Krell KSA-200S power amplifier Measurements

Sidebar 3: Measurements

An idling power consumption of 125W was noted, close to spec and confirming this amplifier's moderate standby consumption, especially in view of the high power and current capabilities. That massive transformer was mechanically silent even on the UK's 50Hz AC supply.

There was no denying this amplifier's big heart. The KSA-200S is rated as a 23dBW design, which is equivalent to an output power of 200Wpc, with one channel driven. The output is claimed to double with each successive halving of load impedance, right down to 1 ohm. Such was the reserve present in its power supply, the Krell cared little whether it was driven stereo or mono, or with 8, 4, or 2 ohms—double monoblocks would show negligible advantage in power terms. With the normal wall outlet supply, within 3% of the rated line voltage, the amplifier happily pushed out a generous 280Wpc into 8 ohms (24.5dBW) over essentially the whole 20Hz–20kHz power bandwidth. The specified power doubling was achieved into 4 and 2 ohms. Into the latter, for example, the Krell would deliver 23.4dBW, or a true 852W! With the available wall current, I could not quite hit the 1600W/1 ohm jackpot; I had to be satisfied with 1560W, a trivial shortfall of precisely 0.1dB.

In addition to a peak program output close to 300Wpc, the peak current capability approached 50A, 96A peak–peak. These ratings should be more than sufficient for any application or load, save a 0.1 ohm ribbon missing its matching transformer!

Output impedance was not fantastically low. It was similar, in fact, to that of the KSA-100S at 0.126 ohm, this pretty constant over the frequency range. Although the specification quotes a figure of 80, this measured output impedance is equivalent to an 8 ohm damping factor of 62. (Values above 40 have little effect on sound quality.)

The Krell KSA-200S's frequency response (fig.1) was wide and flat, from near-DC to 180kHz, –3dB. Within tighter +0, –0.5dB limits, the response ran from below 0.5Hz to 61kHz. Investigating stability, I judged that this was one of Krell's most stable designs to date. The 400Hz squarewave response was first-rate, with a short risetime. On the graph shown (fig.2), the amplifier was barely troubled by having to drive a naked 2µF capacitor: overshoot was well-controlled, ringing negligible. Tough electrostatic speaker loads will hold no fears for this amplifier.

Fig.1 Krell KSA-200S, frequency response at 2.83V into 8 ohms (1dB/vertical div.).

Fig.2 Krell KSA-200S, 400Hz squarewave into a 2µF capacitor.

Channel separation was excellent, at better than 85dB at 20kHz, and beyond 100dB below 5kHz. With results like these, who needs monoblocks?

Noise was low, the S/N ratio measuring 108dBA referenced to full power and better than 80dB, unweighted including hum, referenced to 1W (IHF). Channel balance was excellent at 0.02dB. Easy to drive, the amplifier's input impedance was 47k ohms (100k balanced), with approximately 80pF shunt capacitance.

An input of 150mV, unbalanced, gives you 1W output; 2.5V, unbalanced (5V differential balanced), hits full-power clip—close to 300W. DC offsets were very low: 5mV or less for either channel.

Distortion was on target at rated power, and reminiscent of a well-designed tube power amplifier at a moderate 0.1% (–61dB) at 1kHz. It was a little better at 20Hz, but 10dB poorer at 20kHz, measuring 0.33% (–50dB). (This was to spec as quoted in the Krell manual, but not to that in their brochure, which claimed 0.2%.) The distortion levels improved at lower power levels, measuring –70dB, or 0.033%, at 1W into 8 ohms (0dBW). The total harmonic distortion was pretty consistent from 20Hz to 20kHz.

A representative graph (fig.3) is given for distortion vs frequency at a moderate power level: 20W into 8 ohms. I tried both balanced and unbalanced connections, with the same fine results. The wide intrinsic bandwidth was shown in the relatively slow rate of distortion change at the frequency extremes, especially above 20kHz: 0.3% distortion at 100kHz at 20W is not easy!

Fig.3 Krell KSA-200S, THD+noise ;wvs$w frequency at 20W into 8 ohms, unbalanced input.

