Recording - December 2015 - 62

ike
By M

River


All electronic devices, other than the oft-mentioned "straight
wire," are nonlinear to a degree. This means that the waveform
of a signal coming out isn't exactly the same shape as what
went in. Distortion, which comes in many flavors, is the term that
we use to describe a difference between input and output.
Usually, when asked to describe distortion, the first thing that
comes to mind is "fuzzy-sounding," although non-flat frequency
response and dynamic compression are also forms of distortion
that are very common in audio devices. Distortion is usually
thought of as being a bad thing, but often what we cherish as
"warmth" or "fullness" is distortion that sounds pleasing to our
ears. It's distortion that "colors" the sound of a device, whether
in a good or bad way. In this chapter of SRTM, we'll look at a
few kinds of distortion and how it's measured and specified.
1

THD+N
The type of distortion that you'll see in a spec sheet most frequently is Total Harmonic Distortion, abbreviated THD. If a spec sheet simply reads "distortion" without describing what kind, you can be pretty sure it's THD or its close cousin THD plus noise (THD+N).
THD is measured by applying a pure sine wave-a single frequency with no harmonics-to the input of the device. The output
voltage is measured, and then a very narrow bandpass filter
tuned to the test frequency is inserted between the device's output
and the voltmeter. With the original test signal removed from the
output, the voltmeter indicates the total voltage of any harmonics
of the original test signal that the device's non-linearity causes.
Since nothing is perfectly quiet, the filtered voltage measurement
also includes the device's residual noise that's present whether a
signal is passing through it or not. The measured voltage is the
THD plus noise. It's the ratio of the filtered to the unfiltered output,
expressed as a percentage or sometimes in dB.


A couple of damn lies (or more accurately, sins of omission)
associated with THD are that the test frequency and/or its amplitude aren't always specified. Since residual noise is independent of the test signal (an exception is digital quantization noise,
but that's worth another article), testing with the highest amplitude signal possible can make the THD+N spec look better than
what you might find in normal use. THD measurements are typically made with the device operating fairly close to its maximum output level just for this reason.
For solid state circuits, harmonic distortion usually remains
quite low until just before clipping occurs. At low levels, noise
tends to dominate the level of any harmonics of the test signal
that the device generates, so the folks who write the ad copy look
for the "sweet spot," generally a couple of dB below maximum



output level, where distortion is still low and the noise is a small
percentage of the test signal level.
Another thing that can make a THD+N measurement look better is to limit the measurement bandwidth. By not including anything above 20 kHz in the measurement, you eliminate the contribution of any high-frequency noise that (hopefully) won't be
audible. The analog output of a digital device is inherently
bandwidth-limited to half its sample rate (a requirement of the
sampling law), but occasionally a stray clock signal or other
"digital" noise will find its way to the analog output and (if it's
strong enough) can eat into your system headroom. It's best that
the device takes care of filtering it out so it's not hidden. Pure
analog devices are often able to pass signals well above 20
kHz, and there are those who believe that this wide supersonic
bandwidth is important for maintaining the analog sound.
A cheat that's more common than I'd like it to be is to use an
A-weighting filter in the measurement path. Not only does this

Table of Contents for the Digital Edition of Recording - December 2015