There are 3 main components involved when measuring audio products - the signal generator, the product being measured, and the detector.

 

 

Figure 1. Measuring audio equipments

 

The measurements are carried out by generating a set of reference signals at the signal generator and then letting it pass through the product, after which it is detected and any differences to the original are considered.

 

For instance, if a 1V sinusoidal wave at 1kHz is fed into an amplifier, and the voltage at the output of the amplifier is measured, the output level can be determined.

 

There are a variety of options for the signal generator and the detector - the graphs below were obtained by conducting Lynx L22 (Line In, A/D Converter) on the E-MU 1616M Audio Interface (Line Out, D/A Converter) using RMAA (Rightmark Audio Analyzer) as the generator and the detector.

 

 

Figure 2. Using RMAA to measure an Audio Interface

 

 

 

The frequency response is a measure of how well a particular product reacts to a signal at a given frequency. There are a variety of options in measuring this, contingent on the product, but generally, white / pink noises or sweep signals (signals where a single sinusoidal wave is continuously changing frequency) are fed into the input, and are analyzed at the output against the original input at each frequency range. A typical frequency response will look like the following:

 

Figure 3. Frequency Response

 

The horizontal axis shows the pitch / frequency in Hz (left is lower), and the vertical axis shows the volume in dB (higher is larger).

When we say that a product 'has a good frequency response', we mean that the output level is consistent across the audible range. If the left side of the graph is raised this would mean that the bass is overemphasized compared to the original (imagine an equalizer), while a raised right side results in overemphasized treble.

 

The frequency response is best presented as a graph since this allows a in-a-glimpse snapshot of the performance, but this has a problem that it complicates numerical comparison. Thus, the graphs are usually supplemented with a decibel measurement of the deviation from the reference in a given frequency range (usually 20Hz-20kHz), as in Chart 1.

 

Chart 1. Frequency Response

 

This simple measure is great for comparison between products.

 

Summary

Graphs - Flatter is better

Numbers - Smaller is better

 

 

 

The noise level is a measure of how much noise is introduced during playback.

 

Figure 4. Noise Level

 

 

Chart 2. Noise Level

 

The noise consists of signals that are in the output but were never input.

 

Glossary

* dBA: A-weighted decibels. A measure of loudness that has been weighted according to the A-weighting curve to better reflect the perceived loudness.

* RMS: Root Mean Square. In AC (alternating current) signals, such as that used in audio, the direction of the current is constantly changing, and RMS values are commonly used to measure the absolute size of the signal without reference to polarity.

* Peak: The ceiling of the signal - as it is based only on the largest part of the signal, it tends to measure significantly larger than the RMS value.

* DC offset: Electronically speaking, this is a measure of how much DC (Direct Current) component is in the output. Amplifiers rely on a bias current (DC) to operate, and if this current is not properly taken care of, it may leak into the signal. The DC component must be avoided at all costs because it creates distortion and may damage other equipments in the chain.

 In terms of audio, if this value is positive, then the signal will be played back at levels above what it should be, and vice versa - this may lead to clipping. For instance, if woofer of a loudspeaker moves forward with a positive signal and back with a negative signal, a positive DC offset would make the woofer lean forwards on average, which creates distortion.

 

Summary

Graphs - Lower is better

Numbers - Smaller is better (for all values)

 

 

 

The dynamic range refers to the difference in levels between the smallest and the largest playable sounds. The graph below shows the dynamic range at 1kHz.

 

Figure 6. Dynamic Range

 

 

Chart 3. Dynamic Range

 

A larger dynamic range allows for more variance in the loudness, and in turn a clearer sound. Equipments with higher dynamic ranges present crystal clear sounds out of complete silence, while those with lower dynamic ranges are dulled and somewhat foggy.

 

Summary

Graphs - Graphs are usually not used

Numbers - Larger is better (Dynamic range) ; smaller is better for the DC offset

 

 

THD is an abbreviation of 'Total Harmonic Distortion'.

 

Figure 6. THD+N (-3dB)

 

 

Chart 4. THD+N (-3dB)

 

The THD is the ratio of the sums of harmonic signals created by the reference fundamental frequency (1kHz in this case) to the signal itself. Expressed mathematically,

 

 

For more information on harmonic signals, consult 'Basic Acoustics for Listening'.

