Post-Mortem: Apple 60W MagSafe Power Adapter (A1344)

Unlike most of my other posts which are relatively straightforward, this post has a little story behind it. In March, Matson of Chicago, Illinois, USA contacted me about his “dead” Apple 60W MagSafe Power Adapter and wondered if I’d be happy to take a look at it.

At the time, I felt that it would be silly to post a dead item internationally, because the postal charges from the US to Australia were not cheap. Besides that, some of the failures are more obvious and so I encouraged him to explore a little on his own to try and see if he could work it out with a few pointers to common issues (e.g. open primary fuse, bad capacitors, open windings on the transformer, PCB trace problems). Anyone who knows me closely will know that I am not a fan of Apple equipment in general and the only Apple gear is the stuff that I’ve won. Even if I had the full brick, I wouldn’t have had anything to test it with – the MagSafe adapters are likely to be a little more sophisticated than a straight DC supply.

Five months later, in August, I got another e-mail from Matson, this time with an offer of a very useful piece of equipment. Of course, I was interested, but at the same time slightly hesitant. After a few back and forths, it was decided that I would accept the piece of equipment on one condition – that I also accept the broken power adapter and do a post-mortem on it. From my point of view, it was a win-win situation – I would have a new toy, and a subject for another blog post which might be somewhat informative.

Earlier this week, the goodies arrived, so the first thing I do is get down to business. The other bits and pieces … well, I’ll get to that later when I have some more time I suppose.

The Deceased Power Adapter

As international postage is expensive, the adapter is devoid of it’s plug-portion. It also has the MagSafe lead removed, which is a good thing, because that lead has a very special plug at the end which is worth its weight in gold if you want to hack another power supply for your Mac. Both help to shave down the shipping weight and costs.

As its owner had already decided to try their hands at some exploration, the case was “nicely” opened, which saves me the trouble of prying it myself and maybe hurting myself in the process. The unit is model number A1344.

Despite all of the Apple vs Samsung rivalry, it’s nice to see that this power adapter comes from Dongguan Samsung Electro-Mechanics. Yep. Your Macbook was being powered by a Samsung power adapter … fancy that. Its rating is 16.5V at 3.65A for a power rating of 60.225W (round down to 60W for convenience).

Separating the halves shows just how well built genuine Apple power supplies are. In this case, both halves of the outer shell have metal plate on its interior to help spread any generated heat and prevent hot-spots which can be discomforting to the user.

The interior board is covered by a wrap-around multiple-layer aluminium shielding which both acts as an RF shield to stop interference leakage and works as a heat spreader being thermally bonded to the heatsinks on the board. Copious amounts of tape were used as insulation around sensitive equipment. The metal was even screwed together to heatsinks to improve heat transfer.

The only sign of distress I saw was some melted plastic near the power plug and slight blackening of the PCB. As it turns out, this was a consequence of a first attempt by Matson to desolder the shield grounding connection (a tab soldered to the board and spot-welded to the shield). As this supply will never be used again, I just decided to tear the trace off the board hoping to see something underneath that could explain its lack of function without needing to get out the soldering iron.

With the cover removed, we can see just how crammed these power bricks are. They are chock full of components to the point that the board has many plastic shims inserted to maintain isolation, and rubberized self-adhesive pads to keep components from vibrating against the enclosure and damaging their insulation. There really isn’t much space for anything extra in here.

Even the reverse side is well covered with components. Some helpful silkscreens are provided on the rear – but we can see in the centre, a row of three optoisolators for feedback control, and just to the left and slightly below, the output current shunt resistor. There are rubber feet on the underside as well. The slots for the shims to fit are visible, and they actually weaken the PCB slightly.

If we look near the output, we can see where the original cable had been cut.

The Autopsy Begins

All this was pretty much just equivalent to “taking off someone’s clothes” and doing a visual inspection. Sometimes, at this point, the failure becomes obvious because you can see it, or you can smell it. Unfortunately, in this case, the power adapter actually looks quite fine – assuming that the melted corner near the fuse and input is from the result of an attempted disassembly. To find the cause of death, we must go deeper.

Because of its compact size, it becomes difficult to access components to probe them directly. Probing in-circuit also potentially results in false readings due to interconnected circuitry. As a result, the strategy has now changed to a more time-consuming and in depth teardown. With a soldering iron. And some desoldering braid.

First step was to remove the heatsinks from the PCB, as that will free the semiconductor components.

On the primary side, I removed the heatsink that was shared between the primary switching MOSFET (right) and bridge rectifier (middle). One thing I noticed was that the primary MOSFET wasn’t screwed in very tightly to the heatsink, so was probably having poor thermal transfer. The suspicion seems to be backed up by what appears to be a “thick” layer of thermal grease on the rear – this normally spreads out to a thin amount when securely mounted to the heatsink. Even the bridge rectifier next to it with no fastening hardware has a thinner looking layer. This would be a candidate for a suspect component.

