Sephiroth Kwon
GGRRMMAA
20-05-2009
OUTLINE
• “all dots”
• “00”
All dots
• What is the “all dots”?
“AAllll ddoottss” mmeeaannss tthhaatt tthheerree iiss oonnllyy ppoowweerr ssuupppplliieedd ttoo
DEBUG CARD, but no RESET# or CLK signals. All the
LEDs will dispplayy “dot”. The pproblems should be:
– We call it as “no frequency, no reset”.
• When there are a few of main power abnormal which lead to the
CCLLKK_PPGG iinvalliidd. TThhen tthhe chhiips whhiichh needd CCLLKK siignall can’’tt workk,
and they can’t generate reset signal.
– The other condition is “there is frequency, but not reset”.
• Some parts of the reset circuit are destroyed or running conditions
are not all supplied. Because there is a lack of the reset signal, the
debug card displays “all dots”.
All dots
• NOTICE
–– SSoommeettiimmeess, tthhee DDEEBBUUGG CCAARRDD oonnllyy rreefflleeccttss tthhee pprroobblleemm oonn tthhee
PCI Bus. The status DEBUG CARD can’t completely indicate the
problem on other bus. So, sometimes we need to measure the
ootthheerr bbuuss’ss rreesstt ssiiggnnaall.
For example, although the DEBUG CARD displays “all dots”, the
board can power on. Then you would find the basis voltage and
frequency are good. When this happened, you should measure
the other bus’s reset signals, CPURST#, and LPCRST # and so
on. The other reset siggnals are all OK,, it jjust is that the PCIRST#
is failed. We usually call this “fake all dots“.
All dots
• How to confirm this kind of problem
Visual Inspection to check whether the board or
Visual Inspection
components are damaged. Check whether there is
trace broken.
Measure “3VSB” Measure the frequency of crystal resonator, and check
aanndd 3322.776688KKHHzz tthhee vvoollttaaggee ooff “33VVSSBB”.
Put on debug Put on the CPU & Port 80 card, then power on. “FF”
card display on the card. Then put on the DEBUG CARD
(the big one), “….” display on the card.
All dots
• Example – M2A-VM “all dots”:
Visual Inspection
VCORE,
Measure the SOCKET940 signals with oscillograph ：
frequency, 1.+1.8V_DUAL, VTT_DDR; 2.VDDA_2.5, VCORE; 3. +1.2VHT;
4.C_CK_HT_CPU; 5.H_CPU_PWROK
PWROK
All of above signals are OK. But no PCIRST# is low means this is a
NB voltage,
“there is frequency, but not reset” case.
frequency
1.+1.8V;; 2.+1.2VHT;; 3.+1.2VNBSB;; 4.C_HTREF_NB_R;;
5.C_14M_NB_R; 6.C_PCIEGFX_NB_R
Measure the NB_PWRGD,, and found it is low. M2A-VM’s
NB_PWRGD
NB_PWRGD is generated from VRM’s ASIC_CPUPWRGD.
IO_PWROK (namely NB_PWRGD, SB_PWRGD) come from
SIO, and is sent to SB, then the SB send out the various
RESET# siggnals. On the M2A-VM, the SB__PWRGD and
NB_PWRGD come from the same signal.
All dots
RRSSTTCCOONN##,,
SB_PWRGD
Measure the PANEL PIN: 17 RSTCON#(this
ssiiggnnaall sshhoouulldd bbee aabboovvee 33VV)).. TThhiiss ssiiggnnaall ccoonnnneecctt
to SB directly. Here the NB_PWRGD should be
high. If the SB work well, the SB would send out
various RESET# signals PCI_RST#, IDE_RST#,
IOO_RSST##,, BIOOSS_RSST##,, CCPUU_RSST##,, eettcc.
Well, we found the NB_PWRGD is controlled by
1.2VHT, ASIC_CPUWRGD and O_RSTCON#,
and convert to NB_PWRGD & SB_PWRGD via
QQ18 and OQQ21. Meanwhile the SB_PWRGD &
NB_PWRGD are also controlled by PSON#.
Solution:
Measure the OQ21 & PQ810’s PIN: 1. If the all of them are low, these signals’ status is
correct. We focus on the load of NB_PWRGD, for example, SU1, SC70, OQ21, PQ8101,
NU1, etc. Remove the components SC970, OQ21, PQ81 in turn until the failure vanished.
CChhaanngg aa nneeww PPQQ881100, aanndd rreebboooott. TThheenn tthhee bbooaarrdd iiss ffiixxeedd.
All dots
• Notice :
– The example above is not a standard way to
repair “all dots” solution. It is just a example.
“00”
• What is “00”?
– There are two kinds of “00”:
• CPU doesn’t work, because there are not VCORE
voltages, CPURST# or CLK, etc. Some failure
caused by SB or NB’s bus halting.
• TThhe otthher conddiittiion iis tthhatt CCPPUU hhas workkedd, bbutt iitt
is halted, cause of losing data or address in
ttrraannssllaattiinngg.
”00”
• How to confirm this kind of problem
Visual Inspection to check whether the board or
Visual Inspection
components are damaged. Check whether there is
trace broken.
Measure “3VSB” Measure the frequency of crystal resonator, and check
aanndd 3322.776688KKHHzz tthhee vvoollttaaggee ooff “33VVSSBB”.
Put on debug Put on the CPU & Port 80 card, then power on. “FF”
card display on the card. Then put on the DEBUG CARD,
“00” display on the card.
“00”
• Example – P5VD1-X “00”:
Visual Inspection
MMeeaassuurree tthhee VVCCOORREE vvoollttaaggee,, CCPPUU_CCLLKK aanndd CCPPUURRSSTT.. TThhee VVOOCCRREE
should be 1.5V, the CPU_CLK should be 100MHz, and the CPURST#
“00” confirm
should be above 1.2V. When assert the RSTBTN#, their voltage should
be changed, and then restore normal status promptly.
VCORE 1.5V VCORE is OK.
CPU_CLK is OK.
CPU_CLK,
CPU_CLK#
We found the CPURST# is only 0.3V, and when we assert the
RSTBTN#,, CPURST# didn’t changge. So we can jjudgge that it
belongs to first condition of “00”.
CPURST#
There is no CPURST#. CPURST# is translated form SB to NB,
then NB to CPU. So,, SB & NB’s runningg voltagge and freqquencyy
are necessary.
“00”
NB’s frequency, Measure the voltage of NB circuit:
HCLK, HCLK# VTT_CPU is 1.2V, VTT_DDR is 1.25V. All of these
vvoollttaaggee aarree OOKK.
100MHz
NB’s voltagge, 1.5VSB is onlyy 0.3V.There are two kinds of ppossibilityy.
VTT_CPU, 1.5V, 1. Q23 burned，so there is no 1.5VSB voltage sent out.
VTT_DDR, 2.Load is too big, make the 1.5VSB voltage to be low. (For
example, the load IC is burned, make the impedance
1.5VSB
veryy small between VCC and GND)).
Solution:
1.5VSB and
Measure the
LAN IC
iimmppeeddaannccee bbeettwweeeenn
emitter (1.5VSB) and
GND.
And the value is 10 ohm. So it looks like the load
is too big. Then we short the Q23’s PIN2 and
PIN3 (namely, 3VSB and 1.5VSB) for a while.
Found that the U30 (LAN IC) is hot. So we
ccoonnffiirrmm tthhaatt tthhee LLAANN IICC iiss ddeessttrrooyyeedd.
Change the LAN IC and Q23, reboot again. Fixed.
Thank You!
English 
العربية