As follow up to my blog Low Power Isolated Power Supplies I thought I might continue by considering low power DC to DC converters that don’t use an inductor. After all, I am just a dumb digital engineer messing around in an analog world. Magnetics scare me.
I have often needed to generate a negative voltage to provide a bias to an LCD screen or a negative supply voltage for an op-amp. Most often these are for a few milliamps and so there are some alternatives to using an inductor. Aside from some external capacitors, there are single chip solutions and lately even the capacitors may not be necessary. Of course many analog devices today are single-supply, but every now and again a bipolar device raises its head.
The granddaddy of inductor-less parts is the ICL7660 switched capacitor converter, originally from Intersil. It required two rather hefty capacitors (10uF) (and another if you include decoupling) and it also had a fairly high source resistance, which corresponded to a relatively low output current. The output voltage is approximately the supply voltage inverted. In some of my applications I found that the maximum input voltage was too low for those +/-12V LM741s (a very popular dual voltage op-amp for you youngsters). Apparently others did too since Intersil subsequently brought out the ICL7662 (actually I seem to recall that it was pre-empted by the Si7661 from Siliconix, but that is water under the bridge). The family was duplicated by several others, but as design and technology have evolved there are several other, and dare I say, better options.
Let me pause in my description of inverters and mention the Maxim developed MAX680 which works on much the same principle as the 7660. It is a switched capacitor doubler and inverter so the output will be +2 x Vsupply and -2 x Vsupply and generates up to +12V and -12V from a 6V supply. I suspect that this was an offshoot from their groundbreaking work on the standalone RS232 driver, the MAX220-249. Although there are variations of the MAX232 that only use 0.1uF capacitors, you are stuck with 10uF components for the MAX680.
There are dozens of switched capacitor possibilities in converting positive supply to negative. The selection tables from either Digi-Key or Mouser are so large that they probably confuse rather than help. One aging possibility is the LM828 which uses standard ceramic capacitors and is in a tiny package.
One of the problems of switched capacitor devices is that they have a relatively high source impedance and as a result the output voltage is a function (Ohm’s Law) of the current. To solve this problem several manufacturers hit on the idea of using a linear regulator in series so that there is a low impedance and constant voltage at the output. They have certainly improved a lot – there is the LTC1983 from Linear Technology or how about the LM27761 from Texas Instruments that has an adjustable output from -1.5V to -5V and a current of 250mA.
And if you don’t want an inverter, but just a simple high efficiency DC/DC converter without any magnetics, how about LTC3245, a buck/boost device with 2.7 to 38V input, programmable output 2.5 to 5V and 250mA output.
My favourite part of these blogs is when I get to deal with the exotics. As I have mentioned before, I work a lot with 4-20mA current loops. The 4mA offset was partially introduced to power the receiver and/or transmitter of the loop. Forgive me for stating the obvious- the 4-20mA is provided by a current source and in order to do anything useful (like powering an IC) this current must be used to produce a voltage while the available current is limited to 4mA. Reflecting on Kirchhoff’s first law we can see that the voltage dropped across the current source and the developed voltage across the receiver (plus the drop over the wires) must be equal to the supply voltage of the loop. In some industrial applications two or three receivers can be placed in series, so usually a designer tries to minimize the input voltage drop. Hold on to that thought for a moment.
Even with modern low power micros, 4mA can be constraining and a very smart person came up with the idea that if you have a certain input voltage and current- that is power. And power converters transform power, so if you can find a compromise input voltage, you can use a power converter to drop the internal supply voltage and boost the current. Linear Technologies has a part that does just that- the LTC3255.
Part of an analog engineer’s learning curve is discovering that despite the name “rail-to-rail” some op-amps still have issues at the extremes where the output is up near the positive rail or down near the negative rail. Usually the op-amp doesn’t quite get to zero, or can’t handle any current when it gets there. Enter the LM7705 from Texas Instruments that generates a -0.232V which is applied to the Vss and so the 0V output is actually in the operational range of the op-amp.
Instead of generating a negative voltage you can power the bipolar op-amp between the supply rail and ground and split the voltage to generate a midpoint. This is an old technique which involves a resistive divider and a unity gain buffer to create a low impedance point. As long as your current requirements are modest, the TLE2426 is a single chip solution. I know this approach is slightly off topic, but you can find some more on it in this article.
Of course none of these techniques provide any isolation. However I came across this ingenious design idea from EDN’s Michael Dunn which details a voltage doubler that has a small degree of isolation.
I haven’t discussed another voltage converter alternative which is the voltage multiplier, using diodes. Although it normally converts an AC waveform to DC, several variations (some known as the Dickson multiplier) will convert DC to DC. I leave it to you to research the topic, since I feel unqualified to expound on it.
Charge pumps are not a new idea. Intel upgraded the multi-supply 8080 to the single supply 8085 (those are some early and very popular microprocessors for you analog types) by including the charge pump on chip. Texas Instruments at one point integrated a charge pump onto a quad op-amp to provide a negative voltage, but for the life of me, I can’t find any reference to it. Despite its age, apparently there is still a lot of life in charge pumps. I am sure I missed many of your favourites, and possibly even some design techniques. Please add your comments below.