2. EVMK2HX Rev 1.1 Known Issues

EVMK2HX Known Issues
1. Software and Firmware Version List
2. EVMK2HX Rev 1.1 Known Issues
2.1 EMAC Link Down Issue
2.2 VCC0V85 Voltage Drop Issue
2.3 I2C SDA/SCL Reserve Issue
2.4 Incorrect UCD Setting
2.5 SOC Power Fail
2.6 USB3.0 Performance
3. EVMK2HX Rev 1.1 Hardware Modifications
3.1 EMAC Link Down Issue
3.2 VCC0V85 Voltage Drop Issue
3.3 I2C SDA/SCL Reserve Issue
4. EVMK2HX Rev 2.0 Known Issues
4.1 EMIF and NAND Test cannot be synchronized
4.2 USB Device not work
4.3 SoC PWR stop on Power-up
4.4 Warning Messages in TFTP test on NAND Flash
5. EVMK2HX Rev 2.0 Hardware Modifications
5.1 Design Changes
5.2 EMIF and NAND Test cannot be synchronized
5.3 Power up will happen SoC PWR stop
6. EVMK2HX Rev 3.0 Known Issues
6.1 Warning Messages in TFTP test on NAND Flash
7. EVMK2HX Rev 3.0 Hardware Modifications
7.1 Design Changes
1. Software and Firmware Version List
EVM Revisions could be identified by sticker on RTM Connector, and the PCB Revision is
located on bottom of board near GPS Antenna Input
The initial firmware versions on Rev 1.1 EVMs are shown as below:
EVM Revision SN Range
Rev1.0
Rev1.1
Rev2.0
Rev3.0
ESE0075202 –
075236
EPD0082118 –
0082227
EPD00856820090157
EPD00911970093719
LINUX-
BIOS-
MCSDK
MCSDK
Rev
(NAND)
(NOR)
A102
1.0.1.3a
N/A
Alpha7
A102
1.0.2.5
N/A
Alpha7
PCB Rev
BMC
DSPM8305E
13-Feb PG1.0
22-May PG1.0
Ship Date
Silicon
Rev
3Q 2013
PG1.1
A103
1.0.2.5
N/A
Alpha7
1Q 2014
PG1.1
A104
1.0.2.5
N/A
3_00_03_15
Updated firmware/software could be available on Advantech website or TI MCSDK website.
2. EVMK2HX Rev 1.1 Known Issues


2.1 EMAC Link Down Issue
Description: Noise on MDCLK signal impacts stability of the MDIO link. When Linux
kernel sees MDIO link go down it tears down data link as well. Have not seen an issue
with using MDIO link to configure external PHY.
Workaround: Different options under investigation.
1. Modify code not to depend upon MDIO link remaining active for EMAC
communication
2. Add stronger pull-up resistors on MDCLK and MDIO signal, by replacing R399/R400
from 4.7 kohm with 220 ohm resistors (0402 sized)

Plans for a fix: Will clean up routing of MDCLK signal on Rev2.0. BOM will also be
changed on future builds to incorporate 220ohm resistor pull-up changes.


2.2 VCC0V85 Voltage Drop Issue
Description: The trace width of the VCC085 signal is too narrow and causes a voltage
drop at the pin of the SOC. Since this supply is used for SerDes interfaces it could
cause SerDes interfaces to lose link. This issue has only been identified and observed on
the SGMII.
Workaround: A couple different options are still under investigation for their
effectiveness.
1. Add external wire to provided adequate patch for expected current from pin 1 of
“C503” (net VCC0V85) to pin 1 of “C564, C574” (net VDD0V85). (Wire length = 5 cm)
Note: That the rework should not link to GND and cause short circuit.

Plans for a fix: The trace width will be fixed on Rev2.0. The filter element on AVDDALV
requires an increased current capacity (suggest up to 4A) to allow for margin on
worse-case power as per latest Snowbush power estimates when all SERDES are active.
Increase of current capacity on filtering elements for AVDDAHV also recommended
(suggest up to 3A). Please see latest estimates below. CVDD power budgeted already
on the DSP AVS supply. Both of these nets shall be implemented as copper pours up to
the filters and as a plane afterwards to minimize resistance and inductance and
maximize plane capacitance.
AVDDAS (0.85V)
Total per Kepler
2.951016043
AVDDHVS (1.8V)
1.274949495
CVDD (AVS)
0.005754011
Unit
A
2.3 I2C SDA/SCL Reserve Issue

Description: SCL and SDA are swapped on U54 bus repeater. Impacts I2C going to 120p
Expansion Connector and Zone 3 RTM connector.

