I guess, you forgot, that there is a non-negligible saturation voltage at output of ULN2803? Due to the Darlington topology you won't see a turn-on voltage down to 0V, but something arround 1V at your desired currents. So, 5V minus two diodes' voltage drops minus 1V Ucesat gives about 2V and not 3V, as Per already mentioned. This can mean that the LM (laser module) might not turn-on at all and refuses to draw current.

As you didn't tell what laser module you have, allthough I explicitely asked you to tell us, I can only guess now:

1. You have a LM that works within a Vcc range of 3V...>=5V, and you only want to limit the supply voltage, because of reducing the heat dissipation. Then, I would recommend you to omit one diode and to drive the ULN2803 much harder, by the help of an additional 74HC04 gate, like shown in the picture below.

2. You have a LM that expects exactly Vcc=3,0V (+/-0.2V, or so). Then, forget the idea with the Darlington (ULN2803) and diodes to reduce the supply voltage, but use a logic level N-MOSFET like the IRLZ24N, like shown in the picture below.





If you use a Darlington and diodes here, you will suffer from great changes of voltage drops across these components due to current changes and temperature effects, which can become critical for your LM and eventually damage it.

Take note, that the 74HC04 inverter is also needed to prevent the LM from being turned-on during reset conditions, when the micro will emit logic high level!

The 3V supply of second example can be either generated by a linear regulator like the LM317 (or better) or by a switcher like LM2675 (or better). Eventually, you can also use an On/Off option of switcher to turn-on the laser module, without the need of an additional N-MOSFET. But have a look at the switching characteristics then...

Kai Klaas