Audio Research Reference 5 SE line preamplifier Measurements

Sidebar 3: Measurements

I measured the Audio Research Reference 5 SE with Stereophile's loan sample of the top-of-the-line Audio Precision SYS2722 system (see and the January 2008 "As We See It," ). There was one problem: During shipping, one of the four large, 10µF Teflon output-coupling caps at the right rear of the printed circuit board had broken off. ARC sent me a replacement, along with the appropriate solder. However, although my soldering iron is temperature-controlled, it dates from a time when solder was lead-based and melted at a lower temperature. That, coupled with the fact that the Ref 5 SE's printed circuit board traces are very large and thus have significant thermal mass, meant that I was unable to do a professional job of installing the new capacitor. I rigged a temporary solution with clips that appeared to work correctly. But I do wonder if some of the distortion measurements of the left channel had been compromised, so keep that in mind as you read on.

The maximum gain of the Ref 5 SE, with the volume control set to its maximum ("103" indicated on the front-panel display), was close to the specification at 12.33dB for fully balanced operation, 6.26dB single-ended. The volume control operated in 0.5dB steps, with the unity-gain settings between "78" and "79" balanced, and between "90" and "91" unbalanced. (Props to Audio Research for designing a display that can be easily seen from across the room.) The preamp preserved absolute polarity in both modes, the balanced XLR jacks being wired with pin 2 hot. At low and middle frequencies the input impedance was slightly but inconsequentially lower than specified, at 114k ohms against the specified 120k ohms balanced, and 55k ohms against 60k ohms unbalanced. The input impedance dropped slightly at the top of the audioband, to 100k ohms balanced and 42k ohms unbalanced.

At high and middle frequencies the Ref 5's output impedance was slightly higher than specified, at 628 rather than 600 ohms balanced and 322 rather than 300 ohms unbalanced. However, at 20Hz the impedance rose to 1447 ohms balanced and 637 ohms unbalanced, which, with an extremely low load impedance of 600 ohms, rolls off the low bass by 3dB at 15Hz (fig.1, cyan and magenta traces). Into the more realistically high 100k ohm load, however, the Ref 5's low-frequency response is flat to below 10Hz (fig.1, blue and red traces). This graph was taken with the volume set to "103"; the upper-frequency response extends to –1dB at 200kHz. (Note also the slight channel imbalance in this graph: the right channel is 0.1dB higher in level than the left.) There was some dependency of the upper-frequency response on the volume-control setting. The most restricted response, –1dB at 40kHz, was with the control set to "98"; at "66" it was –1dB at 60kHz, and at "55" it was –1dB at 90kHz (fig.2). This variation will have no audible effects. The single-ended frequency response was the same as the balanced.

Fig.1 Audio Research Reference 5 SE, balanced frequency response with volume control at "103" at 1V into 100k ohms (left channel blue, right red), and at 1V into 600 ohms (left cyan, right magenta) (0.25dB/vertical div.).

Fig.2 Audio Research Reference 5 SE, balanced frequency response with volume control at "98," "66," and "55" into 100k ohms (0.25dB/vertical div.).

Channel separation (fig.3) was excellent at >100dB in both directions below 4kHz, and still 85dB at 25kHz. The wideband, unweighted signal/noise ratio, ref. 1V output with the input shorted but the volume control set to its maximum of "103," was very good in the right channel, at 85.6dB, but less so in the left, at 79.8dB. Switching in an A-weighting filter improved these figures to a superb 102dB right and 91.5dB left. I'm not sure why the left channel was noisier; the spectrum of the Ref 5's balanced noise floor while the preamplifier reproduced a 1kHz tone at 1V into 100k ohms is shown in fig.4. The left-channel noise floor (blue trace) is not much higher than the right (red), while in the latter can be seen some very low-level odd-order harmonics of the AC supply frequency.

Fig.3 Audio Research Reference 5 SE, channel separation (10dB/vertical div.)

Fig.4 Audio Research Reference 5 SE, FFT-derived spectrum with noise and spuriae of 1kHz sinewave, DC–1kHz, at 1V into 100k ohms (left channel blue, right red; linear frequency scale).

