Make a variable power supply symmetrical with a 3-leg regulator

Surprising facts, like “old fangled” 3-leg linear regulators (LM317, LM337, LM350, et al.) can be used in shunt regulator topologies can sometimes inspire “new fangled” circuit designs. Figure 1 shows a useful example.

Figure 1 Symmetrical-output variable power supply uses single DC source with classic 3-leg regulators, one regulator (U3) operates in shunt mode as “rail-splitter”.

Wow the engineering world with your unique design: Design Ideas Submission Guide

Variable voltage regulated power supplies are handy fixtures on most every well-equipped electronics lab bench, and the symmetrical variety that produce equal voltage outputs of opposite polarity, are even more so.  Figure 1’s version of a symmetrical supply uses a thrifty trick in which just one power source input, instead of two, supplies both polarity outputs. 

The trick, often called “rail splitting” works by “floating” the power source. This makes it possible for the source’s positive terminal to provide current for the positive polarity supply output, while its negative terminal supplies the negative. This method can be extra cost-saving given the available plethora of mass-produced regulated AC adapters. These commodity items are actually cheaper than a similarly capable simple transformer plus rectifiers and filter capacitors would be. The 24VDC/1A item used in Figure 1, for example, is sold by a well-known mass-retailer for less than $8! 

Here’s how Figure 1 works.

 The few parts surrounding U2 comprise just a plain-vanilla textbook application straight out of the maker’s LM317 datasheet, with a voltage adjustment range spanning,

Vmin = 1.25 V(165/165 + 1) = 2.5 V
Vmax = 1.25 V((1000 + 165)/165 + 1) = 10.08 V.

 Totally standard stuff. But don’t bother looking for U3’s circuit in the usual run of datasheet application notes. You probably won’t find it there. U3 is configured as a shunt regulator that inverts U2’s +2.5 V to +10 V to produce the complementary negative -2.5 to -10 V output.  Yes, it’s weird.  The LM337 isn’t supposed to do shunt. But it works.

U1 is needed to compensate for the fact that U3’s adjust terminal operates at +1.25 V, which puts (+Vout – 1.25V) – (1.25V – -Vout) = -2.5 V less differential across the upper 1k resistor than appears across the lower 1k.  This difference would spoil output voltage symmetry if allowed to persist, but U1 subtracts a matching 2.5 V and thereby fixes it.

If you’re considering using this design, please be aware it assumes that current drawn from the positive output will be always at least as great as current taken from the negative side. If this relationship is reversed and the negative side is significantly more heavily loaded than the positive, then U3 will fall out of regulation and the negative voltage output will droop and cease to match the positive. Happily, if this problem does happen, no actual damage will result. Moreover, it can be worked around by adding a suitable “dummy load” resistor to the positive side.  

Also be sure to provide adequate heatsinking for U2 and U3, especially for U2 which can be called upon to dissipate nearly 20W if the supply is run at minimum voltage and maximum current.

Meanwhile, Figure 1’s maximum +10 V output limits happened simply because 24 V was the highest voltage I could find on the cheap power adapter market. Of course, this is far from a fundamental limitation, as Figure 2 proves with its “classic” +15 V limits, made possible by switching to a similarly “classic” 25 V filament transformer, diode bridge, and filter capacitor.

It just costs more.

Figure 2 Symmetrical rail-splitter supply design extended to +15 V max output voltage by a different choice of DC source and the change of a few resistors.

Stephen Woodward’s relationship with EDN’s DI column goes back quite a long way. Over 100 submissions have been accepted since his first contribution back in 1974.

Related Content