Supply filters

Hah, 2.2 million in stock at Digi-Key. That's $80k worth, at 3.6 cents each. And they have 880k of the 10-volt version. The datasheet says there's a 10uF version - no inventory.

Power input protection is a chronic problem. Lately I use a 24 volt sacrificial TVS and supply a current-limited 24 volt wall wart and assume disasters will be infrequent.

Polyfuse and TVS is pretty good. It can be teased to fry the TVS, but that takes creative malice.

What's your voltage and current? How much voltage can you lose?

One possibility is to use a depletion fet or an equivalent LED current limiter and a coupled thermistor to shut off the load if it gets too hot.

I have used a thermal-limiting 3t regulator as a current limiter, with the adjust pin open. An RRO opamp would work too for modest currents. But I prefer to not have the first power input device be an IC.

We have used the TI e-fuse things, but I don't like them.

Or use a real fuse!

Careful with LP29xx. I've had them go berserk under certain conditions, such as if the input (drive side) impedance gets too high.

TVS have a large tolerance. In critical situations I use a TL431-based shunt regulator. If currents are low enough then a barefoot TL431. This is often important because some high-speed logic and RF parts can go PHUT when it's just a few hundred mV above abs max. Or worse, they only go partially kaputt and have compromised noise behavior or something.

The product is an improved plug-in replacement for a commercial PMT module widely used in biomed instruments, which is why the weird supplies. We'll probably do a follow-on that runs off +5 only. The modules have voltage-programmable gain, which they get by changing the PMT bias. We do it using the MPPC bias voltage and a PGA made from an AD5273BRJZ dpot (1k, 64 taps of which we use 33), an ADA4899 (first stage, gains 1, 2, 4, 8) and an ADA4860 CFA output stage (gains 1 or

Our faster MPPC box for the cathodoluminescence application uses a

The MCU owns the power supplies, so that wouldn't be hard to do. My circuit is all super tiny packages, except that I'm splurging space on

That would need a couple of gigundo TVSes that I don't have space for, and I've only got about half a volt of headroom before I hit the abs max spec on the TMUX1511s I'm using in the PGA. (Remarkable part BTW.)

Cheers

Phil Hobbs

Interesting. I've never had a problem with one--I nearly always power small micros off the LP2951-33DRGR, with an NPN booster if necessary. It's good enough for a voltage reference for the sort of undemanding jobs that MCU ADCs are good for--checking the other power supplies, current limits, sloppy digital feedback loops, lowish-performance temperature controllers, and such like.

The TVS voltage isn't critical, because it's across the inductor rather than across the supply. I just want it to be less than 5V on the nastiest peaks. That way the circuit won't blow up if the power supply opens up or we turn off the PFETs suddenly. (The plan is to PWM them during turn-on and turn-off, but it'll probably happen during software development at least.)

Yeah, that would be bad. The TL431 is schloooww, though. I suppose I could just crowbar the supplies if they go more than half a volt out of spec, but that would make me quite unpopular--some instruments will use

I thought about just using analogue control of the back-to-back FETs, but they're so tiny that they won't stand much dissipation--the SOA curve is practically indistinguishable from the Y axis. ;)

Cheers

Phil Hobbs

Too slow for a power rail? Then you seriously need more bypass capacitance :-)

With a shunt regulator set a few hundfred mV above 3.3V there wouldn't be a crowbar effect, it just swallows the peak and then goes back into its guard house.

If the spike comes from a simple connection issue there shouldn't be a SOA concern. Unless of course the cause is a loose contact.

I don't have control over the input supply. People sometimes power our stuff from truck batteries. It's the peak of the ring waveform I'm most concerned about.

The PFETs are DMP4047LFDE's, which actually aren't super delicate, so you're probably right that I'm being a nervous Nellie about that. I'll see if I can magic up enough board space.

Cheers

Phil Hobbs

These ought to be small enough:

Just don't try to hand-solder them :-)

All this almost screams for a shunt regulator. Let it shunt brief spikes and if you suspect longer overvoltages you could do this:

Detect whether the shunt draws current which can be done with a simple BJT plus small resistor. If it draws current for more than xxx microseconds cut the power by shutting off the series FETs.

They're all 5%, though. :(

I've already got voltage dividers so the MCU can look at the input supplies, so I can just use a dual OC comparator to overrule the MCU signals. Look, ST makes an LM393 in 2x2 DFN!

Thanks for the steer.

Cheers

Phil

I use a resistor across the L to damp ringing but thanks for the TVS idea - cool!

piglet

They're in huge so8 packages, there are some nice sot-23-6 duals, e.g., BSL314PE.

Maybe I have the wrong p/n. The ones I'm using are in 2x2 DFNs.

Cheers

Phil Hobbs

A capacitor coupled to the input with a diode and a pull up on the cap to keep the diode reverse biased normally

During a pulse the cap clamps the surge

Add a resistor to bleed the cap

The Yageo line comes in 1% but long lead times:

No space unfortunately.

Thanks

Phil Hobbs

Aw, I had the wrong suffix. Three really quite different parts, by changing a few letters at the end.

er ESD

onably

rse.)

at inrush,

wever,

oes

ny

Is there any difference between a TVS and a couple of power zeners in series? (Well besides the obvious.. one versus two parts.)

George H.

A unidirectional TVS is a power zener. Bidirectional ones look like two zeners back-to-back.

Cheers

Phil Hobbs