Fig.5 plots the THD+noise percentage against the left channel's balanced output voltage into 100k ohms. Below 1V output the actual distortion is buried beneath the noise floor, indicated by the upward slope of the trace with decreasing voltage. The distortion slowly rises above 1V, but is still just a low 0.05% at 18V RMS, when actual waveform clipping starts. Defining clipping as 1% THD+N, the Ref 5's balanced output clips at 32V into 100k ohms. The single-ended output clips at 8V into 100k ohms (fig.6), which is still well above any level needed to drive a power amplifier into overload. The distortion increases dramatically into low impedances; I recommend the Ref 5 be used with power amplifiers having an input impedance of at least 10k ohms, when the preamplifier will be at its most linear at typical operating levels.

Fig.5 Audio Research Reference 5 SE, balanced distortion (%) vs 1kHz output voltage into 100k ohms.

Fig.4, for example, plots the THD+N percentage against frequency at a balanced output level of 2V into 100k ohms, which is both sufficiently high to ensure that I'm measuring THD rather than noise, and close to the maximum that will be required of the preamplifier in practical use. The left channel's distortion (blue trace) is slightly higher than the right's (red), but both are extremely low in absolute terms.

The spectrum of the Ref 5's output at 2V is shown in fig.5. The difference between the two channels' THD+N percentage in fig.4 can be seen to be due to the left channel's (blue trace) having more second harmonic than the right (–90 vs –112dB). This might be due to tubes being mismatched—when the tubes are perfectly matched, balanced operation will completely cancel even-order distortion—but the level of the distortion is low enough not to be heard, and in any case, its character is subjectively innocuous. Dropping the load impedance starts to increase the level of the third and fifth harmonics (not shown), confirming that the Audio Research preamp needs to be used with higher-impedance power amplifiers. The Ref 5's performance with the demanding 19+20kHz high-frequency intermodulation test was superb, all intermodulation components in the right channel (fig.6, red trace) lying at or below –112dB. Even in the less-linear left channel, the second-order difference component lies at a very low –96dB (0.0015%), and the higher-order distortion products are the same as in the right channel.

Fig.6 Audio Research Reference 5 SE, single-ended distortion (%) vs 1kHz output voltage into 100k ohms.

Fig.7 Audio Research Reference 5 SE, THD+N (%) vs frequency at 2V into 100k ohms.

Fig.8 Audio Research Reference 5 SE, balanced spectrum of 50Hz sinewave, DC–1kHz, at 2V into 100k ohms (left channel blue, right red; linear frequency scale).

Fig.9 Audio Research Reference 5 SE, balanced HF intermodulation spectrum, DC–30kHz, 19+20kHz at 2V into 100k ohms (left channel blue, right red; linear frequency scale).

Despite the small disparity in the two channels' linearity and noise floor (and remember my caution about the temporary repair I had to do), Audio Research's Reference 5 SE is a superb-measuring preamplifier.—John Atkinson

Audio Research Corporation
3900 Annapolis Lane N.
Plymouth, MN 55447-5447
(763) 577-9700

seank's picture


"The Ref 5 SE's performance at the extremes of the audioband, usually tube electronics' bàte noir, was superb."

I had to search the internet for bàte noir.

bàte noir is french for black beast

I assume from the context you meant "weakness".



John Atkinson's picture

The phrase should have been bête noir - blame my occasionally suboptimal HTML coding :-)

I have corrected it.

John Atkinson

Editor, Stereophile

Markus Sauer's picture

It's la bête, so bête noire, I'm afraid, Mr. Copy Editor.

Anyway, I like bàte noir, it's almost black bat, and every review in the Halloween edition of Stereophile should mention black bats at least once.

audioware's picture

These pres are outstanding. However the problem is their poor reliability. 06 (six) of them were already on my services bench this year to fix problems with the transformer that feeds the dgital, the remote control and the on-off switch, circuits. It is a small xformer below the circuit board that burns out frequently. Another very serious problem is with the 6H30s filament regulators that do not stand for the increased current demanded by the new 6H30s intead of the old 6922s. ARC has to pay more attention to these problems!

laywingsun's picture

Editor ,my englishi is poor,but I'd like to ask a question :Can I replace ref5se's 6550 by kt88? thank you.

John Atkinson's picture

Can I replace ref5se's 6550 by kt88?

These tubes are very similar and used as the series-pass element in the voltage regulator, I don't expect them to behave differently. However,  I wouldn't do so until you have checked with Audio Research.

John Atkinson

Editor, Stereophile