Krell Full Power Balanced 600 power amplifier Measurements

Sidebar 3: Measurements

When you buy a big, expensive power amplifier, you expect it to deliver. Krell has a tradition of generosity when it comes to power, and their specifications are often some of the most conservative in the business. Having said that, it's common for most conventional amplifiers to have inherent power reserves. With line voltage varying from region to region and country to country, a reserve has to be incorporated to help meet the published spec, regardless of such difficulties.

In theory, this reserve is unnecessary for an amplifier that has regulated output stage supplies; the regulation holds the rated output power regardless of line variation. Indeed, when I measured the KAS-2's maximum power output, this was very close to its specified level: within 0.3dB. However, it appears that the FPB design team decided not limit the available power to that specified.

For example, when I tested the FPB 300, this nominal 300W amplifier actually produced 470Wpc into my 7.5 ohm high-power test load. The '600 matched this achievement by measuring 945Wpc continuous into the same load (935Wpc into a scaled 8 ohms). This is a huge output, well nigh on 30dBW, and sounded it! Rated output level is 28dBW, 600W, 8 ohms—and this amplifier could hold to this level at all loads and frequencies, 20Hz-20kHz, 8 ohms down to 2 ohms.

I wasn't able to run my long-term continuous testing at 2 ohms, but compromised with five-second bursts—long by peak-measurement practice (eg, 20ms). The FPB 600 could sustain a 29.3dBW level into this load, corresponding to 3.4kWpc—an extraordinary figure.

Driven on a toneburst equivalent to peak program duty at 8 ohms, it reached to touch the 1kW line, while at 4 ohms it attained 1.85kW, and for 2 ohms 3.53kW. And for 1 ohm—wait for it—an amazing 6kW! These are single-channel results, but, measured as short-term ratings, they should be available from both channels simultaneously.

For the record, when both channels were operated into 4 ohm loads the output fell by typically 0.3dB from the singly driven 8 ohm result, 20Hz-20kHz, confirming the superb power bandwidth.

This amplifier obviously needs to be used with some caution. While it is well known that loudspeakers are more readily destroyed by smaller amplifiers when overdriven, the fact is that the FPB 600 sounds very clean when run loud—it's all too easy to reach life-threatening levels for your speakers! Further, while I found the plastic-shrouded binding posts a nuisance, I'm well aware of the relatively high voltages this amplifier can provide. In clip, the AC level on the output terminals was 86.4V RMS. Some thought should be given to the other end of the speaker cable; perhaps it should be shrouded in insulation, except at the point of contact to the speaker terminal.

The FPB 600 can vaporize some of the smaller speaker cables if inadvertently short-circuited at the speaker end. I measured a huge peak current of >115A into 1 ohm and stopped there, though more might well be available. If you give up listening, you could always use the amplifier for welding!

I wanted to examine some aspects of measured performance that could give some insight into the improvements in sound quality evident as the amplifier warmed up, and also when it was run-in from first switch-on after shipment.

First, I ran a high-resolution sweep of distortion vs frequency for an input level equivalent to 10W into 8 ohms, from first switch-on, for loads of 8 and 2 ohms (fig.1). Then, after the amp had been on for a few hours and was well exercised and thermally stable, I repeated the test. First, there was a clear difference in distortion level between the 8 and 2 ohm loads, typical of most amplifiers and no doubt associated with subjective observation of mild losses in amplifier sound quality when faced with loudspeakers of lower impedance. There's no free lunch where loudspeaker impedance is concerned: Artificially boosting the voltage sensitivity of a speaker while compromising the load impedance exacts a price in amplifier sound quality.

Fig.1 Krell FPB 600, THD+noise vs frequency at (from top to bottom at 10kHz): 40W into 2 ohms, new/cold; 40W into 2 ohms, warmed-up; 10W into 8 ohms, new/cold; 10W into 8 ohms, warmed-up.

When tested for 8 ohms before running in, the harmonic distortion remained at the -89dB level, improving a few dB above 5kHz. For the 2 ohm condition the distortion was some 25dB poorer, essentially stable at -65dB, or 0.06%. After the amplifier had been broken in and was warmed up, both sets of readings improved by several dB, the 2 ohm result by 6-7dB, though with less obvious improvement at the highest frequencies. Clearly, there is a difference; some electronic mechanism was at work here which could be directly or indirectly associated with sound variations.

Another factor, separately analyzed, was the effect of transient bias levels—those plateau bias conditions. For example, if high-frequency two-tone intermodulation was measured at 1W, the result was -85dB—fair enough, and sufficiently low to allay concern. Yet when preconditioned to a 40W level for a moment, in order to seek a higher bias plateau, the 0dBW intermodulation figure improved to -90dB, 0.003%, with the higher bias level clearly providing still greater linearity at the output stage. At this higher bias level, the intermodulation distortion was in fact load-independent, still -91dB into a more demanding 2 ohms. At full power, the high-frequency intermodulation remained excellent at -93.45dB, or 0.0021%.

At 50W into 8 ohms, total harmonic distortion in the midband was -82dB—essentially second and third harmonic. If the next bias plateau was engaged using a brief power kick, the distortion then fell to -95dB. By implication, a crazy route to produce a still more linear amplifier would be to quadruple the heatsink capacity and crank up the baseline bias levels. Conversely, the customer would have to pay heavily in increased electricity consumption and excess heat.

