Bernhard Boser comments:
When I arrived at Stanford University in fall of 1984, Bruce Wooley just had joined the faculty. He introduced me to the topic of oversampled A/D conversion by handing me the draft of a now-famous paper on the topic by James Candy (“A use of double integration ...”). I must have read the paper at least ten times and still did not understand the thing. So I wrote a little program to check out the ideas and, low and behold, this thing WAS a converter. But how did it work?
It took me many months (more precisely, years) to answer this question. On the path there I met many engineers who flat-out stated that getting 16-bit accuracy out of what seemed to be just a 1-bit converter was plain impossible. I even found a doctoral thesis giving a ‘proof’ that higher-order modulators could not work!
In good engineering tradition I proved the skeptics wrong not with a piece of computer code but with an actually working silicon example. In designing that part I was faced with another set of questions: Only a 1-bit ADC (that is, a simple comparator) is needed, but what about its accuracy? Are the offset, hysteresis, and noise specifications at the 16-bit or 1-bit level? What about the requirements for the other components? By now well researched, these questions were most confusing in the mid-1980s.
When I got my sigma-delta modulator working I saw an opportunity to help the community by presenting my experiences and conclusions in writing so that others would not have to take my arduous (and rewarding!) path. I hope I succeeded in helping some readers grasp the concept of sigma-delta modulation a little more quickly than I did. Of course, the results presented in my article have long been exceeded, refined and extended.
I blush when I think that my paper also contains at least one inaccuracy when it states overly modest gain requirements for the amplifiers. Fortunately, engineers never stand still and much has been written on the topic since. Still, it is rewarding to see that authors still cite (and read?) my publication. It is a big honor to appear on the Journal’s list of most-cited papers.

Bruce Wooley comments:
The invention of the transistor in the mid-20th century was followed by the rapid growth of interest in digitizing information for communications, storage and signal processing. In particular, attention began to focus on the use of digital transmission and switching in the voice telephone network. Analog-to-digital conversion thus became an increasingly important function in the communications infrastructure in both Europe and North America.
Prior to the emergence of large-scale MOS integrated circuits, A/D converters for applications in communications were typically expensive, rack-mounted systems that were shared among multiple channels in order to amortize their cost. However, the emergence of MOS VLSI technology and the ability to integrate thousands of digital gates on a silicon chip motivated efforts to find a means of exploiting this capability to reduce the cost of digitizing analog signals.
Oversampling approaches to digitizing analog signals, which combine sampling at well above the Nyquist rate with feedback and digital filtering, offer a means of exchanging resolution in time for that in amplitude. Thus, they are an effective approach for exploiting the component density and speed of scaled VLSI technology to avoid the need for complex, precision analog circuits.
Oversampling modulation for analog-to-digital conversion first appeared in the form of delta modulation, which was introduced at about the same time as the invention of the transistor. Ten years later an approach referred to as sigma-delta, or equivalently delta-sigma, modulation was introduced. As opposed to delta-modulators, which encode the rate of change of a signal, sigma-delta modulators digitize the signal itself while shaping the resulting quantization noise in frequency so as to push most of its energy outside the signal band where it can be removed by digital filtering. Consequently, they are an especially robust means of digitizing signals that require only a very limited amount of analog circuitry.
Bernhard’s research focused on devising architectures and circuit implementations best suited to implementing precision A/D converters in scaled digital VLSI technologies. As described in the JSSC paper, his work focused on the design and realization of second-order sigma-delta modulators, which avoid the large spurious noise tones inherent in first-order architectures, as well as the stability concerns associated higher-order loops. He established behavioral models for each of the functional building blocks in a second-order modulator that included the impairments associated with practical circuit realizations. Extensive simulations were then used to establish the circuit design criteria for each of these blocks, and an experimental modulator was integrated to verify the validity of these criteria.
Because the performance of oversampling modulators can only be assessed by examining long data traces, they cannot be efficiently simulated at the circuit level. As an alternative, Bernhard developed an event-driven functional simulator for mixed digital and analog sampled-data systems (MIDAS) to accomplish this task. This simulator was a critical component of Bernhard’s research and is still in widespread use today.

The Impact
Noise-shaping modulators now play a dominant role in the realization of A/D converters, not only for voice telephony, but also in a broad range of digital audio systems. Moreover, they are beginning to make inroads into applications with higher bandwidths.
The papers resulting from Bernhard’s work continue to serve as a foundational reference for insight into the simulation and design of noise-shaping oversampled A/D converters at both the architectural and circuit levels. His work clearly established the critical circuit requirements for designing such modulators, and it pointed the way to the effective modeling of such systems at the behavioral level. The second-order modulator design he proposed and demonstrated has subsequently served as the key building block for the realization of higher-order, cascaded sigma-delta modulators that have been used to digitize both low-pass and bandpass signals with bandwidths of several megahertz and center frequencies as high as 20 MHz. Moreover, the software he developed remains a valuable tool for assessing the behavior of mixed-signal systems in which feedback plays an important role.