What's Inside![[Toronto AES Logo]](../graphics/logo.gif){kind=logo}
![[line]](../graphics/hline.gif){kind=small}
PLEASE NOTE: THIS EVENT REQUIRES AN ADMISSION CHARGE
For Tickets: phone Glenn Gould Box Office at (416) 205-5555 Hours: 11:00 am - 6:00 pm
$20 Members/Students/Seniors; $35 Non-Members
For More Information: phone Janine at Motion Picture Video (416) 863-6414.
What's Inside
This Month's Meeting
Preview
by Denis Tremblay
The ongoing digital revolution in Audio has expanded the boundaries of our Art/Science beyond anything im-aginable only fifteen years ago. Like all revolutionary change this upheaval has created new interpretations of old rules and assumptions: this is the much discussed "paradigm shift". The digital world has also created new assumptions, often on the bones of the old. One of the most interesting shifts in viewpoint has been on the subject of resolution. Resolution in audio is a sort of amalgam of signal to noise ratio, distortion and "granularity".
In the not so recent past, a system capable of 70dB signal to noise and having distortion at standard operating level of less than 3%, was considered a blessing. Today, comments in the audio press about the inadequacies of systems capable of better than 93dB signal to noise with distortion figures in the .001% range are so commonplace as to be accepted at face value. There are those who still maintain that a step backward was taken when we left 70dB S/N at 3% THD behind. Conversely, there are those who maintain what we really need is 120dB S/N at .0005% THD to achieve true happiness.
The common points and perhaps the most interesting things about these arguments are the unexamined assumptions regarding resolution. The advocates of analogue tape assume that their favoured medium is superior, despite having more noise and distortion than digital systems, because it has "infinite" resolution. Those who would prefer more bits assume that this would increase resolution enough to satisfy everyone.
There are two unexamined assumptions in the argument about analogue tape. The first being that the analogue system is not quantized. Analogue tape systems can be considered to be quantized by the distribution of particles in the tape coating. This quantization will have a random characteristic and will allow resolution of "infinite" voltage steps in a statistical sense only. If such a system is measured in fine enough a slice of time, so as to sidestep the statistical process, it becomes frustratingly difficult to find the superior resolution.
The second assumption is that digital systems are not capable of resolution finer than the quantized steps. Both Mr. Lipshitz and Mr. Vanderkooy have gone to great lengths to examine and to debunk this myth. In a purely theoretical sense, digital systems have no resolution between the quantization steps. However, the addition of a ridiculously small amount of noise with the correct statistical distribution, gives these systems "infinite" resolution in exactly the same sense as the previously discussed analogue system. With the addition of noise shaping, which effectively wraps a feedback loop around the Converter/Filter, the added noise is moved out of the audible range. In effect, the noise is replaced by a series of very fine corrections occurring at (over)sampling speeds.
Perhaps the most fascinating thing about the "more bits please" school of thought is the unspoken belief in a world of purity and electrical cleanliness. In this world, it is always possible to resolve to finer and finer levels of detail, because there are no physical forces interfering with the process. This belief system is mostly observed in the marketing departments of converter manufacturers.
A converter may have the architecture of an 18 or 20 bit device, but, because of manufacturing tolerances, have such a scatter in the steps of the finer bits that it is not capable of more accurate conversion than a very good 16 bit device. These scattered bits are often referred to with more than a little cynicism as "marketing bits." It is quite possible to have a 16 bit converter with more resolution, lower distortion and more pleasing sound than a 20 bit converter with lots of marketing bits.
For those of you with design experience, the difficulty of building anything capable of 120dB of signal to noise (20 Bit equivalent) is obvious. Interfering signals come at you from all sides to spoil the beautiful work you have done. Radio Frequency Interference formerly buried in the noise of your 90 dB S/N system is now plainly audible. You can hear broadcasts that even your nitrogen cooled tuner with the hand polished transistors has never found. Harmonics from AC power lines as far away as Leningrad are now measurable. Noise from digital devices is so bad that even thinking of turning your computer on pegs the needle on the measuring equipment.
I can remember reading a technical paper from National Semiconductor titled "THE FINGER AS AN ANALOG DEVELOPMENT TOOL". I kid you not. The main thrust of this paper is that high gain low noise circuits are very susceptible to external noise sources. Enough electrical noise is literally at your finger tips, so that valuable information about circuit stability and noise immunity can be obtained by poking around in an informed manner.
Electrical noise is a systems problem. It is possible to design a stand alone box that has very low noise and very high resolution, but quite another thing to preserve this performance when it is incorporated into a system. If great pains are not taken to properly cable and to properly ground the system, all that beautiful resolution will be lost to electrical noise. It is very important that the internal workings of this hypothetical device be carefully designed. When combined with other devices this box must not become sensitive to new noise sources, and equally important, it must not create a noise source when integrated into a system.
