Contents - Index

About Distortion

A perfect system could reproduce sound waveforms -- exactly -- in only one way.  But there are infinite ways a real workd system could reproduce sound imperfectly, which is the subject of distortion. Overall distortion could be detected using something like an "Audio DiffMaker" tool, which can determine whether or not some particular signal has been changed at all by what you are testing.  But that is an extremely stringent test and doesn't provide much informatiom about causes or what someone might do to prevent the distortions -- or whether it might even be needed to correct them at all.  Distortion can be categorozed as either linear or non-linear distortion and, to be practical, refers only to what a system does when presented with some sort of test stimulus (you can't play music at 0.1% THD, which would be a meaningless concept -- only test signals).

Linear distortion in a system refers to its volume scale, frequency response, time response, or phase response effects, and these can be investigated with OmniMic using the Frequency Response tabpage.  Basic frequency response is the most common property of a system that OmniMic is used to find, and is done by playing either a sweep (a.k.a. "chirp") signal or a pseudo-noise signal and displaying the resulting curve on OmniMic's main graph..  You can display phase response characteristics  on the same graph.  Various types of combined time-and-frequency response characteristics can be displayed as waterfall graphs.  Frequency responses can vary toward different directions from a loudspeaker, and you can create and view that information in the form of a polar display. Since a loudspeaker and room will always modify the overall system's frequency and time response, desirably or not, OmniMic software provides additional tools for Reverb/ETC echoic room analysis.

Non-linear distortion
When people refer to distortion, though, they usually mean some kind of non-linear distortion.  These are (mostly) changes to sound signal than can't be corrected beyond frequency, scale (volume), and simple time related changes.  OmniMic software provides for three types of non-linear distortion testing.

Harmonic Distortion (HD), is measured by applying a stimulus of a singe-frequency tone (or more conveniently, a log-sweep) to the system and seeing what other tones at frequency multiples of the applied "fundamental tone" get created.  Total or individual harmonic distortion is familiar to most people as it is simple to understand, repeatable, and quantifiable.  Numbers can be assigned to results for comparasons, though except in extreme cases, they seldom are found to much relate to perceived sound quality.  However the simiplicity and standardisation of HD tests can show that something is wrong (and sometimes suggest what is wrong).  Harmonic distortion is particularly helpful as an aid to determining or specifiying amplitude capability -- you can state how much SPL or how many watts can be produced before some (usually gross!) amount of HD starts to appear and compare that to results from other power producing devices such as power amplifiers.

Intermodulation distortion (IM) characterization is similar to harmonic distortion, but it uses two or more tones applied and looks for multiple sum and difference frequencies of those to be created.  There are far more variations possible for intermodulation distortion measurement, both in the relationships between the frequencies and what product frequencies to look at.  There is still not significant correlation however between simple intermodulation distortion numbers and what listeners hear.  A more extensive and general type of intermodulation distortion is Spectral Contamination, in which very many (usually logarithmically spaced) frequency tones are applied at once to the system and the comnbined level of energy resulting at any other audio frequencies is analyzed.  There are of course numerous possibilities for the frequencies of applied tones with Spectral Contamination, but at least there's just one result characteristic being collected for all the possible orders of intermodulation products!

While HD, IM, and Spectral Contamination all look at energy produced at frequencies other than those being applied to the device being tested, a Compression Test is a non-linear distortion test that looks at only how output at the same applied frequency changes as level is increased.  A compression test can only be done in terms of output level changes in comparison to expected levels.

Quantifying most non-linear distortion results will be complicated by the frequency response of what is being tested, including the effects of the room on the actual frequency response of loudspeakers.  A frequency response valley or a room null that occurs at a harmonic frequency of interest could make the harmonic distortion of a loudspeaker look far better than it actually is.  Omnimic's ability to easily measure over wide frequency ranges helps with this, but when testing at low frequencies in particular, room effects must be avoided.  This can be done outdoors (though noise levels become difficult to suppress in most areas) and/or by close-mic'ing within an inch or so of a loudspeaker driver.  Then, of course, the SPL levels at the microphone will be much higher than out in the room so it is important that the microphone not be overdriven or you'll only see its distortion.  OmniMic40k hardware is capable of operating well above 140dBSPL to avoid such problems.

Non-linear distortion behavior typically gets worse as signal level (SPL or volume) is increased, and so the related level applied will be a part of most meaningful non-linear distortion results.  For example, stating "0.2% total harmonic distortion" really doesn't mean very much, but "0.2% total harmonic distortion at 135dB SPL at 40Hz and 1 meter" could be relevant.  So, when measuring harmonic or spectral contamination distortion, you typically want to find out how the subject behaves as you increase the volume rather than make just one curve.  The OmniMic40k hardware is particularly well suited for non-linear testing because of its inherent high linearity and SPL handling ability.

In general, keep in mind that non-linear distortion tests will require that background noise levels be low enough that distortion products of interest are detectable.  That is, the noise level will deternine how well you can verify something acoustic isn't distorting.