Classé CP-800 D/A preamplifier Measurements
Sidebar 3: Measurements
I measured the Classé CP-800 with Stereophile's loan sample of the top-of-the-line Audio Precision SYS2722 system (see www.ap.com and the January 2008 "As We See It"); for some tests, I also used my vintage Audio Precision System One Dual Domain.
Looking first at the CP-800's performance via its digital inputs, with the volume control locked to unity analog gain or "0.0" with the Input Passthrough setting and in Analog Bypass mode, the AES/EBU and S/PDIF inputs successfully locked to data with sample rates ranging from 32 to 192kHz. However, as can be seen in fig.1, the frequency response with 192kHz data was no wider than with 96kHz data, both being down by 6dB at 42kHz. With lower sample rates there was a fraction-of-a-dB rolloff at the top of the audioband before the steep drop in output due to the reconstruction filter. The Mac USB Prober utility reported that the USB input operated as claimed in isochronous asynchronous mode, and handled sample rates of 32, 44.1, 48, 88.2, and 96kHz, with 24-bit word lengths. USB Prober identified the product as the "CP-800" from "Classe Audio Inc," with a serial number of "2144108."
With the volume at "0.0," a full-scale signal at 1kHz resulted in a level of 3.67V at the balanced outputs, 1.83V at the single-ended outputs; both preserved absolute polarity. Channel separation via the digital inputs was superb, at >120dB below 2kHz. With 16-bit data representing a dithered 1kHz tone at 90dBFS, the USB, S/PDIF, and AES/EBU inputs gave a spectrum with the tone at the correct level and a noise floor free from supply-related or harmonic spuriae (fig.2, cyan and magenta traces). Increasing the bit depth to 24 dropped the noise floor by 10dB or so (fig.2, blue and red traces), implying resolution of around 18 bits. This graph also indicates that the left channel (blue trace) was slightly noisier than the right (red) at low frequencies, this confirmed by 1/3-octave analysis (fig.3). Even so, the CP-800's reproduction of an undithered tone at exactly 90.31dBFS (fig.4) showed excellent differentiation of the three DC voltage levels. With 24-bit undithered data (fig.5), the CP-800 produced an excellent if slightly noisy sinewave.
Although the level of distortion with a full-scale digital input signal was low in absolute terms, it was higher than I expected, the third harmonic lying at 77dB, or 0.014% (fig.6). Reducing the output level by 10dB with the volume control preserved the level of the third and other harmonics (fig.7), whereas reducing the level of the signal data by the same 10dB dropped the third harmonic to 87dB, or 0.005% (fig.8). This suggests that the slight nonlinearity occurs in the D/A conversion circuitry ahead of the volume control. But, as I said, the distortion was still low, and with an equal full-scale mix of 19 and 20kHz tones, the intermodulation products were also low in level (fig.9).
The CP-800 offered superb rejection of word-clock jitter via all of its digital inputs. Fig.10, for example, shows the spectrum of its analog output while it decoded a 24-bit version of the Miller-Dunn J-Test signal presented to the USB input. The noise floor is free from any jitter- or supply-related sidebands, and the picture was equally superb via the S/PDIF and AES/EBU inputs.
Turning to the CP-800's performance as an analog line preamplifier and setting it to Analog Bypass, the maximum gain from the balanced outputs was 14dB with the volume control set to "14.0"; with an unbalanced input, it was 7.6dB from the unbalanced jacks. The volume control operated in accurate 1dB steps, with superb matching between the two channels at all levels. The output was polarity correct from both outputs, the XLRs being wired with pin 2 hot. The balanced input impedance was 48k ohms at low and middle frequencies, which is close to the specified 50k ohms, dropping slightly but inconsequentially to 35k ohms at 20kHz. The impedance of the unbalanced input is specified as a high 100k ohms, but I got around 10k ohms at all frequencies. This should not be a problem. The output impedance was close to 600 ohms across the audioband from the XLRs, and an even lower 100 ohms from the RCAs.
The CP-800's frequency response in Analog Bypass mode and with its volume control set to "0.0" was flat and wide (fig.11, blue and red traces), lying at 0.8dB at 200kHz. This was not affected by load impedance, but the ultrasonic rolloff did increase slightly, to 1.5dB at 200kHz, with the volume control set to "14.0" (not shown). Switching to EQ mode, but with all tone controls and EQ parameters set to flat, gave the cyan and magenta traces in fig.11. As all of these functions are implemented with a digital signal-processing chip, analog input signals must be converted to digital. The sharp rolloff above the audioband, reaching 3dB at 45kHz, suggests that the conversion is done at 96kHz, which in turn suggests that this is the sample rate at which the DSP is performed. This would explain why the frequency response with 192kHz digital data is no wider than with 96kHz datathe digital input data are sample-rate-converted to 96kHz in order to be compatible with the CP-800's DSP section.
Once engaged, the Classé's EQ functions are comprehensive, offering both conventional, Baxandall-type tone controls and Quad-like Tilt controls, as well as parametric equalization, with full control of center frequency, Q, and boost/cut. Fig.12 shows the effect of the Treble and Bass controls, set to their maximum (+6dB) and minimum (6dB) positions, with corner frequencies of 200Hz and 2kHz. The latter frequencies conform to the control's Ò3dB frequency, and the maximum boost and cut are Ò6dB, as specified. Fig.13 shows the effect of the Tilt control, set to its maximum positive and negative slopes.
The maximum input level the CP-800 can accept depends on whether it is set to Analog Bypass or EQ. Fig.14 shows how the THD+noise percentage in the preamp's output changes with the output level of a 1kHz tone into 100k ohms with the volume set to its maximum. The output stage clips at 20V RMS, equivalent to an input voltage of 3.9V. (Clipping is defined as when the THD+N percentage reaches 1%.) With a very low load impedance of 600 ohms, the maximum output drops to just under 10V. Backing off the volume control allows the input voltage needed to produce clipping to increase, meaning that at unity gain, the CP-800 will handle more than the Audio Precision's maximum output of 15V RMS without clipping. With the volume control set to "0.0" and with EQ engaged, the balanced input clipped at 5.7V. This is somewhat higher than the specified 4V, and more than enough to cope with any real-world analog sources. The maximum unbalanced input voltage with EQ engaged was half the balanced figure.
Fig.14 suggests that the measured THD at typical signal levels is dominated by noise, so I haven't shown how the measured THD+N percentage changes with frequency. (It doesn't.) With the input shorted and the volume control set to its maximum, the unweighted, wideband signal/noise ratio was 71dB ref. 1V output. Switching in an A-weighting filter increased this to 85dB; reducing the volume control to unity gain increased it to 99.2dB.
With the CP-800 in Analog Bypass mode and the volume control set to unity, the distortion harmonics are not much higher than the residual levels in the Audio Precision's output, at 120dB (0.0001%) or below (fig.15), though the slightly higher level of noise at low frequencies can be seen. This spectrum was taken into the benign 100k ohm load; dropping the load impedance to 600 ohms and readjusting the input level to give the same 2V output increased the third harmonic from 124 to 120dB (not shown), but otherwise the spectra looked identical. Intermodulation levels were similarly very low, even into the punishing 600 ohm load (fig.16).
The Classé CP-800's measured performance in its Analog Bypass mode is beyond reproach. While its digital input offers about 2 bits' worth less resolution than the current state of the art, this didn't prove much of an impediment in my auditioning.John Atkinson