by Eric Vanhaelen, free lance sound engineer († 2000) Designer of Instant Calibration Converter and Arithmetical Serial Direct Converter ©Copyright 1997 Eric Vanhaelen | |
Antwerpen, July 17th 1999. Brief historical review. This review is intended for those who were born later and missed part of the fun. The High Fidelity movement started shortly after the World War II. For years the quality of electronic equipment stayed bad, and that was due to several factors.
Valves had excellent transient response, due to the anode voltages of 250 V. The almost a-material electrons escaping from the cathode were subjected to such accelerations that the anode of end valves glowed deeply red. For the rest non-linear valve characteristics, provided distortion. The first major step in improvement was the Williamson-amplifier with 20 dB of negative feedback. The THD was thereby reduced to 0.1% which has been a standard for many years. The appearance of LP's and 45 RPM singles, made another improvement in recorded music, but even the best PU cartridges who were excellent in frequency-response, showed bad transient response. The Baxandall tone controls tried to give a solution to this phenomenon. Within the scope of this it must be told that bass response in that time was also impaired. Due to the fact that many thought that a stabilized supply was redundant. But it must be stated that in constant velocity conditions of the applied signal the basses were prevailing in amplitude and made the supply go through its knees; resulting in bad bass response. Also the loudspeakers were deficient in that aspect. We had to wait till E. J. Jordan (Goodmans Industries) proved in his paper with differential equations and I quote: "Hence it is the velocity and not the displacement which is directly responsible for the radiated power". From that moment on loudspeakers all over the world began to get better We are still benefiting today from the know-how of Mr. E. J. Jordan and owe him a tribute. Because if it weren't for him, it wouldn't be possible to enjoy a tremendous sound experience in a Dolby surround equipped theater. In between time the transistor appeared, first the germanium then the silicium. It was a practical improvement in circuitry design, but they were hard to stabilize in temperature and had bad Slew rates, which affected transient response. As time went by all these inconveniences were overcome. And as a sort of nostalgia for the performances of valves, birth was given to the field effect Transistor, the FET. Nowadays in the second generation of IC's a Fet is included mostly in the input and towards the years 1980 slew-rates of operational amplifiers had been raised to some 12 V / microsecond. Which was quite satisfactorily for audio purposes. Actually for special purposes op-amps are available with slew-rates as high as 500V/ microsecond. In the year 1980 at the London Convention, AES gave much attention to digital technology. The CD appeared and this technology was in tended to improve noise, distortion and transient response. But it must be emphasised that before the digital Boom came in 1981 Analogue had already kept up with the situation. Proof of this is given by the 16D technology we are going to introduce you to now. The heart of the system is a square wave generator modulated in Pulse width by 16 varicap diodes in parallel. For good operation the varicap diodes should be matched, but this is not strictly necessary; the varicap diodes of the prototype weren't. Note that the transfer curve is not very linear but the presumable curve of 16 Diodes in parallel shows to be more interesting as drawn here by on the same scale. The curve is very wide and can be attacked at several points. This will determine whether you are modulating in pulse width or in frequency. Near Cd = 38 pF per unit, with a bias of approximately 800 mV DC PWM is prevailing in conjunction with the other parameters of the circuit. Further on the curve with more DC bias you slide towards a condition where FM is prevailing. But bear in mind that theoretically the two modulation forms are present at the same time and are complementary. In the PWM condition the frequency of the carrier will be about 250 kHz (BB204) and the shape of the square wave will be symmetrical. This is the best point to operate with regard to THD. Also this frequency, which is quite low in comparison with the frequencies involved with a 16 Bit sampled data procedure at 48 kHz; offers attractive possibilities to record on a modified DAT or CD writer. A 16 Bit sampled data signal puts an unnecessary strain on the bandwidth requirements of the recording medium (actually we can talk about an bandwidth explosion) resulting in higher speeds required for the carrier or restricted play time for a DAT or CD for the same format. Further comparing this 100% analogue system with a digital system, it should be born in mind that the heart of an A/D converter is a comparator and that a comparator is a HYBRID. It is binary on its output but analogue on its inputs. Its architecture inside is for 80% analogue and non- linearity's in the analogue parts can cause quantisizing errors. And furthermore, the input sensitivities of a comparator nowadays are still low and A/D conversion is still restrained to the High end of the line. In general it must be stated that A/D converters need architectures that makes it expensive and unnecessarily complicated and vulnerable. Bearing in mind the saying: "Simplicity is the mark of the true", I developed my system with an extremely simple architecture. When you understand thoroughly our system (see diagram) you will see that increasing the amounts of diodes in parallel over 16 will result in an increase of input sensitivity and improve the THD figures, with no increase of noise, because Varicap diodes are virtuaIly noise immune. With 16 diodes the input sensitivity is 100 mV for 1V P-P at the output of the decoder, that means that the modulator provides an amplification of some 20 dB, since the demodulator operates at unity gain. In practice this procedure is a challenge for the manufacturers of varicap diodes for developing one diode with the necessary capacity and even improved characteristics with regard to THD figures which would get completely under control. When you apply for a completely detailed diagram and build it, you will hear that it operates perfectly as it is now and that it exceeds the performances of a 16 Bit sampled data digital system already. To support the validness of this system I will use a mathematical expression:
The definition of a mathematical limit is: a figure tending towards another figure, but never reaching it, except in infinity. Some manufactures of the Far East published some years ago a stroboscopic analysis of one of their A/D converters. At a certain value of the analogue input it showed on an oscilloscope a ONE BECOMING ZERO and a ZERO BECOMING ONE. This situation which can be clearly observed on an oscilloscope shows the need for an intermediate value, which is neither ONE nor ZERO. Thus exposing clearly lack of resolution, which only analogue can provide, assuming that analogue is achieved with the same discipline that was demanded for a digita1 approach of the problem.
