Cian,

You may not be really aware of this but the catch diodes inside the ULN2003a are put there primarily to protect the IC chip. Unless the IC chip is placed almost on top of the relay coil contacts (i.e. _very_ short connections) I would suggest that you add an additional catch diode directly across the relay coil connections. This can avoid unintentional current spikes in the traces of your PC board that could couple to adjacent circuits.

Of course you have to also use design caution with the current travelling the other way in the trace at the time the ULN2003a output goes low. In any case the diodes are way lower cost than any unexpected interrupt generation causing product problems once deployed to the field.

< side note >
Many years ago I worked on an 8035 (8051 with external EPROM) design that used a chip similar the your ULN2003a. In that design I had placed the GND pin of the driver on the same leg of a routed power bus as the GND pin of the EPROM. Every time the ULN2003 type device would energize the GND bus would locally get a bounce and glitch on it and cause the concurrent code fetch from the EPROM to get corrupted thus crashing the 8052 program execution. At first I had only the symptom that the code crashed unexpectedly maybe once in a few days. It wasn't until I made a special version of the code that toggled the relay driver at a 10 Hz rate that I was able to isolate the glitch that caused the problem. That experience from a long time ago really taught me the value of a full GND plane in PC boards!

Michael Karas