For the PDF version of this article, click here.
In hindsight, it sometimes seems that a great invention evolved so logically from the existing technology and so clearly answered pressing needs that its development was “inevitable.” That very word is the one Bob Mammano uses to describe the creation of the first pulse-width modulation (PWM) controller — the SG1524 — that he designed at Silicon General in 1975.
After all, switching power supplies (also known as switchers or switched-mode power supplies [SMPSs]) were being built using discrete components at least as far back as the 1950s. And by the early 1970s, some of the individual analog and digital functions needed to perform PWM — the control technique commonly used to regulate the output of the switching power supply — were becoming available in IC form.
“An accident in time” is another way Mammano describes the development of the SG1524. “Everybody knew that switchers were a potential marketplace,” says Mammano. “It's just that [Silicon General] managed to get there first.”
Mammano recalls that both Motorola and Fairchild had early versions of ICs that might be used in the switching power supply application. Another chipmaker, Signetics, had introduced the 555 timer and some engineers were using that part in switching power supply designs.
However, Mammano observes, “No one had actually put all of the blocks for a switching power supply together on one chip. The main reason we did it was that we had a customer that was interested.”
That customer was Teletype Corp. of Skokie, Ill. Teletype was in the process of converting its big mechanical teletype machines into electronic versions. The new teletype designs didn't have much space for a power supply, so they needed a switching power supply. But even then, they didn't have room for all of the discrete components required in the complicated control circuitry of a switcher. Although Teletype had tried to build a switcher using a hybrid product, that approach was very expensive. “They were convinced that if someone could do it all for them in a single silicon chip, it would help them a lot,” says Mammano.
So Teletype asked Silicon General to build a single IC that would integrate the control functions for a switching power supply. When presented with this request, Mammano and his colleagues determined that there were no inherent reasons why these functions could not be integrated on chip. “All the individual blocks seemed compatible. That's why I say it was inevitable,” explained Mammano.
However, such an assessment belies the fact there were serious technical challenges in putting all the PWM control circuitry on a single chip. At that point in time, integrated circuit design fell into two neatly divided camps — analog and digital. The techniques for building logic circuits like flip-flops in a gold-doped digital process were well understood. So too, the industry knew how to build op amps and other analog functions in a bipolar semiconductor process. But to develop a PWM controller on chip, it would be necessary to build both analog and digital functions in the same semiconductor process.
The voltage and current levels required of the analog functions demanded that the chip be built in a bipolar process. So Mammano explains, the main challenge in developing the controller became, “How do we build the digital circuits on an analog process? To do that we had to go back to some pretty archaic digital logic technology that had been done with basic bipolar technology.”
To build the SG1524, Silicon General used what's described as a 10-µm bipolar process with a single metal layer. This process could handle high voltage and relatively high current, and could implement both npn and pnp transistors for level shifting. “Another benefit of using the bipolar process was that we learned to combine two different junction diodes to get a good (for that time) temperature-compensated zener diode reference, a critical part of any voltage regulator product,” says Mammano.
And while designing each of the blocks within the controller was not easy, there was the additional challenge of putting all the functions together on one chip with just 16 pins of I/O. That challenge fell to IC layout designer Steve Jensen whom Mammano credits as making a major contribution to the development of the SG1524. Mammano estimates that it took about a year to go from concept to first silicon.
In the end, the SG1524 broke new ground in several ways. By integrating the PWM control functions on chip, the controller removed much of the complexity from SMPS design. Switching power supplies had long been valued for their greater efficiency, their lighter weight and smaller size when compared with existing linear-transformer-based power supplies. However, SMPSs had the drawbacks associated with greater complexity. That is, they cost more and were less reliable than linears. (They also generated more noise than linears, but that aspect of performance would be addressed in the application.) As the first monolithic PWM controller, the SG1524 greatly lessened the cost burden of building a switcher, while also improving reliability over discretely implemented PWM controllers.
Because the SG1524 demanded the monolithic integration of analog and digital functions, it became not only the first PWM controller, but on a broader scale, what Mammano believes to be the first truly mixed-mode integrated circuit. Mammano's controller design also pioneered the implementation of protection, as well as control, on chip. This combination of protection and control features became a mainstay of all future power controllers.
Although the specific feature of current limiting had already been used in linear regulator ICs, the SG1524 was unique in that it limited output current on a pulse-by-pulse basis, a requirement for switching regulators. And while the concept was first applied in the SG1524, Mammano notes that the techniques used to perform pulse-by-pulse current limiting were much improved in later devices.