To the uninitiated, it may sound very odd indeed that a device could be well behaved on the test bench and a disaster in the equipment rack. To those in the know, it is an all too frequent problem. Mr. Muncy, Mr. Windt and Mr.Giddings are most definitely among those in the know, and have confronted the problem for many years. From the macroscopic view of building power and grounding, through the systems viewpoint of interconnecting and configuring equipment, to the internal architecture of electronic devices, these gentlemen have plenty to say: it would benefit us all to listen.
Virtually everyone will agree that a quiet recording environment is essential for high quality work. It is perhaps not always understood that noise and poorly controlled room effects in the studio are all part of the "noise" component in the "signal to noise" of the finished product, and will limit final resolution as effectively as any electrical disturbance. For those with access to high resolution playback systems, the audibility of these acoustical flaws in commercial recordings is a common occurrence. The low distortion and wide dynamic range of digital recording media are merciless in exposing these problems. Mr. Barman and Mr. Paige will enlighten us all on this subject while surrounded by their own work for the CBC Broadcast Center.
The available resolution of any system is limited by the masking effects of the noise present with the signal, this is basic information theory as explained by Shannon over forty years ago. A perfect digital converter will resolve whatever is fed into the device: it will not discriminate against noise. A dirty signal will be resolved into the digital equivalent of a dirty signal. If you are trying to exploit the last few bits of your system and these bits are fully occupied with AC line harmonics and/or thermal noise from microphone preamps and/or air conditioning rumble and traffic noise from the studio environment, they just aren't available to you. God help you if the converter itself is flawed with non-linearities, quantization noise and distortion.
Once over the fence into the deterministic digital world, it is a simple matter to increase resolution by increasing bus widths; in the real world, on the chaotic analogue side of the fence, resolution is a very complex issue. We feel that it is time to dispel some of the myths and legends. We propose to organize a one day seminar on the subject. We have decided that the title of GROUNDVIEW is very appropriate. This is a ground up look at the issue of resolution in audio systems where grounding plays a prominent role.
Last Month's Meeting
Review
The AES tour of Yorkville Sound was quite a success, with approximately sixty people turning out for it.
Yorkville Sound definitely has assembly line manufacturing down to a fine art. Starting off with the design lab, three design groups work utilizing CAD\CAM packages; their design is passed on to a Product Engineering team that turns the design into a working prototype through assembly drawings, material lists and CAM instructions for the various automated machines and production jigs.
Yorkville Sound has over thirty heavy duty machines used in the manufacturing process. Everything from indexing panel saws to I.C. insertion and programmable wave soldering systems.
The factory is divided into three main departments: electronics, cabinets and metal shop. Each department carries out specific tasks in the production of speaker cabinets ranging from 50 to 1000 watts, power amplifiers from 150 to 2400 watts, mixing consoles from 4 to 20 channels, and several lighting products as well.
The latest addition to Yorkville Sound's product line is a digital console, the DMX2400. This DSP-based digital console has built-in A/D and D/A converters. The format is 24 channel inputs and 8 bus outputs with the main monitor and effects buses included in these 8. The console is intended to compete in the 4-bus console marketplace.
It is equipped with a pretty extensive 4-band EQ section, balanced XLR and 1/4" connectors on the ins and outs, and an AES/EBU digital output at the mains. The DMX2400 is an impressive all-in-one digital mixer that will definitely compliment Yorkville Sound's product line.
The new 66,000 square foot facility was also very impressive and encouraging from a Canadian manufacturer's perspective.
GroundView: The Agenda
9:00 am Introduction
THE MORNING SESSIONS
9:05 am Noise in Acoustics: Acoustic Background Noise... How Low Can You Go?
-Mohan Barman (Aercoustics) and Tom Paige (Vibron)
10:30 am [Coffee Break]
10:45 am Noise in Digital Systems: Digital Noise... Coping with Quantization
-John Vanderkooy and Stan Lipshitz (Audio Research Group, University of Waterloo)
12:15 pm [Lunch Break]
THE AFTERNOON SESSIONS
1:30 pm Noise in Electrical Systems: Power, Grounding, Shielding, and Interconnections in Analogue + Digital Signal Processing Systems... Understanding the Basics
-Neil Muncy (Neil Muncy and Associates), Phil Giddings (Engineering Harmonics), John Windt (Windt Audio)
3:15 pm [Coffee Break]
3:30 pm Noise cont'd
5:30 pm [Closing Remarks]
GroundView: Presentation
Abstracts
Acoustic Background Noise: How Low Can You Go?
Presented by Mohan Barman (Aercoustics) and Tom Paige (Vibron)
Recording facilities work best with very low background sound levels, so that full advantage can be taken of audio equipment dynamic range. However, studio construction costs increase dramatically with the degree of sound isolation and noise control to be provided. Inevitably, compromises are made. This process can be done systematically to achieve optimum results.
