There’s an issue that plagues technicians and consumers alike. When a
Microsoft Surface Pro 4 suddenly no longer powers on. Sometimes this
happens seemingly randomly, other times, immediately following a repair.
Parts of this puzzle have already been answered elsewhere on the internet. But what happens when its “neither of those”?
With
no available schematics, it’s difficult to determine the fault. We had a
few of these in the queue, so we decided to reverse engineer one. We’ve
put together a list of the 3 most common failures to check for, in
order of likelihood.
A bad Schottky Diode
If you made it here from a search engine, you’ve probably already seen this failure 10 times over.
A
diode is effectively a one-way valve. It allows power to travel in one
direction, and blocks it from traveling in the other. Unfortunately, a
diode also causes a drop in voltage as power passes through it (known as
the “forward voltage”). This is generally not ideal or efficient, and
the voltage that is lost results in heat dissipation. This effect
worsens the more current the diode passes.
Diodes also have a
rated maximum current. Exceeding that current (say by overloading or
short circuiting) can result in failure. Diodes fail in one of two ways.
Either the forward voltage increases substantially (sometimes
infinitely), or the diode effectively turns into a wire, allowing
current to flow in both directions.
A Schottky Diode is a
slightly “better” diode. These are commonly found in all kinds of
electronics. Schottky diodes have a lower forward voltage compared to
their silicon counterparts, are available with higher current ratings,
are more efficient, and as a result allow for higher switching speeds.
The
Surface Pro 4 has an abundance of Schottky Diodes on the motherboard.
In reality, any of them have the potential to fail. In most cases
though, one of the diodes pictured here will fail. These diodes allow
either the battery or the charger to power the device. The use of diodes
means that whichever input is higher voltage will feed power to the
unit, without feeding power backwards into the other input.
For
example, with the charger connected, the charger will power the unit,
without feeding the battery and potentially over-charging it. Conversely
if you leave the charger connected but not powered on, the charger
won’t cause the battery to drain.
If you have a lower end Surface
Pro 4, you may find that the “charger” diode is not populated. This is
because the dedicated charging IC has enough capacity to charge the
battery, while the battery simultaneously feeds power into the
motherboard. In either case, a failure of the diodes will present with
the following symptoms:
Battery: A failure of this diode will
result in the device immediately turning off when the charger is
disconnected. This won’t effect charging the battery, or the battery
reporting its state to the operating system, but will prevent the device
running from the battery. In the case of a lower end unit (without the
“charger” diode), the device will not function at all.
Charger: A
failure of this diode (on higher end devices that have it) will result
in the unit running from the battery, but not the charger. This may mean
the device says it is charging, but the percentage decreases as it gets
used. Alternatively (depending on the mode of failure), the device may
not charge at all.
Spare: Funnily enough, on all of the devices
we have seen so far, Microsoft was kind enough to provide us with a
spare Schottky Diode. On the boards we checked, both sides of this diode
connect to ground, and as such, it does nothing at all.
We
aren’t quite sure what happened here. Either there are variations of the
SP4 that have this diode in use, or the part was accidentally left in
the pick and place queue and nobody noticed. In either case, it’s handy
to know in the event of a single diode failure, there’s probably a spare
waiting for you in the device.
Testing the Diodes
So if a diode is bad, how can we tell?
With the battery isolated (this is important) use a multimeter in forward voltage test mode (diode mode).
Measure across the diode leads. You should have no more than 0.250V forward voltage, and no less than 0.100V.
With the probes switched around, you should get a very high voltage (like 2.800V), or “OL“.
If you get close to 0.000 in both directions, your diode is shorted and needs to be replaced.
If
you get a high reading or “OL” in both directions, your diode is blown
and needs to be replaced. A burn or hole is also a pretty good
indication.
2. A bad Fuse
If
the Schottky Diodes are good, or you replaced them both and the device
does not work as expected, the next thing to check is the “fuses”. There
are actually a number of these on the board. So many that we can’t list
them all. We’ve circled the most common “fuse” to cause no power.
There
are 3 variations of these fuses we have seen. In most cases they are a 0
ohm resistor, either marked “0”, or “000”. We still call them fuses,
but in reality, they are 0 ohm resistors, aka “links”, aka “jumpers”.
Less commonly, Microsoft chose to populate these with actual fuses,
typically marked P or L.
If you locate the fuse we have circled, whatever it looks like is likely going to be the same for all the others.
Testing the Fuses
If
you’ve ever used continuity mode on your multimeter, you will be quick
to grab it out and check for the “beep”, indicating a connection.