One would expect distortion to increase with reducing load impedance; generally, an amplifier's distortion is proportional to its output current, as it is in a loudspeaker. A 1 ohm load draws 8 times the current of an 8 ohm load—about +18dB. Roughly, a –60dB distortion figure with the amplifier driving its rated output power into the higher impedance would degrade by at least 18dB when that load dropped to 1 ohm. I found this indeed to be the case with the '200S. Low-order output-stage distortion did set in with loads below 3 ohms, and the maximum pre-clip output level into 1 ohm, 22.9dBW, was achieved with a fairly high 3% (–30dB) of distortion. On program peaks where adverse loading actually makes its mark, such momentary moderate distortion values are likely to be inaudible.

When I reviewed the less powerful KSA-100S recently for HFN/RR (footnote 1), I looked for any evidence to support the proposition that small changes in sound quality might track the amplifier's output-stage class-A bias level. On an average-grade AC supply, the '100 did show subtle shifts in transparency according to the bias plateau. Fortunately, with a good dedicated supply, this effect decreased into virtual inaudibility. Whether due to its larger supply or its greater power envelope, I didn't notice this subjective effect with the KSA-200S.

It was still possible to measure such differences, however. While there was no significant relation between distortion and bias level at 1W output power, some dependence became apparent by 20W. For example, with the appropriate two-LED bias, the measured THD lay at –70dB; when an earlier musical peak had increased the bias to the four-LED level, the measured distortion reduced to –80dB.

Fig.4 shows the amplifier's output spectrum driving 200Hz at a 1W level into 8 ohms, with a bias level equivalent to three LEDs lit. The benign second harmonic, at 400Hz, can be seen to be desirably dominant. An evenly decaying progression of negligible higher harmonics is marked with crosses. (For the sake of clarity, the fundamental has been raised by 10dB in this graph; the components at 100Hz, 250Hz, and 350Hz are due to a trace of UK 50Hz mains leakage.)

Fig.4 Krell KSA-200S, spectrum of 200Hz sinewave, DC–2kHz, at 1W into 8 ohms (linear frequency scale). Note that the second harmonic at 400Hz is the highest in level, at –80dB (0.01%).

At a higher power, 75W (fig.5), the third harmonic at 600Hz becomes dominant, in this case lying at –60dB with respect to the fundamental level. Higher bias levels reduced the level of this harmonic somewhat. While the second harmonic was still present, the remaining higher harmonics were held to a low level: –75dB or less. This is a positive result for those who hold that distortion spectra correlate with tonal quality.

Fig.5 Krell KSA-200S, spectrum of 200Hz sinewave, DC–2kHz, at 75W into 8 ohms (linear frequency scale). Note that the third harmonic at 600Hz is the highest in level, at –60dB (0.1%).

Supply modulation—the ability of an amplifier to reject power-supply components under heavy drive—is explored in fig.6. Driven with a 38Hz fundamental at two-thirds the '200S's rated output into a 4 ohm load, the amplifier had a benign distortion spectrum with no trace of unwanted 50Hz AC line components above the –90B baseline. This is a very good result.

Fig.6 Krell KSA-200S, spectrum of 38Hz sinewave, DC–500Hz, at 267W into 4 ohms (linear frequency scale). Note that the third harmonic is again the highest in level, but there are no 50Hz components visible.

Two results for high-frequency intermodulation using a 19+20kHz drive signal are presented in figs.7 (1W) and 8 (full power). The 1kHz difference component was measured at –76dB for 1W output (improving to –86dB with a higher bias level), and –71.7dB at full power. In other respects, these spectra are clean.

Fig.7 Krell KSA-200S, HF intermodulation spectrum, DC–25.5kHz, 19+20kHz at 1W into 8 ohms (linear frequency scale). Note that the 1kHz difference component lies at –76dB. This drops to –86dB with a higher bias level.

Fig.8 Krell KSA-200S, HF intermodulation spectrum, DC–25.5kHz, 19+20kHz at full power into 8 ohms (linear frequency scale). Note that the 1kHz difference component lies at –71.7dB.

All in all, this is a great set of measured results from an experienced designer.—Martin Colloms

Footnote 1: Vol.39 No.4, April 1994, p.31.
Krell Industries
45 Connair Drive
Orange, CT 06477
(203) 799-9954