Pre-meeting "Dutch Treat" dinner, 6:00 pm Montreal Bistro, corner of Adelaide and Sherbourne
Join Jim Burgess, Rick Waye, Carl Stone and other Product Specialists from Saved by Technology for an overview of recent innovations in Macintosh-based MIDI music, digital audio, desktop video and multimedia production systems. The presentation will include:
-short demonstrations of some key new products including the new PowerPC Macs,
-the lastest generation of hybrid MIDI/digital audio sequencers,
-the new wave of virtual mixing/signal processing software, and fully integrated digital audio/video editing systems.
The evening will also include an introduction to the growing phenomenon of electronic communication as it relates to our industry. We'll go on-line to Saved By Technology's own BBS and explore the Internet.
We hope you can join us!
In Toronto, Jim Burgess, Saved By Technology.
by Anne Reynolds
With the introduction of Computer Assisted Acoustic Simulation in recent years, CAD now provides the audio world with a valuable tool for designing and predicting the performance of sound systems for any acoustic environment.
On March 15th, approximately 32 people were on hand to hear Ron Sarau, Engineering Systems Manager for Renkus-Heinz Loudspeakers and John Radul of SF Marketing discuss and demonstrate three examples of the acoustic simulation and auralization software developed by one of Europe's leading acoustical consultants, Dr Wolfgang Ahnert of ADA: EASE, EASE Jr. and EARS.
EASE: Electro-Acoustic Simulator for Engineers - acoustic design and research tool for system and room designers.
EASE Jr: developed specifically for the system designer who does not need all the advanced acoustic analysis features of EASE.
EARS: Electronically Auralized Room Simulation - achieves binaural auralization by electronically combining a dry audio signal with the acoustic characteristics of the room.
Sarau explained that before developing these programs, Dr Ahnert did extensive research to determine exactly what sound system designers are looking for. Since the introduction of these programs, the manufacturer has introduced approximately two software updates per year and Sarau predicts that soon these updates will be available on Bulletin Boards.
At this time, these programs are only available in the PC environment (DOS 3.3 or better) and will run without conflicting with Windows. Sarau suggests a minimum hardware requirement of a 486 with 33 MHz processor or faster and 120 MB hard drive or larger, with EGA or VGA graphics.
When designing a speaker system for a room, you can either draw the space yourself in CAD or import the drawings directly from your client's disc.
Sarau described EASE's excellent presentation function... "it helps you compete with the big guys." For hardcopy output, the programs print out on almost any printer from Epson to laser in a variety of formats including spread sheets, 2D and 3D display, "scan" (to "paint it" or focus in different elements of the display), "speaker view", "spectator view" (to see what the cluster is going to look like and what sound can be expected from that position).
Along with their design capabilities, these programs are also excellent tools for testing the specs of equipment.
The EARS program was demonstrated on Renkus-Heinz's "CoEntrant" Loud Speaker system. The CoEntrant Loud Speaker is the result of their attempt to design a powerful, true broad-band point source speaker with seamless transition from mid to high frequencies. The CoEntrant Loud Speaker accomplishes this with an advanced coax or compound throat design that couples the outputs of multiple mid and high frequency drivers and feeds the combined output into a single shared horn. The CoEntrants are designed to work well together in both small groups and in large arrays and come in a variety of designs including 2 and 3 way, wide dispersion, long throw and subwoofers.
If you missed GroundView: Noise in Audio (and at least 150 people didn't), here's your chance to find out what happened. The Bulletin presents...
Mohan Barman and Tom Paige opened the day-long workshops with their presentation Acoustic Background Noise... How Low Can You Go.
Mr Barman, now of Aercoustics Engineering Ltd of Toronto, began the day by stressing the importance of low background sound in a recording studio. With the use of digital recorders, electrical noise has been replaced by acoustical as the limiting factor in noise level of recordings He went on to describe the design process both for the Glenn Gould Studio in which the seminar was itself being presented, as well as the Canadian Broadcasting Centre building overall.
The inherent mechanical noise of a room (primarily HVAC), as well as the transmission of sound created within adjacent spaces, must be controlled to attain the desired result. To specify the desired result in such a way that measured performance can be quantified, design criteria must be selected. Barman described the original NC curves, initiated by Leo Beranek, as used for the CBC building design (eg the Glenn Gould Studio space being specified as NC-15). He related how improvements to the NC curves are desirable, as a sound meeting NC is unpleasant to hear, especially when later amplified. Later curves addressing this as well as the need to extend low frequencies through the 16 and 31 Hz bands are the RC curves (by Blazier) which strive for neutral sound, as well as the NCB curves by Beranek himself.
