Archive
A Case Study of Three Selected Recording Studios:
Using an Integrated Approach to Determine the Influence of Control Room Ergonomics Upon the Mixing Environment and the Role of Pre-Mastering in Balancing these Effects
by Pedro A. SilvaSeptember 19th, 2005; Revised May 20th, 2006
Table of Contents
-
Key Concepts
-
Content and architectural programming
These are, according to Storyk (1989:1), elements of acoustic design, with the other two being acoustical concerns and human-machine interfacing factors (in this study referred to as "ergonomics"). Content programming is related to the constraints posed on the acoustic design of a studio, as a result of the specific programme material that is to be reproduced in its premises. This is obviously determined by the specified function of the studio. Architectural programming is the product of the architectural needs of the venue. In other words, the existing architectural conditions influence the way the acoustician approaches the design process.
-
Control room ergonomics
In the context of this paper, control room ergonomics relates to all factors which, not falling into either acoustic or architectural considerations nor specific content programming, are the direct product of human-machine interfacing. Examples of this interfacing are near-field monitoring, placement of large mixing consoles and other large rigid surfaces, cockpit-like positioning of outboard equipment, use of multiple computer displays and large communications windows. The ergonomics factors that constitute variables in this study (desktop reflections and near-field monitoring) were chosen on the basis of observation, and are thought to have the most important impact in the monitoring environment. The particular use of the word "ergonomics" in this paper should not be confused with sound ergonomics, as it is commonly used. That area specifically deals with spatial auditory displays, speech intelligibility, recognition and synthesis, noise and hearing protection, and bears little relevance to the study at hand.
-
Mastering or pre-mastering
The term mastering is commonly used in the audio recording industry to refer to the set of creative and or corrective processes of post-production applied to programme material after the mixing stage, prior to duplication. Pre-mastering would be the more technically accurate term, as defined by Bob Katz in his "The Art of Mastering" (2002). Although the use of the term mastering should only be applied to the glass mastering that takes place in a replication plant, in the context of this paper, both pre-mastering and mastering will be used in relationship to the technical-creative process that happens after the mixing stage, unless noted otherwise.
-
-
Introduction
Form ever follows functionality
Louis Henry SullivanThis aphorism, developed in the late nineteenth century by the American architect Louis Henry Sullivan (1896), established the fundamental principle of Functionalism, an "aesthetic doctrine", in which the "interior program dictates the outward form" (Columbia Encyclopaedia, 2001-4). Functionalism can be found in modern recording studio design, in which its validity is obvious, if "neutrality" is to be desired (Spikofski, 2000:1). In fact, this area has advanced to such an extent in the last decades that it can now be considered as one of the most important aspects of music recording. In particular, the live-end dead-end (LEDE) concept by Chips Davis and Don Davis (1980) was a major breakthrough in acoustical design, as was the development of the reflection free zone by D' Antonio (1985). However, some practices still persist in the music recording studio, in an apparent contradiction of current design philosophies. As an example, these include the widespread use of large format consoles with their consequent effect on the reflection free zone, thereby negating its advantages (Walker, 1995/4:25). Other ergonomic-related practices can be found in the use of near-field monitoring, a compromise in sonic accuracy; and cockpit-like placement of outboard equipment, which significantly adds to the noise floor in the critical operating environment.
This study aims to identify one such practice - the primacy of control room ergonomics over acoustics - and find evidence that this seemingly irrelevant inequality of both studio design elements - acoustics and ergonomics - does have, in fact, an important effect on the downstream production processes: that is, post-production processes, such as mixing and pre-mastering. It is particularly concerned with the way some control room ergonomics factors interact with the monitoring environment, and in what way operational decisions made during the mixing process carry consequences to the pre-mastering stage. Its scope is limited to the music recording production, particularly with regards to acoustic music genres because, as technical recommendations have highlighted (EBU, 1997:3), these are more susceptible to the types of problems under investigation.
It is formally hypothesised that ergonomics have primacy over acoustics in current control room design implementation. Because the mixing stage is usually part of the production process, is therefore accomplished within the recording studio. This leads to a fundamentally flawed monitoring environment and, consequently, to an inaccurate operational decision process by the mixing engineer. Thus the need for an otherwise redundant corrective post-production stage often arises in the form of pre-mastering. By using a combination of objective measurements of acoustic parameters potentially related to ergonomics, subjective assessment of these parameters by a listening panel, observation of a test subject's operational decision process when under the given conditions, and finally connecting the results with a survey of mastering engineers, it should be possible to establish a pattern to all three studios, and thereby sustain the hypothesis.
Upon reviewing the relevant literature in acoustic design, control room ergonomics and pre-mastering, a general research methodology is presented, followed by a technique overview, in the "Materials and Methods" chapter. The experimental sections detail the aims, materials and methods, observations and results, and discussion of results for each of the conducted experiments. Unless a new or adapted, and therefore undocumented, technique is used, no detailed explanation is provided. Instead, a simple reference to the relevant literature is given. This approach is endorsed by Chandrasekhar (2002:14). This is followed by a general discussion, which aims to support - or review - the hypothesis in the context of the experimental data. I conclude by addressing the deeper reach of this research within the music recording production process, and what it means for its operators.
-
Literature Review
In investigating previous relevant research, the most successful approach is, necessarily, one of breaking the study into smaller components, and treating those separately. As such, one can identify acoustic design, ergonomics and pre-mastering as the most obvious elements. Within these, acoustic design concerns should go no further than control room design for the recording of music; control room ergonomics must exclude sound ergonomics, concerned with speech intelligibility, spatial auditory displays, noise and hearing protection. Instead, it must focus on human-machine interfacing factors that do not directly relate to acoustic accuracy; mastering should be considered in its role of a technical-creative corrective set of processes, thereby excluding the mainly industrial setting of duplication preparation. While extensive research has been made on various aspects of acoustic design, little has been said regarding the effects of ergonomics, as defined in point 0.2, upon the monitoring environment of recording studios. In fact, there is virtually no previous research on control room ergonomics. In the same manner, pre-mastering has been consistently disregarded by the academic community, and thus continues to be an essentially professional area. A chronological perspective of the literature on acoustic design is presented, from concert hall considerations in the nineteen-fifties to modern control room theories in the nineteen-nineties. In parallel, the works of John Storyk and Robert Walker are analysed. Storyk is an architect and acoustic designer with more than 900 designs to his credit, and what little investigation into control room programming of content, ergonomics, architectural and acoustics exists is due to him. Working out of BBC's research and development department, Walker has produced research on early reflections and control room design very relevant to this study. Finally, some insight is brought into mastering practices, through anecdotal investigation made by three authors: Bassal, Katz and Owinski. Taking these factors into consideration, this study draws upon, and attempts to build on, the following works.
-
Acoustic design
The acoustic parameters related to ergonomics in this study are those associated to the presence of important discrete reflections close to the direct sound. The thin line between early and late reflections and psychoacoustics was first crossed by Haas (1951), whereby he developed the important theoretical fundamentals of how the human auditory system processes differently timed delays. Of particular importance were the concepts of early reflections integration, early-late reflections effect in directional information, and late reflections' interpretation as discrete echoes. Beranek (1962) extensively researched this matter, albeit mainly with regards to concert halls, auditoriums and other forms of large acoustic spaces. Specifically, the notion of an initial time-delay gap (ITDG) of about 20 milliseconds (ms) as being subjectively desirable was the product of this ground-breaking research. Davis and Davis (1980) developed the concept of a live-end dead-end room (LEDE), in part due to Beranek's investigations. Their work explored a well-defined time-delay gap, by mixing absorbent front surfaces and reflective back walls. This model was questioned by Wrightson in 1986, because it relied on specular reflections produced by the back-wall surfaces. This, he argued, destroyed the localization cues so important for correct sound assessment. This seems to be confirmed by earlier research done by Cable and Enerson (1980), who addressed Haas' well-known fusion principle, and concluded that the "early late arrivals" (ELA) do not transfer coherent information, but rather add to the diffuse field. If these ELA have significant energy (less than ten decibels (dB) below the direct sound within the ELA window), as typical specular first reflections have, directional content may be lost, by destructive interference. Instead, Wrightson proposed, "high-amplitude, discrete reflections should be avoided by optimization of room geometry" (1986:793). At about the same time, D' Antonio and Konnert (1985) presented research in control room design using reflection phase grating acoustical diffusers. These were product of Schroeder's work on the limitations of maximum-length versus quadratic-residue diffusers. They successfully created an appropriate reflection-free zone (RFZ) where spatial information was maintained. In 1985, Voelker conducted a series of experiments across four different types of control rooms (reverberant, LEDE, damped and active/loudspeaker-based). His conclusions pointed towards the adequacy of different control room designs towards different types of music. Storyk (1989) built on the concepts of LEDETM and RFZ, and developed is own MODEL (MOdified Dead End Live end) design, where additional requirements were employed. Specifically, he introduced the concept of "venting", where low frequency content should be absorbed from the room faster than mid and high frequencies "to insure good stereo imaging in the listener's position". Working from BBC's R&D department, Walker (1995/3/4/5) developed another type of design, improving even more on existing designs: the controlled image design (CID) employs reflective material where absorption has traditionally been used, as a way of directing reflections outside of the RFZ. This, it was argued, improved stereo image, while making the stereo effect less dependent on the room. This design was partly based on the Archimedes project which has produced valuable data on thresholds of detection of reflections. Incidentally, the EBU Tech. 3276 and ITU-R BS.1116 standards describe very similar specifications, perhaps led by Walker's investigation. Naqvi and Rumsey (2005) propose both the controlled image design and EBU and ITU's recommendations as adequate "for stereo and surround production". Finally, in 2003, Lee showed how auditory load significantly affects localization accuracy, withstanding even repeated practice of subjects.
