Voltage regulator unstable when attach input cap

[sebmadgwick]

I have a battery device drawing ~300 mA with an on-board ADP3338 (3.3 V) regulator.  I found that when I attach a 1000 uF tantalum cap to input while the device is running, the regulator then gets very hot and outputs a voltage higher than 3.3 V.  This state persists even after I have removed the cap and will only disappear if I remove and reconnect the battery.  The problem does not occur if I power up the circuit with the cap attached.

I know that input/output caps and output load can be critical for closed-loop stability but I'd like to understand more about what is going on.  I'd be very grateful for any comments/links you can share.

[Marco]

Some type of latchup if the input voltage drops below the output voltage?

[dannyf]

Not sure whose "input" you are talking about. High capacitance on the regulator's input is generally not an issue - read the datasheet to be sure. High capacitance, coupled with low ESR on the regulator's output (load's input) can be deadly.

Low ESR is generally evil and should be avoided unless you cannot.

[Marco]

Input and output of the regulator.

My guess is that when he plugs in the cap it pulls down the input voltage of the regulator below it's output, with quite a high slew rate at that, causing it to latchup (which explains why it needs to power down before resuming normal operation).

[dannyf]

I doubt that's the case.

maybe the polarity is inverted, or a wiring or soldering issue. Or some silly issues if that caliber.

[sebmadgwick]

There are no soldering/wiring issues and the capacitor polarity is correct.  Furthermore, I have 4 identical arrangements (battery-powered device + removable-cap) and the problem can be replicated on each.

[SirNick]

Try adding a series resistor between the cap and the Vin to avoid that temporary short to ground when connecting.  Might also try a reverse diode between Vout and Vin to test that latchup theory as well.  It doesn't sound unreasonable to me.

[AndyC_772]

Connecting the cap momentarily causes the input to fall to 0V, and a very large current spike flows the wrong way back through the device. I'd bet on it being latch-up, the clue being that the condition persists until power is removed, which is absolutely characteristic SCR behaviour.

[wraper]

Connecting the cap momentarily causes the input to fall to 0V, and a very large current spike flows the wrong way back through the device. I'd bet on it being latch-up, the clue being that the condition persists until power is removed, which is absolutely characteristic SCR behaviour.

Or starts to oscillate. That would describe why it becomes hot. I would attach oscilloscope to the output and watch what happens there.

[sebmadgwick]

Having read a bit more, I'm wondering if the problem could be latchup.  I am working on location and will not have access to a scope until next week; though I would be very interested to plot the transient voltages and currents.

If the problem is latchup, what voltage difference should I expect to see between output and input and why?  Is the behavior when current flows the wrong way back through the device defined?  If not, then what could be speculated?

[T3sl4co1l]

Some type of latchup if the input voltage drops below the output voltage?

Seconded...or, fourth'd?

Try a schottky diode from output to input and see if it still happens.

Lesson learned: plan for transient conditions and add protective diodes and such where useful.

Tim