, baconbadge, gadget73 and 3 others like this.

     

    Please register to disable this ad.

  12. golana

    golana analog Subscriber

    Messages:
    1,836
    Location:
    Akron, Ohio
    Thanks Dave. Yours are always a good read.
     
  13. rufleruf

    rufleruf AK Subscriber Subscriber

    Messages:
    1,125
    Looking forward to the next installment - I've got one of these on my "sometime this decade" list.
     
  14. dcgillespie

    dcgillespie Fisher SA-100 Clone Subscriber

    Messages:
    9,707
    Location:
    Ball Ground, GA
    Anatomy of a Modification:

    Adding EFB

    The SA-300 is so carefully designed that modifying it is like tugging on a loose thread. You pull on it at all, and ultimately end up going through most of everything -- but since many will do that with a full rebuild anyway, it really doesn't matter.

    Any effort to modify the SA-300 must start with the output stage, since it, in conjunction with the power supply, is where most (but not all) of the distortion (and certainly the heat) is generated. Fisher's approach to this problem was to use a lot of global NFB around the amplifier (nearly 26 db), which, it turns out you need the most of anyway not only to lower distortion, but also to sufficiently lower the internal output impedance of the amplifier so as to allow for good speaker damping. Pentode output stages (unlike triode and UL output stages) inherently have a very high internal impedance, so it takes more than the typical 20 db of global NFB with this type of output stage to bring the impedance down to a respectable level (say, about 1.0Ω at the 16Ω tap). Also, the higher NFB helps to linearize the core of the OPT at the lower registers to minimize the distortion it contributes as well. But of course, it was the very use of that much feedback that lead to all the complexities of the circuit, and the collateral damage that incurred from it along the way.

    It was originally hoped that with the addition of EFB™, it would reduced distortion enough that the NFB could be dropped down to a more comfortable 20 db, which would allow the circuit to be greatly simplified, and stability to be achieved without all the complexities required by the original design. This approach worked very well, with mid-band full power distortion dropping from originally 0.6% THD (@ 1kHz) down to just 0.10% THD. This shows the extreme effectiveness that EFB provides in this case: With basically 1/2 the amount of NFB as used in the original design, overall distortion has been reduced by a factor of X6. Actually, when the reduction of NFB (over that of the original design) is considered, this really breaks down to the NFB connection providing a reduction of just X10, and EFB a reduction of X12, for a combined total distortion reduction of X120. This can be seen by the fact that in the original design, 26 db of NFB (which is a X20 distortion reduction) started from an open loop (no NFB) distortion of about 12%. Reducing that figure by a NFB factor of X20 then brings it down to the 0.60% measured with the stock design. The application of EFB however brings the raw output stage distortion w/o NFB down to just 1%, with the X10 reduction of 20 db of NFB then bringing that down further to the 0.10% recorded when EFB and 20 db of NFB were implemented together. If anything, this exercise shows just how much effective output stage distortion Fisher engineers were dealing with from the output stage/power supply combination they were dealing with, which then mandated the use of so much NFB being applied.

    The encouraging results this combination produced received a stiff jolt however when distortion was measured at 20 Hz, where it had now risen to nearly 4.3%, or nearly double the original value -- this owing to only 1/2 the original amount of NFB in place. This spotlights a basic operational point about EFB: while EFB can bring about significant performance improvements, it can only do so by maintaining the linearity of active devices (like tubes) -- not passive ones like a transformer. Only NFB can help an OPT's performance, where at low frequencies, its characteristics dominate the distortion picture, masking the gains made by the application of EFB. So the reduction in NFB -- while not a problem at mid-range frequencies, was very evident (and therefore problematic) at low frequencies.

    Overall however, EFB was clearly shown to be one of the needed answers, because the performance figures noted above were all generated with the output stage now idling at a quiescent current of just 40 mA per tube, or basically half that of the original design (78 mA per tube). And this wasn't just some arbitrarily chosen lower figure, either. This was the true low distortion operating point for EL34 tubes when operated under the pentode conditions given by Fisher, that EFB simply allowed to surface by insulating the stage from the effects of power supply fluctuations. So once again then, EFB was shown to lower distortion by (in this case) a factor of 12, cut the required output tube quiescent current (and largely the plate dissipation as well) in half, extend output tube life as a result, promote cooler operation, and even increase power output slightly due to the rise in B+ resulting from the reduced quiescent current drawn. And, there is another huge benefit in the SA-300 with the addition of EFB: The "Heavy Duty" power supply that Fisher promoted, truly is a heavy duty supply now! As originally designed, the power supply was in fact heavy duty -- but based only on comparison with other designs in the same power class. When it is realized just how much current the original design required to be drawn from the power supply for proper operation, any real pretense of being truly heavy duty just sort of fades away. In other words, it was massive not because it was really overbuilt, but because it had to be that big. With the application of EFB, the power supply -- now -- truly is overbuilt, so the benefits that provides can now be realized. But even so, the LF characteristics of the amplifier showed that even with these benefits, the original level of NFB is still required to secure good LF performance.