Barman went on to describe how the CBC building design dealt with mechanical noise by burying noise sensitive spaces (ie studios as opposed to offices) within the structure, by placing chillers and pumps on the slab on grade in the lowest basement, and by using structurally isolated floor slabs for mechanical cores adjacent to studios. The control of vibration from outside sources (including a proposed subway outside the front door) used large blocking masses in the footings, along with soft rubber pads. The isolation of studios from sounds in adjacent spaces used sound locks, high performance windows, and the room-within-a room concept, modified where necessary to accommodate heavy lighting grids in TV studios. He concluded by discussing the basic principles of controlling the noise of ventilation air systems.
Tom Paige of Vibron Ltd expanded upon Barman's presentation by detailing some of the systems of both the Glenn Gould Studio and the Broadcasting Centre. He also presented data indicating how well the actual spaces as built conformed to specifications.
The ventilation air supply for the Glenn Gould Studio (a 162,000 cubic foot, 350 seat recording concert hall) was described as a constant-volume system, as opposed to the variable-air-volume system used in the control rooms and most other spaces. The need to isolate a large electrical transformer 60 feet away on the floor below was dealt with by means of neoprene-in-shear vibration isolators on the transformer, as well as absorptive construction of the transformer room. The noise in the Glenn Gould Studio as measured was shown to conform with the NC-15 specification, and very nearly with the RC-15 and NCB 15 figures.
Paige related how the concept of acoustic dynamic range of recording spaces in the building was derived. CBC provided a noise source curve (flat to 500 Hz, then 2 dB/octave rolloff above), whose dB level corresponded to dBA-3. When plotted at the desired average SPL along with the background noise, the distance between the two curves at a given frequency gives the dynamic range. The figure for the Glenn Gould Studio was shown to be 89 dB, corresponding nicely to the assumed figure of 90 dB for digital recording systems. He concluded by discussing isolation of a studio from adjacent spaces. The mechanical system background noise and predicted external intrusions should both be 3 dB below the design criterion, so that when they add, the NC spec will still be met.
In Toronto, Tom Shevlin
Drs. Stanley Lipshitz and John Vanderkooy presented Digital Noise: Coping With Quantization at GroundView. I've seen this show in various incarnations at least eight times in about as many years and I will never tire of it. Each time a few new details are revealed . . . and a few more of the old ones finally sink in!
There is at least one "detail" this lecture/demonstration makes abundantly clear from the very start: dither works. More precisely, the right dither properly applied works, but more on that later.
The first order of business is proving that dither is necessary. Hearing is believing, and it is the use of numerous audio examples which makes the Lipshitz/Vanderkooy presentation so persuasive. The main prop in this portion of the show is an ADC, the wordlength of which can be varied.
The first source material is a male speaking voice. As the wordlength is switched from 14 down to 4 bits resolution there is a painfully obvious increase in distortion and modulation noise. With fewer bits to describe the original waveform there is an increase in the quantization error: no big surprise there.
Next up is some orchestral music and it is subjected to similar indecencies. It becomes clear that when the input signal is loud and harmonically complex, we hear less distortion; the quantization error is more noise-like. As the signal becomes soft and/or more simple in harmonic structure, we hear more distortion. Hmmm . . . .
Now the same demonstration is given but this time dither noise is added to the signal before quantization. More dither noise is required for lower bit rates so that by the time we reach 6 bits we hear mostly noise. It may be noisy but there is no distortion or noise modulation even at 4 bits. Wow! This is not a matter of noise masking distortion. The dither actually linearises the transfer characteristic in a manner analogous to bias in an analogue recorder.
In case you don't believe your ears (and just what are you doing in this business?!) we are shown an 8 bit black and white (greyscale) picture followed by the same picture at reduced bit rates, with and without dither. Without dither there are harsh transitions between one grey level and the next. With dither more video noise could be seen but the transitions between greyscale levels were smooth: again, noisy but no distortion.
Time for another audio example- the audio practitioner's proof for the mathematicians' theories! An 8 bit recording of a piano followed by a five second fade is played with and without various types of dither. The wrong kind of dither, or no dither at all, produces varying degrees of nasty noise modulation during the fade while the right kind allows for a fade completely free of artifacts. One LSB peak-to-peak of simple Rectangular Probability Density Function (RBDF) dither removes distortion but still allows noise modulation. Triangular PDF not only removes all distortion, but assures that there will be no noise modulation.
TPDF dither causes about a 5 dB noise penalty but with a 16 bit system you still have 93 dB of dynamic range; and we all know that a constant noise is much more tolerable than a modulating one. In practice it is difficult to create TPDF dither in the analogue domain. So in an analogue-to-digital converter, Stan and John say that Gaussian noise comes acceptably close; it will remove all distortion and leave very little noise modulation. It is a relatively trivial manner to compute TPDF so in a digital system such as a workstation or mixer, TPDF should be generated and added to the signal before requantization.
