Integrated Amp Reviews

I measured the Classé Sigma 2200i with my Audio Precision SYS2722 system (see the January 2008 "As We See It"). Out of the box, I ran into problems. Despite Classé's excellent manual, I couldn't at first define one of the Sigma 2200i's digital inputs to accept USB data. Eventually, I got the input to behave by defining the input, then turning the amplifier off and back on again. Then, I couldn't get the Sigma 2200i to accept a balanced signal at the input marked "XLR L1/R1." But then I remembered that, when I was setting up the USB input with the onscreen menu, I'd had to select a specific input connector for that input, in that case USB. Checking the setup menu for the XLR input, it had been set by a prior user to accept a signal from the unbalanced "Line 2" RCA jacks. Once I'd set it to accept a signal fed to the balanced XLR jacks, all was well—except that the amplifier now went into standby after running for 19 minutes. The manual says that "A power-save feature is enabled that puts the Sigma 2200i into standby after 20 minutes without a signal at its input." In my case, even with a signal present, the amplifier continued to turn itself off every 19 minutes.

I measured the Sigma 2200i's performance via its digital inputs using either S/PDIF signals fed from the SYS2722, or WAV and AIFF test-tone files played with Pure Music 3.0 by my MacBook Pro running on battery power and fed via USB. All measurements were taken at the speaker terminals. For the measurements taken to characterize the Classé's performance as a conventional amplifier, I fed a single-ended analog signal to the Line 2 inputs or a balanced signal to the XLR inputs.

Looking first at the Classé's performance in the digital domain: Apple's USB Prober utility identified the amplifier as the "Sigma 2200i 22A5A7" from "B&W Group Ltd," and confirmed that its serial number was "2270631." The USB input operated in the optimal isochronous asynchronous mode. Apple's AudioMIDI app indicated that the Classé operated at all PCM sample rates up to 192kHz via USB, with integer bit depths of up to 24. With the volume control set to "0.0," the amplifier's output level into 8 ohms with a 1kHz sinewave at –20dBFS was 9.415V, meaning that it would clip with full-scale digital data. I therefore set the volume control at "–13.5" for most of the digital input tests, increasing it to "0.0" for only the low-level tests.

The Classé's impulse response with data sampled at 44.1kHz (fig.1) indicates that its reconstruction filter is a conventional linear-phase type and that the amplifier as a whole inverts signal polarity. Tested with 44.1kHz-sampled white noise at –4dBFS (footnote 1), this filter can be seen to rapidly roll off the output above the Nyquist frequency (half the sample rate, shown by the vertical green line in fig.2), but there is an increase in ultrasonic noise above that frequency, this presumably due to the Sigma 2200i's PWM output stage. The harmonics of a full-scale 19.1kHz tone (fig.2, cyan and blue traces) are a little higher than I was expecting, with the third harmonic lying at –54dB (0.2%).

Fig.3 shows spectral analyses of the Classé's output while it decoded dithered 16- and 24-bit data representing a 1kHz tone at –90dBFS. With 16-bit data (cyan and magenta traces), the spectrum is dominated by the dither noise. However, with 24-bit data, while the noise floor has dropped by 24dB, suggesting resolution of at least 20 bits, harmonics of the fundamental tone are visible. This graph was taken with TosLink data; USB data gave an identical result. But with its low noise floor in the audioband, the Sigma's reproduction of an undithered 16-bit signal at exactly –90.31dBFS was superb (fig.4), with the three DC voltage levels described by the data and the ringing of the reconstruction filter clearly visible. Note that even though the traces in fig.4 were taken at the speaker terminals, there is zero DC offset visible. This is rare with amplifiers having a conventional output stage.

Tested for its digital inputs' rejection of word-clock jitter with 16-bit, 44.1kHz data, the Sigma 2200i didn't accentuate or attenuate the odd-order harmonics of the J-Test's low-frequency, LSB-level squarewave (fig.5, green line). However, the spectral peak that represents the high-level tone at one-quarter the sample rate has widened at its base, which suggests the presence of random low-frequency jitter. Also visible is a patch of spurious low-level tones between 9 and 10kHz that were present with 24-bit J-Test data (fig.6). A wider spectral analysis reveals these more clearly (fig.7), as well as an idle tone of unknown origin at 15kHz.

