For the PDF version of this article, click here.
More than two decades ago, I was responsible for developing the two ac-dc power supplies that would ultimately be used in the AN/ALQ-184 Radar Jammer Pod aboard the F-16 fighter plane. Though the application seems somewhat exotic, many of the design goals would be considered mainstream today. What's more, some of the control concepts used to build these supplies bear surprising resemblances to those now being used in advanced dc-dc converters.
In the Radar Jammer Pod application, size and weight were everything. There were many technical challenges in this new project, but a few of them seemed bigger than others. I carefully considered all of the technologies that existed at that time. There were no current-mode controllers and, if my recollection is correct, feed-forward controllers were just coming into play.
I desperately needed a way to provide an isolated feedback system while maintaining a wide operating bandwidth and superior ripple rejection. These criteria were critically important so that the size and weight of the input filter could be minimized and so the 2400-Hz ripple, resulting from the input three-phase rectifier, would not feed directly through to the outputs. Output noise also was critical.
My solution was a magnetic modulation scheme that offered many benefits, including synchronized switching frequency, an inherently feed-forward architecture for outstanding ripple rejection and nuclear radiation tolerance. The modulation scheme also could be modulated by any number of isolated ports (windings) and was easily implemented in all standard power topologies. In 1985, I received a patent for this new technology, but recently, I have wondered why it never went beyond the AN/ALQ-184 project.
I chose to implement my new magnetically modulated control circuit in a standard two-switch forward converter. It was a modular design, so I chose to use four forward converters, each phase shifted by 90 degrees to minimize the EMI signature (see the figure).
The operation of the magnetic modulator was very simple. A narrow positive-going clock pulse coupled through T2 to the gate of Q1 initiating a cycle. Once the cycle was initiated, winding 5-6 of transformer T1 provided regenerative action through CR1, a constant-current diode. Transformer T2, which was fed a voltage proportional to the input voltage saturated, collapsing the field and turning off Q1. MOSFET Q1 remained off until the next clock cycle. Without any current through Q2, transformer T2 began the cycle at BR (remanant flux density at zero oersteds) and traveled to BSAT (saturation flux density) at which time Q1 turned off. A forcing current through Q2 created a higher effective remanance moving BR toward BSAT shortening the conduction time of Q1 and therefore reducing the output voltage.
The flux path of T2 was proportional to the volt-seconds product across winding 5-6 of T1, which was also proportional to VIN. This accounted for the improved ripple rejection of the converter. In practice, the inherent ripple rejection of the magnetic modulator was approximately 40 dB. The feedback loop increased that by an additional 12 dB at 2400 Hz.
This simple scheme allowed control from the secondary side. Additional windings could be added to the saturable transformer, allowing control relative to other grounds as well.
The saturable transformer used was a miniature permalloy tape core, using very thin tape (1 mil). Operating this transformer from BR to BSAT rather than from -BSAT to +BSAT minimized the core loss effects of the saturable transformer. The final design used a two-switch topology, along with a driver circuit to square the signals from the saturable reactor.
Recently, I thought about the phase-shifted operation of this circuit after reading an advertisement about a new seven-phase buck regulator. I also have recently seen articles and data sheets for hysteretic regulators. I recall successfully using the topology in the mid- to late 1970s. Maybe it isn't that my magnetic modulator never really took off; maybe its time just hasn't come yet.