Continuing the testing of the USB inrush current protection, I
now tried a real-life capacitative load. From previous tests I
already know that lowering the reset voltage alone removed inrush
current issues for all but one device, namely the iCon i-Creativ.

I now tested it with the M1pre-rc4 that has the current-limiting
USB switch. The test setup was intentionally unfriendly, with the
lab power supply providing a voltage near the bottom of the
permissible range and enforcing strict current limits.

I could still produce a significant drop, but it wasn't bad
enough to cause a reset:

http://projects.qi-hardware.com/index.php/p/wernermisc/source/tree/master/m1rc3/usbpwr/usb-icreativ-50us.png

The oscilloscope probe used here has a linear error of a few
percent, so the real voltages are about 100 mV higher. So this
means that we're about 200 mV above the threshold voltage of
the reset chip.

With a power supply similar to the one shipped with M1, the
results were even better, as expected:

http://projects.qi-hardware.com/index.php/p/wernermisc/source/tree/master/m1rc3/usbpwr/usb-icreativ-50us-psu.png

More details can be found here:

http://projects.qi-hardware.com/index.php/p/wernermisc/source/tree/master/m1rc3/usbpwr/README2

Conclusion: I think lowering the reset chip's threshold voltage
combined with the current-limiting USB switch does provide
adequate protection of the 5 V rail under reasonable operating
conditions.

With the USB power switch, the M1 will withstand a certain amount
of abuse. Excessive loads on a USB port (i.e., a short) may make
it reset but cause no lasting damage.

What's next: we still need to verify that the USB power switch
doesn't cause trouble for properly behaving USB devices. I don't
expect to find any trouble there besides electromechanical
instability (due to the way this prototype was made), but I'll
try to use the M1pre-rc4 for USB protocol work, so it'll get
exposed to a number of devices.

- Werner
_______________________________________________
http://lists.milkymist.org/listinfo.cgi/devel-milkymist.org
IRC: #milkymist@Freenode

Reply via email to