As early as the fifty's following modulation forms ware known:
All these modulation forms were purely analogue and I chose the last one for my system. The a-material nature of the carrier plies itself to the finest details of the modulation, whether it be speech or music. In the 80's finally came the sixth modulation form: DIGITAL SOUND, which was called officially PCM - Pulse Code Modulation. It was a samp1ed data technique. This means that several thousands of samples per second were taken and transferred into a binary code. The modulation appeared in terms of sequences of ones and zeros. 12 years ago when we were intensively occupied with A/D converters, it appeared already to us that modulating a FM transmitter in PCM would be very attractive with regard to bandwidth saving aspects of the procedure. But we gave it further no thought. Nowadays the topic is DAB: Digital Audio Broadcast on short wave. It seems to offer attractive perspectives to combat fading in short wave. But it should be borne in mind that once the fading is 100%, you can forget about DAB, for hours or even the whole night; until the ionospheric conditions return to favorable again. So far as for DAB. We can finally conclude this chapter with the statement that digital sound is a system that gives and demands. Our system, limited to recording, on condition that it receives enough attention to overcome some difficulties with regard to adaptation to DAT and CD, is a generous system that doesn't DEVORE bandwidth. Since with 16D PWM there are no side bands appearing. Note that there is no question of aliasing problems at all. For completely detailed diagram and easy tuning procedures as well as supplementary information, apply in writing to:
Lange Vlierstraat 18, B-2000 Antwerpen, Belgium. Or by e-mail: eric@vanhaelen.be
to a DAT carrier or a CD medium. Since the 16-D processor is a non-sampled two-track device with excellent left/right crosstalk figures, it requires a two-track recording medium. DAT is a one-track machine with sampled left and right data successively recorded in a multiplexed way. It appears that a DAT should be completely redesigned to carry two independent tracks. Brilliant spirits among you may be found to give this problem a practical solution. It would be worth the effort since the idee of a DAT for recording high pulse rates is excellent and further more the format is very attractive. We are looking forward to suggestions of some among you as a return favor. Actually this is a call for cooperation and teamwork. For recording on CD, the problem could be solved more easily. CD has only one multiplexed track. Incorporating the 16-D processor will call for two simultaneous tracks. Another solution would be to record only one track in MONO and record L-R on a subcarrier of 19 kHz like it is done nowadays with Stereo FM. For demodulating the classical FM demodulating matrixes would do. Even redesigning both media for the purpose, couldn't be a major problem with the actual state of the art. Don't forget the advantages of 16 D Technology
Short description. 16-D technology is a 100% analogue audio processor, with better performances than 16 bits digital devices as well on frequency range as on dealing with big and small signals. It has been accomplished on purpose with active components of the 80's generation (LF353, LM365 ...). The modulator is a square wave generator modulated by sixteen varicap diodes in parallel. A in frequency or pulse width modulated square wave generator can be used to write on CD's with behold of 80% of the manufacturing technologies who are actually current. Only the digital hardware is not compatible and has to be replaced by amplifiers proper to the system. The demodulator is a fourth order Butterworth Low Pass Filter. Demodulating is achieved on a 24 dB per octave slope. This is like FM flank demodulation like it existed in the pioneer days of FM. The demodulator doesn't discriminate but this is not necessary because at 200 kHz (momentarily) there is no noise. Connecting the demodulator directly to the modulator provides a sound image to make a CD blush. Technical specifications:
|