[Solved] Agilent E3611A plug-in voltage overshoot

[mbless]

I have two E3611A that produce a voltage overshoot when I plug in the banana plug. This can be seen in the attached pictures where set point 5V can reach 8.5V and set point 12V can reach 21V.

Since the E3611A doesn’t have an output enable switch I have three ways of using it.
1. Turn it on with the voltage pot at 0V and increase the voltage to the desired point.
2. Turn it on with the load disconnected, set the voltage, turn it off, connect the load, and turn it back on.
3. Turn it with the load disconnected, set the voltage, and connect the load.

I use method 1 for first-time testing a new board, but after that I don’t go that route since I don’t want to wear out the 10-turn pot and get carpal tunnel. I use method 2 for large loads to avoid contact sparking. For low loads (~100 mA) I use method 3. Sure it’s not ideal due to contact bounce, but I haven’t had any problems until yesterday.

Yesterday I was testing a TI LM5134 low-side gate driver, which I have already tested and was working. When I plugged the 12V line into the PSU (method 3), the current limit was instantly hit at 0.2V! The 12V rail was shorted to ground. After pouring some IPA on the board I found the culprit to be the LM5134. I though it was a fluke, so I switched to another working board, confirmed VDD was not shorted and plugged in the power. Boom, another one bit the dust.

At this point I figured it was the power supply causing the chip to internally short. After removing the chip, I probed across the decoupling cap using the probe ground spring for minimal ground loop and got the attached images while plugging the banana plug into the PSU. For a set 12V, the voltage overshoots to >20V. The datasheet for the LM5134 states a recommended max of 12.6V and absolute max of 14V. So the chip getting >14V for ~3us is enough to kill it.

I investigated this issue a little more to make sure I wasn’t crazy. I tested a purely resistive load and there was no overshoot there; it only happens with some capacitive loading. Turning on the power supply (method 2) results in a smooth voltage rise to the set point with no overshoot. Interestingly, probing at the PSU binding posts instead of on the PCB only shows ~0.2V fluctuation. The datasheet does show load transient measurements with that configuration.

Before anyone asks both E3611As were built in 2009 with higher serial numbers than all of the recommended modifications, meaning all modifications were done at the factory. None of the service notes talked about power-on or plug-in overshoot.

So my questions concerning this issue are:
1. Is this the expected behavior of a PSU without an output enable switch? Am I supposed to use the power button as the enable switch (method 1)?
2. Why does probing at the PSU show minimally overshoot? The power wires to the PCB are 1m in length, can that have something to do with it?
3. If this is unusual, is there a modification that can be done to prevent the overshoot?

[mbless]

I played around with this a little more to no avail. I wired the power supply to a breadboard with a resistor and capacitor in parallel across the power rails. I plugged in another capacitor across the power rails and I got 2V of overshoot. Again I don't see this overshoot when probing directly at the output of the power supply which makes me think it's not real, although I am measuring it and it killed 2 ICs 

It seems like I can't hot swap parts anymore and always have to turn the power supply off. Yes, that is good practice but quite annoying.

[Omicron]

What you see is not due to the power supply. The leads you use to connect your power supply to your breadboard form an inductor. Together with the capacitor and some resistance on your breadboard you get an RLC circuit. You scope simply measures the response of this circuit to a voltage step. You don't see it when you use the power switch of your PSU because in that case the PSU will ramp up the output voltage (relatively) slowly. When you just plug in the banana's plugs on the other hand you get a fast step and the classic RLC response results. There is a Linear Tech application note that talks about this:

http://www.analog.com/media/en/technical-documentation/application-notes/an88f.pdf

[mbless]

What you see is not due to the power supply. The leads you use to connect your power supply to your breadboard form an inductor. Together with the capacitor and some resistance on your breadboard you get an RLC circuit. You scope simply measures the response of this circuit to a voltage step. You don't see it when you use the power switch of your PSU because in that case the PSU will ramp up the output voltage (relatively) slowly. When you just plug in the banana's plugs on the other hand you get a fast step and the classic RLC response results. There is a Linear Tech application note that talks about this:

http://www.analog.com/media/en/technical-documentation/application-notes/an88f.pdf

Thank you, this is exactly what I needed to read! I must have been using rather insensitive chips before since I typically plug usb cables into power blocks that are already in an outlet.

A follow up question. The application note talked about changing the capacitor and maybe adding a voltage suppressor to damp the circuit. Would it be more effective to add a diode to clamp the voltage since different power cables would change the RLC circuit?