    As applied in a single global loop in the original design however, the 26 db of NFB had a notable side effect I had mentioned earlier: This much global NFB required use of the partial direct coupling networks employed between the driver and output stages to help produce good LF stability. But while these networks solve the stability problem, they also introduce the timing issues noted earlier, where large changes in dynamic signal levels cause changes in DC operating conditions, that don't stabilize again until a fractional second later. By then however, the initial transient may be long gone -- creating the classic problem inherent with direct (or partially direct) coupled designs. The only way that could be addressed in this case is to lower the amount of NFB used so as to be able to eliminate the need for the partially direct coupled networks -- but reduced NFB then elevates LF distortion, so you get caught in a circle. The partial direct coupled networks cause other practical problems with the design as well, whereby an aging tube in the phase inverter location causes a shift in output tube bias, or a simple tube swap in the same position changes the output stage DC Balance setting. Fisher tried to minimize these effects by adding a partial amount of cathode bias to the output stage, and while this helps, it also creates its own problems by causing 20 kHz THD to rise significantly.

    What is needed then is an EFB-like answer, that would effectively allow 26 db of NFB to remain in place, but without requiring the need for any time displacing partially direct coupled networks, or partial cathode bias remedies for that matter as well. Thankfully, there's an answer for that too. Next time.

    Dave

    The EFB circuits as developed for the SA-300. Not noted on the schematic is the fact that M1, R7, R69, R70, C5, and C6 all go away as well.
    Fisher SA-300 EFB.jpg

    The EFB Control Grid Regulator. The perf board is about 0.75" X 1.5", and is mounted where the old cathode bias resistors resided. I never really found a good answer for making the bias adjustable from the topside of the amplifier so the bottom must still be removed to make the adjustment. However, it's one heck of a lot easier to adjust than the original single control was, and of course, there's an individual bias control for each channel now. The controls are color coded for easy adjustment: Black for the Channel A, Red for Channel B.
    SAM_2478.JPG

    The EFB Screen Grid Regulator. The Mosfet is mounted where the old selenium rectifier once called home, with the attending circuits taking up residence where the old screen grid dropping resistors once lived. As noted on the schematic, each output tube socket is now outfitted with a 100Ω 0.25 watt Screen Stability Resistor, connected between pins 4 and 6 at each socket. Pin 6 then becomes the daisy chain. The EFB Screen Grid Regulator supplies power to the screen grids of all four tubes, so the dual screen grid supply sources of the original design goes away.
    SAM_2479.JPG

    The new silicon bias rectifier moves over to the T-strip where both of the old bias filter caps once resided. Only one primary bias filter cap is used in the modified design.
    SAM_2480.JPG

    The power supply circuits of the original design. I wish there was a better source of the schematic to post.
    Fisher SA-300 Power Supply Page.jpg
     
    Last edited: Jul 13, 2018
    Pio1980 likes this.
  15. SoCal Sam

    SoCal Sam Lunatic Member

    Messages:
    18,656
    Thanks for the professional, magazine grade read. Looking forward to your results.
     
    Pio1980 likes this.
  16. gadget73

    gadget73 junk junkie Subscriber

    Messages:
    36,761
    Location:
    Southern NJ
    Which Sams is it? If I have it, I can scan it large size if that would be of any help.
     

     

    Please register to disable this ad.

  17. heyraz

    heyraz AK Subscriber Subscriber

    Messages:
    1,906
    Location:
    Long Branch, NJ
    Last edited: Jul 13, 2018
  18. AlTinkster92

    AlTinkster92 AK Subscriber Subscriber

    Messages:
    3,257
    Location:
    NC
    Very good read Dave, will watch with interest!
     
  19. heyraz

    heyraz AK Subscriber Subscriber

    Messages:
    1,906
    Location:
    Long Branch, NJ
    Dave,
    Very good read and breakdown of the circuit and power supply.
    I haven't had a chance to really look at the (Fisher Service Manual) schematic as I read your analysis but am left wondering...how does this amp "sound" with EFB vs Fisher's original design?

