Relays present an inductive load, this means that they resist change in current.

Switching on the coil should not be a problem, the coil current should ramp up over a microsecond or so.

When you switch off the coil, the current cannot stop instantaneously, hence the development of back EMF.

Assuming the coil is connected to +V and the driver is a transistor between the other coil connection and 0V, it is normal to connect a diode with cathode to +V, anode to the other coil terminal.

This stops the back emf from rising higher than +V + 0.7V. The diode should be able to handle the peak current, a 1N4148 is normally OK, for bigger relays a 1N400x.

But if your diode is actually in the ULN driver chip, check the PCB/wiring layout, look at the current path that that current pulse is going to take (ie through the coil, back down to the ULN, into the power rails, and back to the coild again). Check for two things:

- where does the current create a voltage pulse due to residual resistance? If that pulse gets onto your power rails for internal circuits, that can be your problem.

- does the current path run near to and parallel (or nearly so) with any internal signal lines? If so you may be getting induced signals.

In any case, I would always put a diode on the relay itself, right close to the coil terminals ... for exactly these reasons ... and diodes are very cheap. Just solder one on and see if your problem goes away.

If this doesn't work try disconnecting the relay load and seeing if your problem is still there ... remove relay and see if problme is still there ... usual process of elimination stuff.

Let us know how you get on.