In order to determine at what reset voltage we'd be safe with a design monitoring only the 5 V rail, I've tested what effect connecting USB devices has on the 5 V rail.
In all these experiments, the M1 was either sitting at the desktop or rendering. I didn't see the 5 V rail change noticeably between desktop and rendering. I measured 5 V with a clip directly at the reset chip and with the ground reference on a capacitor nearby. The clip has a ~30 cm unshielded cable, so we pick up a bit of noise. http://downloads.qi-hardware.com/people/werner/m1/rst/rst-5v-tap.jpg My M1 is powered by a Minato 5 V 2500 mA switching power supply. This is not the official M1 power supply and I don't know how it compares to it. In standby my scope sees the 5 V rail around 4.97 V. Under load (desktop or rendering), it drops by about 200 mV. Let's begin with an old generic USB mouse. It has "Kensington" written on it. lsusb calls it "Novatek Microelectronics Corp". Rubberized ball, three buttons, no scroll wheel: http://downloads.qi-hardware.com/people/werner/m1/rst/m1-acmemouse.png The bottom cursor is at the nominal threshold voltage of the reset chip. We can see that the 5 V rail drops a little but all is well. Next, my Rii RF mini keyboard with integrated touchpad. The RF dongle is a little greedier but still doesn't cause trouble: http://downloads.qi-hardware.com/people/werner/m1/rst/m1-riirfkbd.png Then I tried a USB-MIDI controller, the Korg nanoKONTROL2. This is a full-speed device. This time we got close to the limit: http://downloads.qi-hardware.com/people/werner/m1/rst/m1-nanoKONTROL2.png Finally, I connected the flagship of my collection of input devices, a HHKB Professional. This is also a full-speed device. http://downloads.qi-hardware.com/people/werner/m1/rst/m1-hhkb.png This time things went too far and the reset chip had to do its duty, kicking the M1 back into standby. (Or at least that's what I suppose happened - I didn't check the reset signal.) This happened each time I connected the HHKB while none of the other USB input devices I tried had any issues. Outside the competition, I've also tried my atusb: http://downloads.qi-hardware.com/people/werner/m1/rst/m1-atusb.png As we can see, it has a hidden talent as master assassin. Its direct capacitative load on VBUS is 10 uF, with 4 uF waiting on the secondary side of the main regulator and another 2 uF behind additional regulators. Conclusion: since I used a 3rd party power supply, the results are only indicative. They show that the 5 V rail of M1 can drop by about 400 mV when connecting USB devices that can be considered light power users. This means that a reset circuit operating from the only externally regulated 5 V input with a 4.4 V threshold voltage is likely to experience resets when connecting USB devices. If we stick with the idea of monitoring only a single input, which would then have to be the 5 V rail, I would recommend using a reset chip with a threshold voltage of 4.0 V. This will still let the regulator of the 3.3 V rail produce a 3.0 V output under worst-case conditions, while providing a safety margin for USB devices with high inrush current and for power supplies providing a low "5 V" voltage. These experiments should be confirmed with other power supplies, including an official M1 power supply. - Werner _______________________________________________ http://lists.milkymist.org/listinfo.cgi/devel-milkymist.org IRC: #milkymist@Freenode
