Krell KSA-300S power amplifier
As he entered his listening room, Sam recoiled in horror. Psycho-esque strings crashed into the silence of his imagination as the chill of the room hit him. On the floor between the loudspeakers was not his trusty Krell KSA-250, but a new KSA-300S. But how? A belated visit from a practical-joking Santa and eight tiny, but persistent, reindeer? A swap made by a weary felon who just couldn't carry the new 300S's 185 lbs any farther and decided that a 145-lb 250 would be a relief? An unannounced home trial from trusted dealer/friend and part-time Olympic dead-lifter Jim, who only the other day had nearly convinced Sam that he just had to try out this big, new Krell amplifier?
Sam gingerly approached the interloper, its blue pilot light a cold reminder of the drafts he had to navigate as he crossed the room. He felt the heatsinkshardly even warm. His shivering spirits were briefly calmed by holographic images of cool, summer listening sessions. "But where's the fun in that?" he blurted.
As our hero has just discovered, owning a Krell amplifier may never be the same. The new KSA-300S reviewed herealong with its companion piece, the KRC preamplifieris, in at least one respect, a dramatic change from earlier Krells: it no longer doubles as a space-heater. But Krell has not abandoned the concept on which it built its reputation: class-A operation.
Class-A amplifier operation is extremely inefficient, in terms of both build cost and power consumption. While class-A is often put forward as a sonic touchstone, not all experts agree that the gains are worth the trouble and expense. In class-A operation, the output transistors are conducting at all times, the current through them only dropping to zero (one per cycle) at maximum power. This requires large amounts of power dissipation in the amplifier's heatsinks, even with no input.
The class-A amplifier does, however, operate with the maximum linearity of which it is inherently capable, and little feedback is generally required to achieve acceptable distortion levels. Most audio power amplifiers operate in class-AB, in which each half of the push-pull output stage conducts for slightly more than a half cycle. This allows them to operate in class-A up to some small percentage of their total rated power. Even the Krell KSA-250Krell's top, last-generation stereo amplifierwas subject to this limitation. (See JA's footnote to Lewis Lipnick's and Robert Harley's review of the KSA-250 in Vol.14 No.1, pp.179180 for more on this.) To operate in full class-A up to 250Wpc output at 8 ohms (the KSA-250's power rating) would require unthinkably large heatsinks. But the earlier Krell amplifiers still operated deeply enough into class-A to run very warmeven hotat idle. When, just before the model was discontinued, Stereophile's early KSA-250 was exchanged early in 1993 for the most recent version, the new sample ran decidedly hotter than the oldan indication that it was biased more deeply into class-A than the earlier samplethe one we measured for our 1991 review.
When they designed their new Audio Standard monoblock amplifier, Krell realized that a new approach was needed to maintain class-A performance while at the same time increasing efficiency. The so-called sliding bias designin which the bias fluctuates rapidly up and down in response to the input's requirementshas been available for a number of years. Krell rejected this approach in favor of a related, though unique and perhaps more elegant, concept, which they have dubbed Sustained Plateau Bias. Following its successful implementation in the $32,500/pair Audio Standard amplifier, the technology has been incorporated into Krell's new "standard line," S-Series amplifiers. Each of these, the KSA-300S included, is designed with five bias "plateaus." Circuitry within the amplifier analyzes the input signal and selects an appropriate bias level to keep the amplifier operating in class-A.
The most obvious question here is: How is the amplifier able to respond fast enough to keep it from being "fooled" by the input into setting too low a bias and thereby dropping into class-B? According to Krell, the bias is set by what they call an "Anticipator" circuit, whose fast (1800V/µs) slew-rate is said to be at least ten times faster than the maximum slew-rate of any musical signal being handled by the output stage. By the time the output voltage has risen, therefore, the bias has already been increased to the appropriate level.
While an upward jump to a higher bias plateau can occur at any time the input demands, a drop to a lower level does not occur immediately upon a drop in demand; plateaus are sustained, by predetermined time constants, long enough to ensure that the drop is likely to be prolonged. While a restoration of higher demand will "reset the clock" and maintain the present (or higher) bias, there are bound to be instancesgiven music's unpredictable naturein which the bias is dropping just as a new musical transient is coming along. In this case, it would appear that the fast slew-rate of the amplifier will again come into play, reversing the drop and reestablishing the higher bias. Krell's promotional literature states that "downward changes are made only when the next lowest bias level would accommodate the input for approximately 2030 seconds"this would clearly be "anticipation" more worthy of a psychic than an electronic circuit. The KSA-300S's owner's manual statesmore reasonably, I thinkthat the plateaus are held for 1520 seconds unless continuing demand requires otherwise.
In any event, the new design topology makes the KSA-300S a cool operator except when called on to deliver high power, and its heatsinking is more than up to the latter. If, however, under high demand its heatsinks reach a temperature of 80°C, the top two bias levels are disabled until the heatsink temperature drops. In this case, the amplifier will still put out its rated power, but in class-AB, not class-Awith the bias rising no higher than the second plateau. Indicator lights on the front panel show both power-on (in blue) and plateau levels for each channel (the plateau changes are performed independently for both). According to the owner's manual, the first bias level indicates approximately 25% of rated power, the second 50%, the third 75%, and the fourth 100% (or 300W into 8 ohms). Krell must have changed this subsequent to the manual's printing, however, as our measurements showed that the bias-plateau changes occurred at significantly lower levels (see "Measurements"). In fact, the top bias level never came on in any of my auditioning, and the third level was triggered only occasionally (footnote 1).
Footnote 1: Listening to the excellent sound of Apogee Divas being driven by the similar but lower-rated (200Wpc) KSA-200S at Hudson's Audio in Albuquerque, I did note that four of the bias levels were being exercised at loud but not-unreasonable sound levels.John Atkinson