    I see that you've split the left/right bias supplies ( as I did by a different method) and have referenced plate dissipation (whereas I used cathode test sockets to measure cathode currents) to set the quiescent bias currents for each output tube.

    Before I modify my spare amp to your design, I have to ask....how does EFB "sound" compared to Fisher's original design?

    (I'm only asking because my recollection is that you compared the SA100 "before and after" results after implementing EFB).

    Rich
     
    Last edited: Jul 13, 2018
  20. dcgillespie

    dcgillespie Fisher SA-100 Clone Subscriber

    Messages:
    9,707
    Location:
    Ball Ground, GA
    Anatomy of a Modification:

    NFB Games

    Back in the day of the SA-300, negative feedback was looked at as the elixir that could cure all amplifier ills, with the more, the merrier. And as the SA-300 has shown, that would certainly seem to be the case under steady state (i.e.sine wave) dynamic conditions, where gross amounts of distortion have been brought into relative compliance. But the SA-300 also shows that past a certain point, collateral issues can start to develop when the signal becomes more transient in nature, where real world coupling networks can cause instability and speed issues to surface. What constitutes too much NFB in a given scenario depends on the characteristics of the particular circuit and the quality of components used, meaning that the negative aspects of NFB can appear at different levels in different scenarios. I've seen designs where just 10 db of NFB (a little over a X3 reduction) represented the absolute maximum amount of NFB that could be applied, beyond which the negative side of NFB started to rear its ugly head. Again, that the stock SA-300 -- with its 26 db of NFB -- is as stable as it is and performs as well as it does under transient conditions, is a credit to the Fisher engineers that designed it. And yet good as it is, when operating properly, it is also clearly operating at the limits of the NFB applied to it, as demonstrated by the delayed settling effects observed with changes in transient signal (square wave) amplitude. But there's one other important point to understand in all of this.

    Besides the amount of NFB used, how it is applied is equally important as well. Fisher chose to apply all of it around all three stages of the amplifier in one giant global loop. This kind of NFB application is always the most problematic with regards to maintaining LF stability, because the loop encircles all of the coupling networks within the amplifier. In the SA-300, there are 5 points of potential LF phase shift, that each act to affect LF stability in their own way:

    1. The input coupling cap.
    2. The input stage cathode bypass cap.
    3. The phase inverter grounded grid bypass cap.
    4. The output tube cathode bypass cap.
    5. The output transformer.

    With this many points of potential LF phase shift within the design, it basically mandated that the design use as much direct coupling as possible.

    The partial direct coupling networks (PDCNs) -- of themselves -- don't actually introduce any LF phase shift at all (hence, their use) at the frequencies of primary interest relative to LF stability (1-2 Hz). However, because of the direct coupled aspect of these networks, the time constants of the various power supply decoupling networks can now impact the LF stability of the amplifier, which is what is really going on here. This is why the amplifier displays excellent stability under steady state conditions (quiescent or dynamic), but begins to hiccup under changing transient conditions during Class B operation. With steady state conditions, the power supply operates in a steady state condition as well. But under changing transient Class B conditions, the output of the power supply changes accordingly, generating its own LF transient "signal" that the output stage also reacts to, as surely as it reacts to the transient signature of the applied signal to begin with -- except that the transient signal from the power supply is not time aligned with that of the applied signal due to the time constants of the power supply's decoupling networks.

    Theoretically, this type of behavior happens in any Class AB amplifier. But in designs employing purely R/C coupling, any LF transient reactions of the power supply are filtered out by the coupling circuits, and so the problem is basically all but absent. Now the first AF amplifier stage is also direct coupled into the phase inverter stage of the SA-300. However, this particular connection doesn't really cause any problems, because both of these stages operate with largely a constant current characteristic. The AF amplifier stage employs a very large plate resistor and a dedicated, isolated B+ supply source for each channel, all of which acts to isolate the stage from the effects of any power supply modulation. The phase inverter stage operates nearly constant current as well due to the large tail resistor employed. But the phase inverter tube plates, operating with low value plate load resistors and nearly as much B+ as the output stage operates with, "sees" a portion of any power supply modulation created by changes in Class B output stage current draw. This modulation is passed through the PDCNs back to the output tube control grids, which ultimately impacts -- in delayed fashion -- the bias on the output tubes. In this way then, the PDCNs help to form a mini DC coupled NFB loop around the output stage.

