Distortion measurements for Audio Research SP-8 line-stage

I've been asking questions about tubes and capacitors and performance, where the answer always seems to involve "Does it meet specifications?". So, I figured I'd better check.

Tests were performed with a Scarlett Solo into a Mac Mini running Room EQ Wizard. Input voltage for all tests was ~1.76V

The high-frequency rolloff appears to be in the interface, as it appears in loopback measurements.

Frequency response
Manual States: +/- 0.25dB, 5Hz-30kHz
Result: 0.23dB total deviation, maximum to minimum (excluding high-frequency rolloff)

Manual States: Less than 0.01% at 2V RMS output, 20Hz to 20kHz
Result: 0.007% THD maximum between 20Hz to 20kHz

Intermodulation Distortion
Manual States: Less than 0.002% at 2V RMS output
Result: 0.02% at 2V RMS output*
Note: Measurement was obtained using 7kHz and 13kHz twin-tone


I've included slides for the frequency and 2V distortion measurements. I've also included a 5v output static distortion slide for interest, although it is not relevant to any of the measurements discussed.

Thoughts: The frequency and distortion numbers seem to be inline with specifications. The IMD is high, but I don't understand the potential causes or ramifications of this enough to discuss it. For all measurements, I consider both my tools and techniques suspect, although I am making ever effort to be accurate, with the equipment I have. I allowed the amp to warm up for approximately 15 minutes before taking these measurements.

For full specifications: https://www.arcdb.ws/model/SP8

*Note: Slide for this measurement was lost


  • SP-8 distortion.jpg
    SP-8 distortion.jpg
    248.9 KB · Views: 34
  • THD @ 5v.jpg
    THD @ 5v.jpg
    195.9 KB · Views: 31
  • Frequency response 1:2 dB scale.jpg
    Frequency response 1:2 dB scale.jpg
    127.4 KB · Views: 34
I think you forgot to specify a decent window function on the FFT - Try Hann window in the first instance, or a flattop window for accurate peak heights. I suspect that spectrum is polluted with spectral leakage from using no window function (aka rectangular window).

The IMD test is much more challenging as it involves frequency of 13kHz compared to 1kHz for the THD figure.