PPBus_G3H is the main power rail of the Mac. It is powered from either the battery or DC in. Its voltage will be based off how many battery cells are in series. The MacBook Pro series uses a 3S2P battery so on the charger it has a PPBus of 12.6V (12.2V before SMC boosts). The Air has a 2S3P battery so on the charger it has a PPBus of 8.6V (8.2 before boosting). This is to charge the battery, which will have a lower voltage than PPBus
PPBus voltage on the battery will equal the current voltage of the battery. It is enabled from a MOSFET controlled by an ISL62xx through a current sensing resistor. This circuit controls how quickly the battery charges. If any of those components are damaged it may not run off battery or charge.
The ISL that creates PPBus is powered from PP3V42_G3H. This rail is created either from the charger or battery. Since it is G3H rail it is always on when the Mac has power. All the power rails are created from this. Always check for #PPBus_Shorts when you do not have full PPBus voltage.
PPBus Current Sense
Measure resistance of the current sensing circuit at the PPBus Supply and Battery Charging chip. It is typically an ISL62xx.
Pins 17/18 measure the battery charging/discharging current.
Pins 27/28 measure the charger input current.
Use the schematic to verify total resistance by adding the current sense resistor value and the value of the resistor on each pin.
If the resistance is off by more than 10% check the individual passive components. Remember to account for the resistance of your leads
injecting 5V on PPBUS_G3H (I think that's what you meant) with a shorted high-side MOSFET = melting CPU/GPU/PCH
So no, it is ABSOLUTELY NEVER safe to inject a high voltage on this kind of rail. First you have to check exact resisntace to ground, can give an indication if it's a full short on the rail, or a partial short through a MOSFET.
If a partial short through a MOSFET is suspected, you have to find it first, so check resistance to ground on all the inductors, if there is a similar value check resistance between that inductor and the main power rail. If that gives an almost 0ohm resistnace there is a high-side MOSFET shorted that you have to remove first
Otherwise always start at 1V and check if any large BGA is heating up first. If 1V is not enough to generate heat (remember that it's the power which is voltage times current that generates heat, not the voltage or the current alone) then you can try increasing little by little, checking again if any large BGA heats up.
Better be on the safe side, even if a shorted high-side MOSFET almost always means a dead BGA chip on newer platforms.
[5:31 PM] piernov: Let's take a concrete example: I'm working on an Haswell platform, I suspect a short to ground on main power rail because it is missing, I measure resistance to ground and I get 50 ohms. What can I conclude? This is a partial short, injecting voltage is not necessarily helpful in those cases, and it sounds like the resistance of CPU VCore. So I check resistance of CPU VCore to ground and I get 50 ohms as well. Then I check resistance between main power rail and CPU VCore and I get 0 ohm. Only then I can rightfully conclude I have a shorted high-side MOSFET on CPU VCore and I should not inject voltage but replace all high-side MOSFETs and their driver/controller (CPU is probably dead anyway though). [5:34 PM] piernov: Now, same platform, still suspecting a short to ground, but I measure 0.3 ohm to ground. Then this is a full short to ground, and it is unlikely any other power rails is measuring 0.3 ohm short to ground, so it is unlikely that I have a shorted high-side MOSFET. I don't have to waste my time measuring resistance between main power rail and each inductor. I can start by injecting 1V on the main power rail, monitoring current to know what power it is dissipating, and trying to locate the component that heats up, but still checking CPU/GPU/PCH first, you never know… If the power is not enough, I can slowly increase the voltage until I get a satisfying power dissipation (let's say 3W). [5:35 PM] piernov: This 2 to 5 basic measurements with the multimeter to know what's happening, before injecting voltage, before any soldering, before destroying the board. And it takes a minute to do at worst.