As thorough as this overview of improvements in control room design may be, little has been said with regards to the place of ergonomics within a functional perspective - that of absolute sonic accuracy - of the control room. This study intends to show how this specific evolution of acoustic design of the control room could be further improved, by eliminating the final element detracting from sonic accuracy: control room ergonomics. Specifically, the effects of the widespread use of large format consoles and near-field monitoring in the critical listening environment of a control room are examined. Walker's research on BBC must be acknowledged at this point as a very strong inspiration for this study, as the author has produced extensive literature on the subject. In fact, appendices I and II of "The management of stereophonic image quality" (1995/4:25-6) are one of the most important sources of information for this research, as they specifically deal with the subject of "diffraction and desktop reflections", and "the measurement of time-frequency responses", which are fundamental in the objective evaluations conducted here.
-
Control room ergonomics
Previous research by Storyk (1989) on the topic of architectural, content and ergonomics-based programming in studio design, is the only reliable source of information on ergonomics within the control room. In it, Storyk argues "acoustics is an equal partner to ergonomics and architectural programming in studio design". Furthermore, in "many instances (...) acoustics simply react to this programming" (p.2). The content programmatic requirements define the ergonomics of the control room, with these influencing the acoustics, which the architectural elements must then either accommodate, or themselves influence the content programming choices; and as early as 1976, Børja noted that the "listening response" of a control room indelibly marked its productions in an inversely proportional manner. The research at hand builds upon Storyk's notion of acoustics in a passive role in modern acoustic design, as well as Børja's early claim that the mix-down process is inevitably marked by the room response it is accomplished within. In doing so, it relates ergonomics preponderance over acoustics with the need for post-production processes at later stages - the case of pre-mastering.
-
Pre-mastering
Concerning mastering, the scenario is more complicated, with no significant research whatsoever on the topic. Aside from Bob Katz (2002) and Bobby Owinski (2000) excellent books on the practice and an unpublished article by Bassal (2002), little else has been documented. Bassal's article presents some useful, albeit anecdotal, research on mastering. Two surveys were conducted on the practice of commercial studios and musicians regarding mastering. Although not academic, his results can be related to the work at hand, and may help establish the difficult relationship between the experimental process and the pre-mastering stage. Specifically, one similar survey is implemented here, with a number of mastering engineers being asked to express their views on the problems with source material generally encountered in their work. Owinski's "The Mastering Engineer's Handbook" is heavily focused on a number of interviews with well-known mastering engineers. Katz's "The Art of Mastering" tends to have a more holistic approach to the area, and is generally well regarded by the industry. One very specific influence of his work on the researcher is the notion of pre-mastering as a separate process to mastering. This view is used throughout the paper, although intermittently conflated with the word mastering, for simplicity.
-
-
Materials and Methods
-
Rationale for a case study approach to the research
Because it employs " a family of research methods having in common the decision to focus on inquiry about an instance" (Adelman et al quoted by Bell, 1999:10), the present research is considered to be a form of case study, as proposed by its title. It should stand to reason that a multi-method approach to characterizing a specific element (the "instance") of three cases, chosen for their representativeness of the typical industry high-end recording studio, fits this category. Again resorting to Bell, the study "is concerned principally with the interaction of factors" (p.10). Its mainly experimental approach - albeit through fundamentally different experiments - is explained by Nisbet and Watt's assertion that "sometimes it is only by taking a practical instance that we can obtain a full picture of this interaction" (quoted by Bell, 1999:10). Elaborating: the decision to drive the case study by means of a survey, as described previously and exemplified in diagram 1, has been applied before (Bell:11). The use of three similar studios, provided "the researcher [...] obtain[s] data on the significant features [...] in general, and then demonstrate[s] where the case study example fits in relation to the overall picture" (Denscombe, quoted by Bell, 1999:10), allows the generalization to other cases. Obviously, the "particular instance" under inquiry - two ergonomics-related factors - is fairly narrow, and therefore cannot be considered in light of a traditional case study in social sciences for example, as its basis is experimental, and focuses on a specific aspect of control room acoustic design. Although Yin argues against the application of case studies to implementation processes (as is the case in question), he also asserts that "the more a study contains specific propositions, the more it will stay within reasonable limits" (quoted by Bell, 1999:12). That is certainly the case.
-
Methodology overview
An integrated methodological approach is used. In that sense, there is a "logic of triangulation", because "the findings from one type of study can be deduced against the findings derived from another one" (Blaxter et al., 2001:85). It could be said that the general methodology is also hybrid in nature, as it uses "qualitative research in a quasi-experimental quantitative study" (ibid). The choice of specific techniques must be closely related to the hypothesis. This helps focus the study in what effectively matters at all times. Thus, the hypothesis presented in this study is broken down into the following premises:
- Ergonomics have primacy over acoustics in control room design implementations.
- This creates a flawed monitoring environment for the mix-down process.
- Operational decisions are proportionally affected.
- Additional post-production in a neutral environment - the pre-mastering studio - is required to correct those mistakes.
HERE GOES DIAGRAM 1
Premise one can be demonstrated by the assertion that any interfacing considerations not directly related to acoustic accuracy are error-inducing, with respect to sonic accuracy. The second premise creates the first research problem: to show that there is a relationship between ergonomic-related factors and monitoring flaws. Premise three demands correlation between previously identified monitoring flaws and specific operational decisions. Finally, the last assertion - the fourth research problem - requires closing the circle, by relating, in a meaningful way, ergonomics, monitoring problems, operational decisions and commonly encountered problems in pre-mastering. As a way to focus the research experimental activity, it would be optimal to know, as much as possible, what parameters should be looked at in the first place. The ideal way to do this is through a survey of pre-mastering engineers. Thus a questionnaire is developed to address the fourth research problem.
This knowledge allows a direct approach to the experimental phase. Thus the first research problem may be tackled. By experimental research, divided into two phases, it is possible to both quantify and qualify ergonomics-related parameters and associated problems. The first experiment objectively measures survey-identified parameters, through electro-acoustic analysis techniques. The second experiment relies on a listening panel to subjectively assess the measured parameters. To correlate these perceptions with specific operational decisions, a third experiment is conducted and recorded, in a short-observation setting where variables are manipulated and an operators' associated decisions and actions are logged. These experiments are carried out in three studios at SAE Byron Bay, chosen for their resemblance to common industry high-end studios, and for their inherent characteristics: the three studios employ large scale consoles, different sets of monitors and abundant outboard equipment.
HERE GOES DIAGRAM 1.
CAPTION: This diagram exemplifies the integrated nature of the research. Each technique's function is three-fold: to gather primary data, to help validate the previous step and to direct the focus of subsequent investigation efforts. The survey is the research's starting point. It drives the case study, by focusing the initial experimental investigation to specific parameters. Its other role is to act as a general framework from which meaningful conclusions can be drawn, with regards to the role of pre-mastering in correcting the product of the effects of ergonomics in the monitoring environment and consequent operational process. Both the survey and the case study employ qualitative and quantitative techniques. Particularly so in the case study, the triangulation of methods may be observed. -
Techniques overview
Survey
The choice of the survey methodology is deemed an effective way of relating the expected results from the experimental process to the hypothesised role of pre-mastering as a function of the observed factors. However, a static cause and effect relationship should not be seen here. The hierarchy, or order of methodologies, varies according to the perspective being chronological or methodological. By looking at the research methodology map in a chronological way, it can be seen how the survey instrument drives the case study. This is done by identifying factors which may potentially be interesting to focus in during the experimental stage. However, a methodological look at the same flow-chart would indicate precedence of the case study over the survey. That is, the experimental process and short term observation are the main data acquisition techniques, and the survey is used later in discussing these results.
Questionnaire
The choice of the survey methodology is deemed an effective way of relating the expected results from the experimental process to the hypothesised role of pre-mastering as a function of the observed factors. However, a static cause and effect relationship should not be seen here. The hierarchy, or order of methodologies, varies according to the perspective being chronological or methodological. By looking at the research methodology map in a chronological way, it can be seen how the survey instrument drives the case study. This is done by identifying factors which may potentially be interesting to focus in during the experimental stage. However, a methodological look at the same flow-chart would indicate precedence of the case study over the survey. That is, the experimental process and short term observation are the main data acquisition techniques, and the survey is used later in discussing these results.
Further discussion possibilities
Although questionnaires are a notoriously difficult way of establishing causes (Bell, 1999:14), they can also provide a first overview of qualitative data, by the introduction of a number of causal relationship questions. This allows for the contact of the respondent, and subsequent discussion, by e-mail, telephone or in person. Using this approach, in-face formal interviews are deemed irrelevant, especially when the nature of the survey instrument in the present context is emphasised: more than a primary data gathering method, it is meant as a general initial direction, that might help close any gaps later on, after the experimental process is concluded. In other words, "qualitative research facilitates quantitative research", by providing context (Blaxter et al., 2001:85).
Case study
Experimental process
The same authors also assert that "quantitative research facilitates qualitative research", by providing direction and focus (ibid). That perspective is implemented here.
Objective measurements
Because of the nature of its independent variables, the first experiment poses very specific problems: it would be impossible, in the context of the available logistics and time, to modify ergonomics-related factors manually. Thus, the only option left is the partial auralization or modelling of the experimental set-up. That is, the processing of the originally measured and recorded room responses in order to model what the system's characteristics would be like with the experimental variables. This is possible because an impulse response contains all measurable linear characteristics of a system. The removal of important early discrete reflections can be effectively accomplished through time-delay spectrometry. This technique was developed by the late Richard Heyser (1967) as a means of conducting tests requiring an anechoic environment in the diffuse field. While the system under test is injected with a linear sine-sweep signal, the receiving transducer is tuned to a specific sound path, through the use of a tracking filter. Developments to the technique have been introduced in later years by Niedrist (1992), Farina (2000), and Richert (2004), amongst others. Because the removal of one specific reflection, or set of reflections, differs fundamentally from the tuning to one sound path, the experimental design requires the use of a modified tracking filter. That is, instead of a sweeping band-pass filter tuned to the frequency of interest in maintaining, this new filter must be of a band-reject nature, in order to remove the unnecessary sound path, within which the unwanted early reflections travel. Another option is also tested: the manual removal, via a sound editor, of first reflections within the first 20 ms window. This technique might prove more effective in limiting the artefacts expected to result from the TDS technique.
