Linear voltage regulator... fire story!

[Al (astral highway)]

I was getting a variably variable scope trace on a low frequency, high pulse-width signal - the 'scope timebase wouldn't lock on to it - so I took a look at decoupling on the board producing the signal.

I added some decade decoupling caps where there were the bog-standard 0.1uF

So one place was the input of the linear regulator. It was seeing 9V unregulated in and popping out 5V regulated.

Normal configuration of 0.1uF on the input, 0.1uF on the output, plus a 50uF smoothing electrolytic.

Shortly after I added the 10nF, it caught fire! And then of course, the linear regulator began to overheat and fail.

So what's going on here?

There were no long board traces to form an inductor - rules out an L (parasitic) + C combination forming a parasitic tank circuit - the additional capacitor was soldered onto the track-side of the board, right beneath the 0.1uF.

[Radio Wrangler]

Linear regulators for power supplies inevitably drive capacitively decoupled rails on their outputs.

The capacitance as seen by the reg, along with the reg transistor's output impedance forms a lowpass pole, which gives a phase shift. The feedback internal to the reg is taken from the output pin, the load C produces its extra phase shift here too. So the C loading beggers up the loop design of the reg chip.

So the chip designers did heroic measures to try to make the beastie stable into as wide a range of loads as they could. But they couldn't do everything.

So with regs don't just look at recommended C values, look at the types too. Some regs rely on the ESR of electrolytic capacitors to keep them stable. Dropping in low ESR at high freq caps can initiate disasters.

The trend to replace electrolytics with high value new type ceramics isn't plain sailing.

David

[Al (astral highway)]

Originally Posted by Radio Wrangler

. Some regs rely on the ESR of electrolytic capacitors to keep them stable. Dropping in low ESR at high freq caps can initiate disasters.

Thank you, David, that's a really clear explanation of what's happening. It was counter-intuitive when I was using a military-grade component - definitely 'low ESR at high frequency.'


What's a workaround when there are instability issues because of inadequate decoupling?

[Radio Wrangler]

There isn't a work around. There's no avoiding the issue, you have to read the data sheet carefully, looking for info on what should and shouldn't be used. Not all data sheets are good in this area, you may have to take their apps note circuit and not stray very far from it.

THis is exactly the same problem run unto by opamps driving capacitive loads.

There is a similar issue even with emitter followers driving capacitive loads... a great many UHF oscillators (of the intentional kind ) are designed this way, with just a few nH of track as the inductor in the emitter circuit and in the base circuit.

David

The problem is rarely not enough decoupling, at least on the output side, it is sometimes too much microfareds, sometimes the wrong type of capacitor.

[Al (astral highway)]

Originally Posted by Radio Wrangler

The problem is rarely not enough decoupling, at least on the output side, it is sometimes too much microfareds, sometimes the wrong type of capacitor.

Ahah, just to be clear, do you mean the instability problem I mention - and the one that was causing me the initial headache - is rarely caused by not enough decoupling on the output side, but rather by too many uF or the wrong type of capacitor?

I am using high quality multi-layer ceramics (MMLC's).

I'm aware that a transmission-line model is a good one for the situation we're looking at...

[Radio Wrangler]

Originally Posted by astral highway

I am using high quality multi-layer ceramics (MMLC's).

Forget words like 'quality' there are so many parameters it becomes meaningless, just marketing puff.

I think you are using low ESR multilayer ceramics. They are probably excellent, so good that the lossy effective series resistance that a lot of regulator designs rely on, isn't there.

You've hit the counter-intuitive world where some parts are too good to be useable.

We are now at the point where seemingly simple bits of DC circuitry need the designer/builder to be a bit streetwise about RF and control-theory matters. "We're not in Kansas anymore!"

David

[cmjones01]

There are some low-dropout linear regulators from TI, I think, which I used to use regularly. Their data sheet specified clearly the minimum ESR of the output capacitor. If the output capacitor was a ceramic, you had to put a 0.47R resistor in series with it to avoid exactly the problems David describes.

I have found that old-fashioned "high dropout" regulators like the 7800 series and LM317 are fairly tolerant of capacitive loads. More modern low-dropout ones with PNP or P-channel MOSFET pass elements, are much more sensitive to them.

The workaround is to read the data sheet and see what the manufacturer says will make the regulator happy.

Chris

[orbanp1]

Hi Al,

There is a National Semiconductor app note (now published by TI) discussing linear regulator theory and compensation techniques:
http://www.ti.com/lit/an/snva020b/snva020b.pdf
It is a worthwhile read.

Regards, Peter

[Al (astral highway)]

Originally Posted by orbanp1

Hi Al,

There is a National Semiconductor app note (now published by TI) discussing linear regulator theory and compensation techniques:

Thanks for the suggestion and link, Peter. I've taken a look and have also put a bookmark for it. It's thorough and very helpful.

[Al (astral highway)]

Originally Posted by cmjones01

I have found that old-fashioned "high dropout" regulators like the 7800 series and LM317 are fairly tolerant of capacitive loads. More modern low-dropout ones with PNP or P-channel MOSFET pass elements, are much more sensitive to them.

Thanks, that's a good insight. I've finally managed to read the absolutely tiny writing on the device and it's and MCP-1702, so a CMOS (hence also P-channel MOSFET) device.

Originally Posted by cmjones01

The workaround is to read the data sheet and see what the manufacturer says will make the regulator happy.

And curiously, the circuit designer ignored the data-sheet advice suggesting 1uF ceramic capacitor on both the input and output and instead put a a 0.1uF on the bill of parts, followed by a 50uF reservoir capacitor.

I noted in my OP the instability issues I was getting before I started to try and resolve them, and I wonder that this is the explanation.

[Al (astral highway)]

Originally Posted by Radio Wrangler

We are now at the point where seemingly simple bits of DC circuitry need the designer/builder to be a bit streetwise about RF and control-theory matters. "We're not in Kansas anymore!"

Incredible, really!

I knew I wasn't quite in Kansas. Now it all makes sense!