Solid State Preamp Reviews

I tested the Rotel Michi P5 with my Audio Precision SYS2722 system (see the January 2008 "As We See It"). Looking first at its behavior via its line inputs, I performed most of the testing with the balanced inputs and balanced outputs and with the volume control set to its maximum. I repeated some of the tests with the unbalanced inputs and outputs with the volume control first set to its maximum, then at 77, which is equivalent to –10dB. (The volume control operates in accurate 0.5dB steps.)

The maximum gain for the balanced input to the balanced output was 18.55dB and for the unbalanced input to the unbalanced output, it was a little lower, at 16dB. For unbalanced inputs, the maximum gain at the balanced output was 23.9dB. The maximum gain at the headphone output was a high 33.5dB. The preamplifier preserved absolute polarity (ie, was noninverting) at the headphone jack and with balanced and unbalanced inputs and outputs. The XLR jacks are wired with pin 2 hot, the AES convention.

The P5's unbalanced input impedance is specified at 47k ohms; I measured 85k ohms at 20Hz and 1kHz, dropping to a still-high 52k ohms at 20kHz. The balanced input impedance, specified as 100k ohms, was 42k ohms at 20Hz and 1kHz, 39k ohms at 20kHz. The balanced output impedance was a low 215 ohms at high and middle frequencies but rose to 1195 ohms at the bottom of the audioband. The unbalanced output impedance was the specified 470 ohms across the audioband. The headphone jack's output impedance was a fairly high 150 ohms from 20Hz to 20kHz. Low-impedance headphones will not be driven optimally by the P5.

The preamplifier's frequency response in both balanced and unbalanced modes was flat from 10Hz to almost 200kHz into the high 100k ohm load (fig.1, blue and red traces), with superb channel matching. Into a more demanding 600 ohm load (fig.1, cyan, magenta), the balanced response rolled off in the bass due to the rise in this output's source impedance with decreasing frequency. The Michi P5 will not be a good match for those few power amplifiers that have an input impedance below 1000 ohms. This will not be an issue with Michi's own amplifiers, including the matching S5, which has a specified input impedance of 12.5k ohms (unbalanced) and 100k ohms (balanced).

Fig.1 was taken with the Michi's volume control at its maximum setting; both the response and the superb channel matching were identical at lower settings of the control. The headphone output offered a more restricted ultrasonic response, rolling off by 0.3dB at the top of the audioband (fig.2).

Fig.3 shows the effects of the treble and bass controls, set to their maximum and minimum positions. The low-frequency boost or cut reaches 14dB, those at high frequencies reach ±12dB. These are on the extreme side—the P5's manual notes that "A properly setup Hi-Fi system should not require changes to the Bass or Treble setting. Use these adjustments sparingly." Channel separation via the line inputs (not shown) was excellent, at >100dB in both directions below 3kHz and still 80dB at 20kHz.

The Michi preamp offered extremely low noise. From unbalanced inputs to unbalanced output, the wideband, unweighted signal/noise ratio, measured with the input shorted to ground but the volume control set to its maximum, was a high 82.25dB ref. 1V output, average of both channels. Restricting the measurement bandwidth to the audioband increased the S/N ratio to an excellent 94.1dB, while switching an A-weighting filter into circuit further improved this ratio, to 98.2dB. Power-supply–related spuriae in the P5's output were very low, with those at 120Hz, 240Hz, and 360Hz highest in level between –90dB and –100dB (fig.4).

Fig.5 plots the percentage of THD+noise in the P5's balanced output into 100k ohms. The THD+N is very low below 20V. (It rises below this voltage, due to the fixed level of noise becoming an increasing percentage of the signal level as the latter drops.) The balanced output clips (ie, when the THD+N reaches 1%) at a very high 23V, while reducing the load to a punishing 600 ohms reduced the maximum balanced output level to 15V (not shown). The unbalanced output clipped at a very high 18.5V into 100k ohms. All these voltages are much higher than will be needed to drive a power amplifier to its maximum power.

To be sure that the reading was not dominated by noise, I measured how the Rotel's distortion changed with frequency at 10V output. Nevertheless, the THD+N percentage was very low throughout the audioband both into 100k ohms (0.0015%, fig.6, blue and red traces) and even lower in the midrange into 600 ohms (cyan, magenta traces).

I examined the spectrum of the distortion at the same high balanced output level, but the only harmonic visible was the third, at –120dB (0.0001%) (fig.7). Even when I reduced the load impedance to 600 ohms, the third harmonic only rose to –100dB (0.001%, fig.8). Tested for intermodulation distortion in balanced mode with an equal mix of 19 and 20kHz tones at peak level of 10V into 100k ohms, any distortion products were vanishingly low in level (fig.9).

Turning to the phono input, in moving-magnet mode the maximum gain was 52.2dB at the unbalanced outputs and 6dB higher at the balanced output. In moving-coil mode, these maximum gains were 21dB higher. The phono input preserved absolute polarity at all outputs in both MM and MC modes. With the phono stage set to MM, the input impedance was an appropriately high 44k ohms at 20Hz and 1kHz, dropping to 33k ohms at 20kHz. The MC mode's input impedance was 101 ohms across the audioband. The P5's RIAA correction offered very low error and was well-matched between the channels (fig.10).