Next we learn that it is possible to deliver the audible equivalent of 19 bits on the CD 16 bit format. Dr. Lipshitz adds the caveat that the new "20 bit" recording systems offer a real challenge to the manufacturers of microphones, preamplifiers, mixers, and recording spaces. In fact in a related paper handed out to supplement the authors' presentation, it was pointed out that when a number of commercially available noise shaped recordings were analysed, not one was quiet enough to take advantage of the noise shaping!
How is this apparent slight of hand achieved? If you somehow manage to get 20 bits worth of dynamic range stored on tape or disc you need to keep it at this extended wordlength throughout any intermediate processing (did you remember to dither?) or storage stages. At the final mastering stage when it becomes necessary to truncate to 16 bits, the signal must first be properly dithered (see above!) and then passed through a noise shaper.
A noise shaper works by feeding back the digital error signal of the requantization operation through a filter. This changes the spectrum of the error noise such that noise is minimised where the ear is most sensitive (around 4 kHz) and moved to where the ear is least sensitive (around 20 kHz). Although there is a small increase in wideband noise, the audible noise is significantly reduced.
John and Stan use the modified E-weighted curve as a model for determining the audibility of the noise floor. They note that there is not total agreement on which curve should be used and that in fact the precise curve would vary from person to person because of physical differences. A simple 2nd-order filter can produce the general shape while a 50th-order filter can match the modified E-weighted curve almost exactly.
More audio examples, this time to remind us that the most elaborate and expensive noise shaper in the world can't do a thing for the modulation noise introduced by truncating without properly dithering first.
Those of us who work with audio in the digital domain must be very careful how we choose and operate our equipment.
Stanley Lipshitz and John Vanderkooy's research has led directly to improvements in the quality of the equipment we can buy today. Their wonderful presentation clearly illustrates to users the importance of choosing equipment carefully and using it properly.
In Toronto, Peter Cook
The afternoon sessions, entitled Power, Grounding, Shielding and Interconnections in Analogue and Digital Signal Processing Systems... Understanding the Basics were presented by Neil Muncy (Neil Muncy Associates), Phil Giddings (Engineering Harmonics), John Windt (Windt Audio), and Tom Shevlin (CBC), all seasoned professionals in audio engineering and systems noise solutions.
Part 1, Introduction, was presented by Neil Muncy in a personable, humorous style. Electrical safety history started when the lightning rod was invented: it made buildings safer. Later, however, lightning rode in on AC power grid wires. Therefore, after observing that lightning can be readily diverted to earth by arcing to one grounded grid wire at some safe point outside a building, the grounded neutral wire became official, and our troubles began.
Some observations: Mother Nature makes the rules, and basic physics underlies all problems, when we choose to ignore it in our safety arrangements.
Some of the means of coupling power grid noise into audio systems include: direct induction, direct connection or injection, capacitive coupling, shared current path, (common impedance coupling), common-mode coupling. Fixes involve avoidance and the inverse-square law nature of inductive and capacitive coupling, used to advantage.
Part 2, Technical Ground Systems vs. the Electrical Code, was presented by Phil Giddings, and accompanied by an 11 page handout, a reprint of his seminar paper of the same name, which also lists 7 references for further study, plus a reference diagram. He began with the observation that there are no technical (i.e. equipment as well as safety oriented) grounding codes, nor common practices related to technical grounding. This situation has led to many different approaches to the basic safety requirements and EMI/RFI avoidance, which are incompatible in an audio system built from different and separate components.
About half of all semi-pro and professional audio system components have fundamental problems, and therefore any complex system will have a high probability of a grounding/EMI/RFI problem. Basic avoidance manoeuvres include: true balanced in- and outputs; low-impedance outputs; input CMRR over 80 dB at 20 KHz; isolated ground wiring back to one point in the power distribution panel; take system power only from one AC branch circuit and outlet, where possible; always use 3-prong grounded power cords; broadband AC power filtering and shielding; always tightly twist the power conducting wires; tightly bundle all power and signal cables together (reduce pickup loop area); balanced 60-0-60V or 120-0-120V distribution for large 120V and 240V equipment; large extra grounding conductor, bundled with the original one.
Part 3 was titled Grounding, Shielding, Hums, Buzzes and things that go ZAP! in Your Sound System. Presented by Neil, with help from John Windt, accompanied by a 9-page handout of equipment diagrams and a list of 17 references. A basic observation is that in any unbalanced audio connection system the ground (shield) return impedance is in series with the audio, when viewed as a complete source-to-load loop. This impedance needs to be very low, but cannot be zero. Therefore a balanced circuit is always superior.
