XO1.5 WLAN SR ECO: Difference between revisions

From OLPC
Jump to navigation Jump to search
(New page: =Scope= This describes a change to B3 prototypes of the XO-1.5 motherboard. It fixes a silicon error which will not be fixed. Boards without this ECO will not reliably ma...)
 
No edit summary
Line 8: Line 8:


[[Image:XO1.5_WLAN_SR_Plain.png|thumb|Glitch in Power Control]]
[[Image:XO1.5_WLAN_SR_Plain.png|thumb|Glitch in Power Control]]
There appear to be two problems: a glitch on the power control line for the WLAN card and a drooping WLAN_RESET# line.
There appear to be three problems: a glitch on the power control line for the WLAN card, a drooping WLAN_RESET# line, and finally, too slow a rise time on the +3.3V rail when resuming.

===Power Control Glitch===


[[Image:XO1.5_WLAN_SR_Wcap.png|thumb|Driving Force]]
[[Image:XO1.5_WLAN_SR_Wcap.png|thumb|Driving Force]]
The glitch is a surprise, and is present in both OFW and Linux. It happens early enough in the resume process that it is likely hardware related. We take a look at it by disconnecting the motherboard pullup resistor on WLAN_EN#, and instead adding a 0.1 uF cap (see Driving Force thumbnail). This result in a WLAN_EN# signal that is driven by a very high impedance driver, such as the built-in pullup resistor.
The glitch is a surprise, and is present in both OFW and Linux. It happens early enough in the resume process (before PCIRESET# has been deasserted) that it is caused by hardware. We take a look at it by disconnecting the motherboard pullup resistor on WLAN_EN#, and instead adding a 0.1 uF cap (see Driving Force thumbnail). This result in a WLAN_EN# signal that is driven by a very high impedance driver, such as the built-in pullup resistor.


If we instead add a resistor between the VX855 output and the inverter, along with a filtering capacitor to ground, we can slow down the turn-off process enough that legitimate software usage is allowed, but this glitch is eliminated.
If we instead add a resistor between the VX855 output and the inverter, along with a filtering capacitor to ground, we can slow down the turn-off process enough that legitimate software usage is allowed, but this glitch is eliminated.
[[Image:XO1.5_WLAN_SR_Modified.png|thumb|After Modification]]
[[Image:XO1.5_WLAN_SR_Modified.png|thumb|After Modification]]

===WLAN_RESET===


We also need to populate a resistor to pull-up WLAN_RESET# to a sane value. This was predictable, and a resistor position (R120) is pre-wired.
We also need to populate a resistor to pull-up WLAN_RESET# to a sane value. This was predictable, and a resistor position (R120) is pre-wired.

===+3.3V Rise Time===

Finally, if the rise time of the +3.3V rail isn't fast enough, the card detect debouncer in the SDIO interfaces takes very long (> 500 mS) to settle on resume. This rise time is determined by a capacitor in the power switch (C253) which needs to be decreased in size.


=Procedure=
=Procedure=
Line 25: Line 33:
# Add a 0.1uF, SMD-0402 resistor between pins 1 and 2 of Q38.
# Add a 0.1uF, SMD-0402 resistor between pins 1 and 2 of Q38.
# Populate R120 with a 4.7K resistor.
# Populate R120 with a 4.7K resistor.
# Replace C253 with a 0.01 uF SMD0402 capacitor.


[[Category:Hardware]]
[[Category:Hardware]]

Revision as of 02:16, 25 November 2009

Scope

This describes a change to B3 prototypes of the XO-1.5 motherboard. It fixes a silicon error which will not be fixed. Boards without this ECO will not reliably maintain their WLAN upon resuming, and may crash hard instead.

This only applies to XO-1.5 B3 phase motherboards.

Problem Description

Glitch in Power Control

There appear to be three problems: a glitch on the power control line for the WLAN card, a drooping WLAN_RESET# line, and finally, too slow a rise time on the +3.3V rail when resuming.

Power Control Glitch

Driving Force

The glitch is a surprise, and is present in both OFW and Linux. It happens early enough in the resume process (before PCIRESET# has been deasserted) that it is caused by hardware. We take a look at it by disconnecting the motherboard pullup resistor on WLAN_EN#, and instead adding a 0.1 uF cap (see Driving Force thumbnail). This result in a WLAN_EN# signal that is driven by a very high impedance driver, such as the built-in pullup resistor.

If we instead add a resistor between the VX855 output and the inverter, along with a filtering capacitor to ground, we can slow down the turn-off process enough that legitimate software usage is allowed, but this glitch is eliminated.

After Modification

WLAN_RESET

We also need to populate a resistor to pull-up WLAN_RESET# to a sane value. This was predictable, and a resistor position (R120) is pre-wired.

+3.3V Rise Time

Finally, if the rise time of the +3.3V rail isn't fast enough, the card detect debouncer in the SDIO interfaces takes very long (> 500 mS) to settle on resume. This rise time is determined by a capacitor in the power switch (C253) which needs to be decreased in size.

Procedure

  1. Remove R325.
  2. Cut the trace between Q38, pin 2 and R325 (disconnecting Q38, pin 2 from the WLAN_EN# signal).
  3. Add a 47K resistor between WLAN_EN# (available at R325) and Q38, pin 2.
  4. Add a 0.1uF, SMD-0402 resistor between pins 1 and 2 of Q38.
  5. Populate R120 with a 4.7K resistor.
  6. Replace C253 with a 0.01 uF SMD0402 capacitor.
Retrieved from "http://wiki.laptop.org/mediawiki/index.php?title=XO1.5_WLAN_SR_ECO&oldid=224929"