I'm trying to figure out the purpose of a current mirror. I read everywhere that it copies a reference current to another node but why do that? Why can't we just use that current source without connecting them to N-channel transistors? Also, why are 2 transistors used?
Current mirrors take a positive current (source) and generate a negative current (sink).
Current mirrors have low input impedance, so the 'reference' current can be from a poor current source (output impedance 1k ohms) and give rise to a mirrored current that is a better current source (output impedance 100k ohms). They isolate the input from variations in the output, so you can wire together a dozen current mirror outputs to make a summing junction, without causing any signal flowing back into the input nodes.
The illustrated current mirror makes an excellent pull-down load for an NPN emitter follower, sinking equal current at a wide range of output voltages (unlike a resistor pulldown).
The use of two transistors is not required, of course; you could use dozens, and the single reference input can drive two parallel output transistors to make a gain-of-2 mirror, or four to make a gain-of-four mirror (there are D/A converters built this way). Integrated circuit operational amplifiers often use a single current reference and current mirror to generate many bias currents. Matched transistors are very easy to integrate, and are more accurate than resistors in most IC processes.
Current mirrors with sense-resistor inputs are used as current monitors high side current sensor as well, with highly attenuated output current. The low input impedance, like an ammeter, is a major benefit.
Current mirrors have a wide range of uses.
Other possible reasons:
There are many answers to your questions. Regarding why don't we use the current source directly, one of the reasons is that the 'original' current source may not be strong enough to drive the load.
Another reason to use current mirrors is to stabilize the current against changes on temperature.
