Designs evolve because someone thinks of a better way to do something or because the first great idea (and sometimes the second and third) just didn't work. This is the story of a little supply design that evolved into something sweet.
The challenge was to design a laser-diode power supply that charges a 1000-pF capacitor to 270 V thousands of times per second for an average power of about 400 mW. In addition, the supply had to be less that 0.5 in. in diameter with a height of less than 0.25 in., and be at least 70% efficient.
Our first idea was to try a Royer circuit. Advantages of this design are its simplicity, low parts count and low cost. The disadvantage is its low efficiency. To employ a saturating transformer at 200 kHz (our target frequency) would mean astronomical core losses and, according to our figures, significantly more power than our output.
Thus, we backed away from the first idea and made something that didn't require a saturating transformer. We combined an ordinary flip-flop with a transformer and made a self-oscillating converter, which is better than a Royer.
However, try as we might, we couldn't make the circuit operate anywhere near our target of 200 kHz. The transformer with a 1-to-30 step up limited us to about 60 kHz.
We then explored another option. Because the function of this circuit was to charge a capacitor and quit, we thought a crude regulator would be adequate. Again, we couldn't get the circuit to work at a high enough frequency, so we abandoned this approach.
The transformer used in the previous experiments was constructed as follows: For a core, we used Magnetics Inc.'s #40603TC with an 8-turn primary and 355 turns of #53 gauge wire on the secondary.
The Forward Converter
Thinking a more conventional approach was needed, we decided to build a forward converter. This forward converter would have been unusually simple, because we reasoned that with an 8-kΩ resistor in the output, we wouldn't need an inductor to limit current. Once again, we ran into transformer problems, where we couldn't get any significant output power until we dropped the frequency to about 70 kHz. Even then, it still wasn't enough output power. As with the previous design, we were using the #40603 core, but with five turns on the primary and 180 turns on the secondary.
The Tapped-Inductor Boost
Early in our design work, we considered a straight boost converter. But with a 30-to-1 boost ratio, we thought an “off” period of 3% would be impractical. However, flyback power supplies are a natural for charging capacitors, because the core energy is dumped into the capacitor each cycle. Since the secondary is a current source, the start of the output voltage near zero doesn't change the output current. Actually, for most circuits, the parts count with a flyback is the lowest.
Nevertheless, the flybacks present disadvantages. Insulation between the primary and secondary is a little difficult because the two are connected together.
Early on, we realized the tapped inductor approach was superior to the others we had tried. However, we needed to adjust our off time so that the secondary energy could completely dump.
For the transformer (or, more precisely, an autotransformer or taped inductor), we tried two different approaches. First, we used a toroid consisting of the #55028 core with 55 turns on the primary and 65 turns on the secondary. However, because of cost considerations, we eventually switched to a less-expensive pot core. This final version was built with the #41003 pot core with a 30-turn primary and 40-turn secondary.
Initially, we used a compensation network like most power supplies, but we soon determined that we needed more lead to approach the required performance. However, on further review, we found that by decreasing the lag to zero, we needed no compensation anywhere. That, in turn, allowed us to meet the required rapid change in output voltage (see Figs. 1, 2 and 3).
Usually, power-supply designers don't miss so badly on selecting a topology. They pick one that works most of the time, until they run into something such as a manufacturing problem with their high-voltage transformer or other problems down the line. In the early stages of the design process, we seemed to have great ideas, but they didn't work. But the final design is an uncommon topology that lent itself to this application.
For more information on this article, CIRCLE 339 on Reader Service Card