Krell KSA-200S power amplifier Page 2

Early amplifier stages have fully regulated supplies, later ones distributed wide-bandwidth supply decoupling.

Once upon a time, Krell power amplifiers claimed and achieved low distortion factors. This wasn't a specific design feature, but a byproduct of their inherently linear circuitry. As good-sounding modern amplifiers continue to demonstrate, the link between sound quality and low distortion figures measured in the lab with continuous tones is a tenuous one. A low-feedback design will generally produce higher distortion; Krell has no qualms in revealing figures for full-power total harmonic content of 0.1% at 1kHz and 0.2–0.5% at 20kHz. Five years ago, it would have been felt that these figures could well have been 20 times lower. In this respect, and in the low-feedback philosophy, it's interesting to observe the convergence of solid-state and tube designs.

As power amplifiers improve in performance quality, they make growing demands on the other components in the system. Obviously the ancillary components, interconnects, and loudspeaker cables should be sufficiently neutral and transparent to allow the amplifier's performance to be heard. However, one area remains often neglected—the quality of the raw electrical power fed to the amplifier. Those IEC-type detachable power cords may be a convenience, but at best they're rated at 10A continuous. While this rating will not result in a failure mode, it is rather close to the maximum needs of a high-power amplifier. A KSA-200 in full song driving a tough loudspeaker load may well draw over 1kW from the wall, equivalent to 10A "RMS" with US 110V supplies. (The peak current, drawn when the amplifier's power-supply rectifiers conduct, will actually be much higher.)

In the UK, the supply voltage is double that in the US. While this can be lethal when mishandled, it does have the advantage of halving the current values. It's possible that some differences in amplifier sound quality could consequently arise in marginal situations.

Using a high-capacity power cord helps with amplifiers like the KSA-200S. Do not use a substitute for the Krell-supplied cord unless a genuinely better one is available.

The quality of the wall outlet also becomes a significant factor. The KSA-200S certainly gets on with the job when plugged into almost any AC supply. However, once you've used a dedicated power line running back to the main circuit-breaker box, the improvement cannot be ignored. An improvement of around 8–12% of sound quality with a $7500 amplifier is not something you lightly forego.

Plateau Biasing
How is a modern, high-performance, "class-A" amplifier designed to have a low, "green," standby power consumption?

Obviously, the class-A designation requires that the dynamic music envelope fits within a static, class-A current and voltage combination for a rated load impedance. Many "dynamic" or pseudo–class-A designs do not feature true class-A operation for their output stages. Generally, the "dynamic," "sliding-bias," and "Stasis" amplifier designs feature signal-controlled biasing of the output stages. The bias level is wound up according to the signal's current demand, with its minimum value carefully controlled to avoid classic class-B crossover distortion. Complex circuitry is common, with unwanted, signal-related modulation effects sometimes apparent in the output waveform.

In the case of the current Krells, several factors working in combination appear to redefine the topic of dynamic biasing. The key to the design was the use of an ultra-fast bias-control circuit. The technology was borrowed from the high-speed, current-feedback amplifiers used for I/V conversion in digital systems.

Using current-feedback technology, this bias circuit looks ahead at the music envelope, and predicts and then sets the output stage's equivalent class-A bias levels before that same music's leading edge has actually passed through the amplifier's circuit. When it finally arrives at the end of the relatively slower main amplifier chain, the required bias level has been anticipated; the output stage is therefore in a relatively stable, static condition of class-A bias current.

Krell has avoided the worst music-related modulation effects by learning from the psychoacoustic aspects of the design of compressor/limiters. Such limiters have a fast attack and a slow decay. Similarly, the slow decay of the Krell's bias level after a music transient has passed allows the whole decay envelope of the music to reach completion long before any bias shifts occur. A very generous delay of 20 to 30 seconds is provided before any downshifting of bias occurs.

Another factor reducing the potential audibility of such program-related "gear changing" is the concept of "plateau biasing" at one of five appropriate levels. This means that the bias level is not constantly dynamic, but varies with the program power envelope only over a few preset levels.

