From Ear to There - Measurements in Audio
Sony Music was once again the host for the Toronto Section's annual day long extravaganza in May. This year's topic dealt with all things measurement: subjective, acoustic, digital, measuring the performance of mics and loudspeakers and a look at some of the tools of the trade. The following is a short review of some of the presentations and demonstrations.
Subjective Measurementsby Sean Olive of Harman International
Subjective Measurements was the first type of audio measurement in our day long event. The presenter was a man whose job title is probably unique in the world: Manager of Subjective Evaluation. Sean Olive of Harman International is exceptionally qualified to speak on the topic of subjective measurements.
Sean has overseen a stereo listening room where loudspeakers have been evaluated for a number of years but recently the CEO handed down an edict that all Harman products must be subjected (!) to subjective listening tests. Last year Sean developed the Multichannel Listening Laboratory which allowed for evaluation of the growing number of products for the surround sound market. The MLL is an NC5 room which meets all the ITU/EBU standards. It features the world's only automated speaker mover. By moving the speakers into the same position for listening tests, positional effects are removed from the results. The practical result of this investment is that the length of the arduous listening tests can be reduced by a factor of at least 4.
Mr. Olive sighted an earlier paper he had co-authored with his mentor Floyd Toole which clearly showed the difference in results from sighted and blind tests. If you show someone an expensive looking loudspeaker it will score higher than a cheap looking loudspeaker. The results may be the opposite on a blind listening test.
Next Sean explained how they select subjects for their listening tests. Potential listeners are given an audiometric test to check for normal hearing. Those who pass are then tested for consistency in their results. The people who are consistent in their reports are chosen as subjects. They are then trained in what to listen for. Trained listeners give more discriminating, reliable results than untrained listeners. This rigorous selection and training process ensures that useful information is fed back to the engineers and designers of the products.
Finally Sean explained that the frequency response of a product does not tell the whole story about how good it sounds. If we turn up the level the distortion may go up and power compression may kick in. He feels that in their subjective evaluation they pretty much have frequency response down and that over the next couple of years they will be looking more closely at distortion. by Peter Cook, Committee Member
Measuring Acoustical Properties of Materialsby Rob Stevens of HGC Engineering
The acoustical properties of materials comprise one of the two broad categories of parameters that we can control or "tweak" in acoustic design. "Acoustic design" can cover the design of any sound-sensitive room, such as a recording studio, concert hall, an automotive interior, or a loudspeaker cabinet. The other broad category of design parameters relates to the geometry (size, shape, etc.) of the item being designed.
Acoustic material properties of interest include sound absorption, sound transmission, and internal energy dissipation (damping). Sometimes diffusion is considered to be a material property, although it may be more properly classified as a geometric characteristic of the surface of a material. The properties most often needed for acoustical design purposes (particularly for rooms), are the absorptive and transmissive characteristics of the material. Absorption is the amount of sound that a material can "soak-up" (ie: the sound energy dissipating into heat energy through viscous friction). Generally, porous materials such as fiberglass or open-cell foam are good absorbers. Absorptive materials are important in controlling reverberation and reflected sound within an enclosed space. Transmission describes the amount of sound that will pass through a material. Non-transmissive materials are needed to contain sound within a space, or isolate one space from another, such as between a studio and control room. Heavy, solid, limp, air-tight materials are good at preventing sound transmission.
The classical method of measuring absorption and transmission involve the use of one or more reverberation chambers in an acoustical laboratory. Inside the reverberation room, a statistically diffused sound field is created to allow an accurate measurement. There are some limitations to the classical approach, which have motivated the development of newer measurement methods. The newer methods employ variations of the sound-intensity technique, which involve measuring both sound pressure and acoustical particle velocity. This can be accomplished by using a phase-matched pair of microphone, or a microphone/anemometer pair, or successive measurements with one mic and a deterministic sound source such as an MLS signal. When measuring absorption , this allows the decomposition of the incident sound wave versus the reflected sound wave, enabling in-situ amplitude/phase measurements of surface absorbtivity in a real room. When measuring for transmission loss, the newer techniques allow for the rejection of interfering background sound, and also enable multiple simultaneous sound transmission paths to be separated and quantified individually (eg: separation of the transmitted sound energy through the walls versus doors versus windows of an acoustical enclosure).
The demonstration included absorption measurements of a fiberglass sample in an impedance tube using one mic and an MLS source signal, transfer-function measurement of the damping of a steel strip using a force-transduction-hammer and a micro-accelerometer, and the basic configuration of a dual (phase-matched) microphone sound intensity probe.
by Rob Stevens, Committee Member
Measurement: Hardware and Methodby Mike Wolfe of Audio Precision
In the short time available to him, Mike Wolfe of Audio Precision provided an interesting and pragmatic description of new as well as traditional measurement techniques associated with both analog and digital audio. He presented a comparison of the results obtained from the different techniques and highlighted the pros and cons of each. From this, it was evident that multitone testing provided frequency response and distortion results much faster than the traditional analog methods thus making this method highly advantageous for manufacturing production lines. As the number of tones used in multitone testing can be significantly higher than the standard two-tone IMD test, intermodulation distortion measurements are more useful as the test more closely simulates real world conditions.