Workaround: Cut the jumper wire (30AWG) to swapped signal of EXP_SDA2_3V3 and
EXP_SCL2_3V3. Two wire each to similar yellow and green to link from U54.6 to R569.1
and U54.7 to R571.1 .

Plans for a fix: This will be corrected on Rev2.0.



2.4 Incorrect UCD Setting
Description: An invalid ‘over current’ setting in one of the UCD power management
devices can cause the EVM to shut down when the SOC is heavily loaded.
Workaround: Follow instructions on MCSDK wiki to perform a field update to get your
UCD settings to the proper value.
(http://processors.wiki.ti.com/index.php/EVMK2H_Hardware_Setup#UCD_Power_Man
agement_Field_Update)
Plans for fix: The correct configuration will be programmed in future builds of EVMs.
2.5 SOC Power Fail

Description: During power on of the board the UCD controller will detect a fault and will
abort the power on sequence

Workaround:
1. Occasionally a successful “power on” will occur if multiple power cycles are
performed.
2. A script is in development to enable a field update to correct the UCD settings.
(http://processors.wiki.ti.com/index.php/EVMK2H_Hardware_Setup#UCD_Power_
Management_Field_Update)
Plans for a fix: Correct configuration will be programmed in future builds of EVMs




2.6 USB3.0 Performance
Description: Lack of AC coupling and in-signal shunts for ESD protection can impact
functionality by causing the link to intermittently go down or never come up.
Workaround:
1. Remove choke B26 , and add AC capacitors C839, C840 (0.1uf)
2. Connect the EN pin of the USB power switch (U59) to USBDRVVBUS pin of SoC (U24)
through R809 (10ohm). (USBDRVVBUS pin on U59 is used by the USB state-machine
to control the VBUS supplied to the Type A connector)
Plans for a fix: Expected to be fixed in future builds of EVMs
3. EVMK2HX Rev 1.1 Hardware Modifications
Rev 1.1 EVM keeps the schematics and PCB layout same as Rev 1.0 EVM, but following BOM
change and reworks are made in Rev 1.1 EVMs for better stability.
3.1 EMAC Link Down Issue

Workaround: R399 and R400 are changed from 4.7 kohm to 220 ohm resistors (0402
sized) in Rev 1.1 EVM.
3.2 VCC0V85 Voltage Drop Issue


Workaround: An external wire is added to provided adequate patch for expected
current from pin 1 of “C503” (net VCC0V85) to pin 1 “C564, C574” (net VDD0V85). (Wire
length = 5 cm). Also, the wire is glued on heat sink back plate for stability.
3.3 I2C SDA/SCL Reserve Issue
Workaround: External rework wires (30AWG) are added to swapped signal of
EXP_SDA2_3V3 and EXP_SCL2_3V3.
4. EVMK2HX Rev 2.0 Known Issues


4.1 EMIF and NAND Test cannot be Synchronized
Description: SoC could not access NAND flash through EMIF interface, and there will be
one cause data transmission errors.
Workaround:
1. Add inverter (SN74LVC1G04DVR).
2. Add wire (30AWG) from inverter PIN 3 to C843.1 (GND),
3. Add wire (30AWG) from inverter PIN 5 to C843.2 (VCC1V8)
4. Add wire (30AWG) from inverter PIN 2 to R27.1 (SOC_EMIFCE0Z)
5. Add wire (30AWG) from inverter PIN 4 to U66.19 (EMIF_OEz).

Plans for a fix: The rework is installed on all Rev 2.0 board before shipping.


4.2 USB Device Not Work
Description: USB 3.0 port could not work even Linux kernel loaded. The cause is 5V rail
turned off due to overshoot/undershoot protection.
Workaround: A couple different options are still under investigation for their
effectiveness.
1. Increase 5V overshoot/undershoot margin in with UCD9090 XML File
(UCD9090_104_A04.XML) in TI Fusion Power Manager Tool so 5V rail would not be
turned off.
2. Remove C387 (56pf) extra capacitor on 5V rail to solve overshoot/undershoot
symptom. (The solution will be applied from Rev 3.0 EVM and later revisions.)

Plans for a fix: C387 will be removed from Rev 3.0 EVM. User who has Rev 2.0 EVM
could choose to remove C387 or apply updated xml file to solve the symptom.



