(This assumes that any bad capacitor problems have already been fixed.)
1. Turn off power to the computer.
2. Remove CPU, memory, and all peripherals from motherboard. Disconnect power
supply.
3. This step verifies that the board is not damaged, and the VRM is not shorted:
a. Measure resistance from power supply pin (+3.3, +5, +12, -5, -12) to input of VRM.
Resistance should be zero (too low to measure).
b. Measure resistance from output of VRM to load supply pin. Resistance should be zero
(too low to measure).
c. Measure resistance from input of VRM to ground (should be greater than 1 megohm,
usually much greater).
d. Measure resistance from output of VRM to ground (should be greater than 1 megohm,
usually much greater).
4. Install CPU, memory, and all peripherals to motherboard. Connect power supply.
5. Turn on power to the computer.
6. This step actually tests the VRM (all voltages taken with respect to circuit ground):
a. Measure voltage at the power supply side of the power connector (+3.3, +5, +12, -5,
-12).
b. Measure voltage at the motherboard side of the power connector (+3.3, +5, +12, -5,
-12). Anything less than the value obtained in step 6.a indicates a high resistance in the
power supply connector.
c. Measure voltage at the input to the VRM (+3.3, +5, +12, -5, -12). Anything less than
the value obtained in step 6.b indicates a high resistance in the PC board traces to the
VRM, and is probably not repairable, even with a schematic.
Is repairable, by making a bridge with a wire. Usually from one capacitor to the next
one. Never occurred to me unless it is a manufacturer design.
Basically, it is pretty simple. If mainboard does not POST, check voltages.
If PSU voltages are within range, check voltages on motherboard. Cpu has pretty much
standard voltage. About 1.4V is fine for newer CPUs, 1.8V for athlons and P3s, 2V or
greater for first celerons and P2 (Some early P3).
If there is no power (voltage) to the CPU, check if you placed capacitors the right way,
observing the polarity. Same goes for other circuits, RAM, AGP etc. ..
If VRM IC is busted, it would be hard to find a new one. Sometimes (rarely) only one of
the mosfets break, usually the upper one. When it does that, the lower one migh go poof
as well (protecting the CPU).
If one breaks, replace both. Check IC for damage, might be good to replace that one as
well, sometimes it is the cause of this mess.
d. Measure voltage at the output of the VRM. The value obtained depends on the
functions being powered by the VRM, so knowledge of the expected value is required.
Anything different than expected indicates a failed VRM (note that output may be higher
or lower than expected if bad).
e. Measure voltage at the device being powered. This requires knowledge of the device
(which pins are power pins). Pinouts are readily available on the Internet for 30-pin,
72/144 pin, and 168 pin memory, as well as ISA, PCI, AGP, USB, etc. Pinouts for CPUs
are either on the Internet or may be obtained from the manufacturer. Anything less than
the value obtained in step 6.d indicates a high resistance in the PC board traces from the
VRM, and is probably not repairable, even with a schematic.
Ive learned that the Chopper (i like that word), has to ramp up, and the reference is
actually two feedback loops.
The outer loop, for voltage, and the inner loop for current.
My problem with smaller powersupplies has always been finding bad Zeners, little ones.
With voltage mode control, e.g. 3524 or TL494, you only have voltage feedback. It
doesnt respond as well to load transients, though the load transient spec for an ATX P/S
isnt very challenging. With current mode control, e.g. UC3842 (3843, 3844, 3845) you
have voltage feedback and a sample of the inductor current is also fedback. In voltage
mode control, the output of its error amplifier (which compares the output voltage sample
to an internal reference voltage) is compared to a sawtooth ramp. The start of the ramp is
also the start of the switch (chopper) on-time; when the ramp voltage rises above the
error amplifier output voltage, the switch is turned off, and remains off until the ramp
voltage is reset to zero. Current mode control is similar, except the sawtooth voltage is
a sample (scaled by the output transformer in AC-DC P/Ss) of the inductor current. This
lets the P/S resond much more quickly to load transients and O/P short circuits.
Im not sure whether ramping up refers to the sawtooth ramp or to the soft-start
function at turn-on. Soft-start prevents large currents through the switch during the turnon time, while the I/P voltage is still increasing. Basically, soft-start limits the switch ontime and current stress, so that the O/P voltage comes up more slowly. Another feature
that is implemented in the 3842 family PWM ICs is a lock-out that keeps the device from
turning on until its Vcc is high enough to maintain control, during turn-on, turn off, or a
I/P voltage drop-out. I dont think the 3524 has this feature, and Im not sure about the
TL494 (theyre almost 30-year-old designs).
Actually, O/P voltage overshoot can also be controlled or not by the compensation
components in the error amplifier circuitry.
My method, strip the board of any heatsinks and battery and cmos chips if possible,
remove all thermal paste and thermal pads, preheat oven to 384F, elevate the motherboard
off of a pan (i ball up aluminum foil and leave a little pointy end for 4 mounting holes to
support the board) and bake for 5 minutes. I dont suggest pushing past 6 minutes, its
never been beneficial in my cases and stinks like hell. But so long as you can vent the
house afterward hey its fun.
Testing VRMs
Procedure for testing Voltage Regulator Modules (VRMs):
(This assumes that any bad capacitor problems have already been fixed.)
1. Turn off power to the computer.
2. Remove CPU, memory, and all peripherals from motherboard. Disconnect power
supply.
3. This step verifies that the board is not damaged, and the VRM is not shorted:
a. Measure resistance from power supply pin (+3.3, +5, +12, -5, -12) to input of VRM.
Resistance should be zero (too low to measure).
b. Measure resistance from output of VRM to load supply pin. Resistance should be zero
(too low to measure).
c. Measure resistance from input of VRM to ground (should be greater than 1 megohm,
usually much greater).
d. Measure resistance from output of VRM to ground (should be greater than 1 megohm,
usually much greater).
4. Install CPU, memory, and all peripherals to motherboard. Connect power supply.
5. Turn on power to the computer.
6. This step actually tests the VRM (all voltages taken with respect to circuit ground):
a. Measure voltage at the power supply side of the power connector (+3.3, +5, +12, -5,
-12).
b. Measure voltage at the motherboard side of the power connector (+3.3, +5, +12, -5,
-12). Anything less than the value obtained in step 6.a indicates a high resistance in the
power supply connector.
c. Measure voltage at the input to the VRM (+3.3, +5, +12, -5, -12). Anything less than
the value obtained in step 6.b indicates a high resistance in the PC board traces to the
VRM, and is probably not repairable, even with a schematic.
d. Measure voltage at the output of the VRM. The value obtained depends on the
functions being powered by the VRM, so knowledge of the expected value is required.
Anything different than expected indicates a failed VRM (note that output may be higher
or lower than expected if bad).
e. Measure voltage at the device being powered. This requires knowledge of the device
(which pins are power pins). Pinouts are readily available on the Internet for 30-pin,
72/144 pin, and 168 pin memory, as well as ISA, PCI, AGP, USB, etc. Pinouts for CPUs
are either on the Internet or may be obtained from the manufacturer. Anything less than
the value obtained in step 6.d indicates a high resistance in the PC board traces from the
VRM, and is probably not repairable, even with a schematic.