The bias in this Krell series of designs settles at one of the five levels for an extended period, and is not switched up or down unless the longer-term power demand needs it. This means that the bias level is not being switched very often, and then only for an instant of transition. Thus, signal-induced distortions, if present, can only occur for a tiny fraction of the time. The prolonged plateau bias levels do, in fact, imply a very high proportion of true, non-varying class-A operation.

Concerning the relative "speeds" of the power amplifier and the bias anticipator, Krell correctly reminds us that the maximum slew rate found in high-level recorded music is a modest 10V/µs. The "S"-amplifiers have an intrinsic slew rate well above that at 100V/µs, ensuring a generous safety margin. On the other hand, the anticipator circuit operates at a "speed" in excess of 1000V/µs, well in advance of the main amplifier.

Brute Force
A key aspect of all Krell power-amplifier designs since their first appearance has been their large power supplies. A Krell amplifier is heavy—the KSA-200S weighs 100 lbs—its mass reflecting the sheer mass of iron ribbon making up the core of its one or more toroidal AC transformers.

Literally and metaphorically, that iron is the very core of the amplifier: it is the critical path for energy transfer in the magnetic domain from the AC wall supply to the amplifier's output circuit, and the very source of motive power. To draw an automotive analogy, if the AC supply is the gas, the transformer and its iron core are equivalent to the car's engine. (The reservoir capacitors are the flywheel.)

A small amplifier has a small iron core, or "motor"; no matter how hard it's driven, at some point it chokes up, the magnetic energy field of the transformer core collapses, and no more energy can be taken from the supply. More energy or torque can be delivered for a short period by greatly increasing the reservoir capacitance (in car terms, this is equivalent to enlarging the flywheel). Though such an amplifier can have a high peak power, it still runs out of power on sustained drive. In addition, it will only recover slowly from heavy drain conditions. This tends to produce a sluggish effect on musical rhythms. Similarly, a small car engine fitted with an oversized flywheel is slow to reach peak revolutions, and is sluggish in response to the gas pedal.

Contrast this with a motor endowed with generous torque, freely converting the available gasoline energy into easy-revving high power and torque. Near-instantaneous accelerative shove is provided on demand; the power is effortless, regardless of payload or incline. In power-amplifier terms, this means a generously large AC transformer with appropriate, not oversized, reservoir capacitors.

The speed at which those reservoirs are brought up to full stored power condition is also important. The Krell's transformer is rated at a maximum power delivery of 2800VA, with respect to a pure resistive load in combination with a defined level of regulation. (This refers to the permissible dip in output voltage on load. Krell toroids are high-regulation types with little voltage dip.) In practice, the effective transformer rating when it is actually in use in an amplifier is probably 60–70% of this published maximum.

The KSA-200S is specified as being able to deliver 200Wpc into 8 ohms; this implies 800W of raw power being drawn from the wall, assuming a 50% efficiency for pure class-A operation. Yet the Krell's real-life transformer rating is more than treble this. It is, of course, even possible to take a cheaper design and deliver 200Wpc to a load in class-AB from a lower-rated, heavily loaded power supply using a 750VA toroidal transformer. The KSA-200S appears to have a magnetic core four times larger than is required for its specified power output.

Krell explains that their specification defines a pure voltage output which is maintained into successive halving of the load value, to 4 ohms, 2 ohms, and even 1 ohm (where the '200S's rated output power is 1600Wpc!). Only a very-low-impedance power supply combined with an ultra-high-current output stage and low-impedance internal wiring can support such performance. At its rated level into 1 ohm, the amplifier is putting out 3200W, for which it will require at least 4000W from the 2800VA toroid.

At first sight, into a 1 ohm impedance, it might appear that the amplifier is capable of supplying more power out than is actually going in. The amplifier's output stage may have shifted to higher-efficiency class-AB operation, but the toroid is beginning to suffer, and no wonder. But such is the conservative overall rating of the overall design that it can sustain this level of abuse! It isn't intended that a 1 ohm load should ever be applied indefinitely, but such performance helps define the boundaries within which the amplifier will still perform without gross overload.

Krell Industries
45 Connair Drive
Orange, CT 06477
(203) 799-9954