Before you do though, be warned that this method is NOT fool proof, and may lead to misdiagnosis.
While
a lack of continuity would certainly indicate a “blown” fuse, a 0 ohm
measurement or “beep” certainly does not indicate a working one. This
actually applies for all kinds of devices, and we learned this the hard
way on iPhone Backlight fuses many years ago.
To correctly test
the fuse, you need to pass some current through it. Your multimeter uses
a very small amount of current to test for continuity, and it isn’t
enough. There is a common failure mode where the fuse will appear to
have continuity, until you try to pass current through it.
To confirm operation of the fuse, we set our DCPS to 1V, and 0.05A (50mA) to 0.10A (100mA).
The
voltage here really doesn’t matter, but its a good habit to use less
than the circuit is designed for. This prevents damage to components
should you accidentally slip and power the circuit.
With probes
connected to the DCPS, we verify the current limit by touching the
probes together. The “CC” (Constant Current) LED on your DCPS should
illuminate.
Now, place your probes either side of the fuse. The
DCPS “CC” LED should illuminate, and the voltage should drop down to 0V.
If it does, you have confirmed your fuse can pass current.
If
the “CC” LED does not illuminate, if it doesn’t stay illuminated, or if
your voltage does not drop to 0, your fuse is not working properly.
If it’s not the Schottky Diodes, and it’s not the fuses, what’s next?
Unfortunately,
this is where the availability of information on the internet ends. It
would seem that without schematics, nobody has worked out (or
documented) what causes no power if its neither of the aforementioned.
We had a few devices that fit this criteria, so we set to work reverse engineering them to find out why.
To
begin with, we connected the device to a DCPS directly and prompted the
SP4 to boot with the power button. We figured if there is a short
somewhere, injecting power and looking for heat dissipation is a good
way of finding it.
We got no response at all.
This is
unusual. Without schematics, we don’t really know what is responsible
for the power on sequence, what IC is responsible for what rails, if any
of our rails are powering up upon prompt to boot etc.
We set our
sights on the power button. The first step to powering on the device,
is obviously here. We were expecting to find a regulated 1.8V, 3.3V, or
Battery Voltage at the power button. This is because the power button
either needs to start “HIGH” and get pulled “LOW” (power button GPIO has
a pull-up resistor), or start “LOW” and get pulled “HIGH” (power button
GPIO has a pull-down resistor).
Upon probing around, we find
that we have none of the above. Instead, we have a curious 0.157V.
That’s bizarre. We appear to have located an issue.
Now to find out whats causing it..
Luckily,
we have multiple devices in our queue, some of which are now working.
We quickly probe the power button on a working device and find that we
have 3.3V. Now we need to find out where this 3.3V rail goes, where it
comes from, and why we don’t have it on some devices.
We found a
3.3V regulator on the board, but unfortunately, its generating 3.3V no
problem. There must be more than one. We started probing around and
found our mysterious 0.157V in multiple locations, next to multiple
IC’s.
3.3V Rail on the Surface Pro 4
3. The secret answer. A missing 3.3V rail.
IT8528VG Hole Marked
While
investigating our missing 3.3V rail, and mapping out all of the
locations we have a mysterious 0.157V, we noticed a suspicious mark on
the IT8528VG.
Just next to this is a capacitor where we expect to find 3.3V, but instead have only 0.157V
We cleaned the IT8528VG to see if it was an impurity sitting on the surface. Nope. That’s a very, very small hole. Huh…
At
this point, we were thinking this IC might be responsible for our 3.3V
output. A hole, and subsequently internal damage, would certainly
explain why the output is low. We removed the IT8528VG and tested again,
expecting to have 0V. Our 3.3V is back!
It turns out, the 3.3V
rail is actually incredibly low current. So much so that a short on the
rail, or excessive current draw, doesn’t appear obvious when powering
the device from a DCPS. We assume this rail is a reference for digital
I/O. We still don’t know where it comes from, but we now know why it was
low. Our IT8528VG has failed, and developed an internal short.
We
just happened to have a donor device, so we promptly pulled the
IT8528VG from it. Without a stencil, we had to manually reball the IC.
IT8528VG Reball 2IT8528VG Reball 3IT8528VG Reball 4
IT8528VG Reball Ready
With a replacement IT8528VG ready to go, we installed it and checked our 3.3V rail again.
Everything looks good. We have 3.3V in all the places we previously had 0.157V. This includes our power button.
We checked the device on the DPCS. Upon prompt to boot, we immediately see current draw that is consistent with CPU activity.
Time for a final test..
Finally it works!