Mike emphasized the importance of recognizing and dealing with out-of-band interference. Sample rate generators, DC to DC converters, computer clocks, Sigma Delta converter noise as well as the traditional sources of radio interference all can produce peak levels that exceed the lowest useful signals in audio circuits. If measurements are to be meaningful, the bandwidth of the measuring instruments must be sufficiently wide. Mike also covered the characterization of the digital interface signal and the embedded audio, A-to-D converters, D-to-A converters and more.
Mike concluded his presentation with a comparison of acoustical measurement techniques. These included swept sine wave, impulse excitation, time delay spectrometry (TDS) and maximum-length sequence (MLS).
By Jim Hayward, Chairman, Toronto Section
Objective Measurement of the Perceived Quality of Audio Codecsby William Treurniet of CRC
The ITU (International Telecommunications Union) based in Geneva required standards for evaluating the quality of digital audio devices. In particular, there was a requirement for methods to compare audio codecs for broadcast purposes.
The CRC began developing the standard ITU-R BS.1116 for subjective testing, which is now in place. This standard describes procedures for comparing audio devices with very small differences in quality. This method has the listener rating two unknown audio sources against a known reference source. A single digit score is generated that rates the test source against the reference. The procedure was found to be involved and expensive and was not applicable in a number of circumstances when measuring audio quality.
An objective testing method ITU-R BS1387 called PEAQ (Perceptual Evaluation Of Audio Quality) was developed by a collaboration of international organizations including the CRC. The goal was to design a method to predict the outcome of subjective listening tests. The PEAQ method combines the data of known measurement methods into a single figure to predict audio quality. Calibration and trial tests of the method have shown a strong correlation with measurements by subjective testing.
Having an objective testing method available can be useful for the processing of more items per audio device, and can serve as a tool for algorithm developers.
A detailed paper on the project is available in the AES Journal Volume 48 Number 1/2.
by Ron Lynch, Committee Member
Room Acoustics MeasurementBy John Bradley of the National Research Council
Dr. Bradley prefaced his talk on room acoustics measurements by emphasizing that ambient room noise is a more important factor affecting the overall aural success of a room than is it's acoustical characteristics. As a simple elaboration of this fact, he pointed out that it is much easier and more common to wind up with 10 dB too much background sound in a room (a tenfold excess), than to have ten times too much reverberation, for example. Thus, while the topic of room acoustics often receives most of the glamour and attention, ignoring the issue of background sound can negate all of the good efforts in designing for optimum room acoustics. A simple measure of ambient sound is the overall A-weighted sound level. For most situations however, more detailed ambient sound levels are measured in octave bands and rated according to a set of criterion curves such as the Noise Criterion (NC) or Room Criteria (RC) schemes.
The reason we measure room acoustics (rather than just rely on our own "golden ears") is to allow us repeatable, quantitative information about a room for design purposes. Dr. Bradley noted that we need to measure quantities that have some subjective relevance to peoples' response to the acoustical quality of a room, given it's intended purpose such as a lecture hall, a concert theatre, a recording studio, etc. As well, we generally need to measure a good sampling of locations throughout the space where listeners would normally be situated.
As with most audio/acoustic measurements, the acoustic properties of rooms can be made in either the time domain or frequency domain. At low frequencies, in small rooms, the acoustic properties are determined primarily by room resonances or "modes". In this regime, frequency response measurements are useful. However, in most moderate to large sized rooms, such as classrooms or concert halls, the room resonances are sufficiently plentiful and evenly distributed in frequency, such as not to significantly effect the room acoustics. Thus, except for relatively small rooms, such as recording studios and control rooms, time domain measurements (usually in the form of an impulse response) are most informative as they directly show information about reflections and sound decay within the space. These are essentially statistical-energy measurements.
Reverberation Time (RT) and Early Decay Time (EDT) are two common measures of sound decay in a room. Reverberation is the tendency for sound to linger after the sound source has stopped and is measured in terms of the amount of time required for sound to decay by 60 dB. The EDT value represents the sound decay in only the first few moments after the sound source has stopped. Because both speech and music contain relatively quick transients, the EDT value is often more important for speech intelligibility and musical clarity than the RT value. Greater amounts of reverberation add warmth to music, and are desirable in spaces intended for unamplified music.
