Thursday, June 20, 2013

The Mili universal car/wall USB charger, tested in the lab

I received a Mili universal USB charger for review from Mobile Fun. This interesting charger has some features that make it my current favorite travel charger. It runs off both wall power and car accessory power. It comes with swappable plugs for Europe, UK, US, or Australia, and runs on 120 or 240 volts. It has two USB outputs - I thought this was pointless until I discovered how useful it is in car trips if two people can charge at the same time. In addition, one of the ports provides 10 watts for charging tablets (when plugged into AC). The charger also lights up - red indicates charging, and green indicates the devices are charged.

The charger has a few disadvantages. It is a bit expensive with a list price of $49. Measuring about 2 3/4 inches by 2 1/4 inches, it's much larger than Apple's super-compact inch-cube charger - although it has much more functionality. Finally, due to the design, it ends up blocking both outlets when you plug it into the wall.

In the remainder of this article, I test the performance of the charger both in the car and with AC power. To summarize, the power quality is excellent in the car, but has more noise than the average charger when plugged into the wall.

The Mili charger with adapters for different countries.

The Mili charger with adapters for different countries.

The label shows that when connected to AC, the charger is rated as 2.1A for output 1 and 1A for output 2; that is, it is designed to power an iPad from output 1 and a phone from output 2. When plugged in to a car accessory outlet, it is only rated to provide 1 amp, so charging a tablet will be slower. In the measurements below, I find that the charger's power exceeds these ratings when plugged into the wall, which is good, but provides a bit less than the expected one amp when plugged into a car output, which may make charging slower.

Label from the Mili charger.

Label from the Mili charger.

Apple devices can reject "wrong" chargers with the error "Charging is not supported with this accessory"; Apple uses special proprietary voltages on the USB data pins to distinguish different types of chargers (details). I measured these voltages on the Mili charger and verified that it is configured to appear as an Apple 2A charger on ouput 1, and an Apple 1A charger on output 2.

Cars: a hostile electrical environment

You might expect to find 12 volts at your car's accessory outlet, but what comes out can be surprisingly noisy and variable. This voltage will have spikes from the ignition system as well as very large transients due to starting, malfunctions, or jump starting. A car charger must handle this hostile voltage input, and make sure the output to your device is smooth.

Test setup to measure charger performance in a car.

Test setup to measure charger performance in a car.
I measured the voltage in my car to see what happens in a real-world environment using the setup illustrated above. The Mili charger is plugged in just to the left of the gear shift. Above it is the USB interface board, which is connected to the oscilloscope on the dash.

Car voltage drops and rises when the car is started. Car voltage at idle showing ignition spikes.

Car voltage drops and rises when the car is started (left). Car voltage at idle showing ignition spikes (right).

The oscilloscope trace (yellow) on the left shows the large voltage fluctuations when I started the car. At the very left, with the ignition off, the battery provides about 12.5 volts. The starter pulls the voltage down to 8.88 volts until the engine starts. The voltage gradually rises over 6 seconds, settling around 14 volts.

On the right, zooming in shows that while the car is idling, the accessory output has 1/2 volt spikes every 28 milliseconds, due to the ignition firing. Note the voltage on the left is much noisier with the car running than on battery - the line on the left is thin, and the line on the right is thick.

Performance of the Mili charger in a car

The Mili charger has a plug that folds out from the side for use in a car. While this makes the charger larger than a dedicated wall charger, having a charger that works both in the car and with AC is more convenient than I expected, especially when traveling.

The Mili USB charger with car adapter.

The Mili USB charger with car adapter.
I looked at the output of the Mili charger while starting the car, to see if the large voltage fluctuations shown above affected the charger's output. The Mili output remained steady, which is good. I also didn't see any of the ignition spikes in the output from the Mili charger. This indicates that the Mili charger does a good job of filtering out noise from the automotive environment.

I tested the Mili charger with inputs from 0 to 30 volts. 30 volts may seem excessive, but jump-starts often use 24 volts, and car electrical failures can result in a 120 volt "load dump". Fortunately, the Mili survived 30 volts just fine (unlike some other chargers I'm testing). The image below shows that the Mili generates a stable output voltage (horizontal line) for inputs from 7 volts to 30 volts. This is a good thing, showing that the Mili won't overload your phone even if your car is providing too much voltage. As expected, the Mili can't produce the full output voltage if the input voltage is too low (left side of the graph).

Output voltage (Y axis) of the Mili charger as the input ranges from 0 to 30 volts (X axis).

Output voltage (Y axis) of the Mili charger as the input ranges from 0 to 30 volts (X axis).

The oscilloscope displays below show the output and frequency spectrum with 12V DC input and a 5W load. The power quality is very good - the yellow line is thin and has very few spikes. The high frequency spectrum (orange) shows a spike at the switching frequency, but overall the power quality is among the best of chargers I've looked at.

High frequency spectrum of the Mili charger with 12V input. Low frequency spectrum of the Mili charger on 12V input.