Channel separation via the phono input was an excellent 90dB in both directions in the low treble. The phono input's noise performance in MM mode was very good, with unweighted audioband signal/noise ratios (ref. 1kHz at 5mV input signal) of 68.5dB (average of both channels). The ratios improved by 10dB when A-weighted. Set to MC mode, the ratios were all 20dB lower than in MM mode.

The P5's phono input offered good overload margins, at around 29dB ref. 1kHz at 5mV across the band in MM mode and 28dB ref. 1kHz at 500µV in MC mode. Distortion was very low. It primarily consisted of the second harmonic, but this lay at just –100dB (0.001%), even at a level 17.5dB higher than the nominal MM cartridge level of 5mV at 1kHz (fig.11). Intermodulation distortion via the Rotel's phono input was similarly very low (fig.12).

The P5's coaxial and optical S/PDIF inputs locked to datastreams with sample rates up to 192kHz. Apple's USB Prober utility identified the processor as "MICHI USB Audio 2.0" from "MICHI" and confirmed that the USB port operated in the optimal isochronous asynchronous mode. Apple's AudioMIDI utility revealed that, via USB, the Rotel preamplifier accepted 16- and 24-bit integer data sampled at all rates from 44.1 to 768kHz. The P5's maximum output level at the unbalanced outputs was 15.88V with a 1kHz tone at 0dBFS, which suggests that the gain architecture of the preamplifier's digital inputs is well-arranged. All the P5's outputs preserved absolute polarity with digital input signals.

The P5's reconstruction filter had a minimum-phase impulse response with 44.1kHz data (fig.13), with all the ringing following the single high sample at 0dBFS. With 44.1kHz-sampled white noise, the Michi's response featured a fast rolloff above the audioband, reaching full stop-band attenuation at 24kHz (fig.14, red and magenta traces). The aliased image at 25kHz of a full-scale 19.1kHz tone (fig.14, blue and cyan traces) was suppressed by 100dB. Given the superbly low distortion via the P5's analog line inputs, I was surprised to see distortion harmonics in this graph, with the second the highest in level, at –77dB (0.014%).

With data sampled at 44.1kHz, the P5's frequency response rolled off gently in the top audio octave (fig.15, green and gray traces), reaching –0.25dB at 20kHz before the inevitable sharp rolloff above that frequency. At higher sample rates (fig.15, cyan, magenta, blue, and red traces), the responses followed the same gentle ultrasonic rolloffs.

When I increased the bit depth from 16 to 24 with a dithered 1kHz tone at –90dBFS (fig.16), the random noise floor dropped by 22dB, meaning that the P5 offers 19 bits' worth of resolution. With undithered 16-bit data representing a tone at exactly –90.31dBFS, the three DC voltage levels described by the data were well-resolved (fig.17), with the minimum-phase ringing at the waveform transitions visible. The 24-bit waveform was a fairly clean sinewave (fig.18).

As I noted above, the Michi P5's digital inputs offered higher levels of harmonic distortion than I expected. With a full-scale 50Hz tone into 100k ohms, the second harmonic was the highest in level (fig.19). This harmonic was higher in the right channel (red trace) than in the left channel (blue), at –79dB (0.011%) and –97dB (0.0014%), respectively. This behavior was not affected by the volume control setting. Though intermodulation products were present with equal-level tones at 19 and 20kHz with the combined waveform peaking at 0dBFS (fig.20), these were all low in level. However, the distortion products were again higher in the right channel than the left.

Tested for their rejection of word-clock jitter with 16-bit J-Test data, the Michi's optical and coaxial S/PDIF inputs didn't fare well (fig.21). The odd-order harmonics of the LSB-level, low-frequency squarewave either side of the spectral spike that represents the high-level tone at one-quarter the sample rate (Fs/4) were not at the correct levels, which are indicated by the sloping green line in fig.21. More significantly, a large number of power supply–related sidebands are present that were not present with USB data (fig.22). (Pairs of sidebands are present with USB data at ±60Hz and ±120Hz, and the spectral spike at Fs/4 has significant broadening of its base, due to the presence of low-frequency random jitter.)

I investigated whether the sidebands with S/PDIF data were due to jitter or to something else, like supply noise on the DAC chip's voltage reference pin, by performing narrow-band spectral analyses with 24-bit data representing full-scale tones at 5, 10, 15, and 20kHz. If the sidebands were due to supply noise, the levels of the sidebands would remain the same at all four signal frequencies. If they were due to jitter, their levels would drop as the frequency of the signal reduced. This turned out to be the case (fig.23). The sidebands in fig.21 are due to poor jitter rejection at the S/PDIF inputs.

Tested via its line-level inputs, Rotel's Michi P5 offered superb measured behavior—its astonishingly low distortion means that at the output levels at which it will typically be used, it will indeed be a "straight wire with gain," in Peter Walker's classic phrase. The phono input also offers very low distortion and noise, particularly in its moving-magnet mode, and its RIAA correction is one of the most accurate I have measured. However, the Michi P5 is let down by its digital inputs. While their distortion is not high enough to have audible consequences, especially in the right channel, the poor jitter rejection of the S/PDIF inputs is a matter of concern. I recommend staying with the USB digital input.—John Atkinson