Text preview for : 34675C_lxdb800_spec_update.pdf part of AMD 34675C lxdb800 spec update AMD 34675C_lxdb800_spec_update.pdf



Back to : 34675C_lxdb800_spec_updat | Home

AMD GeodeTM LX DB800
Development Board
Specification Update

1.0 Scope
This document discusses known issues of the AMD Note: This is revision C of this document. The changes
GeodeTM LX DB800 development board. The table below between revision B (dated January 2006) are in
provides a summary of the issues. A detailed description of issue #3 and are non-technical. Minor rewording of
each issue, its impact and a recommended resolution/fix the Description and Resolution. In the Resolution,
follow. the application note that is referenced changed
names. Now called "AMD GeodeTM CS5536 Com-
panion Device USB 2.0 Device Linux Workaround"
with publication ID remaining the same (40471).


Table 1-1. Issues Summary
Issue
#(Note 1) Description

1 Limited High speed USB 2.0 functionality
2 Suspend-to-RAM non-functional
3 USB 2.0 high speed (HS) device defaults to full speed (FS) after wake from APM Suspend (S3)
Note 1. Issue numbers may not be sequential since issues are omitted once they are resolved.




34675C - March 2006 - Confidential 1
Specification Update

34675C - March 2006 - Confidential




2.0 Issues

1) Limited high speed USB 2.0 functionality that when the system is put in the Suspend-to-RAM
state it will not power off the ATX supply.
Description: USB high speed data transfers may not
complete. Resolution: The CS5536 signal WORK_AUX needs
to be connected to the inverter at R67 that controls
Implications: A signal integrity issue has been iden-
the ATX power supply main rails. This change will
tified with the USB 2.0 interface that affects all four
cause the system to power up instantly when the ATX
USB ports routed through the ETX connector X1.
power supply is connected for the first time. Power
There is coupling between the PCI AD bus and the
management firmware should configure GPIO24 to
USB data signal lines. This causes unwanted noise
be WORK_AUX resulting in correct system behavior
on the USB signal pairs that manifest themselves as
for the S5 and S3 states. The power button will then
spikes in the amplitude of the differential signaling.
be required to turn on the system.
The USB 2.0 specification states that high speed
signaling is to be 400 mVpp, and the EHCI controller This has been corrected in the production version of
must tolerate signaling with differential amplitudes the board.
<525 mV. Signals with differential amplitudes >625
mV must reliably activate the Disconnection Enve- 3 USB 2.0 high speed (HS) device defaults to full
lope Detector. On the LX DB800 spikes caused by speed (FS) after wake from APM Suspend (S3)
the coupling of the PCI bus to the USB signal pairs
exceed the EHCI controller's signal amplitude limit for Description: In APM Suspend (S3), both the EHC
high speed operation causing the disconnection of and OHC controllers are switched off on the CS5536
the USB device. The OS typically will show the companion device. The EHC spec implies that the
device as being disconnected and immediately CONFIGFLAG and the PORTSC[4:1] registers be
reconnects the device in Full speed mode. If an appli- auxiliary powered in S3. This is not implemented in
cation is in the middle of a High Speed data transfer our controllers, so the BIOS restores the contents of
when this occurs, the transfer will be incomplete. these registers before the BIOS gives ownership to
the operating system during resume from S3.
Resolution: Testing and evaluation has shown that
the addition of 18 pF to 22 pF capacitors on the USB Implications: The Windows EHCI driver does a full
data signals reduces the amplitude of the unwanted initialization and enumeration after resuming from
spikes to within the EHCI tolerable level (<625 mVpp) S3, so no impact was seen.
with a margin of approximately 0-25 mV on USB
The Linux EHC driver does not do a full initialization
ports 2 and 3. USB ports 0 and 1 signaling is not
sequence. Therefore, the ConfigFlag is not set and
significantly affected by the addition of capacitance
PORTSC.PortPower is cleared. A connected device
on the data signals. Testing has shown that USB
is then always routed to the OHC.
ports 0 and 1 may work at high speed, but could fail
with large data transfers. On the LX DB800 ETX Resolution: The workaround is to unload/load the
SOM, 0402 18 pF to 22 pF ceramic capacitors can driver around the sleep cycle. See the application
be added on the bottom of the PCB directly beneath note titled AMD GeodeTM CS5536 Companion
the AMD GeodeTM CS5536 companion device on the Device USB 2.0 Device Linux Workaround (publica-
resistor pads: R317