Integrated Amp Reviews

To precondition an amplifier before testing, I run it at one-third power into 8 ohms for an hour. For an amplifier with a class-B output stage, this thermally stresses the amplifier to the maximum extent. To my alarm, after 15 minutes of driving 67Wpc into 8 ohms, the top panel of the KAV-400xi above the internal heatsinks was way too hot to touch, implying a temperature well above 60ºC; the chassis was also too hot to touch and smelled of hot insulation. More alarming, the distortion level, a good 0.05% when the amplifier was cold, had risen to 0.15% and was continuing to rise. Worried about the amplifier being damaged by thermal runaway, I concluded the preconditioning at that point.

The Krell is one of an increasing number of wide-dynamic-range amplifiers that do not have sufficient heatsink area to allow continuous high-power operation. This tradeoff may permit a product to be competitively priced—using appropriately sized heatsinks for a 200Wpc amplifier would significantly increase its cost—but it needs to be pointed out. While it is unlikely the Krell will overheat in normal use, it may well do so when used for a party.

The KAV-400xi's input impedance at 1kHz was a usefully high 47k ohms, double that for the balanced input. The amplifier was noninverting through both balanced and unbalanced inputs, and the maximum voltage gain was 35.7dB. (Assessed at the preamplifier output, which has a lowish source impedance of 198 ohms, the maximum preamp gain was a sensible 10.2dB.) The 152-position volume control operated in approximately 0.2dB steps at the extremes of its range, these increasing to 0.4dB steps in the middle of the range and 0.7dB steps around the unity-gain setting of "39." The "0" position mutes the output.

The output impedance was a little higher than usual for a solid-state design, at 0.35 ohm across most of the audioband, this rising slightly at 20kHz. As a result, the modification of the amplifier's frequency response reached ±0.25dB, when it drove Stereophile's simulated speaker load (fig.1, top dotted trace). Into resistive loads, the response was flat in the audioband, with a very slight rise below 20Hz and a high-frequency -3dB point of 134kHz, which correlates with excellent 1kHz and 10kHz squarewave shapes (figs.2 and 3), there not being any hint of overshoot or ringing. The response was the same through both balanced and unbalanced inputs, and was not affected by the setting of the volume control.

Fig.1 Krell KAV-400xi, balanced frequency response at 2.83V into (from top to bottom at 2kHz): simulated loudspeaker load, 8 ohms, 4 ohms, 2 ohms (0.5dB/vertical div., right channel dashed).

Fig.2 Krell KAV-400xi, small-signal 1kHz squarewave into 8 ohms.

Fig.3 Krell KAV-400xi, small-signal 10kHz squarewave into 8 ohms.

Channel separation (not shown) was excellent in the audioband, at better than 85dB, L-R, and 80dB, R-L, in the treble; beneath 1kHz, the crosstalk was buried in the noise floor. With the high gain, however, the amplifier's signal/noise ratio (ref. 1W into 8 ohms) with the input shorted and the volume control set to "151" was good rather than great, at 62.2dB, wideband. This increased to 71dB when A-weighted.

The noise presumably contributes to the shape of the THD+noise curves plotted against output power (fig.4), where the downward slope with increasing power is due to the reducing proportion of a constant noise level. However, looking at the distortion spuriae on an oscilloscope revealed there to be distortion harmonics in the noise, and, as I found with the preconditioning, this level increased as the amplifier got hotter. The important point to note from fig.4 is that the Krell comfortably exceeds its specified output power. No fewer than 290W were available with both channels driven into 8 ohms (24.6dBW) at our normal 1% THD definition of clipping, with 350W available into 4 ohms (22.4dBW). However, with one channel driven into 2 ohms, that channel's rear-panel 3A fuse blew at 500W output (21dBW), which is why the trace in fig.4 ends at that point. Replacing the fuse restored the amplifier to normal operation.

Fig.4 Krell KAV-400xi, distortion (%) vs 1kHz continuous output power into (from bottom to top at 1W): 8 ohms, 4 ohms, 2 ohms.

Concerned about the temperature-dependent nature of the amplifier's linearity, I measured the manner in which the KAV-400xi's THD+noise percentage varied with frequency at a moderately high level (16V) into 2, 4, and 8 ohms. The results are shown in fig.5: the audioband distortion is nicely below 0.1% into 8 ohms, with the right channel a little more linear than the left. A rise in THD above the audioband gets more severe into the lower impedances, but this is nothing to be concerned about. I then repeated the measurement after running the amplifier at 67Wpc into 8 ohms for 15 minutes. The results are shown in fig.6. Again, the rise in distortion above 20kHz can be seen, but now the audioband distortion has tripled!

Fig.5 Krell KAV-400xi when cold, THD+N (%) vs frequency at 16V into (from bottom to top): 8 ohms, 4 ohms, 2 ohms (right channel dashed).

Fig.6 Krell KAV-400xi when hot, THD+N (%) vs frequency at 16V into (from bottom to top): 8 ohms, 4 ohms (right channel dashed).

Of more subjective importance than the absolute level of an amplifier's distortion is its spectrum, with low-order harmonics being more benign than high-order. Unfortunately, the Krell's distortion waveform (fig.7) suggests the presence of both, the spikes in the waveform coinciding with the zero-crossing points. This measurement was taken with the amplifier very hot, hence the highish distortion level of 0.47%. The level was lower with the amplifier cold, but the spikes increased in amplitude as it warmed up.

Fig.7 Krell KAV-400xi when hot, 1kHz waveform at 2.3W into 4 ohms (top), 0.47% THD+N; distortion and noise waveform with fundamental notched out (bottom, not to scale).

This is shown in a different manner in fig.8, the spectrum of the amplifier's output while it drove a 50Hz tone at half power into 4 ohms. A regularly decreasing series of odd harmonics can be seen, interspersed with a lower-level series of even harmonics. The fact that the harmonics decrease in level with increasing order might mitigate their audibility, but I must admit to some alarm at seeing this behavior. Finally, the decrease in the circuit's linearity at high frequencies, referred to earlier, results in rather more intermodulation products than I like to see in the high-power, high-frequency intermodulation test (fig.9), taken just below the level that caused visible waveform clipping on the oscilloscope screen. However, this is at a very high output power; the spuriae fell rapidly with decreasing power.

Fig.8 Krell KAV-400xi, spectrum of 50Hz sinewave, DC-1kHz, at 100W into 4 ohms (linear frequency scale).

Fig.9 Krell KAV-400xi, HF intermodulation spectrum, DC-24kHz, 19+20kHz at 350W peak into 4 ohms (linear frequency scale).

Summing up the Krell KAV-400xi's measured performance is difficult, as some of its odd behavior will not be an issue when it comes to playing back music at normal listening levels. It is also possible that our sample was defective, though the fact that both channels behaved similarly is evidence for that not being the case. But I would avoid pairing the amplifier with loudspeakers that drop below 4 ohms—such as Krell's own Resolution 1—and would make sure it had adequate ventilation.—John Atkinson