#9566 NORM Not Tri: Discharge USBVDD rail

[Zarro Boogs per Child]

#9566: Discharge USBVDD rail
-------------------------------+--------------------------------------------
 Reporter:  wmb at firmworks.com  |                 Owner:  wad        
     Type:  defect             |                Status:  new        
 Priority:  normal             |             Milestone:  Not Triaged
Component:  hardware           |               Version:  1.5-B2     
 Keywords:                     |           Next_action:  design     
 Verified:  0                  |   Deployment_affected:             
Blockedby:                     |              Blocking:             
-------------------------------+--------------------------------------------
 We need some way to bleed the charge from the 100 uF filter capacitors on
 the USB power rail.  See #9423 for related information.

 Mitch says:

 While investigating http://dev.laptop.org/ticket/9423 ,
 I realized that the USB power rail USBVDD lacks a discharge
 path in the case when no devices are plugged in.  When devices
 are plugged in, the discharge waveform is unpredictable,
 depending on nonlinear characteristics of the specific devices.  In some
 cases, it can take several seconds for the
 power to decay, with the voltage hovering in "no man's land" for quite
 awhile.

 That was the cause of intermittent USB init failures that I sometimes
 see with the D-Link USB ethernet dongle.  When the EC reboots the
 system, the USB power rail is turned off for about half a second, during
 which time the rail discharges part way, just enough to make the device
 lock up sometimes.  I fixed that particular device by extending the off
 time to 2.5 seconds (overlapped with other OFW startup tasks), thus
 giving the rail time to discharge "enough" for that device.

 I doubt this is a robust solution, as USB devices are all over the map
 with respect to their supply current, and the discharge waveform often
 has breaks at different voltage levels.  The startup delay would have to
 be very long to guarantee dropping to a "low enough" voltage, whatever
 that is.

 I'm not sure what the solution is.  I guess the ideal would be to have
 some form of active discharge to drain off the three 100 uF capacitors
 when you disable the power.  A simple bleeder resistor might be a good
 idea just for safety's sake, but if it is small enough to discharge the
 caps fast enough to help the device-reset problem, it will dissipate too
 much power in the on case.

 Wad says:

 We have active pull-downs on most rails, we can add one
 to USBVDD, but not for B3...

-- 
Ticket URL: <http://dev.laptop.org/ticket/9566>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system