My RPi does not start up!!

It would be good to check the voltage at the test points TP1 & TP2 if you have a voltmeter or multimeter.  If it is below 4.75V, then it might be that the power supply you are using it not capable of powering the Pi.  More info on power problems at: http://elinux.org/R-Pi_Troubleshooting#Troubleshooting_power_problems

Drew, your response suggested a cute idea for Pi version 3, given that the Pi is aimed at technical education ...

Build in to the standard board a very cheap 0-Vcc high impedance buffered voltage sensor with say 2 or 4 channels using the GPIO inputs, and by default have one channel connected to VBUS-VIN, another to TP1, and so on, for all of the important voltages on the board, subject to available sensing channels.

As soon as the VideoCore has started running and initialized its display, it can show these voltages even before the Linux kernel has begun to boot, and long before the ARM's devices and USB peripherals have started up (which is the point at which many power-related problems start to arise).  I don't know whether the VideoCore has the ability to freeze the ARM, but if it does then during Linux startup it could freeze the CPU immediately on sensing supply voltage plummeting to its death.   Using this approach it should still be able to continue displaying the voltage info despite imminent voltage collapse --- very helpful !!!

(The chicken-and-egg issue of the GPIO peripherals being part of the ARM can be addressed in various ways, perhaps initializing them early as part of bootstrap, or doing this entirely in the VideoCore since it has an I/O system of its own which it uses to read the SD card.)

Of course, the VideoCore being undocumented and extremely sekrit and available only to accredited developers, the open community cannot do this kind of thing for itself, a clear example of how closed systems suck.

Still, it's an interesting idea with educational value.  The voltmeter could even be used as an instrument for rough external measurements once Linux has booted successfuly.  Perhaps this could be done for some ARM board that doesn't use the closed Broadcom SoC --- maybe the new Freescale i.MX6?

It would certainly help early diagnosis of power supply issues on Pi, which are very common when a USB charger is used.

Morgaine.

The idea of measuring the voltage is good, but having a board with a wide input voltage range is better.

It's not only chinese adapters that will cause problems. Even the european certified chargers are allowed to go as low as 4.85V under full load. This doesn't leave much margin for the polyfuse voltage drop. While the Pi might still work at lower voltages, It's hdmi signals lower in amplitude as well, so some monitors might refuse to produce a picture untill you give a hdmi_boost in config.txt.

If they want to do it good, they need to redesign the Pi power input circuit, or they need to provide a separate module that allows a wide input voltage range. While they are up to it, they might consider adding some stuff to ensure a proper shutdown as well. I don't think childeren will be pleased if they find out their sd card with all their created stuff became corrupted. Maybe a cheap network drive in the classroom for backups?

If they want to do it cheap, they can leave everything as it is now.

I agree completely, redesigning the power circuitry to accept a wide voltage range is by far the best improvement they could make.  And needless to say, it should also isolate the SoC and the supplies it provides on P1 from fluctuations caused by USB devices as well.  The current lack of voltage regulation is not good engineering.

I agree completely, redesigning the power circuitry to accept a wide voltage range is by far the best improvement they could make.  And needless to say, it should also isolate the SoC and the supplies it provides on P1 from fluctuations caused by USB devices as well.  The current lack of voltage regulation is not good engineering.