The LT1070CT's Linear Technology's old but neat 5 pin TO-220 SMPS IC. By looking at the datasheet, it appears designed for boost and flyback SMPS configurations. It normally runs from 38KHz to 40KHz switching frequency.

I didn't see any SEPIC reference designs so I wonderred what would happen if I put it into such topology. Because the LT1070CT appears to be best suited for boost mode supplies (since its switch is a current-to-GND, low side switch instead of a high side switch needed for buck supplies) SEPIC seems to be a good way to use the IC.

Schematic

Photo of homemade switch mode PSU: All parts were salvaged from broken power supplies, monitors, etc. except the phenolic PCB, the LT1070CT itself, and the 2.2µF capacitor. I have a 10KΩ variable resistor as a the voltage divider feedback, hidden behind the 2200µF capacitor. The "far" winding is L1, and the "near" winding closer to the diode/smaller electrolytic is L2. I salvaged this 1:1 transformer (actually, input filter inductor) from a dead ATX PSU. C2 is the yellow ceramic capacitor. I was lazy and did not trim the leads, which I should - it adds unwanted inductance on the leads.

Calculate R2 and R3 such that the output makes a voltage divider where the divided voltage is 1.24V, accounting input leakage of Vfb (which is very low). For a 5V output, R3=1.24KΩ and R2=3.74KΩ is the reference design resistances for the divider.

This sure is a lot cleaner than the reference design 'buck' converter - it has no minimum current requirement other than the load provided by the feedback network. However this is like any other PS where input and output voltages have limited range. I was able to get stable output of 12V 1A with input ranging from about 6V to 18V. At the lower voltages the LT1070CT started to get quite hot, and the switching frequency went down a lot, down into the audible range and probably saturating L1. The "transformer" that's L1 and L2 is actually quite crappy. The transformer's windings are not very tight and makes noise when the switching frequency dropped into audible range.

D1 is a Schottky or Fast switch diode. It is a switch mode power supply after all.

Coupled with a 12V transformer+bridge rectifier (or 24CT + half bridge) and an input capacitor, this can make a simple, safe (the first transformer provides isolation, and most circuit elements are on the low voltage side), 120V to 12V switch mode power supply that's much more efficient than a 7812, not to mention the minimum in-to-out voltage differential is mostly no longer an issue - the SMPS will handle it! This is also good for car use as a part of a carputer, the dirty 12V battery power is converted to clean 12V.

Unfortunately the LT1070CT is a fairly expensive device, I'm not sure how many people will actually buy one of these to make a SEPIC with it. Oh well.

After stress testing the supply a bit, I destroyed the LT1070CT. Seems that it can't completely protect itself from thermal runaway in a SEPIC configuration ...