Tube Preamp Reviews

The Audio Research Reference 3's maximum voltage gain, with its volume control set to "103," was 11.8dB from balanced input to balanced output and 5.75dB from unbalanced input to unbalanced output. These figures are sensibly suitable for use in practical systems. The preamp was noninverting; ie, it preserved absolute polarity in both conditions. The input impedance was to specification at low and midrange frequencies, at 58k ohms single-ended and 116k ohms balanced, these dropping slightly and inconsequentially to 48k ohms and 106k ohms, respectively, at 20kHz.

The output impedance was also to spec., at 635 ohms balanced and 326 ohms unbalanced in the treble and midrange, but rose to 1437 ohms and 625 ohms, respectively, at 20Hz. This rise in source impedance rolled off low frequencies a little early into the punishing 600 ohm load (fig.1, bottom pair of traces), with a –3dB frequency of 17Hz. As this is a relatively low frequency and the preamplifier will never be used with a power amplifier having an input impedance as low as 600 ohms, this rise in source impedance will not be a factor in practical use.

Fig.1 Audio Research Reference 3, volume control set to "103," balanced frequency response at 1V into (from top to bottom at 2kHz): 100k, 600 ohms (0.5dB/vertical div., right channel dashed).

At the other end of the spectrum, the Reference 3 offered a wide bandwidth, with a –3dB point at 200kHz into 100k ohms with the volume control set to its maximum (fig.1, top traces). There was a slight decrease in ultrasonic extension into 600 ohms, and with the volume control set to unity gain or below, but the effect on the preamplifier's audioband response was negligible. Fig.1 shows the balanced response; the unbalanced response (not shown) was effectively identical. The unity-gain setting of the volume control, by the way, was "79" balanced, "92" unbalanced.

Balanced channel separation, assessed with the undriven channel's input shorted and the volume control set to its maximum, was excellent—better than 100dB below 1kHz—but less good for unbalanced operation (fig.2). You can also see in this graph that separation decreases with increasing frequency due to the usual capacitive coupling, but is still excellent at 20kHz for balanced operation. Unbalanced separation is 60dB at 20kHz, which is good rather than great. The unweighted, wideband signal/noise ratio for balanced operation, again taken with the input shorted and the volume control set to its maximum, was excellent at 80.3dB (ref. 1V output). Unbalanced operation reduced this to 70.2dB, but both figures improved significantly when A-weighted, to 94.7dB balanced and 88.4dB unbalanced.

Fig.2 Audio Research Reference 3, channel separation, from bottom to top: L–R balanced, L–R unbalanced (R–L dashed, 10dB/vertical div.).

The Audio Research Reference 3 could swing very high voltages with very low distortion into loads greater than 10k ohms. This is shown graphically in fig.3, which plots the THD+noise percentage in the preamp's output against balanced and unbalanced output voltage into loads ranging from 100k ohms down to 600 ohms. The actual clipping voltage (1% THD) into 100k ohms was 33V balanced but 8.2V unbalanced. Both figures are significantly higher than the maximum voltage the Reference 3 will be asked to deliver in practical use. However, fig.3 suggests that the preamp not be used with loads below 10k ohms. This is confirmed by the plot of the THD+N percentage against frequency at 2V output (fig.4), where the distortion percentage stays low into loads of 10k ohms or higher. Note, however, that the single-ended output performs significantly less well than the balanced, with a rise of THD at the top of the audioband.

Fig.3 Audio Research Reference 3, THD+noise (%)vs 1kHz output voltage into (from bottom to top at 1% THD): balanced into 100k ohms, unbalanced into 100k, unbalanced into 10k, balanced into 600, unbalanced into 600 ohms.

Fig.4 Audio Research Reference 3, THD+N (%)vs frequency at 2V into (from bottom to top): balanced into 100k ohms; unbalanced into 100k, 10k, 1k ohms (right channel dashed).

The Reference 3's distortion may be very low at practical levels into sensible loads, but is also almost entirely second-order in content (fig.5), which will reduce its audibility. Into impedances much lower than 10k ohms, not only does the second harmonic rise in level, it is joined by the third harmonic, again suggesting that the power amplifier with which the preamp is used have an input impedance above 10k ohms. (Audio Research's own power amplifiers all have balanced input impedances of between 200k and 300k ohms, and the company recommends the Reference 3 not be used with amplifiers having an input impedance of less than 20k ohms.)

Fig.5 Audio Research Reference 3, balanced spectrum of 50Hz sinewave, DC–1kHz, at 2V into 100k ohms (linear frequency scale).

Fig.5 was taken from the Reference 3's balanced output. When I repeated this test from the unbalanced output using a different test set, the Miller Audio Research QC Suite, I got a similar result (fig.6): the distortion is low and almost entirely second-harmonic in nature, and the noise floor is also low. The spectrum of the Reference 3's output while it drove an unbalanced mix of 19kHz and 20kHz tones into 8k ohms (fig.7) is also very clean, though the second-order difference product at 1kHz lies at –72.5dB (0.024%), this correlating with the decreasing linearity seen at high frequencies in unbalanced mode in fig.4.

Fig.6 Audio Research Reference 3, unbalanced spectrum of 1kHz sinewave, DC–10kHz, at 1V into 8k ohms (linear frequency scale).

Fig.7 Audio Research Reference 3, unbalanced HF intermodulation spectrum, DC–24kHz, 19+20kHz at 1V into 8k ohms (linear frequency scale).

Overall, the Reference 3's measured performance is excellent, but it does suggest that it works best in full balanced mode, and into higher impedances.—John Atkinson