I have a bit of an enigma with a Nokia 5110 LCD screen. The basics of it are this. When the back light of the screen is active ADC readings jump by 3 or 4. Very very basically I can demonstrate the problem by simply wiring up the Arduino (It's a Nano) to a 12V source (using VIN of course). Then use a 22K and 10K resistor as a voltage divider between the power supply with the node of those two resistors connected to A0. This should give me a reading of about 768 ish depending on how close the supply is to 12V and the accuracy of the resistors. This is indeed the case, however, if for example I get a reading of 770 with the LCD back light turned off, this reading will jump to 774 when I turn the back light on. Weird. The display is wired to the Arduino with all pins but ground connected by 1K resistors to limit the current (no I do not wire VCC to VIN, VCC is wired to +5V). The back light is wired to a data pin so I can switch it on and off. If I power the LCD display back light from an external source, it doesn't happen, the reading stays steady. So my guess is that using the Arduino power to light the display is causing a slight voltage drop on the +5V rail of the Arduino, making the AVCC drop slightly. If this is the case then the 5V regulator on the Arduino must not be particularly good.
Does this mean that my "box" will have to use its own dedicated 7805 regulator to AVCC to prevent this drop happening?