The next heatsink to some off was the secondary side synchronous rectifier MOSFET. This was insulated from the heatsink with a silicone pad, which is appropriate. Because the heatsinks are secured to the PCB with solder, it was a little chore to desolder them.

That gets us here. Not much of an improvement, but the tear-apart is making progress. At this point, I decide to get serious about probing the primary side …

… only to find that the fuse (encapsulated in brown, a 4A time-delay type fuse) was intact and not blown at all. Whatever failure appears it might actually be secondary side (e.g. controller goes into shutdown for self-protection perhaps). The board has B22E_R16 written on it, dated 11th April 2011, with a thermistor just next to it (normally for soft-start purposes, but this seems maybe to be a thermal overheat protection). Since I wasn’t going to power up the unit, I decided for a full strip-down of all through-hole components ….

… there’s a few taken away …. and if we keep going …

… voila. We’ve reached the base PCB. I like how there’s a spelling error on the silkscreen – the ‘neutral’ connection is marked ‘NEUTURAL’. Interesting, buried underneath all of that, there is F102 – a Littelfuse 125V 5.0A time-delay fuse protecting the output in case everything else fails as a last line defense. I checked this with a meter – it was also intact, so that made things a little curious.

The underside was very much still littered with SMD components, which I wasn’t going to desolder as many of the markings were hard to read and the functions were not easily tested for. There’s a lot of desoldering braid flux (black/brown) left everywhere, but I think I did a good job of taking everything off.

Above are all of the components which ended up being removed. Lets do some component level testing to see if anything here is the problem.

Component Functional Testing

Just like in air-crash investigations, once you recover components from a system, it’s a good idea to see if they still work. After all, if they do, I could add them to the junk box, and we would know they were not the cause of failure. I first started with most of the primary side components, hand-drawing on a sheet of paper as I went along.

Nothing was amiss here, but in case you were interested, here are some datasheets for the identified components:

• Samyoung NFA Series Capacitor
• Samyoung NXH Series Capacitor
• Rubycon YXF Series Capacitor
• Surge Components CMPP X2 Mains Suppression Capacitor
• Hiel SYP Temperature Limit Sensor Thermistor
• POElectronic AH Series Leaded Disc Safety Capacitor

It was somewhat interesting to see the various components used – I would have assumed they might have opted for the Japanese electrolytic capacitors for all of their filtering needs because of their better lifetime reputation, but instead, they went with South Korean products from Samyoung instead. That being said, the Samyoung capacitors are 6000h+ rating at the load temperature of 105 degrees C, so they are long life capacitors (at least, by rating).

Now we move on to the semiconductors and transformers to see if there’s anything wrong with these …

The transformer turned out to be a Li Shin Enterprise (LSE) product, which is no surprise, as they’re involved in many SMPS power products and even power cables for PCs. The transformer had three windings, as many SMPSes do – a primary, a feedback and an output winding. All windings were insulated from each other, and continuous – so the transformer was just fine. The bridge rectifier was a Lite-On Semiconductor Glass Passivated Bridge Rectifier GBL406, rated for 600v peak reverse voltage and 4A forward current (matching the primary fuse). I measured the voltages across legs, and all diodes were still functioning, and functioning well as diodes. It seems like it hadn’t failed.

Now it was down to the semiconductors – the first was the primary side MOSFET which was an Infineon SPP11N65C3 CoolMOS Power Transistor with a 650V/11A rating and an Rds of 0.38 ohms. The gate wasn’t conductive to any other pin, which was good, but the source and drain were shorted through in both directions. This indicates the primary MOSFET had failed a dead short – this would stop the unit from functioning, but would normally result in a blown primary fuse and potentially smelly result with the overheating of the MOSFET.

This didn’t happen in this case, which seemed a little strange. But I have a theory – this transistor wasn’t well mounted to its heatsink, so it might have overheated over a long time and failed shorted due to internal melting of silicon. But maybe (just maybe) the bond wires which connect the legs to the package also separated by thermal expansion, so that the unit is shorted, but as soon as current flows and it heats, it “breaks” the connection due to thermal expansion thus preventing an explosion or blowing the fuse. This is similar to the flickering which can happen when LEDs overheat and their bond wires start making intermittent contact with the LED dies.

Another possibility might be that the MOSFET was static-damaged in the process of removal – but this is unusual and relatively unlikely. It’s never happened to me before.