The post-processing phase, required to generate the multiple experimental impulse responses, adds another constraint: while the measurement of frequency responses requires no more than one channel of omnidirectional measurement, the necessary spatial parameters (such as the inter-aural cross correlation coefficient) depend on a two-channel capture, in some cases, through a binaural recording dummy head. Giusto (2001), Tarabusi (2004) and Farina (2005) have presented research on the use of binaural sound systems in the characterization of spatial attributes - for auditoriums mainly, although nothing is said against its use in control room design. Other authors have extensively researched on the limitations of non-individualized binaural recordings, but Farina asserts that these "are very convenient, in terms of being easy to obtain through room acoustical measurement and computer simulation" (2005:1).
As seen, the requirements of the experiment demand two sets of fundamentally different techniques: omnidirectional single-channel and binaural capture of impulse responses (IR). Because of the unavailability of a binaural recording dummy head, the results of previous research on head related transfer functions (Gardner & Martin, 1994) are used. By convolving this set of binaural impulse responses of a KEMAR binaural recording dummy head recorded at M.I.T.s Media Lab in 1994 with each experimental IR, one obtains an effectively binaural experimental impulse response. Although a compromise, another option is the use of a human test subject, with the two omni-directional microphones placed directly outside the ear canal. This should allow the satisfactory encoding of the experimental impulse responses by the subject's torso, head and pinna. Both options are implemented in this study for comparison, although Farina recommends the later (pers. comm., 01-09 2005).
Subjective assessments
The second experiment, intended to subjectively assess previously measured acoustic parameters, branches into two different types of evaluation: accuracy and preference. Preference has traditionally been the main consideration in subjective assessment of sound. However, as Klepper recently pointed out, accuracy is subjectively tested each time a new pair of loudspeakers, for example, is installed in a studio control room (2004:1060). This certainly applies to the type of testing required in this study. Indeed, in comparing both the control-reference and variable-experimental models of the three control rooms, the first factor to be considered must be the accuracy, albeit in a different form. The evaluation of both models necessarily starts at identifying whether there are any perceivable differences at all - a form of accuracy testing. As such, this first test will employ an ABX double-blind comparator procedure (Clark, 1982).
Assuming a statistically valid accuracy rating by the listening panel in identifying perceivable differences in the control and experimental impulse response-processed test materials, a preference evaluation test is then conducted. This listening session is directed according to EBU Tech. 3286 and AES20-1996. These standards are chosen for their adequacy to the purposes of this test. Both are commonly used in subjective evaluation by both academics and industry professionals, providing guarantees of reliability and repeatability, and hence validity. The specific terminology proposed in EBU 3286 was also employed in the questionnaire presented to mastering engineers, so it stands to reason that the same nomenclature should be used in this setting. The EBU document, in particular, represents the corollary of research started by Lipshitz and Vanderkooy (1981), where the necessity of objectively implemented subjective assessment tests was first raised. Later on, Bech developed methodologies for selecting and training subjects for listening tests (1992). He successfully divided listening expertise into hearing threshold-related factors and previous experience or training. This study lead to the inclusion of on-site methods for training subjects in the EBU standard. Spikofski (2000) and Walker (1997) discussed in detail the means and methods within the EBU's subjective evaluation standards: from those, it is evident that both the standard and the associated test materials (which are obviously employed in this study) are directed at the evaluation of acoustic music mainly. This might be thought of as counter-productive in subjectively evaluating a monitoring environment mainly designed for "pop-rock" styles of music, as the objects of the case study are. However, the use of classical, chamber and other types of acoustic music is the most appropriate one in assessing the particular acoustical parameters in question, such as the spatial-related ones. Indeed, and particularly when confronted with discrete early reflections, if a monitoring environment cannot cope with acoustical, progressive music, that perception should only increase when in presence of the typically more percussive modern styles.
Short-term observation
The first and second research problems, those of establishing a direct correlation between specific ergonomics factors of the control room and the given acoustic parameters, and between the later and the associated perceptions of a listening panel, were first addressed. Assuming the results are as expected, the aim is now to identify, in the work flow of an operator working with the measured conditions, a range of decisions that can be meaningfully related to the the expected margin of error in the monitoring environment that the hypothesis suggests. Short term observation is the primary method for data acquisition. This means that, although a quasi-experimental procedure is implemented and the results recorded, this data is qualified by means of observation and interview of an operator. "Today, it is generally acceptable to study groups for less than six months, provided that the researcher triangulates the research methods" (Colorado State University, 1997). The operational setting to be observed is not a typical one, within the social sciences realm (namely those used in ethnographic studies). However, short term observation can prove to be effective in understanding how the identified problems translate into the normal work flow and decision-making process by the audio engineer. Particularly so in this situation, the triangulation of methods applies, as observational data is cross-referenced with survey and experimental data. It is thereby possible to correlate specific operating errors induced by a flawed monitoring environment with the role of pre-mastering in addressing these issues.
-
-
Data Compilation and Analysis
-
Survey
Questionnaire
Aims
The questionnaire was designed with the objectives of: gathering information about common problems encountered by mastering engineers in their work; quantifying the difference between approaches to mastering (creative input, technical correction, impartiality, etcetera); and highlighting where the main problems could be related to recording-specific ergonomics. That information is then used as a basis for more focused investigation. The output report tabulates typical problems related to incoming source material; identifies the most common requests made by clients of mastering studios; and qualifies the relevance of several identified problems to the research.
HERE GOES DIAGRAM 3
Design
The questionnaire was divided into the following topical sequence: a background control section, where the main purpose was to help validate the respondent. This section gathered the type of programme material the respondent most commonly mastered, and the scope and scale of their work and of their client's. This contextualization allowed a more precise interpretation of the results. Another section dealt with common problems that were both subjectively and objectively dealt with by the respondents. The terminology employed here complied with the subjective assessment standard EBU Tech. 3286. In addition, respondents were asked to provide their opinion on the most likely causes for those problems. This section aimed at focusing and contextualizing subsequent investigation. A single question was dedicated to quantifying the reasons that lead their clients to request their services. This was intended to test the researcher's assertion that most requests are in fact for technical correction, not creative input. Lastly, the respondents were inquired as to whether they would permit their opinions to be used non-anonymously. This served two purposes: the declared one, for further discussion possibilities; and a second one, to gather contact details and personal information, which would simplify validation of the individual questionnaires.
Sampling
The questionnaire was run over ten weeks, from June 19 to August 25, 2005. The target statistical population consisted of pre-mastering engineers, without any further restrictions. It was conducted on-line, through the Mastering WebBoard. Because of the specialization of the target, it was deemed sufficient to have at least 20 questionnaires filled out, assuming a successful identification rate of 50%. Non-probability sampling was used. The reason for this was three-fold: lack of sufficient statistical data to allow a fair estimation of the total population, and thus, impossibility of generalization; impossibility of contacting the full population, which leads to high sampling error; and high level of specialization of the target population, which makes judgemental (or targeted) sampling an adequate technique, if not scientifically sound. Thus, the survey may be characterized by the following: a non-probability sampling methodology (due to lack of existing data on the existing population); convenience and judgemental sampling techniques (product of the impossibility of contacting the full population and of the benefits of this technique in characterizing a highly specialized target, respectively). The sampling frame consisted of registered users of the "Mastering WebBoard", with a public valid e-mail which included the word "mastering". This was subjected to further analysis, lest the word not be used with the intended meaning. To that end, the forum's built-in filters were used, with the end result of 105 potential respondents. These were contacted via e-mail, out of which 20 did not go through. This sampling frame, it was felt, was representative of the defined statistical population, while simplifying the contact process.
Results
HERE GOES DIAGRAM 4
CAPTION: A number of factors influence the sampling methodology. The major one is the non-existence of reliable published statistical data on pre-mastering studios. This renders any valid sampling error calculations impossible. Because a known sampling error is essential to probability sampling, the technique may not be applied in these circumstances. Another factor is the purpose of the questionnaire. It does not seek to enable the statistical generalization of the results, but rather to act as a driving source of information to the case study. In that sense, it is almost informal, and therefore suitable for a judgemental type of sampling. This leads to the last factor, the specificity of the target population - pre-mastering engineers - which requires that very specific non-probability method.From 105 potential respondents, exactly 20 successfully completed the questionnaire, with another 20 answers deemed invalid for various reasons - a success rate of 24%. This is within the initial objective. and given that the identification rate was 60% - more than half of the respondents identified themselves - The outcome is positive.
Background
HERE GOES TABLE 1
Table 1 lists the results from the background control section. An overview of these indicates a slight bias in the respondents towards the higher end of the scale of mastering studios. One can see how the highest source of programme material mastered are small recording studios, which is concordant with expectations, as that type of facility is the most commonly represented in the industry. All respondents master music for independent labels, and only one-quarter regularly deal with major labels. This seems to be in agreement with their scale of operations. Finally, as expected, "pop-rock" is the most common type of music mastered, which, again, does not contradict the expectations. The acoustic genres "classical", "jazz" and "acoustic" make up 60% of all mastered music. This is an important number, as discussed further ahead.
Common requests
HERE GOES TABLE 2
Table 2 shows that generally, clients resort to mastering for technical reasons. One may assume that they find that something could be technically improved in the programme material. Creative input and critique make out 40% of the requests, which could have been expected. Contrary to popular belief, "loudness" is not a common request, or at least, not a direct one.