The first step is to set the background noise design criterion for the studio. Available criteria include NC curves, RC curves, PNC cruves and NCB curves. These will be discussed in terms of relative merit and the two basic variables. The first is the sound generated within the studio by mechanical systems, lighting and technical equipment. The second is the sound from sources located outside the studio which will be transmitted through walls, floors and ceilings.
To achieve low background noise levels, the acoustical consultant must work closely with the other consultants and contractors. Some important considerations include: space planning, proper vibration-isolation of equipment, keeping air-flow velocities low and reducing fan noise from air-handling units. Some fundamental design guidelines will be discussed.
Noise intrusions from sources outside the studio are classified as airborne or structure-borne. Basic room-isolation techniques for reducing sound transmission will be described. These requirements become more complex when studios are located in close proximity with each other (as in the case of the new CBC building). Environmental noise sources such as roadway traffic and aircraft noise can also have an influence on the design of the outside walls and roof.
The Canadian Broadcasting Centre serves as an excellent example to demonstrate the design process. Mohan Barman, as acoustical consultant for the CBC, will explain the process used for design development of the base building as well as preparation of acoustic performance specifications. He will discuss the design review process and the testing procedures used to verify compliance with the criteria. Tom Paige, as acoustical consultant for the developer (Cadillac Fairview) will elaborate on the design process and explain some of the compromises made to control costs yet still meet the CBC criteria specifications.
An interesting development in the performance specifications for the CBC project was the utilization of a new method used for specifying and measuring sound-isolation based on a source spectrum with strong low-frequency components such as music or mechanical-system noise. The advantages of this method over the conventional STC rating system will be discussed.
NOISE IN DIGITAL SYSTEMS
Digital Noise: Coping With Quantization
Presented by Stanley P. Lipshitz and John Vanderkooy
(Audio Research Group, University of Waterloo)
Our presentation will consist of a lecture, profusely illustrated by audio examples, covering the quantization of digital audio signals, and how the artifacts of quantization can be rendered benign by suitable preprocessing of the signal.
We will start at an elementary level, and build up to the most recent developments concerning uniformly-quantized audio. The topics to be addressed are:
How the specific nature of the quantization error depends upon the input signal, taking the form of a steady white noise for large, complicated signals, but becoming distortion-like and displaying noise modulation when the input signal is small. The 6-dB-per-bit dependence of the quantization noise upon the digital wordlength will be demonstrated.
Next, it will be shown how the quantization noise can be decorrelated from the input signal, rendering it free from distortion and noise modulation, by the addition of a suitably-chosen dither noise to the audio signal before it is quantized (or, in the case of digital signal processing, before it is requantized). Both subtractive and non-subtractive dither schemes will be illustrated, and the effect of the dither's statistics upon its performance will be shown. We will also address the use of spectrally-shaped dithers in order to make the dither noise less audible.
Thus far, the dither noise has been optimized to eliminate quantization artifacts, but the (re)quantization noise itself now represents the limiting factor in determining the overall audible signal- to-noise ratio. We now show how, by using error feedback around the (re)quantizer, even the roundoff noise component can be shaped to render it less audible.
With such noise shaping, proper psychoacoustic design of the shaper parameters can afford a 3- bit (i.e., 18-dB) audible benefit, so that a 16-bit CD can provide an apparent 19-bit noise floor. This makes it possible for the lower noise floor and greater resolution of professional 20-bit recordings to be largely preserved when transferred to CD.
The performance requirements that this imposes upon the final digital-to-analogue converter stage will be addressed, as will be the necessity of using proper dither in the noise shaper to avoid the occurrence of low-level (but audibly pernicious) limit-cycle oscillations due to the feedback in the noise shaper. These effects will be demonstrated, and some current designs will be compared.
NOISE IN ELECTRICAL SYSTEMS
Power, Grounding, Shielding, and Interconnections in Analogue + Digital Systems: Understanding the Basics
Presented by Neil Muncy (Neil Muncy and Associates), Phil Giddings (Engineering Harmonics) and John Windt (Windt Audio)
Our presentation will address a significant industry-wide problem resulting from the present lack of meaningful standards pertaining to analogue equipment interconnections in general, and to the matter of ElectroMagnetic Interference (EMI) and Radio Frequency Interference (RFI) immunity in particular.
In the last decade, the Audio industry has experienced spectacular growth. Rapid advances in computer, fibre-optic, Digital Signal Processing (DSP), and other new technologies have revolutionized the way the Audio industry operates, and the products and services which result. The widespread adoption of computer-based design and manufacturing techniques has facilitated a wholesale reduction in the cost-per-feature and consequent selling price of studio equipment. Compared to a decade ago, the worldwide market for audio signal processing equipment is enormous.
