News and speculations about the future from the other side of the ocean
by Dr. John D. Bullock jr., JDB-Systems, 7622 Kedvale Ave., Skokie, IL 60076 U.S.A. (received in Jan-1991)
Many software developers and computer scientists are also audiophiles. With the advent of the digital recording techniques for the consumer market like DAT, CD, and the new DCC (digital compact cassette using audio compression and encoding) the consumer is more than confused. Some hints on selection criteria of audio purchase could be helpful.
I would suggest the following overall stategy for future audio purchases as a guide. I should warn you that most audio stores and retail store salesmen (or women) will not be aware of the philosophy that I will now state. This is because they are generally not up-to-date on the market directions planned by manufactorers. They must respond to a customer in the present (to keep sales rolling) by selling what they have today and not what they will have in the future.
Audio in the 90s will further integrate with video (i.e., a picture), and the standard audio format will be four channels using a Left, Center, Right, and Surround (or back channel) arrangement. The reason is that the generic two channel stereo format has essentially run out of gas from a marketing and creative standpoint, and a multi-channel concept is the obvious evolution. What most consumers, including audiophiles, forget is that stero was a compromise from the beginning. There is nothing sacred or necessarily accurate about stero with two channels. This is the reason HDTV research and development in Europe (and the USA) specifies four channels of audio for the new standards. This is a minimum number with some consoderation given to eight channels as the system evolves into the year 2000.
The format which defines multi-channel audio for the next ten years is called Stereosurround (SS) of which Dolby Stereo for the movies and Dolby Surround for home playback of movies are subsets. The best decode electronics for playback specifies Stereosurround or Dolby Pro-Logic (DPL) decoding. Current versions of the SS format are in matrix form which combines 4 discrete channels into 2 then extract then again to 4 during playback. This can be accomplished because the two channel format of stereo is a subset of Stereosurround encoding assuring compatibility. Sterosurround encoded program 'folds' to stero and then to mono correctly at the user end of chain. And most important, several robust storage and transmission media are already available for stereo. These include video tape and DAT, CD, Laser video disk, BTS broadcast. However, the matrix two- channel form is not required for program distribution and discrete L, C, R, and S program will likely be delivered on specialize media in the future.
In Europe there is a following for 'ambisonics'. This form of encoding can theoretically reproduce the correct sound image (spacial locations of sources in the sound field), but the loudspeaker arrangement is not generally desirable at positions above the listener in the front. In addition, the listening area for correct imaging is very limited as in the case for stereo. This dependance on precise loacations in the listening space makes ambisonics (in general) not very interesting (since it lacks enough channels without using enhancement) nor robust. The LCRS format of SS can be effects driven as in movie program which is intentionally bigger than life, or artistic in presentation such as the new Spies II CD from Telarc.
The Stereosurround format requires three front loudspeaker of the same type (all alike) to achieve uniform sound fields and good imaging. Generally the front speakers should be split into satellite and subwoofer systems since full range systems (30 Hz to 20000 Hz in one unit) are too large for three font speakers in average size rooms, and a large system at the front_center position can be difficult to position with a large screen for video playback. This is the reason 'sub-woofer' systems have evolved in combination with satellite loudspeakers for the 80 Hz to 20000 Hz range. For best performance the subwoofer should cross-over to the full range system between 80 Hz to 100 Hz with fourth order slopes on both high-pass and low-pass filters. The best systems accomplish this cross-over electronically.
Another note on electronics; the better quality decoders provide six output channels: left, center, right, left surround, right surround, and subwoofer, which drive amplifiers and then loudspeakers. A master level (volume) control should work on all channels simultaneously, but each channel should be capable of a small level 'tweak' adjustment to align the system level so that the total SPL over the primary listening area in the room has equal contributions from all channels. DPL decoders require a build-in shaped noise source for this function which rotates between channels automatically. All power amplifier channels are set at equal gain unless room errors can not be controlled by decoder tweaks. The finest SS systems may use room equalization on the individual channels but only after the decoder (LCRS outputs) and never on the decoder inputs.