Problems most regularly found
HERE GOES ILLUSTRATION 1
"Unbalanced frequency response" was the most common problem, with 85% of the respondents rating it negatively (either "bad", "poor" or "fair" - "fair", as defined by the EBU, means that the programme material presents "a number of annoying defects"). "Hyper-compression and loudness level differences", "spatial impression" and "stereo impression" all come second, with 75% and 70% of respondents rating it negatively. Problems such as noise and distortion, lack of transparency, sound balance and timbre divided the classifications, with approximately half of the respondents opposing the other half.
Discussion
Within the given results, the reliability of the questionnaire is considered appropriate. The nature of the respondents allow for a fair confidence in their technical ability and knowledge, and attitude towards the survey. This means that, although the return numbers may seem a bit low, with more than half of the questionnaires properly identified, with name, contact and work place, the confidence margin is much higher, when compared to a traditional unidentified survey of a random sample. Indeed, an analysis of the background of the respondents reveals that, because most work in a high-end facility or at least in a small commercial studio, the general level of technical competence, ability and knowledge is high enough to give reliability to the results.
The results are also valid and appropriate, because they fit adequately in the research. For example, most respondents mainly produce work coming from small or high-end commercial recording studios. The case study within the research is also concerned with commercial recording studios, and disregards the lower extremes of the production spectrum. Work coming from major-labels is more likely to be subjected to variables other than control room ergonomics, such as executive pressure. Most respondents do not work in niches that would render their responses invalid. Finally, the results regarding styles of music mastered seem to imitate a "real world" pattern.
As to common requests made by clients, the learning agreement's rationale that most clients resort to mastering for purely technical reasons - which could be dealt with in a proper environment - is confirmed. Again looking at the learning agreement, the main control room ergonomics factors identified seem to echo these results. An unbalanced frequency response is the natural result of comb-filtering produced by first reflections (coming from large consoles, maximized by the use of near-field monitors over the meter bridge, the use of multiple computer displays in the monitoring path, and the presence of large communication windows). Loudness and hyper-compression problems can be shown to be maximized by the generalized use of near-field monitoring, which is typically worsened by low power outputs. Finally, spatial and stereo impression problems seem to be the obvious result of the meddling with of the reflection free zone by control room ergonomics.
Qualifying process
Aims and methods
As the questionnaire was activated and publicized through the Mastering WebBoard, an informal discussion took place. While a formal interviewing process with mastering engineers was initially planned, the direction these discussions took made that unnecessary. By combining this process with the qualitative section of the questionnaire, the researcher was able to gather enough qualitative data, albeit in an informal setting, that significantly added to the interpretation of the results of the survey. Taking into consideration the four underlined problems in illustration 1, the opinions of the respondents on their perceptions are presented and discussed next. A full transcript is included in appendix E.
Observations and results
Perceived causes for imbalances in frequency response
HERE GOES ILLUSTRATION 2
In an attempt to quantify what the surveyed pre-mastering engineers perceive to be the major causes for problems related to frequency response in incoming programme material, all responses are divided into three main categories: monitoring environment, inexperience and complexity problems. Monitoring problems are ostensibly the main cause for imbalances in the spectral content of incoming mixed down material. Of course, "monitoring environment" includes very distinct factors. Some of the raised issues are:
- Modal acoustics factors: related to the control room's outer shell, these mainly effect the low frequency content. These are most likely brought about by poor loudspeaker positioning, modal coupling, incorrect room dimensions and inappropriate or non-existent acoustic treatment.
- Specular acoustics factors: these are a function of the inner shell of the control room. That is, the characteristics that are dependent on the room's internal configuration.
- Loudspeaker inadequacy: the use of near-field loudspeakers is pointed out as detrimental to the monitoring environment, namely with respect to the low frequencies, but also to the spatial parameters.
Modal acoustics factors are out of the scope of this investigation, as those are hardly connected to control room ergonomics. However, both specular acoustics and near-field monitoring-related problems may well be functions of the factors under study. It is certainly the researcher's expectation that they are, and as such, both will be examined in the experimental process, detailed in 4.2.
Loudness and compression problems
HERE GOES ILLUSTRATION 3
Another category is added: compression settings, which can be related to inexperience, but also to inadequate equipment, hence its own label. The monitoring environment is under-represented here as a cause for problems. Therefore it is unlikely that the hypothesised ergonomics factors - which mainly have an impact on the monitoring environment - have a part in this problem. Although one of the main issues raised in the survey, it falls outside the scope of this study, and is best left for further research.
Spatial impression problems
HERE GOES ILLUSTRATION 4
The four categories for causes of spatial impression impairments are: the monitoring environment, inadequate equipment, loudspeaker type or positioning and inexperience. Most mastering engineers considered inexperience to be the main cause for problems with spatial impression. To clarify, spatial impression is the perception of the original acoustic space in which the sound was recorded. Needless to say, this parameter usually applies to acoustic music, as defined by the European Broadcasting Union (1997:13). Being a generic category, inexperience comprises a number of factors, such as:
- Inadequate artificial reverberation settings
- Inadequate use of stereo "enhancer" processors
- Misunderstanding of "panning" and frequency manipulation to achieve "depth" and "spaciousness"
The above are procedural errors that can be attributed to inexperience. For the purpose of this study, these errors must be assumed as non-existent, because they constitute uncontrollable variables. By assuming perfect operational competency, one may focus on ergonomics factors alone as a way of improving sonic accuracy. This approach leaves "monitoring environment" as the main cause for spatial impression problems. By adding "loudspeaker type and positioning" to the monitoring environment, which is reasonable, these constitute 70% of perceived causes. That is significant, and must be examined closely. Following are specific remarks that fall into monitoring environment problems.
- Exclusive use of near-fields during production
- Use of close microphone techniques to compensate for poor acoustics
- Improper placement of monitors
Phase problems provoked by close-miking techniques are outside the scope, but type and placement of monitors, as well as more general specular acoustics considerations are well within the expectations, and are therefore examined in 4.2. Finally, inadequate equipment is also disregarded as a factor under study, because it is not a relevant variable related to ergonomics factors. As inexperience, it must be assumed that proper equipment is available and in use.
Stereo impression problems
HERE GOES ILLUSTRATION 5
Although standardized by the European Broadcasting Union, the terms spatial and stereo impression appear to have been conflated in the minds of some of the pre-mastering engineers surveyed. This rendered a number of responses invalid, which were therefore not used. Most responses to this question are quite similar to the ones regarding spatial impression. The main difference is the assertion that inexperience has less to do with stereo impression (how well the stereophonic sound stage is represented) than with spatial impression impairment causes. The researcher's expectations were in the direction of an increase of the loudspeaker type and positioning as a major source of problems. The contradictory results may have been provoked by the mentioned conflation of concepts. This would have been solved by a clear distinction between both terms in the questionnaire, and should undoubtedly be implemented in subsequent efforts.
In any case, the combination of monitoring environment and loudspeaker type and positioning still accounts for 82% of the perceived causes for stereo impression problems. These generic categories are comprised of the following:
- Phase anomalies caused by close-miking
- "No perception of reflections"
- "Absence of sound controls" (absorption and diffusion)
- Loudspeaker positioning atop consoles; spread too far or too close apart; too close to reflecting surfaces
- Lack of integration between monitoring devices and room
It appears that most of the hypothesised ergonomics factors have a close relationship to problems appearing in the stereo impression of programme material. Consequently, these will be looked at in detail in 4.2. as well.
Discussion
An overview of the observations extracted from the questionnaire's qualitative section points the experimental process in the direction of the following factors (which have in common not only being considered, by the surveyed pre-mastering engineers, as main causes for problems with frequency response, spatial and stereo impressions, but also can be directly related to control room ergonomics):
- Specular acoustics: specifically, effect of large reflecting surfaces in the sound path to the monitoring environment
- Loudspeaker considerations: Use of near-field monitors; positioning and type of monitors
From this surveying process, the experimental variables may be deduced. Therefore, the independent variables to be tested are: 1) first-order discrete reflections provoked by large reflecting surfaces; 2) inappropriate stereophonic sound stages, and other related issues, created by use near-field monitoring. The way these are manipulated is described in detail in 4.2.1.b.i. The dependent variables, that is, the parameters under test, are: 1) the relevant acoustic ones; 2) those relating to perceived imbalances in frequency response, spatial impression and stereo impression problems. These are explained in 4.2.1.c and 4.2.2.c, respectively.
-
Experimental Process
Objective measurement of acoustical parameters
HERE GOES DIAGRAM 5
Aims
The first experiment aims at establishing a measurable and significant relationship between the chosen factors related to ergonomics of the control room (the independent variables) and the relevant acoustical parameters (the dependent variables). The quantification of these parameters allows a subsequent contextualization of the effects of a given ergonomics factor in the perceptual evaluation of the monitoring environment, in the second experiment. It is expected that a pattern will emerge, commonly to all three studios, in the statistical analysis of the results, where significant first reflections are present within the 20 ms window, due to large reflecting surfaces. It is also expected that these elements will be a strong factor in the calculation of a number of acoustical parameters, namely those concerned with spatial characteristics, but also in the magnitude responses. Another point under investigation is the influence of differing sets of loudspeakers in the calculation of the inter aural cross correlation coefficient. The prediction for these is that there will be an important range variation of cross-correlation coefficients when switching monitoring systems. The effect of this variance on the subjective assessment of sound is later examined.
