Jeff Rowland Design Group Coherence preamplifier & Cadence phono stage Measurements
Coherence: Although the Coherence can be used with single-ended-to-balanced adapters, it is a true balanced design, so all the measurements were done in balanced mode. The Coherence inverted signal polarity, as Jeff Rowland adheres to the older US convention of wiring pin 3 of the XLR to be "hot" rather than pin 2, as the AES now recommends. The input impedance was 11k ohms in the normal setting, 690 ohms in "terminated" mode; the output impedance was a low 55 ohms across the audio band.
The volume control operated in accurate 0.5dB steps up to an indicated maximum setting of "63.5." As supplied, the voltage gain was unity, but up to 20dB can be dialed-in for each input. This was an accurate 20dB, meaning that an indicated "43.5" was the unity-gain volume-control setting when the gain was to as high as it would go. The noise floor with the inputs shorted was low in level, but did vary with the volume-control and gain settings. With maximum gain but the volume control set to minimum, the S/N ratios were 87dB (wideband), 99.8dB (22Hz-22kHz), and 102dB (A-weighted), all ref. 1V output. Keeping the conditions the same but setting the control to its maximum worsened all the above figures by around 19dB.
The Coherence's frequency response (fig.1) was perfectly flat up to 20kHz, with then a gentle rolloff to -1dB at 100kHz. The channel separation was superb, at better than 110dB below 1kHz, with only a slight decrease above the audioband.
Fig.1 Rowland Coherence, frequency response into 100k ohms (top) and 600 ohms (bottom). (0.5dB/vertical div.)
The preamplifier was also very linear. Fig.2 shows the spectrum of a 50Hz sinewave reproduced at a very high level into 100k ohms. Both the second and third harmonics (100Hz and 150Hz, respectively) are around -83dB (0.007%)—ie, very low. In addition, the distortion level didn't increase when the load was reduced to a much more demanding 600 ohms. However, if the THD+noise is plotted against frequency at this high output level—5V is more than twice as high as the preamplifier will be asked to deliver in practice—the distortion increased dramatically in the low bass (fig.3, top trace below 30Hz). This is due to an increase in odd-order harmonics at these low frequencies, presumably due to the input transformer starting to saturate. If the output level is reduced to a more typical 1V (fig.3, bottom trace below 30Hz), the increase in distortion in the low bass is benign. (The higher level of distortion at higher frequencies in this graph is actually due to the apparent rise in the noise floor at the lower output level and should be ignored.)
Fig.2 Rowland Coherence, spectrum of 50Hz sinewave, DC-1kHz, at 5V into 100k ohms (linear frequency scale).
Fig.3 Rowland Coherence, THD+noise (%) vs frequency at 5V (top below 30Hz) and 1V into 100k ohms.
At 1kHz, the Coherence could output around 15V before clipping (fig.4, bottom trace), even into 600 ohms. But at 20Hz (fig.4, top trace) the maximum output was about 8V, as expected from fig.3. Again, this difference will be irrelevant in practice.
Fig.4 Rowland Coherence, distortion (%) vs output voltage into 100k ohms/600 ohms at (from top to bottom) 20Hz, 20kHz, and 1kHz.
Finally, Shannon Dickson made much of the Rowland's rejection of common-mode noise on its balanced inputs. I therefore ran a sweep, examining the Coherence's rejection of a very high common-mode signal, 1V, present on both pin 2 and pin 3 of the input XLR. The results with the volume control set to its maximum and the gain set to 20dB are shown in fig.5 (bottom traces). Below 10kHz, the leakage is below the preamplifier's noise floor. Above that frequency there is a gradual drop in the rejection ratio, somewhat different in the two channels, to 53dB at 200kHz, which is still excellent. This is the first preamplifier whose common-mode rejection ratio (CMRR) I have measured, so I repeated the test under identical conditions for the only other fully balanced preamplifier I had to hand—the Mark Levinson No.380S sample that I purchased in 1998. The CMRR for this all-active balanced circuit is better than the noise floor below 3kHz, with then a 6dB/octave decrease to 48dB at 200kHz. The Rowland is slightly better than the Levinson at ultrasonic common-mode rejection, but I'm not sure what influence, if any, this has on sound quality.
Fig.5 Rowland Coherence, common-mode rejection ratio (CMRR) vs frequency with a 1V common-mode signal, maximum gain, volume control at maximum (bottom), compared with Mark Levinson No.380S (top).
The Coherence not only features looks to die for, but its measured performance is also superb.—John Atkinson