4.3 SoC PWR stop on Power-up
Description: UCD controller detects a fault during power-on and aborts the power-on
sequence.
Workaround: The fault comes from un-used CVDDT rail and FLT pin is floating. Adding a
1K ohm resistor on location C124 will pull-down the un-used FLT pin and avoid false
fault signal.
Plans for a fix: The 1Kohm resistor is added before Rev 2.0 EVM shipping.


4.4 Warning Messages in TFTP test on NAND Flash
Description: When performing TFTP test on NAND in Linux, there are warning messages
frequently. The symptom is related to power phase noise coupling on EMIF signals.
Increasing boot resistor value could mitigate the issue but won’t solve the issue
completely.
Workaround:
1. Change R117 from 0 ohm to 10 ohm (boot resistor for CVDD/U10/UCD74120)
2. Change R468 from 0 ohm to 10 ohm (boot resistor for CVDD/U44/UCD74120)
3. Change R116 from 0 ohm to 10 ohm (boot resistor for VCC1V5/U8/UCD74111)

Plans for a fix: This issue is found in Rev 3.0 and it will also affect Rev 2.0 EVM. The
resistor value change will be installed on all Rev 3.1 board before shipping. Root causes
are still under investigation.
5. EVMK2HX Rev 2.0 Hardware Modifications
The schematics and layout design are modified for Rev 2.0 EVM, and major differences from
Rev 1.0/1.1 EVM are described as below:
5.1 Design Changes:
1. Combine CVDD and CVDDT power rail
2. Change Boot resistor R116, R117, R468 from 10 ohm back to 0 ohm
3. Swap nets EXP_SCL2_3V3 and EXP_SDA2_3V3 on U54
4. Change on-board RAM (U39, U40, U41, U42, U43) from 128Mbit x16 to 256Mbit x16
5. Change EMIF termination resistor to EMIF buffers (U63, U64, U65, U68) for solving
signal reflection issue
6. Connect enable pin of USB VBUS switch (U59) to SoC instead of enable directly
7. Rename “VDDAHV” rail to “AVDDAHV”
8. Rename “VDD0V85” rail to “AVDDALV”
9. Modify PCB stack-up so Top/Bottom layers are 1.2 oz (0.5 oz originally) and the inner
Power/GND layers are 1.0 oz (0.5 oz originally) in thickness for sustaining larger AC
ripple current.
Also, following reworks are installed on Rev 2.0 EVM before shipping:


5.2 EMIF and NAND Test cannot be synchronized
Workaround:
1. Add inverter (SN74LVC1G04DVR).
2. Add wire (30AWG) from inverter PIN 3 to C843.1 (GND),
3. Add wire (30AWG) from inverter PIN 5 to C843.2 (VCC1V8)
4. Add wire (30AWG) from inverter PIN 2 to R27.1 (SOC_EMIFCE0Z)
5. Add wire (30AWG) from inverter PIN 4 to U66.19 (EMIF_OEz).
5.3 Power up will happen SoC PWR stop
Workaround: Add a 1K ohm resistor on to GND.
6. EVMK2HX Rev 3.0 Known Issues


6.1 Warning Messages in TFTP test on NAND Flash
Description: When performing TFTP test on NAND in Linux, there are warning messages
frequently. The symptom is related to power phase noise coupling on EMIF signals.
Increasing boot resistor value could mitigate the issue but won’t solve the issue
completely. (This is the same issue as 4.4.)
Workaround:
1. Change R117 from 0 ohm to 10 ohm (boot resistor for CVDD/U10/UCD74120)
2. Change R468 from 0 ohm to 10 ohm (boot resistor for CVDD/U44/UCD74120)
3. Change R116 from 0 ohm to 10 ohm (boot resistor for VCC1V5/U8/UCD74111)

Plans for a fix: The resistor value change will be installed on all Rev 3.1 board before
shipping. Root causes are still under investigation.
7. EVMK2HX Rev 3.0 Hardware Modifications
The schematics and layout design are modified for Rev 3.0 EVM, and major differences from
Rev 2.0 EVM are described as below:
7.1 Design Changes:
1. Un-mount R812 for EMIF buffer OE control (page 13).
2. Add inverter U69 to solve NAND access issue on EMIF interface (page 19).
3. Add U70 (SN74LVC1G07) for USBDRVVBUS control (page 15).
4. Remove C387 to optimize VCC5 feedback compensation (page 43).
5. Add R815, R816, R817 pull down resistors for clock PDN (page 25, 26, 27).
6. Change R310 from 3K ohm to 200 ohm for USB PHY PVT Sense path (page 15).