Experimental procedure
Independent and dependent variables
Following the flow-chart of the experimental procedure for the objective measurement of the necessary acoustic parameters, the independent variables are: large rigid and reflecting surfaces in the sound path generated early reflections, within the first 20 ms window (Walker refers to this factor as "desktop diffraction and reflection") and use of differing sets of monitors (near-field and far-fields). The impact of these on the following dependent variables will be examined: the initial time-delay gap (ITDG), the inter aural cross correlation coefficient (IACC), and the frequency response. The main analysis tools are the energy time curves (ETC), the energy-time-frequency plot (ETF), and the transfer and coherence functions (magnitude and correlation, respectively).
System under test and on-site procedure
The systems under test are, in each case, the monitoring environment within the control room, at the usual reference listening position for an operator. For each variable that is controllable on-site, the testing procedure is as follows: two small diaphragm omnidirectional measurement microphones (Behringer ECM8000) are set up at the reference listening position, separated 16,5 cm apart, at approximately 1,2 m from the floor and meter bridge. These are connected to the measurement system (a small Behringer UB1002 mixer and Digigram VxPocket sound card, which is plugged into a Compaq EVO N800C laptop). The absolute sound pressure level is then calibrated, and the noise level is recorded in 1/3 octave bands. A slow sweep-sine wave is input into the system under test, and the output recorded through the microphones. The same procedure is repeated, but with a human test subject positioned between microphones.
Post-processing
In the post-processing phase that follows, the impulse response for each measurement is obtained, by deconvolving the recorded output by the input applied to the system. The original IR is then stored, and a copy further processed. This copy should be processed with the modified tracking filter mentioned earlier in order to remove first reflections within the 20 ms window, but that technique proves to introduce severe artefacts which limit the usefulness of the measurement. Therefore, a simple editing task is conducted, and all early reflections contents removed, thereby creating an effective 20 ms initial time-delay gap. This is done by applying gain changes of 10 dB RMS discretely to the intended reflections, while referring to the ETC and ETF plots for accuracy. This ITDG-modified non-binaural IR should also be convolved with the appropriate head related transfer functions, and the results analysed in comparison with the binaural impulse responses recorded on-site with the human test subject. However, as Farina pointed out, the use of a human test subject to produce head related transfer functions is appropriate and "creates very good binaural recordings (2005, personal communication). Thus, the on-site recorded binaural impulse responses are used, and the relevant acoustic parameters calculated for both the experimental and the original control impulse responses.
Observations and results
The most effective way of analysing first order discrete reflections is by means of energy-time curves (ETC) and energy-time-frequency (ETF) waterfall plottings (see sample plots next page), provided that adequate frequency and, most importantly, time resolution are obtained (Walker, 1995/3:5). Because desktop reflections can typically be found 0.8 to 1.2 ms after the direct sound, this means that, at the least, time resolution should be close to 1 ms. This also allows the identification of other rigid surfaces' reflections, namely, computer displays (1-2 ms), ceiling (3 ms), side walls (7-8 ms) and rear wall (15-20 ms) (Walker, 1995/4:26). The ETC and ETF figures in appendix A were plotted within 30 ms and 20 dB ranges. The time resolution is 666.7 µs. The windowing function is half-Hanning, as recommended by Walker (ibid). The FFT size is 1.3 ms, which allows for a 750 Hz frequency resolution. This is adequate, because the lowest frequency that can be reflected by the mixing consoles under study is approximately 286 Hz (the consoles are generally 1.2 m in width across the monitoring path. Therefore, the lowest wavelength potentially reflected is 286 Hz. Frequencies below this value tend to be diffracted by the surface). Since the mixing console is the largest factor under study, no frequencies below that value need to be evaluated.
HERE GOES SAMPLE 1 AND SAMPLE 2
Desktop reflections
Both EBU 3276 and research on the audibility of reflections1 indicate that first reflections during the first 20 ms should have a relative level to the direct sound of at least -20 dB. Figures 1 and 2 (see appendix A) show the time response for the near-field monitors in room 16's SSL G series studio. Important reflections at ~1, 1.5 and 5 ms are evident. Their levels are well into the -20 dB threshold (~-12, -18, -16 dB). The same can be observed in figures 9,10 and 17, 18, which are the near-field control impulse responses for room 3' SSL K studio and Studios 301, respectively. Because of the frequency-averaging effect of the ETC plot, these reflections are not as evident as in the waterfall graphic. When analysing them under this light, one can notice their importance, particularly because they are found between 1 to 16 kHz, and are therefore easily detectable by the human hear. For the modified or experimental impulse responses, these reflections were reduced in gain in a sound editor, and the results are plotted in figures 3 and 4 (SSL G), 11 and 12 (SSL K) and 19 and 20 (301). This was done after identifying the major offenders in the ETC and ETF graphics. A posteriori, no major reflections are found within the 20 ms window, although the removal process is somewhat imprecise, resulting in some artefacts in the time domain graphics. The process is repeated with the far-field set of loudspeakers, with the same results: it is possible, by means of the above described method, to improve the early discrete reflections patterns produced mainly by the desktop. The relationship between differing sets of loudspeakers is analysed further ahead. The frequency responses and spatial measurements for each case are observed next.
Frequency magnitude responses
The presence of severe destructive and constructive interference in the original near-field impulse responses is obvious, in the form of comb-filtering. This is true for all three studios, as may be observed in the blue traces of figures 25 and 26, 29 and 30, and 33 and 34 (see appendix B). The effect is significantly limited in the far-field responses of figures 27 and 30, 31 and 32, and 35 and 36. This is most likely due to the larger distance between far-field monitors and the desktop or console, which has the effect of shifting the respective reflections to larger delays. Consequently, any phase interferences happen at higher wavelengths, and therefore at lower frequencies. This is not measurable here as the minimum frequency resolution is 750 Hz (and, in any case, only the 1 to 20 kHz region is plotted).HERE GOES SAMPLE 3
The modified responses, represented by red traces, are effective in limiting this severe comb filtering, to an extent: particularly in the left-channel near-field responses of all three studios, this is visible. There is a noticeable improvement in the frequency response of the three control rooms (figures 25, 29, 33) However, the respective right-channel effects are not as immediately visible, or are shifted towards non-plotted lower frequencies. A number of factors may have contributed to this: small positioning errors in the measurements set-up; asymmetrical desktop configuration; or, alternatively, deficiencies in the removal of reflections. This is the most likely scenario, as strict measures were taken to ensure near-perfect positioning of the measurement microphones, and no significant differences in desktop asymmetry were found. The fallible nature of the reflection-removal technique is then left as the probable cause for this observation.
Spatial measurements: Inter Aural Cross Correlation coefficient (IACCearly)
Ahnert and Schmidt (2005:36) offer a comprehensive overview of the Apparent Source Width (AWS) subjective perception parameter. Quoting Beranek, they propose the inter aural cross correlation coefficient as an appropriate measure of perceived source width. This parameter is appropriate in ascertaining the influence of first order reflections, or lack of, in the perception of width of programme material, thereby allowing a direct correlation to the results of the questionnaire (namely, those related to the spatial and stereo impression questions). However, because it was initially developed in an effort to characterize concert hall acoustics, its proposed values for different quality categories of auditoriums do not fully apply here. Instead, the coefficient values are used as a simple measure of relationships between different variables. The binaural impulse responses recorded are used here because, as Ahnert and Schimdt assert, "with binaural measurements [...] it is possible to calculate correlation measures" (2005:36). These were subjected to the same process of reflection removal as the omnidirectional ones.HERE GOES SAMPLE 4
After a first glance at figures 37 through 42 (see appendix C), it would appear that the influence of first reflections upon the apparent source width is negligible. A closer look, however, yields the following observations: the sharp notch at approximately 1 kHz is common to all binaural responses, and thought to be part of the head related transfer function. Therefore, it can be safely ignored. Another common pattern to all responses, is the subtle change in coherence of low frequencies, which decreases in the modified responses (red trace). In parallel, frequencies above 1 kHz are rendered more coherent when first reflections are removed. That is to say, the presence of specular rays within the first 20 ms provokes differing levels of comb filtering in both transducers, caused by the distance between them. Generally, the modified impulse responses show more consistent coefficient values, which in turn means that the perceived stereo and spatial impressions do not change as much across the frequency spectrum. It must be said, however, that the subtlety of the described effects may render the interpretation inconclusive, as they are well within a potential instrumental margin of error.
Loudspeaker positioning
The second major variable is loudspeaker positioning. The research is particularly concerned with the widespread use of near-field monitoring, as that was a common factor raised in the questionnaire by pre-mastering engineers as a source of stereo and spatial problems, as well as frequency response imbalances. The frequency responses and IACC values for each control room are presented next. This section differs from the previous one in that the emphasis is now on the differences between near and far-field monitoring responses. As such, the relevant figures should be cross-referenced towards that end.
Frequency magnitude responses
The SSL G studio original responses can be found in appendix B. By comparing figures 25 and 26 with 27 and 28, it becomes readily evident that any comb filtering effects are much more present in the near-field responses. The same is true for figures 29 and 30 when compared with 31 and 32 (the responses for room 3's SSL K studio), and figures 33 and 34 when seen against the far-field responses of figures 35 and 36 (Studios 301's responses). As explained earlier, this is to be expected, as the larger the distance from the desktop, the lower the affected frequencies will be. Additionally, the removal of first order reflections seems to affect mainly the near-responses, while the far-field monitors appear almost unchanged. Far-field monitoring is, therefore, shown to be less prone to the types of frequency imbalance problems that desktop reflections tend to inflict on the monitoring environment.Spatial measurements - Inter Aural Cross Correlation coefficient (IACCearly)
The most striking feature of the near-field IACC responses when cross-referenced to the far-field equivalents is their higher homogeneity across the frequency spectrum: with the exception of the first case, the SSL G control room, there is less variation in coherence. The near-field IACC responses for the three studios (figures 37, 39 and 41 - see appendix C), show a decreasing coherence (that is, higher perceived source width) above 4 kHz. Contrastingly, the equivalent far-field parameters (figures 38, 40 and 42) have the opposite tendency, increasing above 4 kHz. This is generally true for both the original and the modified responses. It is interesting to note that, at 4 kHz, the associated wavelength is approximately 8.5 cm, which is half of the distance between the transducers used. However, this relationship can not be explained in the given context. What can be asserted is that, it should be expected, a listening test panel will find the far-field responses to give a more coherent, and therefore less wide, stereo impression and sense of involvement.Discussion
Two independent variables were under experiment: desktop or mixing console reflections and positioning of loudspeakers. For each, the dependent variables were analysed: frequency response and inter aural cross correlation coefficient. These directly relate to the results of the questionnaire: the reported frequency imbalances are highlighted through frequency magnitude plottings; and the IACC presents a well documented descriptor of stereo and spatial impressions.
