I made my own Arduino-like board. I have a 7805 on it to provide +5V, but sometimes I need to power the board by the +5V/GND pins, bypassing the regulator (using an external one). This works perfectly, but I was wondering: what happens to the onboard regulator when I feed +5V on its output pin? Can it get ruined? Should I somehow protect it? Maybe with a diode?
As a side question: my board will also need +3.3V. I have another regulator for that: is it better to power it in parallel with the 7805 or should I power it through the 7805? Its dropout voltage is low so it can also turn +5V into +3.3V without issues.
I think I have noticed that when +5V is provided on the 7805 output pin, ~7.8V seems to present at its input pin. I'm not sure I measured this correctly (I was looking for some incorrect wiring), so that's why I was in fear of ruining the 7805.
Regulating the 3V3 from the 5V has the advantage that I also get 3V3 when powering through the Vcc/GND pins, so I'll go that way if it causes no issues. But if "protecting" the 7805 is only the matter of adding a diode on its output pin, I might just throw that in.
It depends on the regulator. 1117's for example don't like that much. What you should do is to put a diode from the regulator's output pin to its input pin. Any diode will do.
Interesting conversation, I regularly power my self made boards through the 5V pin bypassing the regulator. I mostly use LM2940's, I have not had any problems so far. Can I ruin the LM2940 that way?
Very unlikely. In general, the ldos tend to have issues with that, and 1117 is probably an extreme example: it gets hot very quickly if powered reversely.
Where I work, we sell a board that can be powered by 5V, or with a higher voltage using an on-board LM2741 regulator (similar to an Arduino).
There is a "regulator bypass" jumper that simply shorts-out the 2741 input & output pins for a direct (externally regulated) 5V connection. Sometimes during test, we bypass and "back-power" the regulator (without the shorting-jumper connecting back to the regulator input). With this chip, there is no problem back-powering in either case.
Some of our other products use different (usually switching) regulators (and without the bypass jumper on the board), and when there's a "power supply problem", it's often helpful to temporarily bypass the on-board regulator with a bench supply (to see if there's a 5V to ground short, etc.) and I've never had a problem back-powering the regulator in those cases either.