Also in fig.7 can be seen the start of the PWM output stage's ultrasonic noise floor. Because of the presence of this noise, when I measured the amplifier's performance using analog input signals, I used Audio Precision's high-order AES17 high-pass filter when measuring distortion, as otherwise the reading would be obscured by the noise.

The Sigma 2200i's overall gain via its analog inputs with the volume control set to "0.0" was 27.4dB balanced and 33.4dB unbalanced, the 6dB difference being in the opposite direction to what I usually find. All the analog inputs inverted absolute polarity, and the unbalanced input impedance was 46k ohms at 20Hz and 1kHz, 35k ohms at 20kHz. The balanced input impedance was a high 90k ohms at 20Hz and 1kHz, 84k ohms at 20kHz.

The analog inputs are converted to 96kHz-sampled digital, so the Sigma 2200i's frequency response measured via these inputs rolls off sharply above 40kHz (fig.8). The output impedance was 0.1 ohm at low and middle frequencies, but rose to 1.5 ohms at the top of the audioband. The response therefore peaks a little between 10 and 40kHz into higher impedances, which in turn results in overshoot with a 1kHz squarewave (fig.9). But with 2 ohms, the high output impedance at high frequencies results in an output that is down by 4dB at 20kHz (fig.8, green trace). Although there is very little change in the amplifier's frequency response below 1kHz with our standard simulated loudspeaker (fig.8, gray trace), the top-octave sound of the Sigma 2200i will vary more than usual, depending on the loudspeakers used. With a 10kHz squarewave, the 96kHz sample rate gives a band-limited response (fig.10) that also reveals the presence of the ultrasonic noise.

The effects of the Sigma 2200i's treble and bass tone controls, set to their maximum and minimum settings, is shown in fig.11. The boost and cut are sensibly chosen. I didn't test the effect of the Tilt control, as I couldn't manage to set this correctly using the onscreen menu. Channel separation was good, at >80dB L–R and >90dB R–L below 3kHz. Though the output was commendably free from power-supply–related spuriae (fig.12), the wideband signal/noise ratio, ref. 1W into 8 ohms and taken with the analog input shorted to ground and the volume control set to "0.0," was affected by the output stage's ultrasonic noise and was just 29.7dB (average of both channels). Restricting the measurement bandwidth to the audioband increased the S/N ratio to 58.1dB; A-weighting further increased it, to 75.75dB.

I then plotted how the percentage of THD+noise varied with output power, using the AES17 low-pass filter set to 20kHz. The Classé offered very low distortion and high power below clipping into 8 ohms (fig.13) and 4 ohms (fig.14). These graphs reveal that the amplifier clipped at 215Wpc into 8 ohms (23.3dBW) and 410Wpc into 4 ohms (23.1dBW), both powers slightly higher than the specified 23dBW.

Fig.15 shows how the THD+N percentage varied with frequency at 20V, which is equivalent to 50W into 8 ohms, 100W into 4 ohms, and 200W into 2 ohms. The AES17 low-pass filter was again set to 20kHz, which is why the graph ends at 5kHz rather than the usual 20kHz, in order not to eliminate the higher harmonics. The distortion is very low into 8 ohms (blue and red traces), with a slight rise in the treble. The THD+N percentage rises more at high frequencies into lower impedances, with the 2 ohm measurement higher than 0.1% above 1kHz (trace incomplete because the Classé went into standby at that point). The distortion at this power level seems primarily to be the third harmonic (fig.16), accompanied by some lower-level, higher odd-order harmonics (fig.17). Assessed with my usual mix of equal-level 19 and 20kHz tones, the intermodulation distortion was very low (fig.18). However, you can see in this graph the start of the rise in the ultrasonic noise floor, examined in more detail in fig.19.

Its use of a PWM output stage doesn't seem to have compromised the Classé Sigma 2200i's measured performance, and the amplifier offers high power in an attractively lightweight package. But I don't think it would be a good match for loudspeakers that would present it with impedances as low as 2 ohms.—John Atkinson