HERE GOES DIAGRAM 6
CAPTION: A "variable matrix" structure was used in the analysis. That is, the effect of each of the independent variables was tested for each of the dependent ones, with four resulting analysis sections. The success, and therefore validity, of the experiment, as described in the aims section, depends on reliably demonstrating each of the four relationships to be significant. That is to say, it should not be inconclusive, due to instrumental errors or inadequate measurement and analysis techniques. The use of a control set-up eliminates most instrumental errors. That leaves instrument resolution, which was appropriate, according to Walker's recommendations. The techniques were also adequate, with perhaps the exception of the reflections removal process. This is undoubtedly the weakest link in the experiment. However, again due to the use of a control and an experimental set-up, the results are most likely valid.The presence of significant reflections in the measured impulse responses was demonstrated through ETC and ETF plots. Generally, important delays were found in the 0.8-1.5 ms range (desktop, computer displays, mixing consoles); 3-5 ms (ceiling and side walls); and 20 ms (rear walls). The effect of reflections on frequency response was ascertained through the observation of severe comb-filtering in the high frequencies of the control IR, but significant smoothing of this effect in the experimental IR. This applied mainly to the left channels, which may have been due to a number of factors not directly observable. In any case, even this oddity is common amongst all three studios, which enables the establishment of a pattern of improvements in comb filtering when first reflections are removed. This comb filtering is shifted towards lower frequencies in the far-field responses, due to the larger distances involved. This, too, is common to all three sets of IRs. Thus, the use of near-field monitoring is more prone to comb filtering problems related to first order reflections. The inter aural cross correlation coefficient was analysed as a function of desktop reflections. The removal of these tends to decrease the apparent source width for frequencies below 1 kHz, and increase it above that point. Additionally, the experimental IRs show more homogeneous IACC values across the audiometric spectrum, although this is a subtle effect that may well be within the margin of error. As it is also common to all three cases, one could expect that the sense of stereo impression and spatial involvement will not change as much in differing frequencies. The final element of the matrix, the effect of positioning of monitors in the IACC values, is also interpreted, and shows seemingly contradicting characteristics: near-field monitoring is generally more homogeneous, in regards to the stereo and spatial impressions. However, far-field monitoring is slightly more coherent, in the sense that the apparent source width is smaller. This cannot be explained with the available data, but may be clarified with subsequent experimentation.
In conclusion, the results point towards the hypothesised premise that certain ergonomics-related factors are, in fact, responsible for artefacts in the frequency response and stereo imaging characteristics of modern control rooms, as well as the sense of spatial envelopment. This influence is next quantified into the perceptions of a listening panel.
Subjective assessment of the acoustic parameters measured in 4.2.1
HERE GOES DIAGRAM 7
Aims
The second experiment's function is to qualify the independent variables objectively measured (that is, each control room's calculated acoustic parameters). Furthermore, it purports to establish a correlation with data from the survey. By using the same terminology in the subjective assessments experiment as the one used in the questionnaire, it should be possible to definitely establish the hypothesised relationships. Because this perceptual evaluation testing aims to identify differences between differing monitoring environments, and one of these is artificially modelled, the tests must be conducted through headphones, using convoluted test material. Although the objective quality of the impulse responses, and consequently the test material, is not very high, because the objective is to compare responses to a reference, the essential information may still be extractable from the experiment.
Experimental procedure
Independent and dependent variables
Following the flow-chart of the experimental procedure for the subjective assessment of the identified acoustic parameters, the independent variables are: frequency response and inter aural cross correlation coefficients for each of the ergonomics-related factors under study (console reflections and monitor positioning). The perception of these is evaluated, by means of a listening panel, in the form of the following parameters: timbre (frequency response imbalances); stereo impression; and spatial impression.
Test material and post-processing
The test material used in this assessment is extracted from the EBU "Sound Quality Assessment Material - Recordings for Test Subjects" CD (1988b). In accordance to the "Suggested applications" chapter, (EBU, 1988a p.13), seven different tracks are chosen. Tracks 61-66 ("vocal & orchestra") are used in evaluating frequency response problems. Tracks 63 and 69 ("pop") are used for "stereophonic image" assessment. Because the SQAM CD only recommends two tracks for stereo imaging assessment, four other samples are used: track 54 ("Section on Stereo-Imaging") of the "Perceptual Audio Coders: What to Listen For" appendix CD (AES & Erne, 2001); another three samples are taken from the EBU CD "Parameters for the Subjective Evaluation of the Quality of Sound programme material - Music" (1997b): tracks 11 ("Acoustical balance"), 20 ("Apparent room size") and 36 ("Sound image width"). These test samples are then convolved with the binaural impulse responses for each of the independent variables for all three studios. Finally, they are level matched for loudness-independent accuracy in the assessment.
Subject selection and on-site procedure
Prior to the actual testing procedure, subject selection takes place. The number of subjects to be used is mainly dependent on previous training. Bech as shown "the number of subjects required [...] can be reduced up to a factor of 7 if highly trained subjects are used" (1992:604). EBU Tech. 3286 (p.5) defines seven as the maximum number of test subjects that may be seated at the same time. It also characterizes the appropriate listening panel as one "composed of listening experts, that is people who understand and have been trained in the agreed methods of subject quality evaluation" (p.4). According to AES20-1996, the best way to determine the expertise of a listening panel is by evaluation of "self-consistency in blind listening tests" (p.10). These guidelines point towards a listening panel of seven "experts", comprised of second year degree students of recording arts at SAE Byron Bay. Incidentally, the first experiment to be carried out is the ABX double-blind evaluation of the panel's accuracy in determining if there are any differences between variables. This experiment enables the assessment of the listeners' expertise, to an extent. The tests are conducted individually, because they require the use of headphones and an individual computer display and input-output devices, for the ABX test. The use of headphones is accommodated by EBU Tech. 3276, as long as its operational frequency response is in accordance with the specifications, which is the case. Therefore, the only requirement for this test is a space with a noise rating of at least NC 15, which is attainable by the room used.
A short ten minute training session is conducted, where the PEQS CD is used in demonstrating the different parameters under evaluation. Extreme and normal samples are played, and the evaluation scale demonstrated by a "dummy" assessment session. This is followed by a brief oral explanation of the test, also given to the subject in a hard copy, with a description of the parameters under assessment. The first experiment is the ABX test, documented in the literature. This test takes approximately five minutes and evaluates the significance of the following variable relationships for both classical and pop music: 1) control near-field v. experimental near-field; 2) control near-field v. experimental far-field; 3) control near-field v. control far-field; 4) experimental near-field v. experimental far-field.
Secondly the preference test is conducted, in accordance to EBU Tech. 3286. Each sample is assessed for each variable. The results are given in a six point subjective impairments rank scale. There are 12 items under assessment, which correspond to the control and experimental impulse responses, for both near and far-field monitoring systems, for each studio. During the actual procedure, the test subject is given 12 playlists, each corresponding to an item and having 11 samples. The subject may switch between any of these tracks any number of times during a maximum period of three minutes per item. 30 seconds are then provided for recording the evaluation of the required parameters on the evaluation form (appendix E). This is repeated for all 12 items, and takes a maximum of 45 minutes. In total, each subject needs a maximum of one hour to complete the test. Together with the initial interview, necessary explanations, short training and questions and answers, this amounts to approximately 90 to 120 minutes.
Observations and results
Illustration 6 presents the findings of the ABX double-blind test. Two generic music genres were surveyed: "classical" and "pop". The tracks were selected recordings from the EBU SQAM CD.1 Each subject evaluated differences in 12 relationships: control versus modified near-field and far-field responses, for each studio, for both types of music. The median values are used in the analysis, as proposed in the standard (EBU Tech. 3286:29).
HERE GOES ILLUSTRATION 6
The G series studio impulse response-processed material presented no variation within different programme material. The other studios showed otherwise, with a tendency for "pop" music to be more easily affected by the different impulse responses. Again with the exception of the G series studio, all other far-field responses were more susceptible to identification of differences than their near-field counter-parts.1
HERE GOES ILLUSTRATION 5
CAPTION: In the second listening test, the panel clearly preferred room 3, the SSL K studio. This was true for most of the parameters, as can be seen in illustration 7. Only the median values are presented. Hoeg et al. (1997:48) recommend the use of a radar or net diagram for generic quality profile analysis, as it provides "a good overview of the quality [...] The bigger and less unbroken the diagram is, the better the quality".Illustrations 7 through 18 (see appendix D) detail various comparisons of interest. Specifically, the following relationships are shown: control versus modified near-field responses; control versus modified far-field responses; and near-field versus far-field responses, for all three studios. The SSL G studio presents unexpected results, as it had already in the ABX test. While both far-field responses show improvements over the near-field ones, the modified responses do not benefit the quality profile in almost any aspect. The SSL K studio, on the other hand, has slightly improved modified responses, but again, the near to far-field differences are more significant. This tendency continues throughout Studios 301, both in the control versus modified and near versus far-field responses. Taking a closer look at the individual parameters of interest, the apparent benefits in the modified response take place in the far-field monitoring of SSL G, where there is an increase in the general main impression (illustration 8); also, the main impression of near-field monitoring in the SSL K studio is benefited, as well as the stereo impression in the far-field monitoring (illustrations 11 and 12); and finally, the transparency of the near-field response and spatial impression of the far-field response in studios 301 (illustrations 15 and 16). In parallel, most far-field responses, control or modified, show significant improvements in various parameters.