High frequency spectrum (left) and Low frequency spectrum (right) of the Mili charger on 12V input.

Next, I measured the voltage the charger can provide under increasing load (details). The horizontal line shows the voltage drops from about 5 volts to 4.5 volts as the load increases. The vertical line shows the charger maxes out around .9 amps with less than the expected 5 volts. This is slightly less than the rated 1 amp the charger is supposed to provide. Both USB outputs provide the same current when plugged into a car outlet.

Voltage vs Current for the Mili charger with 12V input.

Voltage vs Current for the Mili charger with 12V input.

Charger performance with wall input

I also examined the performance of the Mili charger when plugged into the wall (120V AC). One minor annoyance with using the Mili as a wall charger is that due to the position of the USB ports, both wall outlets are blocked either by the charger or USB cables.

The Mili charger.

The Mili charger.

The images below show the voltage the charger can provide under increasing load (details). When plugged into the wall, the two USB outputs provide different maximum currents, unlike when plugged into a car outlet. Output 1 (the high current output) is on the left, and output 2 (the low current output) is on the right. Output 1 reaches about 2.45A before the voltage starts dropping, well above the 2.1A rating. The line for output 1 gets fairly wide above 1A, showing the voltage is not too stable. The line also slopes downwards to the right, indicating the voltage drops somewhat as the load increases. Output 2 reaches about 1.1A before the light starts flashing and the power drops and climbs (the curved lines). This graph shows strange behavior under overload that I haven't seen in other chargers. The lines are all fairly wide, showing the voltage is

Voltage vs current for the Mili charger (output 1) with 120V AC input. Voltage vs current for the Mili charger (output 2) with 120V AC input.

Voltage vs current for the Mili charger (output 1 left, output 2 right) with 120V AC input.
I looked at the voltage output along with the high frequency and low frequency spectrums (below), to examine the quality of the power outputs. The yellow line is much wider than when plugged into the car outlet, showing a lot more noise in the output. The large orange spike in the middle of the high frequency spectrum shows that a lot of the charger's switching noise is appearing on the output. Compared to other chargers, the power quality is lower than average. On the positive side, the flat low-frequency spectrum shows the charger is very good at eliminating ripple due to the 60 Hz power lines.

High frequency spectrum of the Mili charger with 120V AC input. Low frequency spectrum of the Mili charger with 120V AC input.

High frequency (left) and low frequency (right) spectrum of the Mili charger with 120V AC input.

Conclusions

The Mili charger is convenient for travel because it has plugs for multiple countries, works as an auto charger, and has dual outputs. The power quality is very good in the car, but not so good with AC power. This charger is my favorite charger now - while I'd like to tear it apart and examine the circuit inside, I like it too much to destroy it. Hopefully if you get one you'll like it too. And if you found this interesting, check out my detailed analysis of a dozen chargers in the lab.

Thanks to Mili, Mobile Fun, and Mihnea for providing me with the charger and patiently waiting for the review.

Sunday, June 2, 2013

Teardown and exploration of Apple's Magsafe connector

Have you ever wondered what's inside a Mac's Magsafe connector? What controls the light? How does the Mac know what kind of charger it is? This article looks inside the Magsafe connector and answers those questions.

The Magsafe connector (introduced by Apple in 2006) is very convenient. It snaps on magnetically and disconnects if you pull on it. In addition it is symmetrical so you don't need to worry about what side is up. A small LED on the connector changes color to indicate the charging status.

The picture below shows the newer Magsafe 2 connector, which is slimmer. Note how the pins are arranged symmetrically; this allows the connector to be plugged in with either side on top. The charger and computer communicate through the adapter sense pin (also called the charge control pin), which this article will explain in detail below. The two ground pins are slightly longer than the others so they make contact first when you plug in the connector (the same as USB).

The pins of a Magsafe 2 connector. The pins are arranged symmetrically, so the connector can be plugged in either way.

The pins of a Magsafe 2 connector. The pins are arranged symmetrically, so the connector can be plugged in either way.

Magsafe connector teardown

I had a Magsafe cable that malfunctioned, burning the power pins as you can see in the photo below, so I figured I'd tear it down and see what's inside. The connector below is an older Magsafe; notice the slightly different shape compared to the Magsafe 2 above. Also note that the middle adapter sense pin is much smaller than the pins, unlike the Magsafe 2.

A Magsafe connector with burnt pins.

Removing the outer plastic shell reveals a block of soft waxy plastic, maybe polyethylene, that helps diffuse the light from the LEDs and protects the circuit underneath.

A Magsafe connector with the plastic case removed. In front is the metal holder of the pins. Behind it is the circuit board encased in plastic. The power cable exits from the back.

Cutting through the soft plastic block reveals a circuit board, protected by a thin clear plastic coating. The charger wires are soldered onto the back of this board. Only two wires - power and ground - go to the charger unit. There is no data communication via the adapter sense pin with the charger unit itself.

After removing more plastic, the circuit board inside a Magsafe connector is visible. The power cable is soldered onto the board.