    But in this case, the NFB generated is not helpful. Besides acting in time delayed fashion, it's also acting to increase output tube bias voltage as output tube B+ voltage is reduced, which then increases distortion as well. Since the time gap between signal action and power supply reaction appears the moment either channel shifts into Class B operation, the effects of employing PDCNs at the output stage control grids are present throughout most of the power output range of the SA-300. The stock design shifts from Class A to Class B operation at about 8 watts RMS per channel -- thanks in large part to the extremely high output stage quiescent current draw. The output stages are also partially cathode biased to help circumvent the problems of the PDCNs as well. The networks are in fact very effective in helping to stabilize the LF stability characteristics of a high feedback amplifier. But the downside of their use in this location can only be eliminated by using them in fully regulated designs, where no influence on the previous stage's plate supply can upset the operating point of the output stage at any power level. As a second basic point of this modification effort then, the PDCNs as used in the original design must go for the modification to be considered a success.

    So if the PDCNs have to go, then global NFB must be reduced to maintain LF stability. If it is reduced to a more typical 20 db, then it means that another 6 db of NFB must be picked up somewhere else -- and preferably include the output transformer as well within the new loop, to regain the high ground with regards to 20 Hz THD. A number of different configuration were experimented with including:

    1. Push-pull Schade style feedback from the plates of the output tubes to the plates of the driver tubes,

    2. Push-pull feedback from the plate of the output tubes to the grids of the phase inverter (cross-coupled), and

    3. Push-pull partial cathode coupling of the output stage.

    All of these approaches had issues to be addressed for successful implementation, some of which however were insurmountable within the basic design offered. One however -- with proper implementation -- proved to be a highly successful answer to all the issues previously noted. By partially cathode coupling the output stage, each output tube is subjected to an additional 4 db of NFB -- but that's (literally) only half the story. What is not well known is that when NFB is applied in the push-pull connection like this, the distortion reduction received is equal to the square of the feedback factor, rather than just the factor itself. Therefore, 4 db of NFB produces the distortion reduction of 8 db when applied in the push-pull connection. For maximum effectiveness, this type of feedback requires two things:

    1. The use of a very high quality OPT, wherein the total response characteristics of the transformer appear equally across the full secondary winding. Some transformers - and very high quality ones at that (Acro for example) -- are not up to this level of service, and simply become oscillators when configured this way. On the other hand, the SA-300's OPT are of such quality that even if configured for positive partial cathode coupling, the output stage remains completely stable! This indicates that the balance that exists within the windings of these transformers with regards to response and phase characteristics is extremely tight.

    2. The use of high Gm tubes, for which the EL34 certainly qualifies. All of a sudden, the high sensitivity that makes these tubes so problematic with regards to fluctuations in power supply voltage becomes the very advantage that makes this application of NFB so successful.

    One of the really neat things about NFB, is that when an inner loop is applied like this (where it didn't exist before), the open loop gain (OLG) of the overall design relative to the global feedback loop automatically adjusts that loop's feedback level so that overall, the original feedback level is still maintained -- but now spread over two separate loops, improving stability and effectiveness. Therefore, with 8 db of feedback in the inner loop, the outer loop now operates with just 18 db of feedback, which easily allows for very good LF stability using conventional R/C coupling networks: say goodbye to the PDCNs and the resulting quirks of the original design!

    Yes, the modification fundamentally changes the topology of the amplifier, but the change made is so significant, that it simply cannot be discounted. The combination of converting the output stage to a partially cathode coupled design, and operating it under the control of both EFB Control and Screen Grid regulators produces a result that takes the SA-300 (imo) from a generic also ran status, to an ability to take on all comers. The SA-300 becomes one sweet songbird: Gone is the outrageous heat production and the time delay effects of the stock design. As noted earlier, the power supply really does become a truly heavy duty design. What's more, all the original stock performance markers are not only met, but now exceeded. In short, there are no downsides to this change, other than the fact that the speaker Common output terminal no longer operates at ground level. As with so many of Fisher's later designs, it is the 4Ω tap that is now grounded -- but of course for a completely different reason than Fisher did in their later units. In any event, even this change then is ultimately not that un-Fisher like at all.

    And then there's the sound. I'll reserve my detailed comments on that for later, but suffice to say that the modified amplifier certainly sounds at least as good as the original, and imo, significantly better.

    There are details to address for the successful implementation of the revised output stage, which will be covered next time.

    Dave

    New Output Stage Configuration:
    Fisher SA-300 Output Stage.jpg

    The old damping terminals are used for test points, with the new cathode sampling resistors connected to provide partial cathode coupled NFB.
    SAM_2481.JPG

    The modified output stage is far less cluttered for a cleaner look and build.
    SAM_2482.JPG
     
    Last edited: Jul 17, 2018

Share This Page