This article covers several measurement techniques, depending on the amount and type of instrumentation you have available.
Frequency Response For the most "natural" sound, your system should respond uniformly to any frequency in the audible range, generally considered to be about 20 Hz (cycles per second) to 20,000 Hz.
One of the Precision Test Signals is a 20 - 20,480 Hz frequency sweep with duration of 10 seconds. While playing this signal, listen for any increases or decreases in sound level. Room resonances and speaker resonances are a common cause of non-uniformity. If you hear a peak or valley, you may be able to adjust that frequency with your graphic equalizer or other tone controls to eliminate it. To estimate the frequency to be adjusted, note the time from the beginning of the track until the peak or valley occurs. The frequency doubles each second, so a peak 3 seconds into the track would be at 160 (20 x 2 x 2 x 2) Hz.
Another signal that can help evaluate frequency response is White Noise - a uniform mixture of all frequencies. When listening to white noise, you should not hear any particular musical pitch.
To determine if frequency peaks or valleys are introduced by your speakers or the room, try listening to the same signal through a good set of headphones.
Note that extreme frequencies below around 100 Hz and above around 7,500 Hz may appear to decrease in volume because your ears are less sensitive in these regions. It is best to assess your own hearing with a set of high quality headphones and known good audio system if possible.
Centering and Phasing can also be evaluated by ear, using the same test signals. If both speakers are at equal volume and in phase with each other, the sound should appear to come cleanly from the center. If the source shifts from one side to the other during a frequency sweep, or sounds like "stereo" when listening to white noise, the two channels are not exactly matched in amplitude and phase.
If this happens, you can try adjusting each channel individually for flat frequency response.
If white noise, and even the frequency sweep sounds like "stereo" instead of coming cleanly from the center, try reversing the polarity of the connections to one of your speakers. Out-of-phase speaker connections are a common problem.
Sound Pressure Level Meters are self-contained and very convenient to use. Inexpensive units from Radio Shack, as low as $39.99, claim reasonable accuracy up to 10 KHz.
For wider frequency range at a similar price, a Measurement Microphone, such as the Apex Ape220N, may be plugged into a metered audio preamp or PC.
Begin with the speaker terminals. Play a mid-frequency signal such as 1 KHz, and adjust the volume to a convenient reference level. If your meter has a decibel scale, you can make your measurements in decibels instead of volts. For rapid testing, play one of the frequency sweep signals and note the amount of variation on the meter. For more precise testing, play as many individual frequencies as you like, and note their voltage levels with respect to that of your reference. If your system is "flat," voltage levels at all frequencies will be the same.
If you find any anomalies, you may want to measure signal levels earlier in the signal chain. A Y-Cable between your CD player and amplifier lets you measure the CD player's output directly, while still listening through your speakers. Your system may have other access points, such as between the preamplifier and power amplifier, which can be individually measured. In this way, you can narrow down the source of any anomalies to one component of your system.
Frequency Response can be measured via individual frequencies and sweeps as with a voltmeter. Alternatively, you may do a spectral analysis on White Noise. The result should be a flat response over the entire spectrum. Any variations should be readily apparent.
Phase Coherency can be measured by observing left and right waveforms in time alignment. Waveform tops and bottoms should coincide at all frequencies. Any discrepancy represents distortion, measured in degrees. (One full waveform cycle is 360 degrees.)
Phase Delay can be measured by viewing square wave signals. They should, as the name implies, appear square in shape. If instead they appear "wiggly," it is because some frequency components (usually higher ones) have been delayed longer than others. Analytically, square waves consist of sinusoids at 1, 3, 5, 7 ... times their fundamental frequency. When all are properly aligned, they add to form a square shape. If they are not properly aligned (i.e., some are delayed more than others) the square becomes wiggly. By analyzing the shape of the wiggles, you may be able to determine which frequencies are delayed and by how much. This capability may also be present in your spectrum analyzer or spectral analysis software. (Note: Square wave analysis should be undertaken only after frequency response has been set flat. If the various frequency components of the square wave are not in proper proportions, the waveform will appear distorted.)
Waveform Distortion When you view a sinusoid of any frequency, it should be smooth and perfect. Small amounts of distortion are difficult to see. On the other hand, common problems such as "clipping," in which the waveform is flattened at the top or bottom, are easy to observe.
Many sinusoids are present on the CD, and all may be analyzed for distortion. A "stress test" is available in the form of a 1 KHz sinusoid recorded at the absolute maximum undistorted level.
To determine the amount of distortion present, it is best to do a spectral analysis on the waveform. Only one frequency (the original) should be present. Additional frequencies, at multiples (or sub-multiples) of the original represent distortion. The combined level of harmonics, as a percentage of the original, is called "Total Harmonic Distortion," or THD.
Inter-Modulation Distortion is a special kind of harmonic distortion in which one frequency interferes with, and "modulates," another. To measure this, a 50 Hz sinusoid at 90% of maximum amplitude is mixed with a 5000 Hz sinusoid at 10% of maximum amplitude. Spectral analysis should reveal only the two original frequencies. However, multiples of the lower frequency may show up above and below the higher frequency, i.e., at 5050 Hz and 4050 Hz. The combined level of sideband frequencies, as a percentage of the originals, is Inter-Modulation Distortion.