The NSLU2 expects a regulated 5V supply at JS1. C160 and 168 are there to remove high-frequency noise. Q1 is reverse polarity protection, although not up to much more than -5V. C13 and C28 are bulk decoupling capacitors.
U6 and its associated components are a bit mysterious. The LM358 op-amp is being run open-loop as a comparator, comparing the reference voltage from the zener on the positive input with the voltage on the 12nF capacitor, which will rise very swiftly to 5V when power is applied. When it rises above the reference Q3 is switched on, applying power to U13, U15, and Q11. The delay imposed by this arrangement is very small, and it does not seem to depend on the input voltage. I have measured it with a scope at 350usec, which is too small to provide much debouncing effect, so the reason for it is unknown.
U13 is unidentified but is certainly some sort of switch debouncing/reset controller. It generates a debounced positive-going signal on pin 4 when the power-on button is pressed.
U15 is a latch that controls power to the rest of the NSLU2. A rising edge on the clock input will clock a 1 from the permanently high D input into the latch, turning on Q10 and thus Q5 which applies power to the rest of the system. Further clocks will have no effect. The CPU, once running, can monitor the clock input via R107, and respond as it sees fit. The state of the latch at power-on is determined by Q11, which is a reset controller IC. It will hold the CL input low, forcing a 0 into the latch, for anywhere between 140 to 500 msec. The CPU can turn its own power off by applying 3.3V to R10. This turns on Q2, reducing the voltage applied to Q11. As soon as the voltage goes below 3V Q11 will trigger and hold CL low for 140-500msec.
U15 and U6 are both dual components and the second latch and second op-amp are unused. Their inputs and outputs are left floating, which isn't recommended practice. Floating gates have a bad habit of forming spontaneous oscillators and using power.
Q5 once turned on applies power to the FP5452 dual switching regulator, which generates 3.3V and 1.3V supplies to the CPU and the rest of the circuitry. The FP5452 switches at about 115kHz.
Always use a regulated 5V supply to power the NSLU2. The NSLU2 has no transient protection other than the decoupling capacitors, and the 5V supply is fed directly to any bus-powered devices attached to the USB ports. Any voltage over 6V falls outside the legal USB range and may damage them. Voltages over 7V at the input will destroy the 74HC74 and/or the TS809. The FP5452 and the LM358 are more robust and should cope with up to 30V, but only briefly.
The 5V 2A supply provided by Linksys is comfortably overrated. The NSLU2 draws only 500mA at 5V, and the single bus-powered USB port is only allowed to draw another 500mA. In most circumstances a 5V 1A supply will be all that is needed.
Contributed by Nigel Spon
U6 and its associated components are just a low voltage cut out - both for power on and power off situations, power is not enabled until the input voltage rises above a minimum.
(Carl Willis 18 May 2008)
It may have been intended as that; if you put a scope on it and feed a slow ramp into the power input you will find that it doesn't actually work as such. This isn't very surprising as the behaviour of the LM358 with both inputs a few millivolts from the supply is undefined.