Discussion
If the first experiment relates ergonomics factors to specific acoustic parameters, the second one directly connects these to common problems found by pre-mastering engineers in source programme material. Thus, the all important relationship between ergonomics factors and the role of mastering in correcting monitoring-related problems may be asserted, or not, with sufficient scientific accuracy. Therefore, the question to be asked is: does this experiment qualify the objective data of point 4.2.1 in a meaningful way, towards an acceptance of the hypothesis? This needs to be broken down into parts: the ABX test is the firs step in understanding the problem. If it can be shown that there is at least a perceptible difference across the different items, then it will be worth to investigate what those differences are. A typical confidence level employed in comparison tests is 75%. However, for scientific purposes, this is usually elevated to 95%. On the other hand, only five trials were tested. For the purposes of this paper, a minimum 75% confidence level is used in validating the testing procedure. Illustration 6 shows that, at 50%, the "classical" programme material cannot be safely assumed to be differentiable. On the other hand, "pop" was generally close to 75%. This implies that perhaps the more percussive characteristics of the genre simplify identification. A problem arises from the classical genre's results, however. Since the preference testing procedure was mainly conducted used this type of material, as recommended by EBU, its results may not be reliable. Obviously, the number of trials has an impact on the final results. On another note, while the SSL G series studio's results may seem nonsensical at first, it should be noted, however, that those responses were generally considered of poor quality. Artefacts may have been used to help identification by the panel. Sampling error may also be considered as a factor in the inadequacy of these results. However, Hoeg presents results of similar tests for an eight person listening panel (1997:47). The most likely source of error for this experiment was the initial technique used to remove the reflected path component. As argued before, the technique was a compromise solution generated by the unavailability of adequate resources. That being said, and in concordance with the first experiment, the results do seem to indicate a subtle improvement in perceived quality, when the reflected path is removed. However, this subtlety falls within potential instrumental error, and therefore cannot be assumed to be fully conclusive. Contrarily, there is a significant and evident benefit in the use of far-field monitoring, both because of timbral aspects (full range loudspeakers), and spatial and stereo ones (flush mounting, better angles of incidence upon the desktop, etcetera). In conclusion, as was said of the objective measurements experiment, while the results do seem to point towards the hypothesis, additional testing with more precise equipment would be required to remove all doubt. As it is, one cannot completely remove instrumental error, signal processing artefacts and even researcher bias from this interpretation.
Short-term observation
Under this new light, the initially proposed observation of operational decisions within the experimental monitoring environments is irrelevant. Because the objective was to show a correlation with the experimental process, which revealed itself to be inconclusive within an unquantified margin of error, this phase must be discarded until the problems are addressed. This is because it fundamentally depends on the accuracy of the experimental results, due to two main reasons: 1) to allow for a meaningful contextualization of the observational results; 2) the experimental IR must be of reasonable quality, because it is used to process the test material for the short-term observation setting.
-
-
General Discussion and Conclusions
-
Practical Problems
Two main factors impede further progression in this study. Both are concerned with the first experiment, and the compromises that had to be assumed, due to lack of more appropriate resources. Although practical problems encountered in the first experiment, they influence the second one at the highest level. This is because the test material used in the listening experiments is processed with the impulse responses produced in the objective measurements phase. Those two factors are instrumental error and unreliability of the reflections removal technique.
Instrumental error may have been caused by the use of a live human test subject, together with two omnidirectional microphones to create a simple solution to a simple problem: the unavailability of a binaural dummy recording head. This equipment was necessary, because the second experiment required the convolution of test material with the binaural impulse responses for each variable, to realistically reproduce spatial characteristics through headphones. This technique is well documented and was discussed in the literature review and methodologies sections. The unavailability of a binaural recording device was due to the extreme high price and rarity of such equipment, usually only available in civil engineering and acoustics laboratories. To the researcher's knowledge, no equipment like this is available for rental or loan in New South Wales. The solution found was given by Angelo Farina, a renowned scholar in the field of convolution and binaural recordings. In private communication, Dr. Farina proposed that "the usage of a real head, instead of a plastic one, creates very good binaural recordings".1 This solution was implemented in all three studios. However, even though the same test subject was used in all experiments, and the same distances were measured, significant inconsistencies in the frequency responses of the left and right channels were produced, if figures 25 through 36 are any indication. This was certainly not helped by the fact that no adequate portable pre-amp was available as well. Additionally, small movements by the test subject during the measurement may have introduced timing issues.
Another issue was the removal of the relevant desktop and console reflections. Initially there were three possible solutions to this problem: 1) physically remove the offending items from the control rooms; 2) use time-delay spectrometry; 3) manually edit out the reflection using a graphical sound editor. Simple logistics ruled out option number one: it would be a very significant undertaking to remove a large format console such as the SSL G or K series consoles from their control room. TDS was considered, but a professional software-based solution is, at the moment, non-existent. Every TDS-capable system is hardware-based, and is therefore out of the reach of the researcher. A simple rudimentary inverted TDS technique was attempted, as described in the methodologies chapter, but the results were severely distorted by assorted artefacts, which rendered an analysis impossible. That left the manual editing option open, arguably the least desirable, by its very nature. This may also explain the left-right inconsistency, and may well be the most important factor in the inconclusive results achieved.
In the second experiment, although a standard was followed, the six-point rank scale could be altered to a 12 point one. This could possibly yield a more visible pattern. However, the mere resolution of the first experiment would render this redundant. Another option would be to simply employ a larger listening panel. In any case, it must be said that the number of listeners used is concordant with the standard. One could question the absolute validity of the recommendation, on the basis that the number seven is arguably derived from the maximum number of listeners one is usually able to fit into a single listening session. It seems that there is no significant statistical reason for the number, other than the aforementioned practical one.
-
Research Process
No data refutes the general research methodology presented. The survey brought forth the expected results, as did the objective measurements one, albeit at a subtle level. The same can be said of the subjective assessments phase, although at a lower level still. The relationship between the various phases should be clear, and perfectly attainable, should the encountered problems be addressed in subsequent efforts. This is supported by the fact that the relevant methodologies and techniques presented in the literature review and methodologies sections were followed, where possible. The hypothesis' premise that acoustics reacts to ergonomics was established, mainly through documentary evidence. Objective measurements allowed to confirm the fundamental flaw of the monitoring environment where mix-down takes place. This was, to an extent, correlated with specific ergonomics-related factors, particularly near-field monitoring and first reflections. However, unexpected measurement inconsistencies were not clarified by means of subjective assessments, as initially expected. Although specific perceptions in the monitoring environment where successfully connected to mastering problems, such as spatial, stereo and timbral impressions, these weren't, in turn, undoubtedly correlated to ergonomics-related factors. As was expected - and addressed in the learning agreement - the main problems encountered were of a practical and technical nature. These were, unfortunately, of such scale that the hypothesis must be considered inconclusive, to the extent that it was tested in this study.
-
Further Research Recommendations
Even though it is customary, in inconclusive hypothesis cases, to either confirm the null or propose alternate ones, this research's hypothesis is maintained, although the results do not yet confirm it. This is due to the fact that instrumental precision and general resource unavailability were in the genesis of its inconclusiveness. As said in 5.1, there is no reason to believe that the general research process was flawed in any way. With that assumption in mind, the hypothesis must be maintained, and recommended for further study, with the proposed solutions to the practical problems taken into consideration.
Within the limitations of this inconclusiveness, and taking into consideration what was, in fact, established with some degree of certainty, particularly by documentary evidence analysis, the survey and the objective measurements, I propose: the overlap in production and post-production processes is to blame for the observed fundamental flaws in the monitoring environment. The mix-down process is undoubtedly a post-production one, and as such belongs to downstream facilities. In fact, it could be argued that it is highly symbiotic with the pre-mastering stage. This is supported by the fact that most pre-mastering studios now master a high number of stem mixed works (Katz, 2002). Considering pre-mastering had its genesis as an absolutely technical step of preparation of masters for duplication, its path towards creativity in the last decades is obvious. Accepting that, along with the increasing number of accessible tools for mastering in the home studio, means that mix-down and pre-mastering are converging. A solution to this paradigm would be the hybridization of post-production processes into an unique post-production studio. The recording studio would therefore be free to focus on the correct tools for its purpose, as would the post-production studios. At the same time, other solutions can be investigated. Possible research topics in this area of study are: the sound scattering properties of mixing consoles and their diffusion effect in the sound path; acoustical-conscious console design; effect of angling of loudspeakers in the reflections path; and influence of non-ergonomics factors in creating the need for a pre-mastering phase (such as inexperience and equipment inadequacy), among others.
