Let this be a lesson why you should always fully condition inputs and outputs to your boards.
On your last schematic I still see that 24V line going directly to an OPA2180 amplifier, with some small capacitors attached. If you're confident those won't create nasty spikes during hot plugging (cable = L, cap = C, LC tank oscillator, as Tim pointed out already), then keep it like it is.... But I would connect that opamp after your 22R/1uF filter. It also will filter against mid-frequency ripples on your power supply.
Input protection (ESD and TVS diodes) plus EMI filtering seems to be absent. Although this board only contains linear components, you don't want any oscillations or incoming switching noise be amplified to the outside world.
The idea is to test this as close as possible to what will be used in the final implementation. But probably, I'll have to use a PI filter after the DC-DC converter. The 22R resistor is there, as will be in the final application, just to protect the regulator. I think the issue might be caused by inrush current, and not by inductive spikes alone. In summary, a PI filter to smooth the voltage going to the whole system, and a local 22R resistor just for the regulator.
The 22 ohm resistor provides damping for the LCR setup that includes your board, wiring and input capacitor of the regulator.
You can view this as the inrush current of the capacitor charging up it's voltage also simultaneously "charging up" the inductor's current. Namely, the inductor current will rise because the input potential is higher than the output. Once the capacitor voltage has stabilized, the inductor current has "charged up" to it's maximum, since both sides are now equal. However, for the "charged up" inductor current to dissipate, the inductor needs a negative voltage in order to do so. Since your bench power supply will probably be incredibly low impedance (it's not going to change), the voltage on the capacitor will increase. This "discharge cycle" of the inductor can cause the capacitor voltage to spike up to 2x it's nominal applied value. This is somewhat lessened by any quiescent loads applied on the circuit, so it's rare to see an actual 100% overshoot, but nonetheless it's something to take into account.
If you hot plug this 30V (35V max) regulator into a 24V supply, you can expect that it will damage the regulator. Trust me, I've been bitten by this bullet very hard once that costed me about half a dozen fancy regulators and 2 prototype boards.
The poles for a series LCR circuit are -R/2L +/- sqrt(R^2 - 4L/C). If you want a over-dampened system (i.e. no overshoots), you'll want to have R^2 - 4L/C > 0. With L=3uH (guessing you used roughly 3 meters of cable) and C=1.2uF, you get R^2>10, so a 3.3 ohm resistor should be sufficient. I read that you tested a 4.7R resistor also worked fine in protecting the regulator.
See the attachments for the circuit in LTspice, then simulation with all series resistors 100m ohm (voltage source, inductor and capacitor), then final simulation with L1 having a series 3.3 ohm resistance. R1 simulates any small load that might be present, but honestly does little in this case.
Counterintuitively, another solution would be to not use a ceramic capacitor, but a "crap" electrolytic capacitor instead.
I don't think that the in-rush current on the output of the regulator is going to damage it. Namely the absolute max specs says that the output current is "internally limited". The device is thermally protected against long overloads. In my experience you will need to very grossly overload a part for it to fail instantly, and applying over 10V it's nominal rating can certainly cause it to very quickly degrade (i.e. a few power cycles) or instantly fail.
Home
Help
Search
Login
Register
