Page 47 - Analog Circuit Design Art, Science, and Personalities

On Being the Machine

“zero“ voltage. and the net current to be integrated in C1 other than from the detec-
tor source is ”zero.“
With the realization of this circuit nearly two decades ago, it was quite possible
to produce, in large-scale manufacture and at quite low cost, a low-level pulsed
integrator whose effective input current was about 0.01 picoamp! At the same time,
the effective output voltage drift was less than 0.1 microvolt per degree Celsius!
But, perhaps more important, this elegantly simple circuit enabled the system de-
signer to realize significant savings in detector costs. Due to the ability of the circuit
to cancel the effects of detector leakage, the construction of the detector could be
based on substantially lower-cost materials and testing.
Now consider a modern digital signal processing system requirement. Figure 5-7
shows the sequence of functions in what might be a high-speed, high-accuracy
mathematical waveform generator. The sequence of events is that a primary com-
puted function undergoes secondary digital filtering, the output of which is applied
to a high-speed digital-to-analog converter, whose output is transferred to a holding
circuit, the output of which is applied to a recovery filter, and thence to an output
amplifier. The problem is to determine the “optimum characteristics” of each of the
building blocks. Presume that a project engineer were to write a specification for
each of these building blocks. Is there, in fact, a set of individual specifications
which. independent of each other, can provide a reasonably optimum technical
economic result?
Assume that the computed function is to cover a frequency range from 0 to 100
MHz, that the rate of computed words is to be about four hundred million per second,
and that analog integrity is to be preserved to the 12-bit level. A multiplicity of
interacting, conflicting requirements arises if the architect project engineer attempts
to assign to a group of designers the task of preparing a specification for each indi-
vidual piece. Characteristics of the recovery filter depend on the characteristics of,
at least, the holding circuit, the digital prefilter, and the computing function. The
characteristics of the holding circuit in turn certainly depend on the D/A converter
and the output filter. The nature of the computed function depends on the algorithms
and the computing capability, and this in turn will be directly related to the capabili-
ties and characteristics of the other building blocks.
How, then, will the project engineer realize a superior solution unless an integrated
sequence of operations can be visualized and internalized? One part of the engineer
must be the computing function, linked to another part which is the digital filter, in
turn linked to yet another part acting as the D/A converter, and so on. The designer
must, in a certain sense, interactively play these internal-self parts, balancing off in
various combinations the possible capabilities and limitations of the individual
functions. The designer must fully understand the error-producing effects of these
interrelationships. The design decisions for each function cannot be explicitly com-

Figure 5-7.
Functional
sequence of a
mathematical Computed - Digital DIA - Holding - Recovery
wave form Function Prefilter - Converter Circuit Filter
generator. I I

28