One final note on loudspeakers for SS and/or stereo: the front speaker system architecture can be important. By architecture is meant the physical arrangement of transducers on the front baffle and crossover topology. The preferred layout is that of the vertically symmetric array which in a two- way system means mid-woofer--tweeter--mid-woofer in a vertical line. This tends to make systems which are usually referred to as 'towers', although all tower systems are not necessarily designed correctly. With properly designed 4th order crossovers between the loudspeaker system drivers the sound field beam remains on axis throughout the frequency range. The power response (acoustic energy into the space) may change depending on other design criteria but the primary beam will remain on axis.
The above criteria produces superior image performance, with the best results achieved by positioning the left, center, right loudspeakers in an arc (horizontally, viewed from the top or room ceiling) pivoted on the radius from the center listener location, and at the same position for all three speakers vertically up from the floor (at equal heights). The vertical height should be ear level for the sitting listeners or slightly above (between 5 to 10 degrees). The left and right loudspeaker for SS should be positioned wider than the stereo standard of 60 degrees total to about 35 (or 40) degrees for each loudspeaker from the center line drawn to the center channel loudspeaker. This depends on the room size, and do not move the left or right speakers closer than four feet to the side walls. Sound absorbing panels on the front and side walls near the front speakers can be useful; in reducing unwanted reflections in the 10 to 30 msec. range which can be detrimental to good image performance.
Some final notes on digital electronics. The CD and associated optical technology only improved on the weakest link in the high fidelity audio chain, the vinyl phonograph record and stylus. Even the best performing cartridge devices and arms are 1% distortion and 60 dB dynamic range. The CD can easily achieve 90 dB dynamic range and better than 0.05% distortion. But digital audio is still not good enough for the full dynamic range capability of recording microphones and well designed analog electronics. Digital must maintain average levels at least 20 dB to preferably 30 dB below 0 dB (all ones) to contain peaks, which moves the recording media back to 60 dB above the noise and grunge using 16 bit A/D converters. This dynamic range is not enough.
The digital optical interface whith 16 bit quantization is a Japanese invention for marketing reasons not necessarily delivering great quality for the cost by eliminating the alleged failings of analog connection technology. Standard RCA phono connectors and well shielded cables provide inaudible interfaces. Gold on connectors is a good idea for practical reasons, but outrageous expensive interconnects, especially for loudspeakers are marketing and sales hype (i.e., lies). The hype is not substantiated by scientific facts. In other business the perpetrators would go to jail for fraud.
When all of professional audio moves to the AES/EBU professional digital interface standard by 1994 and better than 20 bit A/D and D/A converters are available for reasonable cost, then digital audio will make good on its promise with better than 110 dB dynamic range. There is still room to grow since the AES/EBU digital standard can handle 24 bits for audio into the 21st century. The CD must be updated, but DAT can handle more bits now. The moral of the story: do not spend much money on digital audio in the early 90s.
In the USA, Windows is where the marked is moving. The primary business software players, namely Lotus and WordPerfect, will deliver Windows compatible versions by Summer '91. Microsoft will also take a substantial share of the marketplace with their "Windows suite" of applications programs in spreadsheet, word processing, presentation graphics, and design. This must represent 80%+ of the computer marketplace in the USA. Software on DEC VAX hardware pales in comparison to the sales to be generated by 1995 using the Intel architecture CPUs (i386/486/586).
More good news for PC based users. It now appears likely that AMD (Advanced Micro Devices) will make a compatible i386 processor by Spring '91 which will solidify the market and decrease the PC/AT and EISA systems cost 20 to 30% by Fall 1991. Intel has sued them of course, but they appear likely to lose since Intel cheated on a second sourcing deal with AMD made in the early 80s. A 33 MHz 386 with 64 KB cache, 8 MB, and a 150 MB ESDI disk, has about 10 times the processing power of a MicroVAXII for less than half the cost. (This basic system is $2400.00 without video in the USA.) Its all over now but shouting.