With regard to speech intelligibility, early arriving reflections (within about 80 milliseconds of the direct sound) are desirable, as they tend to enhance and reinforce the direct sound. Late arriving reflections (after about 80 milliseconds) are generally detrimental, as they act as an interfering sound, similar to background noise. Therefore, it is common to measure ratios of early-to-late arriving energy in rooms intended for speech. C80 is one such measure; it represents the ratio in dB of the energy arriving before 80 ms to that arriving after 80 ms. To account for the detrimental effect of ambient noise, we can include noise in the ratio by adding the ambient sound energy to the late arriving reflected energy, resulting in the U80 value, which represents the ratio of useful to detrimental sound in the room. As it turns out, U80 correlates very well with the Speech Transmission Index (STI) rating of a room, but is much quicker and easier to measure, by making a simple impulse response measurement.
The "relative level" or room gain (G) is a parameter which is important for both speech and music rooms. G represents the degree to which a room amplifies or attenuates sound versus free field conditions (amplification occurs through contributions of reflected sound, while attenuation results from sound absorption by the room furnishings). Louder is better for both speech and music. Higher G values result in improved speech intelligibility and more positive subjective response to the musical quality in a room.
Dr. Bradley briefly touched upon two more esoteric measures relevant particularly to music spaces: Apparent Source Width (ASW) and Listener Envelopment (LEV), both of which involve measuring not only the time domain characteristics, but the also the spatial/angular distribution of reflected sound arriving at the listener.
To conclude, Dr. Bradley listed some typical acoustical design goals for several common types of sound sensitive spaces, as follows: A small to medium classroom or meeting room [^] ambient sound less than 35 dBA or about NC/RC-30, RT/EDT about 0.5 seconds; A teleconferencing room [^] ambient sound less than 30 dBA or NC/RC-25, RT/EDT about 0.3 seconds; A concert hall [^] ambient sound less than 25 dBA or NC/RC-20, RT/EDT about 2 seconds, G about 5 dB and C80 about 2 dB.
by Rob Stevens, Committee Member
Measurement Problems for Loudspeakers
by Don Keele of EVI Audio
Don Keele's lecture focused on the problems faced in measuring loudspeakers in less-than-ideal conditions and various options available to work around these obstacles. Most testing is carried out in semi-reverberant rooms because anechoic chambers are very rare, expensive and limited in low frequency resolution or free-field measurements are subject to the vagaries of weather and noise. Diffuse-field measurements require much averaging but can be useful for measuring total sound power radiated.
Measurements in a semi-reverberant situation are valid as long as no reflections from room boundaries are included. To measure lower than that limit, the very-near-field technique must be used. The microphone is placed as close as possible to the cone or aperture. The results track anechoic far-field measurements within a decibel as long as the diaphragm is operating in its pistonic range, typically below 800 Hz for your average woofer. If more than two sources are to be measured, such as the woofer and port, the two may be measured separately and vector summed or a location for the microphone may be chosen between them.
To help relate measurements to what the ear hears, spectral averaging or smoothing can be used. While this ignores very narrow resonances or dips, wider-band tendencies are seen more clearly.
Ground plane measurements, where the loudspeaker and microphone are situated very close to one boundary have some attractive features. By being so close to the nearest surface, reflections from that surface are eliminated or minimized. In an outdoor situation this can leave a very long distance for the first reflection if the site is not close to any obstacles. As long as the surface is smooth [^] Styrofoam sheets, for example, and high frequency elements are kept nearer the boundary, valid results up to 16kHz may be obtained. One must realize that the source appears as a mirror image to the microphone and apparent sensitivity is increased on the order of 6dB. To measure the high frequency source on axis, tilt the enclosure.
Shaped tone bursts are a useful measurement tool for determining the short-term power handling and SPL capabilities of speaker systems without destroying them in the process. This technique, originated by Siegfried Linkwitz, uses a 6-[product] cycle burst shaped by a Hamm or Hamming window.
Crest factor is about 7 dB compared to 3 dB for a sine wave or 0dB for a square wave. When cycled every second or two this allows inputs of several kW into a dome tweeter with no fear of vaporizing the voice coil. In practice, the power is raised until the driver sounds audibly distressed, the power is backed off slightly and then power and SPL measurements are taken. Shaped tone bursts are also useful for hearing LF room reverberation and detecting driver resonances when the fundamental is filtered
by Paul Reibling, Treasurer and Recording Sec, Toronto Section
At this time the Toronto Section would like to extend a heart felt thank you to the two gentlemen from the University of Waterloo: Drs Lipshitz and Vanderkooy for their guidance. Thank you to our hosts, Sony Music and Denis Tremblay for organizing the venue. Also thank you to Jim Norris of Norris Whitney Communications for their help with the PR. And of course a thank you to all of the presenters for giving of their time and knowledge and helping to make the day such a success.
Return to the Front Page