Fig.2 shows the harmonic spectrum for 20W into 8 ohms (upper trace) and 20W into 2 ohms (lower trace). (The 1kHz fundamental has been notched out in these graphs.) Note that the proportion of odd-order harmonics (3rd, 5th, 7th) increases with the 2 ohm load. Loaded with 8 ohms, the amplifier's output spectrum has a "sweeter" balance of even and odd powers. The actual distortion waveform, at 1W into 8 ohms, is shown as the upper trace in fig.3 (the fundamental is the graph's lower trace). Of significance here is the lack of crossover artifacts; the waveform is of a class-A nature.

Fig.2 Krell FPB 600, spectrum of 1kHz sinewave, DC-10kHz, at 20W into 8 ohms (upper trace) and 20W into 2 ohms (lower trace) (linear frequency scale, 10dB/vertical div.).

Fig.3 Krell FPB 600, 1kHz waveform at 1W into 8 ohms; distortion and noise waveform with fundamental notched out (top, not to scale).

Tested at rated power into 8 ohms, the distortion was typically -80dB or 0.01% in the bass and midrange, rising to -66dB or 0.05% by 20kHz—still inconsequential. At the 0dBW level (1W, 8 ohms), the residual noise floor of the test prevented measurement below the thresholds of -86dB at 20Hz, -88dB at 1kHz, and -82dB at 20kHz. These are all very fine results, requiring no qualification.

Harmonic analysis for the distortion "character" at a 1W level, 8 ohms, with a 170Hz signal (fig.4) proved difficult, as the relevant distortion fell below the baseline set by the residual level of the power-line harmonics (50Hz based at -95dB relative to 1W). It's just possible to see second-harmonic distortion at -110dB, and I have highlighted the third harmonic, also at -110dB, with the marker. Clearly, at lower power levels the FPB 600's distortion is vanishingly low, despite the absence of global loop feedback encompassing the output stage.

Fig.4 Krell FPB 600, spectrum of 170Hz sinewave, DC-1kHz, at 1W into 8 ohms (linear frequency scale, 10dB/vertical div.).

Fig.5 shows the Krell's output spectrum driving a 37Hz fundamental at 2/3-rated power, both into 4 ohms (dashed trace) and with no load. The graph shows the variation in distortion from no load to nearly full load, yet there was no detectable intrusion of power-supply harmonics—in view of the advanced power-supply regulation, just as you'd hope. Also worthy of note was the clipped waveform, which was very clean, free from power-supply ripple noise (as it should be for a regulated supply), and showing no latching or delayed recovery.

Fig.5 Krell FPB 600, spectrum of 37Hz sinewave, DC-1kHz, at 800W into 4 ohms (dashed trace) and at the same output voltage but unloaded (solid trace) (linear frequency scale, 10dB/vertical div.).

The FPB 600's output impedance was somewhat lower than the '300's, the latter reading 0.04 ohm, the bigger model 0.03 ohm, both of these essentially constant with frequency and entirely negligible in the context of speaker and cable loadings. This guarantees that there will be no shifts in frequency response related to speaker impedance.

With a very big amplifiers such as this, electrical background noise will be a consideration with sensitive speakers. First, mechanical transformer noise was very low, if not as superbly low as my FPB 300 sample. (This was on the UK's 50Hz lines; it is likely to be still quieter with the US's 60Hz supply.) Second, even with my ear on the speaker grilles, there was no audible hiss or hum from the amplifier. Of course, I can't vouch for some of the electronics with which it may be used.

For the noise measurement, the IHF 1W, 0dBW reference level was a good starting point. I got an excellent 87.3dB unweighted S/N Ratio, 90dB A-weighted. Near-state-of-the-art figures were obtained referred to full power, 116.6dB unweighted, with only 2dB worth of hum at this level. A-weighted, the FPB 600 hit the magic figure of 120dB for noise relative to full power.

DC offset at the output was negligible, the meter readings hovering at the 1mV level, reacting to a trace of infrasonic noise.

When run balanced, a 141mV input gave 1W output, but for full power 4.3V were required; ideally, you will need a source or preamp capable of 8V output or more, and with at least 12dB of gain for line inputs. On the plus side, the FPB 600 has a kind input loading of 85k ohm, with around 100 picofarads of capacitance, suitable for tube, FET, and solid-state sources. I got equally good results with transistor and tube sources.

Channel balance held to within ±0.01dB, while the frequency response was dead flat, almost to DC, reaching to below 1Hz for -0.5dB and to 78kHz at high frequencies. The -3dB point was set high at 200kHz, if a little lower than the 300kHz bandwidth recorded for the FPB 300.

Thumped with fast squarewaves, the '600 behaved very well, a pure 2µF capacitance load giving rise to a minor 15% of overshoot accompanied by fast ringing that lasted only 4% of the cycle time (fig.6). No visible effect whatsoever was evinced by 0.1µF alone. Once the usual 8 ohm resistor was added to the capacitance loading, the result confirmed the high stability margins achieved by this amplifier.

Fig.6 Krell FPB 600, small-signal 10kHz squarewave into 2µF.

Channel separation was excellent: better than 100dB in the lower-frequency range, and still an excellent 93dB at 20kHz (with my present test arrangement).

All in all, this most excellent set of test results describes a highly neutral, superbly linear, highly load-tolerant, and very powerful amplifier.—Martin Colloms

X