The secondary side MOSFET was an Intineon IPP12CN10N OptiMOS2 with a 100V/67A rating and an Rds of 0.0124 ohm. Testing of this MOSFET resulted in the appropriate isolation of gate, body diode between drain and gate, and an open circuit in the other direction as expected.

As mentioned earlier, checking the PCB’s fuse didn’t detect any anomalies. To round out the checks, I decided to check some other components on the PCB:

As it turns out, no anomalies were detected with the diodes on the boards, nor the LEDs in the optocouplers. It would be expected that such controllers would fail to output if the feedback from the optocouplers is not received, and LED failure is a major cause of this. Aside from that, another form of feedback is the current sensing shunt, and the resistor came up okay as well.

Bonus: Transformer Teardown

A point of contention when it comes to cheap Chinese transformers is the poor insulation of the transformer, often held together with just tape and barely one-wrap of tape to insulate primary from secondary windings. Seeing as I wouldn’t have much use for the transformer, I thought taking it apart would be educational and beneficial for comparison purposes.

Step one was to remove the tape on the outside. Then, the transformer was found to be encased in a plastic holder, and was carefully prised out. Already, we can tell it’s something better than most cheap Chinese efforts which fall apart at the first layer of tape.

The outside of the transformer is wrapped in two rings of copper tape at right angles, soldered to one end of a winding. This is probably used to sink any stray leakage.

The transformer is wound on a bobbin with a core that goes around the bobbin, encasing the windings almost completely resulting in a highly efficient coupling. This results in a higher efficiency transformer with less magnetic flux leakage. The two halves aren’t quite in perfect alignment, and are varnished together. Unfortunately, I couldn’t easily separate the two halves with a few taps of the screwdriver or the pliers … so I went down to the garage, got a hammer and smashed the ferrite to bits.

In the process, one side of the former cracked away as well – but this does go to show how the unit is constructed winding-over-winding for improved efficiency, but with copious layers of insulation.

Here, we see a multi-stranded copper wire winding end soldered to an insulated leader wire that connects to the pins that go to the PCB. The other end will be connected directly to another set of pins. Unwrapping each and every layer brings a few surprises, such as the secondary winding.

The secondary winding is actually a tri-filar winding with all windings in parallel. The difference is that the winding itself is made of insulated wire that is like hook-up wire with some sort of plastic-like insulation. This is addition to the wraps of tape between each layer which adds to the insulation. Its clear based on this how much they care about primary to secondary isolation.

There is also an internal screening layer, likely to reduce interference between primary and secondary coils.

In all, that was what was recovered from the transformer – a total of four wires – a secondary, what appears to be two primaries and a thin winding which is probably a screening winding. The scrap of bobbin and former is shown, minus all the ferrite fragments which now litter the garage, and also the remains of the layers of tape within the former.

In case you were wondering – the copper in the transformer is about 7.95 grams – but probably a little less due to the enamel and plastic insulation.

Conclusion

Through a long and dedicated afternoon, I basically desoldered all the through-hole components off the board and checked them each for their functionality. The only anomaly spotted was in the primary side MOSFET being a dead short – this would normally trigger a cascade failure which would be smelly and blow the primary side fuse. Interestingly, this didn’t happen – and I theorize it may be due to intermittent internal wire bonds as I’ve seen in some LEDs causing them to break the connection as they heat up. But who knows, if it had been plugged in and turned on another 10-20 times, maybe it would have blown a fuse.

Another possibility is that the MOSFET was damaged from extraction, but I find that unlikely. Supporting evidence includes extracting the other MOSFET without damage, and the poor heatsink connection on the failed MOSFET which may have been the primary cause.

Why this particular MOSFET may have failed is a bit of a mystery. While I list heat as a major contributor, it may have not been the only contributing factor. The supply doesn’t seem to have any surge protective devices on its input (e.g. MOVs). Without this protection, transients from lightning or switching events on the network are likely to get through to some degree, and over time, they could overstress the MOSFET and cause it to fail prematurely. Some other larger power supplies and even LED downlights have a MOV as a low-cost insurance policy against surges – basically having a surge protector within the PSU itself. Not seeing one inside the Apple supply is a little bit of a disappointment.

Of course, no other anomalies were detected in the components I probed, but equally possible is a failure in a controller IC, or surface mount transistors which were not tested.

While it is unfortunate that the power supply had failed, its design seemed fairly decent. Isolation was ensured from primary to secondary even with its compact size by using lots of anti-tracking slots and insulating plastic sheets. Capacitors, while not my preferred quality Japanese make throughout, are at least long-life units with a good temperature rating. The number of inductive filters was also good, and the shielding was excellent, doubling as a heatsink. The insulation in the transformer was excellent as well, as expected.

I hope this satisfies your curiosity Matson, and thanks for your contributions.