However, the ground return impedance is still important in a balanced system, since many currents will flow through the shield and, via an XLR connector's pin 1, into equipment chassis. If this pin 1 goes directly to a metal chassis - OK! If longer than a centimeter, or not to a metal chassis, you've let EMI and RFI into your box. First rule of chicken farming: Don't let the fox into the henhouse! (Neil Muncy).
Induction into various cable types was then demonstrated. A sampling of shield currents: Induced, from lighting circuits; RFI, EMI, ESD; differential, from interconnection of different AC ground wires (ground wires are NOT usually at zero Volts due to leakage, induction or high neutral current) or from grounded loudspeaker wiring, which should have been left floating.
Typical values: from a few mA, to Amperes (in fault conditions). Eight sample system diagrams with XLR pins 1 yes/not directly connected to chassis were presented and traced. The conclusion was obvious, yet a lot of professional equipment is not built accordingly. Therefore, Neil and John developed trouble-sniffers which can, by using basic physics, identify a grounding or unbalance problem.
1. The 100 mA 60 Hz "Hummer" consists of a 6V AC, 300 mA wall-plug transformer, a 68 or 75 Ohm 1 Watt current limiting resistor (or 100 mA, 6V lamp) and long clip leads.
Connect its output across all suspect ground connections with volume control near max.
No extra noise is good. Any extra noise is not good.
2. The "Adaptor" for use with the "Hummer" consists of an XLR male, wired through to a female (pins 2, 3 and shell) with pins 1 brought out as two thick, short alligator clip leads.
Plug into suspect input. The near-chassis pin 1 lead clips to chassis and the far one to the "hummer", so it can drive a grounding or chassis loop or excite shielding with 6V AC. Increased noise during testing means a weakness in design or construction of the system.
3. The third is the "Balanced Output Tester", same as #2, except pins 1 are wired through, no clip leads. An SPDT center-off switch is wired so it can short pins 2 to 1, or pins 3 to 1, or none (open).
If, while listening to steady sound, a balanced output thus shorted causes an audible level difference, it is poorly balanced. If the level drops over 6 dB it is simply unbalanced. Using just these three testers, you can qualify a $50 gadget or a $50,000 console for proper grounding and shielding design, and then check out the cabling.
Part 4, Why is it that almost every time I plug it in, it doesn't work right? was presented by John Windt. Some results of using the "Hummer" and "Balanced Output Tester" were presented. One 24- channel EQ gave a 50 dB decrease in S/N when "hummed" from input to output jack shields. Many electronically balanced output circuits (i.e. not using a transformer) are inadequate and use 3 to 6 op- amps. Yet, the simple 2-amp (inverting and non-inverting arrangement) is superior. Also, the recent all- integrated circuit driver IC is excellent.
Another observation was that some equipment usually tests quite OK in its native land (Japan, England) and then less so here. Blame our unbalanced AC power.
Part 5, Questions and Answers, led to some lively interaction on imported vs. domestic equipment and how basic physics underlies some apparent black magic. Also, if you want manufacturers to improve their products, you must first test these, then vote with your wallet.
Part 6, Noise in digital interconnections, was presented by Tom Shevlin. During tests in the new CBC headquarters he found that the available digital format audio interconnect cables had neither the required accuracy and consistency of impedance, nor adequate shielding/termination characteristics, resulting in reflections and noise egress/ingress on long runs. A custom cable was designed, ordered and used throughout the building, with (so far) excellent results.
In Toronto, John Fourdraine.
Industry visionary Leonard Feldman succumbed to a two-year battle with cancer February 14 at his home in Great Neck, NY. He was 66 years old.
Feldman was past AES VicePresident Eastern Region, and was hoping to visit the Toronto Section when he learned of his illness. He was a consulting engineer, writer and lecturer in the field of professional and consumer electronics, authoring seven books covering various audio and electronic topics.
He was perhaps best known as the senior editor of Audio magazine, where his reviews of consumer electronic equipment helped making meaningful comparisons of competing brands possible in a market-driven industry. He was also a columnist for EQ magazine and had several articles published in Popular Electronics, Popular Science and many newspapers. He contributed to a number of consumer electronics industry trade journals
He is survived by his wife, Rayma, two children and their spouses and three grandchildren.
Forward to May 1994
Articles may be used with the Author's Permission. Contact the Bulletin Editor: firstname.lastname@example.org
Editor: Earl McCluskie Assistant Editor: Anne Reynolds Layout Editor: Lee White
The Bulletin is prepared in print by Lee White, and on Horizon and the Internet by Earl McCluskie.