Inverting buck or inverting cuk?

[carveone]

Hi all,

I've been tasked with creating a +/-12V supply as part of the bias voltages for some opamps (AD820 I think). I haven't been told the current requirement but if it's just biasing it's likely small enough. I'd have a tendency to assume low noise is a requirement given the input is from a sensor. The project uses a Li-Ion 7.4V battery (at least I'm told that much!).

I haven't generated negative voltages before but looking around I see circuits which use an buck converter in an inverting configuration:

• https://www.eevblog.com/forum/projects/layout-review-inverting-buck-converter/
• http://electronics.stackexchange.com/a/18174

but if I search the manufacturers, I see specific inverting chips like the LM2611 and the LT1931. These use two inductors in a Cuk configuration.

Is there an advantage in taking the Cuk route over an inverting buck route? As I hadn't seen the latter in datasheets I assumed it might be a "less approved" route for generating negative voltages.

[tom66]

My favourite technique when I want to generate a split supply that's relatively low current from a boost/buck converter is to use a charge pump.

One method you can try is to generate +16V rail with a boost converter. A charge pump circuit will generate -14V from the +16V, using the boost converter node as its switch node. This can then be regulated to ~12V using a 79L12 type regulator, as can the 16V rail.

Pros: extremely simple; two diodes, two capacitors. Low noise, if designed correctly and fly capacitor is sized right. Very low cost, as MLCC capacitors <10uF can be used, and SOT23 dual diode is usable up to ~70mA.

Cons: output is unregulated (load dependent somewhat) - usually not an issue for unloaded/lightly loaded opamps as quiescent current is typically constant throughout operating range, only works if the positive rail is loaded, and output current is limited to the 100mA range before components get too big to be practical. No current limit, a short circuit will either cause the buck/boost to current limit itself, or it will damage the diode... or both potentially.

Attached shows a 5V buck solution with a -11V generator. The -11V is later regulated to -5V for opamps. The negative rail follows the input voltage for a buck circuit. In this case the input voltage is fixed, so this does not present an issue. For a boost circuit, it follows the output voltage.

[carveone]

I guess I was hoping to go from 7V or so directly to -12V in the simplest way. But I get your point - if I needed the +12V as well (I have an LMR64010 producing +12V for other uses) then producing +16V and converting to -14V then to -12V would make sense. I do worry about the noise inherent in some charge pumps though.

Oh - I'm looking at your circuit. That's very interesting. I haven't see that done before. I had assumed the charge pump was a separate item like the TL7660 Dave refers to in one of his video blogs. Pretty cool, I like it!

I also find it interesting that you are using the TPS54331 as the buck regulator. I looked at this and the TPS5430 and chose the latter solely because I thought "hey, internal compensation. Less components." Not sure that's always a valid way to choose components (and one is voltage mode the other current) but there's so much choice out there!

Edit: Actually I should see what the +12V is being used for as I could possibly do it your way if the current isn't too high. Makes sense to use what I have...

[Marco]

It seems rather common for LCD biasing circuits, lots of app notes and datasheets out there which describe these combined boost/charge pump circuits (sometimes with a doubled positive output as well for high side gate drive).

I've seen it said that the only time cuk really makes sense is if you need a buck-boost type circuits for constant current loads (ie. LEDs)

[filssavi]

Hi
in a design of mine not long ago (few months ago) I used the TPS5430, so i can give you mt advice, first of all it's a great little device very simple to use, and can provide a high(ish) current i have tested it up to an amp and it wasn't even feeling it

That said i dont know if it's the right part for you, it has a maximum current of 3A that os really huge for biasing opamps, and to prevent it from melting it has a power pad heatsink on the bottom, and that's not the easiest thing in the world to solder, there is the big via tecnique, but i dont particularly like it, so i made a pad that came out from a side, with no solder mask

As for the compensation, it sames you some parts (well not really much since they are a couple of passives) and some calculations, but you lose a lot of flexibility and i had some noise on the output, nothing drammatic since i was only driving a servomotor but for your application could be a problem so the passives you save could very well be you have to use them to smooth the output voltage

All in all a great part though...

[carveone]

The more I get into power supply design the more I realise how big a field it is and I know nothing :-)  I'll do what Marco said and start looking closely at the datasheets. I did see one where the technique was shown (can't remember where now!). I think there's nothing wrong with using boost/charge pump or inverting buck instead of the more complex cuk.

@filssavi: The TPS5430 is being used for the 5V rail which goes to an ARM board which, along with its camera, can pull a decent amount of power. More than 1A anyway. I appreciate what you say about flexibility but I'd need to learn more in order to use that flexibility. I bet there's a book on this stuff somewhere (Horowitz and Hill doesn't really cover this I think).

My colleague found that the given layout recommendation in the TPS5430/5420 datasheet wasn't particularly good as regards noise. I think the layout was awkward because of the pin configuration - generally the inductor and the catch diode need be to right on the PH pin... It worked out OK eventually though and has been pretty good.

[filssavi]

As for  compensation i think it's not that critical difficult, you it's treated in just about every analog electronics textbook (gray meyer/jager blalock) you basically have to shape the frequency response of the thing to reject as much HF switching noise as possible by adding poles and zeroes (for example in the same design I had a LM5001 in boost configuration and that converter is compensated with 2 poles and a single zero; a LF pole to tame the high DC gain of the error amplifier, a zero in proximity of transition frequency to provide phase margin for stability and a higher frequency pole to filter out HF noise) also in the datasheet you have all the formulas needed to calculate the filer components values knowing the desidred frequency

As for the layout I totally agree with your collegue, i did a full blown 3A design and there is no way to fit the huge inductor and output cap in the suggested layout, also my diode is much bigger than theirs so...

[Mad ID]

generally the inductor and the catch diode need be to right on the PH pin... It worked out OK eventually though and has been pretty good.

This is not correct. Current through the inductor is not switching so inductor location is of low importance. The loop carrying the switching current has to be minimized.

See
AN-1229 SIMPLE SWITCHER ® PCB Layout Guidelines OR/AND
Reducing Ground Bounce in DC-to-DC Converters—Some Grounding Essentials

You will see that inductor location is of low importance in comparison to the components carrying the switching current

[carveone]

This is not correct. Current through the inductor is not switching so inductor location is of low importance. The loop carrying the switching current has to be minimized.

Yes, I stand corrected there. I don't know enough about layout of SMPS. I hope to know more - there's been a number of interesting PCB layout analyses done on eevblog that end up going through why A is good and B is bad. It's very useful.

[carveone]

Again, lots of thanks to Tom66 for that suggestion. This looks to be a great solution and one that saved me a great deal of effort.

I tried it out yesterday as I had a +12V supply from a boost converter (LMR64010) at 1.6MHz. So I tapped between the inductor and the diode with the charge pump.

I just used a scrap (just 5 x 4 holes!) of veroboard and I didn't include the Pi filter, mostly because I ran out of veroboard room! So just the coupling cap (4u7 smd), two 1N4148 diodes I had handy and two smoothing caps (22u smd) on the output. Hey presto -12V out (a bit less).

Looks good so far and I'll see what load it'll take. It'll be interesting to see how one rail affects the other in load terms...