Curiously, and with only a few notable exceptions, most of this equipment exhibits surprisingly little EMI/RFI immunity. As a result, it is becoming increasingly difficult to successfully assemble complete signal processing systems which offer adequate immunity to interference from nearby EMI/RFI sources. Previously existing system noise problem have become even more apparent due to the introduction of low-cost digital recording devices whose inherent noise floor is too low to "mask" the noise contributed by other system components.
This situation seems to have resulted from various combinations of the following:
-the apparent lack of systems experience, and/or unfamiliarity with effective EMI/RFI prevention techniques on the part of equipment designers:
-the apparent lack of system experience and awareness concerning the fundamental need for EMI/RFI immunity on the part of sales and marketing departments
-price competition from other manufacturers
-the perception that this is a low-tech problem which does not deserve attention
-entrenched corporate policies which tend to ignore feedback from technicians in the field, and/or do not acknowledge the importance and consequences of this issue
-the long established industry-wide tradition of perceiving this entire subject as more a matter of "witchcraft", "voodoo", "black magic", and other unspecified occult practices, than of misunderstood and neglected applied science.
-a lack of meaningful standards
-a general lack of appreciation of the fundamental role of basic physics
-the presence in the industry of at least one whole generation of people who have never encountered an audio system which did not hum, buzz and pick up nearby broadcast stations, and as a result have come to accept such performance as an inevitable part of the "game".
The present situation is also partly the result of the promotion and marketing strategies currently employed to sell this equipment, which clearly imply:
-that assembling a system should involve nothing more complicated than plugging-in a few cables, and
-that purchasers possessing minimal technical skills should be able to do this work.
The fact of the matter is that many current purchasers of studio equipment (who are quite rightly preoccupied with other more serious matters) simply do not have either the technical skills or the time to do much more than match the plugs on ready-made cables with the appropriate jacks on their newly- acquired equipment. To make matters worse, many new installations are now of the "home studio" nature, where technically trained support personnel are simply not part of the business plan. When problems arise, these purchasers are essentially helpless. Complaints to manufacturers and vendors often produce the response of "...gee,, we never heard of that one before", and entirely too frequently "...it's not our problem". Ultimately, many purchasers find themselves forced into the position of having to devise (and often pay for) a "fix", which in many cases may only partly "smother" the symptoms, and often severely compromises the basic electrical safety of the entire installation.
The present situation represents a massive (and expensive) waste of time and resources on the part of all concerned. Repeat sales, which are the long-term lifeblood of any company, are at risk. There is no basic technical reason why this situation should exist at all. All of the methods and techniques required to minimize RFI/EMI problems in audio equipment have existed for decades in the technology of other electronic instrumentation and communications systems. A vast body of prior work has been published on this subject, and for the most part, is still in print.
An on-going study conducted by members of the AES Standard Committee Working Group 03 has identified an easily demonstrable cause-and-effect relationship between a commonly employed equipment design practice and the consequent lack of EMI/RFI immunity in audio signal processing systems.
During this presentation, this relationship will be described and explored. Easily implemented procedures for quickly identifying equipment responsible for system noise problems, and techniques for minimizing these problems in existing systems, will be demonstrated. It will be shown that the present situation is not the result of some mysterious or difficult to solve "high-tech" problem... there is no need to pursue a search for a "rocket science" solution, and that relatively minor and inexpensive changes to existing equipment design and manufacturing practices could result in a significant reduction in future system noise problems.
Classifieds
Downtown Toronto, established client base, Equipped for music recording, automated mixdown and audio post production for video (4 machine SMPTE lock).
Call (416) 868-0713 for details
Henning Bog, graduate of The Technical University of Denmark, 1987 specializing in Consumer Electronics, Electroacoustics and DSP design.
Worked for Bruel and Kjaer, Denmark on their electroacoustic test systems and audio analysers, and new applications development.
Moved to London, Ontario and presently works for Microtronix System Ltd (telephone testing equipment) in the design of acoustics test fixtures and transducers and development of new test and measurement techniques.
Member of AES and IEEE, and co-author of two technical papers on the measurement of loudspeaker parameters using laser velocity/displacement transducers.
Welcomes inquiries, referrals or hints about possible opportunities. Phone (519) 455-8590
For information on workshops on the AVID 8000 Non-Linear Video Editing System, please contact Jim Cox: Phone (905) 845-4620 Fax (905) 815-4067
Previous and Next
Issues
Back to January 1994
Articles may be used with the Author's Permission. Contact the Bulletin Editor: earlm@hookup.net
Editor: Earl McCluskie
The Bulletin is prepared in print and on Horizons and the Internet by Earl McCluskie.
What's Inside 
Forward to March 1994