Audio Research LS8 line preamplifier Measurements
Generally, the engineering design of Audio Research products is beyond reproach; the LS8's very good lab results for response flatness, distortion, and noise hardly came as a surprise. There is hardly any point in printing a frequency-response graph that is perfect from 10Hz to 30kHz, 0.5dB, at better than 5Hz and 45kHz, 3dB down, or better than 1Hz and 118kHz (a bit less than claimed, but of course these results are output-cabledependent).
I did investigate the output impedance variation with frequency to explore the dynamic range of the feedback-enhanced output coupling capacitor. At 1V output the entire loss at 20Hz relative to 1kHz for a cruel 600 ohm loading was less than 1dBno significant weakness here. The output impedance was on the button at 200 ohms, conferring a strong load and cable drive potential. Input impedance was also correct at just under 60k ohms for normal settings.
Unlike designs with conventional passive potentiometer inputs, a preamplifier like the LS8 that has an "electronic" volume control, input overload margin becomes an issue. The LS8's input section began to distort at 3.5V RMS, and clipped at 4V. Although only a few signal sources deliver as much output as this, I would have liked more headroom; some digital sources, such as the Theta models, can have an output in the 36V range.
Channel separation was very good, measuring some 100dB midband, still fine at 91.9dB for 20kHz, and 115dB at 20Hz. These were orders of magnitude better than for ARC's older-generation, hardwired designs. An IHF 500mV output required 125mV of input. A typical power amplifier will require 2V drive fro clipping; this is achieved with s a 500mV input to the LS8a sensible gain for modern sources.
The volume control is extremely well balanced, the two channels typically within ±0.0035dB of each other. This is helpful in view of the absence of a balance control. As with many electronic controls, the step size varied with level setting. Nearer the top, the gradations were just under 0.5dBusefully fine, even if there is no verifying display or physical scale readout of the volume setting. At 15dB attenuation and lower, the gradations increase in size to around 0.8dB. The total range was 60dB of attenuation, below which point the output was muted to 100dB or better. The last dB settings were pretty coarse, in fact, and ran approximately 60dB, 54dB, 49.2dB, 43.7dB, 39.6dB, and 37dB, gradually improving in resolution at reduced attenuations. It's just as well that the LS8's gain was a moderate 12dB or 4x linear, and that these coarser low-level steps will get little use.
The total harmonic distortion results were excellent, the meter reading 85dB midband, 82dB at 20Hz, and with some impairment to 70dB (0.03%) by 20kHz. For the record, a spectrum analysis of the 1kHz result for 500mV output (fig.1) revealed the 85dB "meter" reading to be composed predominantly of noise, with the second harmonic actually at 96dB and with the barest trace of third harmonic at better than 114dB downa nice balance of harmonics, and excellently low to boot.
Fig.1 Audio Research LS8, spectrum of 1kHz sinewave, DC10kHz, at 500mV into 100k ohms (linear frequency scale).
Recalling that arduous 600 ohm test loading, the distortion increased moderately to 0.025% as the feedback loop worked harder to maintain the output level. Under normal conditions I would expect very little variation in distortion due to loading.
A high level of linearity was also maintained in the face of high-frequency twin tones: 19+20kHz at IHF level. Local "side bands" were very low, while the in-band intermodulation itself was excellent at around 110dB by spectrogram (fig.2), the meter reading 98dB (including some noise contribution).
Fig.2 Audio Research LS8, HF intermodulation spectrum, DC22kHz, 19+20kHz at 500mV into 100k ohms (linear frequency scale).
Noise levels were fairly low, with negligible transformer hum in the output section (the S/N ratio was 82dB ref. the IHF 500mV input level, 84dB ref. a CD 2V level). Hum excluded (filtered below 400Hz), the residual noise floor at volume zero was 102dB, the IHF noise at normal volume settings was 79.5dB (1kHz ref., CCIR) and 87dB A-weightedsatisfactory if not exceptional, and not up to 16-bit CD standards of typically better than 110dB (for a 2V reference level).Martin Colloms