If your laser diode does not have electronics to regulate the current, then the current through the laser will vary violently depending on very, very small voltage changes. This is similar to a LED. It has a sharp knee where it for low voltages doesn't draw any current at all, and then suddenly the current goes through the roof.

A laser module with proper electronics on, should be fine as long as you have enough voltage in. A tiny bit too little voltage so the internal regulation can't do the job, and you will see the laser dim down greatly. In yuor case, you have 5V sent thruogh two diodes and one ULN. Each diode may give a voltage loss that will depend on how much current and how hot they are. Depending on diode etc, you may have a voltage loss of 0.6-1.2V for each diode. And the ULN has also a very large loss that depends on temperature and also individually between different ULN chips. So in the end, your voltage loss may be way more than 2V in which case you have less than 3V to the module. If you then load your power supply so it dips 50mV, these 50mV may represent a huge difference in current through the laser diode, and hence in light intensity.

You have always said that the laser module is a 3V module. But you have never once really expanded on that figure. Is it a module that requires a very exact voltage - like 2.95 to 3.05V? Or is it a module that requires around 3V - for example 2.5 to 3.5V? Or is it a module that requires at least 3V but can handle a lot of extra voltage - such as 3V to 6V? Or is it a a module that has an expected voltage drop of 3V when supplied with a 300mA current, i.e. where you must feed it from a constant-current generator?

Where is a link to the datasheet for the laser module? Or do you have a schematics for it? Any other information? Just saying 3V isn't enough, since 3V may mean so many things.


Yet another thing you seem to have forgotten to tell us. You are now saying that you have an 8-bit ULN and that you use four output signals from the processor to drive two ULN inputs/each. That very much affects the load on the processor pins. How can we even begin to consider the needs for pull-up of the processor pins, if we don't know the load on the processor pin?