Boulder 1012 D/A preamplifier Measurements
The Boulder 1012 D/A preamplifier didn't invert signal polarity from either its analog or digital inputs, its analog XLR wiring conforming to the AES "pin 2 hot" convention, while its source impedance was a very low 26 ohms. The volume control operated in accurate 0.5dB steps and the channel balance was superb. With the volume control set to its maximum ("0" on the display), the line-stage gain from the analog output was 20dB, which is perhaps a bit on the high side. Repeating the measurement with a digital datastream representing a 0dBFS, 1kHz tone gave an output of 16.5V. However, the distortion at this level was 0.9%, suggesting the onset of clipping. Backing off the volume control to "-0.5" significantly reduced the THD. (I was surprised to find a difference in this behavior that was dependent on sample rate: the 0.9% THD figure was taken with 44.1kHz-sampled data; changing the data sample rate to 48kHz with the volume control still at "0" dropped the THD to 0.12%.)
Looking at the 1012's performance in the frequency domain, the top traces in fig.1 show the unit's response with normal CD data; it rolls off smoothly by -0.5dB at 20kHz. The bottom traces were taken with pre-emphasized data—the rolloff at 20kHz has increased to -0.75dB, while a very slight boost has appeared in the treble. The bottom pair of traces in fig.2 show the digital input response when fed with a 96kHz-sampled datastream; the top-octave response seen in fig.1 smoothly continues to -2dB at 39kHz, with then a precipitous decline, as expected. Fig.2's top traces reveal that the response of the analog inputs extends to well over 100kHz, with the output at 200kHz down by just 1.75dB.
Fig.1 Boulder 1012, CD frequency response at -12dBFS, without emphasis (top) and with emphasis (bottom). (Right channel dashed, 0.5dB/vertical div.)
Fig.2 Boulder 1012, 96kHz-sampled frequency response at -12dBFS (bottom at 40kHz) and analog input frequency response at 2V output (top). (Right channel dashed, 2dB/vertical div.)
Fig.3 shows the phono input's RIAA error. This is superbly low in the audioband, and the response sensibly starts to roll off above 50kHz, reaching -1dB at 100kHz. The phono input impedance, as supplied, was 185 ohms at 1kHz, with the maximum voltage gain a very high 73.7dB, making the unit suitable for use with very-low-output MC cartridges. However, this high gain meant that the S/N ratio was only okay, at 57.3dB A-weighted, 51dB wideband unweighted (both figures referred to 500µV input at 1kHz).
Fig.3 Boulder 1012, phono input RIAA error at 1mV (1kHz) input (right channel dashed, 0.5dB/vertical div.).
Channel separation via the digital input (fig.4) differed in the two directions. The L-R crosstalk was buried in the 16-bit noise floor below 900Hz or so, but increased above that frequency due to capacitive coupling. By contrast, the R-L crosstalk was slightly higher at low frequencies, but remained below the noise floor at high frequencies. A similar picture could be seen via the analog input (fig.5), though some capacitive coupling in the R-L direction now appears. In both cases, however, the channel separation is good enough that imaging will not be impaired.
Fig.4 Boulder 1012, channel separation via digital input, V/C = "0": L-R (top at 20kHz), R-L (bottom). (10dB/vertical div.)
Fig.5 Boulder 1012, channel separation via analog input, V/C = "0": L-R (solid), R-L (dashed). (10dB/vertical div.)
Fig.6 shows audioband spectral analyses with the Boulder's digital input fed data representing a dithered 1kHz tone at -90dBFS. The bit depth was 16 bits (top pair of traces) and 24 bits. The increase in word length gives rise to about a 15dB drop in the noise floor in the treble, and even more than that in the midrange and below, which suggests a resolution of at least 19 bits, which is superb. However, the right channel's ultimate resolving power at low frequencies is somewhat compromised by the presence of power-supply spuriae at 60Hz and 180Hz, which will probably be due to magnetic coupling into the circuit from the AC transformer; while the left channel's 24-bit spectrum shows both second and third harmonics. (I repeated this measurement with a different, again properly dithered, digital signal generator to ensure that what you see is really the Boulder, and not due to my inadvertently using an undithered source.)
Fig.6 Boulder 1012, 1/3-octave spectrum of dithered 1kHz tone at -90dBFS, V/C = "0," with noise and spuriae (from top to bottom): 16-bit data, 24-bit data (right channel dashed).