The THD+N is more generally presented as in Figure 8; the RMAA (Rightmark Audio Analyzer) is intended to measure soundcards, which means that it only needs to measure the THD+N at a constant output level. However, other equipments such as amplifers change their THD+N based on output power and impedance, hence the need for curves like the one below.

 

 

Figure 8. THD+N

Source: http://www.tripath.com

 

Figure 8 shows the THD+N characteristic of a Tripath chipset. Each point on each line represents a test just like the one whose result was Figure 7, except that the tests are done at incremental outputs so as to produce a line that relates THD+N to power. Each line represents the behavior of the amp under a given load impedance. This provides a three-dimensional knowledge the THD+N characteristic - for example, the graph shows that the chipset has a worse THD+N characteristic at 4Ω than at 8Ω, but also that the maximum output is higher for 4Ω at 12W compared to 6W for 8Ω.

 

Summary

Graphs - Shown as in Figure 7 - lower is better

Charts - Smaller is better

 

 

 

An IMD (Intermodulation Distortion) occurs whenever two signals that have different frequencies meet. It consists of the sums and differences between the original signals, or its harmonics.

 

Figure 9. IMD

 

 

Chart 5. IMD

 

The E-MU 1616M shows some secondary and tertiary peaks as in Figure 9, but I thought that these were examples of harmonics rather than IMD, so I decided to give additional explanation as to why it happens.

 

 

Figure 10. What causes IMD

Source: http://www.mwrf.com/files/30/16649/Figure_02.gif

 

Figure 10 shows the resultant IMD from two signals, at 5Khz and 6kHz, respectively. Mathematically, the distortions are created (the 3rd order distortions were omitted for sake of brevity):

 

     1. 5+6 = 11

     2. 6-1 = 1

     3. 11-1 = 10

     4. 11+1 = 12

 

IMDs are caused by non-linear components (transistors), much like harmonics.

 

Summary

Graphs - Better if there are less second- and third-order harmonics. That is, if there are no signals other than the two original input signals.

Numbers - Smaller is better

 

 

Stereo Crosstalk is, as the name suggests, a measure of how much 'talking' is done between the left and right signals. Electronically speaking, it refers to the interference between different channels - that is, when the signal from the left channel leaks into the right channel or vice versa. Thus, this crosstalk is undesirable.

 

 

Figure 10. Stereo Crosstalk

 

 

Chart 6. Stereo Crosstalk

 

The Stereo crosstalk happens due to electronmagnetic induction - electronically, sound can be described as a combination of the electric potential (voltage) and the direction of the current (polarity). A larger sound corresponds to a higher electric potential, and a higher frequency with a more rapidly changing direction of the current.

 

Whenever there is a current through any circuit, it sets up a magnetic field in the vicinitiy, whose intensity is proportional to potential and the direction is determined by the direction of the current. In order for sound to be played, the signal must change potential and direction - this changes the magnetic field due to the current, and a changing magnetic field induces signal in another circuit, creating crosstalk.

 

Summary

Graphs - Lower is better

Numbers - Smaller is better

 

 

The IMD discussed earlier only considers the distortion caused by two signals. However, music usually doesn't consist of only two frequencies, and thus to obtain a more relevant IMD value, a sweep tone must be used. Measurements made at various frequencies, normally at 5kHz, 10kHz and 15kHz.

 

Swept frequency: As the name suggests, a sweep tone is a sinusoidal signal that smoothly changes frequency ("sweeps") across the audible spectrum beginning in the deep bass.

 

Figure 11. IMD (Swept Frequency)

 

 

Chart 7. IMD (Swept frequency)

 

Summary

Graphs - Lower is better

Numbers - Smaller is better

 

 

 

All of the measurements that we have covered in this article can be put together into a neat chart as below. RMAA evaluates each measured parameter and rates each characteristic into five categories: Excellent, Very Good, Good, Average, and Poor.

 

 

Chart 8. Test summary

 

For a quick look at the results of a review unit, consult the summary chart.