|A PROGRAMMABLE PARAMETRIC ATTENUATOR
and DIGITAL LOGIC CONTROL SYSTEM
|by Andrew Berliner, Crystal Sound Recording Studios|
As high quality audio systems increase in complexity, super-human demands are being made on the operators of these systems. It is natural for the designer of large recording systems to employ computer techniques to aid the operator in his quest for control as well as to provide more flexibility in operation.
The Crystalab system currently under development and in use at Crystal Sound Recording Studios has combined a 40 input channel, 4 output channel mixer with a digital logic system and a 300 megabyte disc storage memory into a high performance, super reliable creative tool.
Modern electronic attenuators are primarily of the voltage control type, i.e., attenuation is a function of control voltage. Such attenuators, using pulse width modulation or analog transconductance methods, are well known, as are their limitations, namely noise, slew related distortion and temperature instability.
The cornerstone of the Crystalab approach is the development of the Programmable Parametric Attenuator.
Parametric attenuators have been with us as long as electronics. A simple potentiometer is a parametric attenuator; the attenuation being a function of shaft rotation; the ratio of the resistances between the series and shunt legs: essentially an “L” Pad.
The Crystalab Parametric Attenuator incorporates and expands this basic concept. Twelve separate “L” pads are connected in series and isolated by an input buffer amplifier and an output driver amplifier. Each pad section contains two switches which insert it or bypass it into the audio signal flow. The values (dB attenuation) of each pad section are chosen in a sequential binary and Grey coded progression. After considerable experimentation, the chosen values are:
An electronic attenuator when full on can be considered a unity gain device. The change is gain is expressed as “dB attenuation”. Careful circuit design insures the accuracy of each pad section individually as well as additive combinations of any and all sections. Each of the twelve pad sections are addressed by one bit in a twelve bit word. 1,280 different combinations of the twelve bit word provice a gain range from unity to 79.9 dB before off in exact 1/16 dB steps. Pad trimpots mean “step size error” can be made arbitrarily small.
Field effect transistors (FETs) were an obvious choice for the switching elements. Their advantages of high speed, linearity and low “on” resistance, are, however, offset by the charge coupled noise as it changes state (on to off and off to on).
A major design effort was directed toward designing a floating FET switch with minimal gate/channel capacitance effect and high signal level handling capability. In 1975, after almost three years of research and development, a patent, describing a circuit, embodying isolated gate pullup and constant current pinch off techniques was issued to Crystal Industries, Inc.
Singularly the most important ingredient in making this attenuator work, the Crystel-FET Driver, switches signal levels up to ± 12 volts from DC to 100kHz in 2 microseconds. Charge coupled noise is ≤ 90 dBV (relative to input) while linearity is primarily dependent on the FET, 0.01% in “on” mode is average.
Each new version of the attenuator solved some problems and created others. For example, in spite of the care taken to provide minimum charge on the gate, the finite noise generated increases with switching speed. When the attenuator slews, (the operator has moved a fader knob…) the least significant bit (1/16 dB) has the highest switching speed. Organizing the pad values such that the highest switching rate occurs at the front of the pad chain allows the noise, so generated, to be attenuated by an amount equal to the total dB attenuation of the subsequent pad sections.
Zipper noise is produced by a discrete step size change in amplitude and is exaggerated by the finite time discrepancies of one pad energizing and another de-energizing. It is completely undetectible on complex waveforms such as program material, however, on sine waves, well… so by incorporating a zero-crossing detector to limit the switching of pad sections in to or out of the audio path, only to the time when the amplitude of the waveform approaches 0 volts, all zipper noise vanished.
The Programmable Parametric Attenuator as the variable gain element in a professional recording/mixing console offers many unique advantages. In its static mode, at any attenuation (or no attenuation) the audio signal passes through no non-linear or noisy elements. Fifty or more attenuators will track within 1/16 dB over the entire 80 dB range. There is no drift. Slew rates of 125 dB/second with a 1 kHz signal to 2,500 dB/second with a 20 kHz signal are inherent in the design.
Another intriguing application of the Programmable Parametric Attenuator is as the control element of a program-controlled gain circuit (limiter / compressor / expander). Such a circuit could peak limit, maintain a constant average level, or expand the dynamic range of an input signal, or any combination. Since 80 dB of control range is not needed, resolution of 1/32 dB or 1/64 dB may be more desirable.
The Programmable Parametric Attenuator is the ideal tool for radio and TV broadcast applications. The revolutionary significance of the Programmable Parametric Attenuator is that it provides the missing link between analog and digital. It is the catalyst that integrates the vast resources of applied digital technology into the creative analog musical systems of modern recording studios.
This is a data acquisition and management system. In contrast to current automated or “automation-ready” consoles, the “computer” is an integral part of the system. Technically it’s “dumb” because the program is hard wired, but its internal 14.2 MHz clock and command time of 70 ns mean the performance and sophistication of a lunar landing.
Its operation is easily understood by examining the function of its four main sub-systems: Input, Processing and Control, Storage, and Output.
— The Input system translates the commands of the operator into its internal language.
The block diagram of the digital system components and the flow of data between them illustrates the entire system concept: The simultaneous control of volume of 44 audio signal channels in time increments of 1 ms.