As a PDP-8 user in the 60s, DEC will always deserve a tip-of-the-hat from me, but there is trouble in Camelot. DEC deserted a large part of their installed base (which will not return) in trying to compete with IBM. DEC will survive only because of their installed base in VAX systems. But they will shrink over the next 5 years at least 20% since this is still small compared to IBM. (That's what they get for not giving RT-11 users RT-32 on MicroVAX for PC prices.) Even DEC has delivered a 486 system, but they have not and will not commit to really competing on the desk-top. This is why they will shrink at least 20%.
Also predicted is that pure RISC boxes will run out of gas. RISC appears to be much more memory intensive than originally predicted. The i860 (not true RISC), i486/586, and 68040/68050 will be the CPUs of choice until 2000 by providing desk-top systems with 30+ MIPS performance for under $10,000. SPARC must be cloned to be relevant. The best buy right now for Unix is the Next cube with color (25 MHz, 68040 CPU). One can purchase a "loaded" Next system from several University computer stores in the USA for about $6000. Going to college at an old age can have several benefits. It is still guess work whether or not Next will survive.
Judging from what goes on in several computer magazins I read, Modula-2 is essentially dead in the USA. C is the dominate with C++ coming up fast to eventually take over new development. The big guys, DEC, HP, and SUN have already delivered C++ compilers and libraries, and standardization process is underway for C++. With Turbo C++ available and Microsoft about to deliver their object-oriented environment by Fall 1991, the die has been cast. These developments and applications on Unix and DOS workstations and enviroments using C account for about 95% of the market in the USA.
If Borland delivers Microsoft Windows 3.0 support by Summer '91, they will probably dominate Windows development under the new MS-DOS 5.0 (May 1991). This is because Microsoft C Version 6.0 with Windows Software Development Kit (SDK) is now very slow using Programmers Work Bench (PWB) environment. Incidentally the rumor is that DOS 6.0 in Fall 1992 will have extended memory support and virtual memory built into the operating system along with specific switches for i386/486 operation. This will be delivered via Windows 3.2 or WIN32 for short. DOS will become what OS/2 never will because three letters from the alphabet will always goof it up. The letters are IBM.
I saw the mention of the ISO-complex extensions to Modula-2 in a Journal of Pascal, Ada and Modula-2 a month ago, and as an engineer/scientist, I was elated. I suspected that Modula-2 compiler writers would not approve. They are wrong of course. Function calls instead of built-in operators are far too complicated and contribute to errors and maintenance problems. (Granted C++ does and will generate much larger programs.) For commercial business applications C will slowly dominate in the future, and C++ will move to dominate scientific applications if Fortran 90 porting proves particularly difficult. One can write "Fortran-like" math code in C++ with many more benefits.
The only chance Modula-2 has to be somewhat accepted for at least scientific applications is to add complex numbers and object extensions in the form of class structures. As much as the Modula-2 camp of computer scientist would like Modula-2 to remain a small, elegant, and relevant language, without these extensions at least, ISO-Modula-2 will have no chance in the marketplace, and it will remain an academic exercise. These changes are required for survival of the language since they will give programmers of the future the tools necessary to build complex but useful software which can be read by normal humans with less work and maintenance. Less work is the name of the game, and to some extent C and C++ fail on this point. It is up to Modula-2 to come to the rescue. Oberon will never be relevant, but I will always have a soft spot in my programming heart for Modula-2.
IMPRESSUM: The ModulaTor is an unrefereed journal. Technical papers are to be taken as working papers and personal rather than organizational statements. Items are printed at the discretion of the Editor based upon his judgement on the interest and relevancy to the readership. Letters, announcements, and other items of professional interest are selected on the same basis. Office of publication. The Editor of The ModulaTor is Günter Dotzel; he can be reached at mailto:[email deleted due to spam]
Home | Site_index | Legal | OpenVMS_compiler | Alpha_Oberon_System | ModulaTor | Bibliography | Oberon[-2]_links | Modula-2_links |