Forums » General Electronics Discussion

Hi All,

I needed to generate -5V, 20mA from a +12V source. I was able to do this with a LT1373H chip. But I was just wondering what other people are doing when they need to generate +- voltages for....op amps..or other uses.

Very much depends on how much power you need (how much current at what voltage?).

Other considerations are specifications for efficiency, line and load regulation, input and output ripple and noise, input current limiting, output short circuit protection, undervoltage lockout and so on.

If you don't need galvanic isolation between input and output then a buck-boost switchmode supply may be the simplest solution.

Switched capacitor converters/regulators are an option for low power.

If you can put up with the current spikes it draws from your source (and may inject into an existing supply if used to boost an existing rail) then for low cost you can sometimes use a simple diode pump, possibly driven from the switching node of another SMPS or maybe a spare logic gate output:

and for example, to generate a low power bias supply above an existing rail:

Following the pumped output with a linear regulator would give you a well regulated, reasonably low ripple, short circuit protected output.

Very good suggestions from signality-- it really all depends on the power level, regulation, ripple, load variations, heat, dissipation, thermal gradients, efficiency, and EMI requirements. Not to mention the availability of pre-existing sources of AC, like the secondary of a power transformer, or a node in a switching regulator.

If you just need a weakly-regulated source, you could piggyback a voltage-doubler off the swinging node in an existing 5 to 12 volt power supply, or even off a strong master-clock source.

If you need better regulation, you could follow that up with a LDO linear regulator.

If you have the room and $ for another IC, a separate switching-mode regulator chip or switched-cap IC will do it, although you might have issues with the EMI from yet another set of E and M spikes at yet another clock rate. Last time I tried this, the difference frequency between the two clock rates ended up as a wavering 400 to 600 Hz buzz in the audio output. Horrible. Had to sync up the two switcher clocks to eliminate this spurious note.

What would be the current capability of the first example?

@Eth, Do you mean this one?

Up to a point, you can design a simple diode pump (what Arduinohacker referred to as a voltage doubler), for any sort of output power.

The first thing limit on output power is the pulsed signal source. It has to have a low output resistance. Drive from HCMOS or faster logic gates should be OK for a few mA output. The drive from the switching node of a switch mode supply (the common node of the synchronous switch MOSFETs or the common node between the switch MOSFET and the rectifier diode) will be lower resistance because the FETs are chunkier than those in a logic gate. It may also have a higher voltage swing if it's stepping down from, say, 12V.

The next limit is the size of the capacitors. Bigger C1 brings the output voltage up faster at the expense of bigger current pulses from the pulse source. Bigger C2 reduces the ripple on the output voltage.

Schottky diodes instead of ordinary silicon diodes increases the o/p voltage due to their lower forward drop.

Higher pulse source voltage swing increases the output voltage.

The easiest way to design a diode pump is probably to do it through simulation. A rigorous design is not easy to specify unless you are designing the pulsed source too and even then the sums are a bit obscure.

Design the simulation to include the on resistances of the pulse source (draw the circuit as an NMOS/PMOS driver or reproduce the switching MOSFETS from the SMPS if that's your source), choose some diodes then play with C1 and C2 until you get a useable output voltage and ripple.

While you do that, keep an eye on the pulsed currents and power dissipations from the source and through the diodes. Change the diodes and pulse driver stage as necessary to accommodate the currents and powers measured in the simulation.

Unless you're switching more than about 50V then the caps should be X7R dielectric ceramics (Y5V has a very poor capacitance vs. applied voltage drop off so you get far less capacitance than you expect with some very strange modulation effects: see

but beware CL still doesn't show the plots properly ...)

Note that you can in fact make a voltage tripler or quadrupler etc., out of diode pumps: the limitation is that the diode drops make higher multiples from low voltage sources impractical.

You can even include a diode pump multiplier in a boost SMPS: I once did a 5V to 80V boost converter for an APD supply based on a 5V to 25V boost converter with a tripler on the output.

An example of driving a diode pump from an SMPS switching node can be seen in Figure 19 on page 15 of:

http://www.ti.com/lit/ds/symlink/tps61040.pdf