Audio Test Stimuli
Steve F. Temme and Steve Tatarunis
There are many audio test stimuli in common use, and selecting the best stimulus for your application can be a daunting task as most audio analysis systems offer a wide range of options. This overview explains the different stimulus types, what the waveforms look and sound like (click the speaker icon to hear the sound), the advantages and disadvantages of each and when you would be likely to select each one.
Stepped Sine Sweep
A stepped sine sweep, as its name would suggest, plays a sine wave at discrete frequencies, step by step, within a frequency range specified by the user. It is not a continuous sweep; the user sets a resolution to determine how many steps there will be. It is one of the most widely used audio test stimuli for basic electroacoustic testing (particularly for testing raw transducers) because it offers good noise immunity and yields a lot of data such as frequency response, phase, harmonic, and various distortion measurements (THD, Rub & Buzz, Perceptual Rub and Buzz). Some test system vendors offer variations on this basic stimulus. For example, the SoundCheck test system from Listen has a unique variant called the ‘Stweep’ which is an optimized stepped sine sweep that uses an integer number of cycles at each frequency step, ensuring a smooth transition from frequency to frequency. This reduces transducer ringing and ensures significantly less transducer settling time which results in faster and more accurate measurements. In measurement systems that support compound stimuli, it can be desirable to configure a stepped sine wave with different resolutions at high and low frequency (more resolution at high frequency) to optimize the trade-off between stimulus duration (and therefore test time) and accuracy.
Log Sine Sweep (also known as Frequency Log Sweep, Log Chirp or Farina Sweep / Farina Method)
A continuous log sine sweep (as opposed to the stepped sine sweep which plays multiple discrete frequencies) from low to high will cover every frequency in the chosen range. It uses the same time (and energy) for every octave, which achieves a good signal to noise ratio for all frequencies in typical electro-acoustical measurements. Since the sweep rate is slow at low frequencies but increases with frequency, distortion is accurately measured at low frequencies while retaining a fast test speed and high production line throughput. This makes it a common choice of audio test stimuli for production line loudspeaker measurements for fast time selective measurements of fundamental response and distortion. It is also widely used in conjunction with a Time Selective Response algorithm for simulated free field measurement. The first commercial implementation of the Log chirp stimulus was SoundCheck, just months after Angelo Farina’s landmark AES paper on the subject in 2000. It is now available in several test systems.
Log Amplitude Sweep
A Log amplitude sweep is similar to a stepped sine sweep except a single frequency is chosen and swept logarithmically through various levels. It may sweep from low to high, or high to low. This test signal is used for testing linearity to see how devices perform when different stimulus levels are presented and for testing devices with automatic gain control (AGC). For example, to see at what level a compressor kicks in and how it behaves at different stimulus levels. It can also be used in conjunction with advanced analysis algorithms to look at characteristics like THD vs. stimulus level, which will produce a curve that has % distortion on the y-axis and stimulus level on the x axis.
Pink Noise, White Noise and MLS Noise
Pink noise, white noise and MLS noise audio test stimuli have all been around for a long time and used for a variety of different purposes. These are used in situation where sine waves cannot be used, usually devices which include DSP. One of the most common applications these days is for testing noise cancelling headphones. Pink noise is noise with a continuous frequency spectrum and with equal power per constant percentage bandwidth. For example, equal power is any one-third octave band. White noise is noise with a continuous frequency spectrum and with equal power per unit bandwidth. For example, equal power in any band of 100-Hz width. MLS noise is the noise signal used in the MLSSA system, a system which although no longer available was widely over the past 20 years. Some audio test systems (e.g. SoundCheck) offer it so that users can directly compare their new measurements to historic data.
A multitone stimulus is multiple sine tones of discrete frequencies played simultaneously with equal amplitudes and random phase. All the frequencies are present at the same time (in contrast to the stepped sine sweep where the frequencies are present sequentially). The advantage of this audio test stimulus is that all frequencies are covered in a very short stimulus, which is useful for production line environments where you just want to test frequency response, for example microphone testing. The use of a multitone signal does place restrictions on other electroacoustic measurements. For example, it is not really possible to measure harmonic distortion because the number of simultaneous frequencies in the stimulus makes it virtually impossible to examine the individual harmonic structure of each one of those stimulus frequencies. Instead, non-coherent distortion (a technique that involves dual channel analysis of the noncoherence between a stimulus and response) can be used instead.
Two Tone Stimulus
As the name suggests, a two tone stimulus plays two simultaneous tones. Two-tone stimuli are used for measuring distortion. When two tones are played in a non-linear system, they generate output frequencies that are not part of the stimulus. These ‘orders’ are different linear combinations of the two original frequencies and are undesirable since they have no harmonic relationship to the original signal and are described as intermodulation distortion. There are two different types of two-tone stimulus – Intermodulation (IM) and Difference Frequency (DF), used for measuring intermodulation distortion and difference frequency distortion respectively. The intermodulation two-tone stimulus is a stepped sine sweep superimposed on top of a fixed frequency tone. The fixed tone is usually a low frequency tone, and the sweeping frequency is usually restricted to frequencies five times greater than the fixed tone frequency. The difference frequency two tone stimulus is composed of two simultaneous sweeping tones separated by a specified frequency interval which may be a fixed difference or a fixed ratio.
Speech and Music Based Signals
Many modern audio devices contain signal processing that requires the device to be tested with a signal representative of what it will see in the real world, and this requires the use of speech or music audio test stimuli. This is often the case for devices with built-in non-linear signal processing such as cell phone codecs and hearing aid compressors. Some test systems allow you to load your own custom WAV file from disk to use as a stimulus. This may be music, real speech, or artificial speech. Artificial speech reproduces the average characteristics of a human voice. It is often used for objective evaluation of speech processing systems or devices. The use of artificial voice as the signal test, instead of real speech, has the dual advantages of being more easily generated and having a smaller variability than samples of real voices. P50 male speech is specified as the test signal in many telecommunications standards. When testing with speech, some measurement systems also offers active speech level control (required for some telephony standards).
Artificial Speech Stimulus
Composite Stimulus Signal
Some tests require the use of a composite stimulus signal. For example, a conditioning signal to warm up the device or open up an activity detector before playing the stimulus signal. Some test systems offer the ability to create and later analyze such signals.