Switching Power Supply


A switching power supply is one that achieves efficiency by chopping the input power at a high frequency, avoiding the power device (transistor) linear region. The resulting pulsed DC current can be passed through a transformer to raise or lower the voltage, provide multiple output voltages, and provide isolation from the source of power.


This power supply is a box that provides stable, pre-regulated voltages to any devices on the telescope that require ±15VDC and +5VDC. Because of size constraints, this box has only two outputs. Each output has +18, -18, +7.5, +12, and two grounds. The ±18VDC is regulated to ±15VDC at the device end. The +7.5VDC is to be regulated to +5VDC. The +12 is the battery voltage, and isn't used for sensitive circuitry.


  • Input Voltage:9VDC to 18VDC
  • Ouput Voltage 1:+18V @ 1.5A
  • Ouput Voltage 2:-18V @ 1.5A
  • Ouput Voltage 3:+7.5V @ 1.0A
  • Ripple (+18V line):
  • Eficiency:>85%
  • Size (L x W x H): 4.25" x 2.6" x 1.5"

Design Goals

The power supply input is 81.4W at 9VDC. Calculating what is required is straighforward. We need 36VDC @ 1.5A, and 7.5VDC @ 1.0A. That is 65W. The diodes are going to drop 2V in a bridge configuration. That needs to be added to the transformer output voltage. 38V and 8.5V. Those 2V are dropped across diodes carrying either 1.5A or 3A. 3W + 6W = 9W lost across the diodes. 10% transformer losses = 74W * 0.1 = 7.4W. The input to the transformer has to be capable of pushing 81.4W. The transformer is good for up to 90W.

The transformer primary, in order to guarantee maximum output at minimum input voltage, is wound to 9VDC. At full load, with minimum input, the power supply is working it's hardest, and it has to work. The transformer turns ratio is then 9:19 for the two high-voltage secondaries, and 9:9 for the lower voltage secondary. At 9V, the current required to put 81.4W into the transformer is 81.4 / 9 = 9.04A worst case. How about the norm? At 14.7V input the current required at full power is only 81.4 / 14.7 = 5.53A.

The controller chip is the LM3524D, which is the latest version of a very old device. It works, is very reliable, and is a known quantity. I run them at 50kHz with this pot core.

The pot core is an Amidon 3622-77. You can get them all over the place, but I get them at Alltronics.

The whole thing is in a little extruded box from Context Engineering.


Winding Voltage Current Turns Wire Size and Count Cut Length
Primary 1 9V 9A 12 24ga. x 5 34"
Primary 2 9V 9A 12 24ga. x 5 34"
Secondary 1 18V 1.5A 30 28ga, x 2 85"
Secondary 2 18V 1.5A 30 28ga, x 2 85"
Secondary 3 9V 2A 13 28ga. x 3 43"

This is the part that scares people, but there is no need for that. It is just wrapping some wire on a form, and trusting in physics to make it work. The most important thing with this specific transformer is to wrap the wire evenly and tightly. There are two reasons. First, it won't fit if you don't wrap it tight. This transformer has nearly 100% space utilization. Secondly, if it isn't tightly wound the wire will vibrate in the very strong magnetic field. Vibrating a wire takes energy. The drivers have to supply that energy.

Start with the primary. There are two, and each has 5 wires. I cut 10 x 34" long 24ga. magnet wires, grouped them in two groups of 5 wires and twisted the ends of each group together. Just the ends, though - not the length of the wires! These two bundles will be the two pirmaries. Use a marker to identify both ends of one of the bundles. When you wind it, remember to keep it tight, but don't worry too much - it is going to take most of the coil form. The secondaries are much smaller.

The main secondaries (there are two) are each 30 turns of 28ga. wire. The length is now pi * 0.9 inches per turn, since the primary now makes the form diameter larger. Times 30 turns that's 85 inches. So I cut four strands 85" long.

Finally, the lesser secondary has 13 turns of three strands of 28ga wire. Since we are now at nearly an inch on the form, the length of one turn is 3.25". 13t. x 3.25" = 42.25" (round up to 43).

When I assembled the power supply and tested it, the ±18V supply was unbalanced by 0.09V. Not bad. Ripple was around 100mV with a frequency of 46kHz. The output filter capacitors got hot after a few minutes. I added a 0.1uF ceramic across the output capacitors, and added one across the outputs of both bridges, and there is now no measurable ripple (on the 20mV scale I see nothing), and the output filters are not getting warm to the touch.