Bruce Carsten: Lifetime Achievement Award Winner

Sep 1, 2006 12:00 PM
By David Morrison, Editor, Power Electronics Technology

Exploiting Active Clamp Reset

In 1977 while “temporarily unemployed” (i.e., consulting), Carsten worked on another design technique that would later become very popular in the industry. In a typical PWM single-ended forward converter, the transformer off-voltage would be equal to the on-voltage. “But what,” Carsten asked at that time, “if we lower the reset voltage to a value just sufficient to reset the core during the OFF period?”

Carsten discovered that this approach, the active-clamp resetting of the transformer flux, would allow the power converter to operate over a much wider (than usual) input range for a given transistor rating. Moreover, it would improve efficiency and reduce output rectifier reverse voltage.

Carsten would need both these advantages when designing a power supply for one of his customers, EPIC Data Industries. That company was developing an inventory control system that demanded a very small power supply with high power density. In addition, the customer wanted the supply to operate over a universal input range (85 Vac to 265 Vac).

The active clamp technique yielded enough efficiency to fit a convection-cooled, 50-W power supply in the space the size of a paperback book. When completed, the power supply had four outputs (±5 V and ±12 V). The converter architecture was a coupled inductor, forward converter where the 5-V output was regulated to ±1%. Efficiency was around 85% despite the use of Schottky rectifiers on the ±5-V outputs.

This power supply, which went into production in 1978, also used a grounded-base BJT on the input, which allowed it to operate up to the 700-V VCBO rating of the transistor. This power-supply design also provides another example where Carsten was able to reuse a technique for driving bipolar transistors he had developed in 1975 while designing a 1-kW sonar amplifier for the Canadian Navy.

The technique involved grounding the base of the transistor and driving the emitter with a transformer-coupled current source. Carsten called this emitter-drive technique “bulletproof” because if anything went wrong, you had a protective feature as collector current could not exceed the emitter current. When applied in power supplies, the emitter-drive technique offered the advantage that a short on the transistor output would trip a “de-saturation” detector and terminate the conduction period. Carsten drew his original inspiration for this technique from the cascode-type of transistor drive used in oscilloscopes in the vertical deflection amplifier.

Looking Ahead, But Too Far

“In the early ‘80s it became self-evident to me that logic voltages were going to drop below the standard 5 Vdc, largely to reduce power consumption in ever larger and higher speed logic chips,” says Carsten. “I had noticed that some 5-V CMOS logic would still function, albeit very slowly, as low as 300 mV, so the lower limit on supply voltage was obviously very low, and the IEEE Journal of Solid State Physics had articles on lab devices operating from 2 V to 1 V and lower.

 “I realized that this would have a major impact on power converter and system design; rectifier losses would become extremely problematic —they were already the largest source of power loss at 5 V—and the conventional central power supply was not viable as voltages dropped to 1 V or 2 V and currents increased to hundreds of amps or more,” Carsten explains. “The power source would have to be located close to the point of load, and thus share board space with the logic, placing unprecedented demands on efficiency and power density. Since it would be unwise to bring the noisy, transient prone ac line power onto the logic board, the only solution was a distributed power system with an intermediate voltage dc distribution bus.

“I demonstrated a proof-of-concept 48-V to 2-V, 50-A, 500-kHz prototype dc-dc converter at the PCI ‘84 conference in Paris, and presented a paper on the development and why I thought that this was the future,” says Carsten. “The converter consisted of a two-stage design with a classic buck converter producing a regulated, but nonisolated 30-V output. The buck converter ran at 500 kHz at a time when 50-kHz switching was the norm.

“The high switching frequency necessitated the use of homebrew surface-mount power components (components in which the die had been removed from the package, or the packages had been modified to shorten interconnects and reduce the associated parasitics,” Carsten explaines. “The second stage stepped down the 30-V intermediate voltage to 2 V, while providing isolation, with isolated feedback from the output to the input regulator.

“The second stage consisted of two single-ended forward converters operating out of phase,” Carsten observes. “Note that the converter uses a preregulator stage followed by an unregulated dc-dc converter, which resembles the approach used in Vicor’s ‘factorized power architecture’ concept, which was introduced a few years ago. Also, at the time this prototype was built, surface-mount components were being used in logic circuits, but were highly unusual in power conversion.

“Like many of my papers, it met with resounding apathy,” Carsten recalls. “When the silence died down, [conference attendees] would look at my prototype, with its home-made surface-mount power components, and say, ‘That's interesting, but logic voltage is 5V. What is it good for?’ I wrote several more papers on the subject, but finally gave up in 1991 since no one seemed to be listening. Of course, today ever-lower logic voltages are old hat, and many engineers don’t remember a time when that was not so.”

More recently, Carsten has been working on advanced algorithmic control methods that provide much faster dynamic response to line and load conditions than is currently possible with power converters using classical feedback techniques.

Carsten explains, “I see such control algorithms as a promising ‘super fast’ control methodology, quite distinct from PWM, voltage regulation in current-mode control and digital control. The inner loop in most current-mode control does qualify as a control algorithm, however. I have used several such control algorithms in the last 20 years, but once again, discussion in one of my seminars and a couple of published papers are meeting with evident apathy. Designers are reluctant to give up what they are familiar with as long as it still seems adequate. But then, I didn’t get much interest in current-mode control at first, either.”

Want to use this article? Click here for options!
© 2007 Penton Media, Inc.