-
References
- AES20-1996 AES Recommended Practice for Professional Audio, "Subjective evaluation of loudspeakers", Audio Engineering Society, 1996
- Ahnert, W., & Schimdt, W., "Fundamentals to perform acoustical measurements", appendix to EASERA, Software Design Ahnert GmbH, 2005
- Barass, R., "Scientists must write: a guide to better writing for scientists, engineers and students", London, Uk: Oxford University Press, 1996
- Bassal, D., "The practice of mastering in electroacoustics", translated by Ian Stewart and David Paquette, unpublished, 2002
- Bech, S., "Selection and training of subjects for listening tests on sound-reproducing equipment", Journal of the Audio Engineering Society, Vol. 40, No. 7/8, 1992
- Bell, J., "Doing your research project", 3rd ed., Buckingham: Open University Press, 1999
- Beranek, L. L., "Music, acoustics, and architecture", New York: John Wiley and Sons, 1962
- Blaxter, L., Hughes, H. & Tight, M., "How to Research", 2nd ed., Open University Press, McGraw-Hill Education, 2001
- Børja, S. E., "How to fool and make bad recordings: the influence of the listening response in sound control rooms on the mixdown of a multitrack master recording", convention paper 1176, presented October 29 - November 1, 1976 at the 55th Convention of the Audio Engineering Society, New York
- Cable, C. & Enerson, R., "Those early late arrivals! Mr. Haas, what would you do?", Journal of the Audio Engineering Society, Vol. 28, No.1/2, 1980
- Chandrasekhar, R., "How to write a thesis: a working guide", Australian Research Centre for Medical Engineering (ARCME), The University of Western Australia, 2002
- Clark, D., "High-resolution subjective testing using a double-blind comparator", J. Audio Eng. Soc., Vol. 30, No. 5, 1982
- D'Antonio, P. & H. Konnert, J., "Recording control room design incorporating a reflection free zone and reflection phase grating acoustical diffusors", Journal of the Acoustical Society of America, Suppl. 1, Vol. 78, 1985
- Davis, D. and Davis, C., "The LEDETM concept for the control of acoustic and psychoacoustic parameters in recording control rooms", Journal of the Audio Engineering Society, Vol. 28, No. 9, 1980
- EBU Tech 3286-E, "Assessment Methods for the Subjective Evaluation of the Quality of Sound Programme Material - Music", European Broadcasting Union, 1997
- Erne, M., "Perceptual audio coders: what to listen for", unnumbered convention paper, presented September 21-24, 2001 at the 111th Convention of the Audio Engineering Society, New York
- Farina, A., Martignon, P., Azzali, A., Cabrera, D., Capra, A., "Reproduction of auditorium spatial impression with binaural and stereophonic sound systems", convention paper 6485, presented May 28-31, 2005 at the 118th Convention of the Audio Engineering Society, Barcelona
- Farina, A., "Simultaneous measurement of impulse-response and distortion with a swept-sine technique", convention paper 5093, presented February 19-22, 2000 at the 108th Convention of the Audio Engineering Society, Paris
- Farina, A., personal communication, 2005
- "Functionalism, in art and architecture", The Columbia Encyclopaedia, 6th ed., New York: Columbia University Press, 2001-04
http://www.bartleby.com/65/fu/functism-art.html
Accessed 09-01-2005 - Gardner, B. & Martin, K., "HRTF measurements of a KEMAR dummy-head microphone", MIT Media Lab Perceptual Computing - Technical Report #280, 1994
- Giusto, A., Tronchin, L., Tarabusi, V., Farina, A., "Spatialization and acoustical simulation in the binaural technology", International Symposium on Musical Acoustics, Perugia, 10-14 September 2001
- Haas, H., "Uber den Einfluss eines Einfachechos auf die Hörsamkeit von Sprache". Acustica, Vol.1, 1951
- Heyser, R. C., "Acoustical measurements by time delay spectrometry", Journal of the Audio Engineering Society, Vol. 15, No. 4, 1967
- Hoeg, W., Christensen, L., Walker, R., "Subjective assessment of audio quality - the means and the methods within the EBU", EBU Technical Review, Winter 1997
- Katz, B., "Mastering audio: the art and the science", Canada: Focal Press, 2002
- Klepper, D., "Subjective loudspeaker testing for accuracy", Journal of the Audio Engineering Society, Vol. 52, No. 10, 2004
- Lee, M., "Multi-channel auditory search: toward understanding control processes in polychotic auditory listening", Haas, E. & Edworthy, J., The Ergonomics of Sound: Selections from Human Factors and Ergonomics Society Annual Meetings, 1985-2000, Santa Monica: Human Factors and Ergonomics Society, 2003
- Lipshitz, S. & Vanderkooy, J., "The great debate: subjective evaluation", Journal of the Audio Engineering Society, Vol. 29, No. 7/8, 1981
- Naqvi, A. and Francis Rumsey, F., "The active listening room - a novel approach to early reflection manipulation in critical listening rooms", Journal of the Audio Engineering Society, Vol. 53, No. 5, 2005
- Niedrist, G., "Echo suppression for loudspeaker-microphone system measurements", convention paper 3245, presented March 24-27, 1992 at the 92nd Convention of the Audio Engineering Society, Vienna
- Owinski, B., "The mastering engineer’s handbook", Vallejo: Mix Books, 2000
- Richert, W., Ahnert, W., Feistel, S., "Merging room-acoustic and electro-acoustic measurement methods", convention paper 6165, presented at the 116th Convention of the Audio Engineering Society, Berlin, May 8-11, 2004
- "Short term observation", C.S.U. Writing Guide, Colorado: Colorado State University
http://writing.colostate.edu/references/research/observe/com3a3.cfm
Accessed 09-01-2005 - Spikofski, G., "Assessment of sound-field parameter differences in studio listening conditions", EBU Technical Review, September 2000
- Storyk, J. M., "Application of varying architectural and programmatic requirements to audio recording studio acoustic design", convention paper 2880, presented at the 87th Convention of the Audio Engineering Society, New York, October 18-21, 1989
- Sullivan, L. H., "The tall office building artistically considered," Lippincott’s Magazine, March 1896
- Tarabusi, V., Tronchin, L., Farina, A., Hotehama, T., Shimokura, R., "The calculation of binaural parameters by means of HRTFs and Ando's method", 18th International Congress on Acoustics, 4-9 April 2004
- Voelker, E.J., "Control rooms for music monitoring", Journal of the Audio Society, Vol. 33, No. 6, 1985
- Walker, R., "Controlled image design: the measurement of time-frequency responses", Research and Development Department, Technical Resources Division, The British Broadcasting Corporation, BBC RD 1995/3
- Walker, R., "Controlled image design: the management of stereophonic image quality", Research and Development Department, Technical Resources Division, The British Broadcasting Corporation, BBC RD 1995/4
- Walker, R., "Controlled image design: results from the first installations", Research and Development Department, Technical Resources Division, The British Broadcasting Corporation, BBC RD 1995/5
- Wrightson, J., "Psychoacoustic considerations in the design of studio control rooms", Journal of the Audio Engineering Society, Vol. 34, No. 10, 1986
Bibliography
- AES-4id-2001 AES Information Document for Room Acoustics and Sound Reinforcement Systems, "Characterization and measurement of surface scattering uniformity", Audio Engineering Society, 2001
- AES-6id-2000 AES Information Document for Digital Audio, "Personal computer audio quality measurements", Audio Engineering Society, 2000
- Booth, W., Colomb, G., Williams, J., "The craft of research", 2nd ed., Chicago: The University of Chicago, 2003
- Bruyn, B., Mason, R., Ford, N., Rumsey, F., "Verbal and nonverbal elicitation techniques in the subjective assessment of spatial sound reproduction", Journal of the Audio Engineering Society, Vol. 49, No. 5, 2001
- EBU Tech. 3253-E Sound Quality Assessment Material, "Recordings for subjective tests - user’s handbook for the EBU-SQAM compact disc", European Broadcasting Union, 1988
- EBU Tech. 3276, "Listening conditions for the assessment of sound programme material: monophonic and two-channel stereophonic", European Broadcasting Union, 1998
- EBUR53-1998 EBU Technical Recommendation, "Test material for the subjective assessment of sound signals", European Broadcasting Union, 1988
- EBUR90-2000 EBU Technical Recommendation, "The subjective evaluation of the quality of sound programme material", European Broadcasting Union, 2000
- Everest, F., "Master handbook of acoustics", 4th ed., U.S.A.: McGraw-Hill, 2001
- Haas, E. & Edworthy, J. (eds), "The ergonomics of sound: selections from human factors and ergonomics society annual meetings, 1985-2000", Santa Monica: Human Factors and Ergonomics Society, 2003
- Henrique, L., "Acústica musical", Lisboa: Fundação Calouste Gulbenkian, 2002
- Rec. ITU-R BS.1116-1, "Methods for the subjective assessment of small impairments in audio systems including multichannel sound systems", International Telecommunication Union - Radiocommunication Sector, 1997
- Rec. ITU-R BS.1283-1, "A guide to ITU-R recommendations for subjective assessment of sound quality", International Telecommunication Union - Radiocommunication Sector, 2003
- Rec. ITU-R BS.1284-1, "General methods for the subjective assessment of sound quality", International Telecommunication Union - Radiocommunication Sector, 2003
- Rec. ITU-R BS.1285, "Pre-selection methods for the subjective assessment of small impairments in audio systems", International Telecommunication Union - Radiocommunication Sector, 1997
- Rec. ITU-R BS.1387-1, "Method for objective measurements of perceived audio quality", International Telecommunication Union - Radiocommunication Sector, 2001
- Rec. ITU-R BS.562-3, "Subjective assessment of sound quality", International Telecommunication Union - Radiocommunication Sector, 1990
- Rumsey, F., "Spatial quality evaluation for reproduced sound: terminology, meaning, and a scene-based paradigm", Journal of the Audio Engineering Society, Vol. 50, No. 9, 2002
- Streicher, R. & Everest, F., "The New Stereo Soundbook", 2nd ed., Pasadena, California: Audio Engineering Associates, 1998
- Turley, R., "Writing Essays", London: RoutledgeFalmer, 2000
Discoography
- Katz, B., "Hear the Magic Vol. II", Digital Domain, 2003
- "Perceptual audio coders: what to listen for" CD, Audio Engineering Society, 2002
- "PEQS" CD - Parameters for the subjective evaluation of the quality of sound programme material, European Broadcasting Union, 1997
- "SQAM" CD - Sound quality assessment Material - recordings for subjective tests, European Broadcasting Union, 1988