Although the SG1524 would take cost out of the SMPS design, the device itself was not cheap. Those first controllers were priced somewhere in the vicinity of $5 to $10 per unit in 1976. However, as semiconductor-manufacturing techniques progressed, the cost of producing a PWM controller fell dramatically. Rich Valley, vice president of Texas Instruments' System Power Management business, notes that the same function found in the SG1524 is available for less than 25 cents today.
As with many IC designs, development of the SG1524 required give and take between the chipmaker and its customer. “One of our tasks in designing the SG1524 was to take Teletype's specific requirements and to generalize them so that the same product could be applicable to the rest of the world.” According to Mammano, the Teletype engineers asked for features such as a switching frequency up to 100 kHz, the ability to drive external power devices (bipolar transistors then) with reasonable amounts of current, reasonable voltage accuracy, and the ability to close the feedback loop with an error amplifier that had a reasonable amount of gain.
In the end, the SG1524 not only satisfied Teletype's requirements, it met the basic performance requirements of many other SMPS and switching regulator applications. But gaining widespread acceptance for the first PWM controller did not happen overnight. Silicon General was a relatively small company, and many customers would not allow the development of their systems to ride on the ability of such a company to deliver a critical single-sourced device.
In time, there would be proven high-volume production on the SG1524 at Silicon General and, later, the sought-after second sources. But before any of that could happen, potential users of the SG1524 needed to test out the device in their designs.
Making a Big Splash
As Mammano recalls, he and his colleagues at Silicon General did not know how great an impact the SG1524 would make. However, they did believe that power supply designers would be interested in getting their hands on the very first PWM controller chips. Anticipating that interest, the company planned a “big, splashy introduction for the SG1524.”
Mammano himself introduced the device in June of 1976 at the PowerCon convention, where he presented the first technical paper describing the chip's operation. Mammano came to the conference with “a bucketful” of controller chip samples in hand, he says. After presenting the paper in which he described the new controller, Mammano offered samples to the conference attendees. At that point, Mammano recalls, “Half the audience charged to the front” to get samples of the chip.
In general, there was a widespread interest in the SG1524 and the company gave out thousands of samples to customers. Years later, many of those customers told Mammano they had built successful power supply designs with those early chips. But Mammano readily admits there were “lots of complaints” about the chip's shortcomings. And Silicon General made many post-production design changes to fix problems that hadn't been anticipated in the original design. In some instances, problems identified by customers led to innovative improvements in the controller design.
“One customer found out that when he turned the power supply on very slowly, you could get into a situation where both outputs, which were supposed to alternate back and forth, came on together,” says Mammano. “And that was a pretty good recipe for blowing up the power supply.”
To prevent this problem from happening, Mammano went back to the lab and designed what became the first on-chip undervoltage lockout (UVLO) circuit. The UVLO did not allow the controller to turn on before the supply voltage had reached the needed minimum level. (The SG1524 was designed to operate off of a supply above 5 V.) Later, Silicon General spun off protection circuits developed for the SG1524 into the SG1543, a dedicated power supply monitoring device.
Convincing engineers of the controller's performance was one matter, but convincing customers to trust their designs with a single-sourced device from a small company was another. In the beginning, Silicon General had just a small initial order from Teletype and small orders from others who used parts for prototyping or limited production runs.
However, acceptance of the SG1524 seemed to reach a turning point when Silicon General convinced the power supply design group within Digital Equipment Corp. (DEC) to adopt the SG1524 in its designs. The order from this major computer maker represented the first big production contract for the PWM controller.
DEC also inspired a second-generation version of the PWM controller, the SG1525. This device incorporated a number of improvements such as a bigger output stage capable of handling more current. This stage consisted of a totem pole output driver, which could be used to drive the power MOSFETs that were just then becoming popular. The MOSFET gates required high pulse currents in both directions to turn them on and off. The totem pole output on the SG1525 could perform that task nicely.
Building the Business of Analog and Power ICs
The opportunity to design a groundbreaking chip like the SG1524 might not have come Mammano's way were it not for his involvement in founding Silicon General. As its first vice president of engineering, Mammano along with Verda Hinkle, as vice president of manufacturing, and James Johnson, as president, formed Silicon General, because they perceived a need for analog products targeting military applications.
