Epilog
With this post, I have provided a schematic of the EFB(tm) modifications installed in the SA-100. I will go over the modifications in a moment, but first, some general comments.
This unit comes from an older generation of Fishers, so any such unit will need to be gone over carefully. I've already documented the ground issue, which should be checked in all units. Also, the general build and small component quality improved in succeeding years, so that aspect of earlier generation equipment should be examined as well. On the other hand, I think the earlier generation's transformers are brawnier, as there are far fewer failures with the earlier models than in the latter years.
The design of the amplifier portion of the SA-100 is certainly good, although could likely benefit from a closer examination of the NFB/stability scheme used. The sq wave performance of the unit was documented previously, and is generally good, but not nearly as good as that of the clone version. Whereas that amplifier displayed a wave form with a very flat top, it also had an extended response to 70 kHz, while the SA-100's response is basically to 1/2 that level, and displayed a sq wave that was somewhat compromised. The stability displayed in both the SA-100 and the clone were absolute.
Another area of opportunity in the SA-100 -- and likely represents the biggest difference between the clone and the EFB modified SA-100, is that of the basic power supply. The power transformer and rectifier are capable enough, but the healthy choke used in the clone version represents a significant advantage in the war on distortion over the power supply of the stock design. While the simple series dropping resistor used in the SA-100 provides filtering enough for normal listening purposes, its limitations show up in the slightly higher distortion levels produced, and particularly so at the lowest frequencies. The level of AC ripple current present under high demand conditions at the OPT CT is significant in the SA-100 (over 10 vac). With the clone, it remains negligible in high demand conditions over the entire audio spectrum, which allows for lower distortion levels to be achieved due to the elimination of this influence.
The problem is space. This little amplifier has no room for the choke I chose for the clone. I then thought of using a smaller choke to supply just the screen and driver circuits, but the EFB screen regulator provides significant filtration by itself, making any gains from this approach nil. To improve filtering any further, it is the filtering for the plates that requires attention. For a choke to be effective under high levels of current draw, it takes size to fill that bill. Still, a small choke could be added in the area under the OPTs which would offer some benefit, as there is a very low level hum attributed to the limited plate filtering used in the original design. A small choke would certainly tamp that hum down, but not offer the distortion reduction benefits at elevated output levels that a full sized choke will provide. But even to install a small choke would require altering the installation of the EFB circuits to allow for the extra space. It could be done, but would be a chore, first requiring the removal of the Channel B OPT to gain access for drilling the mounting holes.
These two areas then represent the biggest difference that exists now between the clone and modified SA-100. Both units sound superb. I do give the edge to the clone version however, as I do believe the approach used in that unit to generate absolute stability allows for a more open sound if not certainly its extended bandwidth. I would again emphasize that both of these amplifiers sound like 6BQ5 amps on steroids. They just have that extra punch enjoyed by the 7591 crowd. Also, the degree to how much cooler the SA-100 operates now cannot be overstated. The reduction in heat output is more than significant -- which is important when cramming so much horsepower into so little space.
With regards to the EFB regulator circuits themselves, the screen regulator is almost a direct copy of the regulator I designed for my clone version, without the voltage adjustment control I used on that version. Its operation is very simple. The output stage in the original SA-100 operates as almost a text book copy of the conditions published by RCA for a 24 watt output level, where in the screens operate at 75% of the plate voltage. While this relationship is established under quiescent conditions in the original SA-100 design, it falls apart as power output increases -- and more so in the SA-100, where an attempt was made to semi-regulate the screens, but not the plate voltage. As power output increases then, the relationship becomes destroyed with only one of these elements regulated, shifting the optimum required load, reducing power, and increasing distortion. Because the EFB screen regulator now holds the proportion of the plate and screen voltages in place, the screens always maintain their optimum 75% operating potential with regards to the plate voltage -- whatever that voltage falls to, and regardless of any increase in screen current due to increasing power output. The result is that the load line of the output stage remains largely in tact even though the main supply is unregulated, and actually drops by nearly 100 volts from 0 to full power output in both channels! The EFB screen regulator is largely responsible for the 25%+ increase in power output available from the modified amplifier.
The small .47 uf cap helps to provide significant filtering so that ripple applied to the screens is in the millivolt region, yet it is small enough to allow the regulator circuit to react quickly with any changes in plate voltage.
As detailed in the clone thread, the Mosfet device is quite over-rated for the job it does being a 500 volt device, and capable of passing some 9 Amps of current. It should last a lifetime.
The EFB bias supply regulator is newest EFB regulator to be published, and consists of a simple two transistor inverter circuit, supplied from a modified version of the original bias supply, and controlled by the screen voltage supplied by the screen regulator.
In a pentode, the screen voltage sets the maximum current that can flow at any given instant, so control of it is important to maintain the load line as discussed above. But if the screen voltage moves at all, then the bias voltage needs to move proportionately as well, or the original low distortion operating point set under quiescent conditions is destroyed, once again causing elevated distortion with increasing power output. Therefore, as the screen voltage changes, so must the bias voltage.
In the SCA-35 EFB installation, the cathode regulator provided fixed bias and adjusted the set point all in one neat package. But in the SA-100, fixed bias is already part of the design. Therefore, in this case it is more appropriate to control the bias supply itself with an EFB bias supply regulator.
In the schematic, the left transistor provides the inversion -- causing a negative bias voltage to go more positive, as a positive screen voltage goes more negative at the screen. The right transistor isolates the load of the standard DC bias and balance controls used so that their setting does not interact with the inverter stage's operation.
The original components R56, M1, C2, and R51 in the bias supply (Sams designations) have been adjusted to provide more bias voltage for the bias supply regulator to operate from. That, together with the installation of the bias supply regulator, cause the actual control settings of the bias and balance controls to change very little from that of the original design -- given the same output tubes in the same socket locations.
As in the clone amplifier, both EFB regulators in the SA-100 come together to produce a superb result. It is common for most manufacturers who set up their output stages to mirror published conditions, to do so with one exception: they increase the quiescent current significantly (when possible) to account for the power supply droop that exists as power output is increased. Fisher did as much in the SA-100, by specifying a quiescent current draw that ran the dissipation levels right up to the limits of the tubes -- the idea being that as voltage levels fell with increased power output, the output stage would then "fall" into the low distortion operating point established by RCA to maintain performance under conditions of high power output. Without any other means of economically regulating the potentials of the tubes, it was really the only means they had at their disposal. Today, enter EFB.
Because the EFB bias supply regulator reduces the bias voltage as screen voltage drops, the bias voltage can then be set for the true low distortion operating point under quiescent conditons, and the supply regulator will then maintain that operating point as the screen voltage falls with increasing power output. In this case, it allowed low distortion set point to be reduced from 31 ma for the stock design, to 24 ma for the EFB modified design. This allowed a significant 15%+ drop in output tube dissipation, resulting in extended tube life, and cooler operation.
Well, that's about that. With three different EFB installations now documented, and between them, all three of the different EFB regulators employed, I can finally turn my attention to my next article on EFB, wherein I will discuss when and where the various regulators are most appropriate to use. A break for the holidays before starting that article (to be published on Tronola), but with the book closed on the clone and SA-100 EFB projects now, I hope to start that effort sometime in December, time permitting.
With three regulators in the EFB family now, there is virtually no application that cannot use EFB to good advantage. Happy listening!
Dave