In his previous job at Autonetics, a division of North American Aviation, Mammano designed circuits for military projects. As a designer, he found that most semiconductor manufacturers were primarily building integrated circuits for the computer market. As a result, their emphasis was on developing logic circuits and they offered little in the way of analog functions.
“We were looking for analog products to go in our power supplies — voltage regulators, amplifiers and various specialized controls,” Mammano explains. “But we weren't getting much support from the semiconductor people, even though we were willing to pay a lot of money for them. We just didn't use many devices.” Based on that experience, Mammano and his colleagues concluded there would be good potential in a company focused on developing analog chips for the military.
However, soon after Silicon General was started in Westminster, Calif., in 1969, the military aerospace industry went into a tailspin, taking with it potential orders for analog ICs. As a result, Mammano and his Silicon General cofounders refocused the new company on developing power management products for commercial markets.
Mammano's experience at Autonetics in the early 1960s not only inspired the founding of Silicon General, it also introduced him to the concept of switching power. One of his assignments at Autonetics was to design a switching power supply that would be used to power a militarized airborne computer. A switcher was needed because a linear supply would have been too heavy.
Although the SMPS Mammano designed was significantly better than a linear in terms of the watts per pound, the switcher itself would not qualify as light by today's standards. The power supply was based on a magnetic amplifier design that employed many toroidal transformers. With the supply operating at a switching frequency of 20 kHz to 40 kHz, these transformers required large, hand-wound cores. In addition to its significant mass, the SMPS was also quite expensive and a look at the semiconductor content in that supply reveals why. Mammano recalls that the design employed silicon bipolar power transistors made by Texas Instruments. Those transistors cost $125 each.
Through this experience at Autonetics, Mammano learned the underlying control principles he would later be asked to implement on chip. But it's also likely that his work on these complicated, heavy and expensive SMPSs drove home the point that here was an application that could truly benefit from silicon integration.
No doubt these same lessons were applied again when Mammano, Hinkle and Arthur Bruno, a former CEO of Silicon General, left that company to create a power IC division within Unitrode Corp. Already established as a manufacturer of power discretes, Unitrode wanted an IC division to complete its power semiconductor portfolio when the threesome set up shop for the IC division in 1981 at a facility in Merrimack, N.H.
At Unitrode, Mammano took on more of a management role. Consequently, his focus shifted from designing new power circuits to helping define them. Mammano guided product development on new power ICs that would, like the SG1524, establish trends in the industry. For example, under Mammano's guidance, Unitrode introduced the first PWM controllers with current-mode control. Two of these controllers, the UC1846 and UC1843, ultimately became two of the most popular power supply controllers in the industry.
From Designer to Teacher
Rather than designing the new chips himself at Unitrode, Mammano served as mentor to the next generation of power IC designers within the company. However, his role as a teacher extended well beyond his company's engineering lab. Soon after joining Unitrode, Mammano began participating in company-sponsored seminars that addressed power supply design from both the power control side and the power device side. When Texas Instruments (TI) acquired Unitrode in the late 1990s, these seminars continued — a testament to their popularity. Through the years, Mammano has continued to teach young power supply designers in these seminars.
At a time when many of his contemporaries have retired, Mammano continues to play an active role in the power electronics industry. He still teaches seminars, and as a staff technologist helps to shape power IC development, even occasionally meeting with customers. Mammano also participates in industry forums such as the energy efficiency committee within the Power Sources Manufacturers Association. This group helped the U.S. Environmental Protection Agency to develop its energy efficiency standards for external power supplies.
Mammano continues to play a role in efforts to advance the state of the art in power supply design. Recently, Mammano has been involved with TI's efforts to develop power ICs that employ digital control techniques. Mammano comments, “This is a new challenge for us. It's an opportunity for us to develop many interesting new products, which we think will mean a lot to customers.” But, he explains that this opportunity comes with some concerns because digital power control adds complexity, which in turn requires use of a more-dense semiconductor process. That requirement adds to the cost of fabricating the chip as well as to the cost of developing it.
Mammano sums up the challenge of digital power control as one of “defining our products in a way that represents significant value to a broad base of customers so that we have both the volume in the product to drive the manufacturing costs down, and the economies that result from living within the process limitations.” No doubt, these are issues Mammano has confronted many times in the past as he worked to develop the ICs that have paved the way for this newest generation of power controllers.