Disassembling the connector shows the spring-loaded "Pogo pins" that form the physical connection to the Mac. The plastic pieces hold the pins in place. The block of metal on the left is not magnetized, but is attracted by the strong magnet in the Mac's connector.

The spring-loaded 'pogo pins' inside a Magsafe connector.

The circuit board inside the Magsafe connector is very small, as you can see below. In the middle are two LEDs, orange/red and green. Two identical LEDs are on the other side. The tiny chip on the left is a DS2413 1-Wire Dual Channel Addressable Switch. This chip has two functions. It switches the status LEDs on and off (that's the "dual channel switch" part). It also provides the ID value to the Mac indicating the charger specifications and serial number.

The circuit board inside a Magsafe connector is very small. There are two LEDs on each side. The chip is a DS2413 1-Wire switch.

The chip uses the 1-Wire protocol, which is a clever system for connecting low-speed devices through a single wire (plus ground). The 1-Wire system is convenient here since the Mac can communicate with the Magsafe through the single adapter sense pin.

Understanding the charger's ID code

You can easily pull up the charger information on a Mac (Go to "About this Mac", "More Info...", "System Report...", "Power"), but much of the information is puzzling. The wattage and serial number make sense, but what about the ID, Revision, and Family? It turns out that these are part of the 1-Wire protocol used by the chip inside the connector.

Every chip in the 1-Wire family has a unique 64-bit ID that is individually laser-programmed into the chip. In the 1-Wire standard, the 64-bit ID consists of an 8-bit family code identifying the type of 1-Wire device, a 48-bit unique serial number, and an 8-bit non-cryptographic CRC checksum that verifies the ID number is correct. Companies (such as Apple) can customize the ID numbers: the top 12 bits of the serial number are used as a customer ID, the next 12 bits are data specified by the customer, and the remaining 24 bits are the serial number.

With this information, the Mac's AC charger information now makes sense and the diagram below shows how the 64-bit ID maps onto the charger information. The ID field 100 is the customer ID indicating Apple. The wattage and revision are in the 12 bits of customer data (hex 3C is 60 decimal, indicating 60 watts). The Family code BA is the 1-Wire family code for the DS2413 chip. Thus, much of the AC charger information presented by the Mac is actually low-level information about the 1-Wire chip.

The 1-Wire chip inside a Magsafe connector has a 64-bit ID code. This ID maps directly onto the charger properties displayed under 'About this Mac'.

The 1-Wire chip inside a Magsafe connector has a 64-bit ID code. This ID maps directly onto the charger properties displayed under 'About this Mac'.

There are a few complications as the diagram below shows. Later chargers use the family code 85 for some reason. This doesn't indicate an 85 watt charger. It also doesn't indicate the family of the 1-Wire device, so it may be an arbitrary number. For Magsafe 2 chargers, the customer ID is 7A1 for a 45 watt charger, 921 for a 60 watt charger, and AA1 for an 85 watt charger. It's strange to use separate customer IDs for the different models. Even stranger, for an 85 watt charger the wattage field in the ID contains 60 (3C hex) not 85, even though 85 watts shows up on the info screen. The Revision is also dropped from the info screen for later chargers.

In a Magsafe 2 connector, the 64-bit ID maps onto the charger properties displayed under 'About this Mac'. For some reason, the 'Customer data' gives a lower wattage.

In a Magsafe 2 connector, the 64-bit ID maps onto the charger properties displayed under 'About this Mac'. For some reason, the 'Customer data' gives a lower wattage.

How to read the ID number

It's very easy to read the ID number from a Magsafe connector using an Arduino board and a single 2K pullup resistor, along with Paul Stoffregen's Arduino 1-Wire library and a simple Arduino program.

The circuit to access a 1-Wire chip from an Arduino is trivial - just a 2K pullup resistor.

The circuit to access a 1-Wire chip from an Arduino is trivial - just a 2K pullup resistor.

Touching the ground wire to an outer ground pin of the Magsafe connector and the data wire to the inner adapter sense pin will let the Arduino immediately read and display the 64-bit ID number. The charger does not need to be plugged in to the wall - and in fact I recommend not plugging it in - since one interesting feature of the 1-Wire protocol is the device can power itself parasitically off the data wire, without a separate power source.

The 64-bit ID can be read out of a Magsafe connector by probing the outer pin with ground, and the middle pin with the 1-Wire data line.

The 64-bit ID can be read out of a Magsafe connector by probing the outer pin with ground, and the middle pin with the 1-Wire data line.

To make things more convenient, the serial number can be displayed on an LCD display. The circuit looks complicated, but it's just a tangle of wires connecting the LCD display. Using a simple program, the 64-bit ID number is displayed on the bottom line of the display. The top line is a legend indicating the components of the code: "cc" CRC check, "id." customer id, "ww" wattage, "r" revision, "serial" serial number, and "ff" family. The number below corresponds to an 85 watt charger (55 hex = 85 decimal).

A 1-Wire ID reader with LCD display. Touching the wires to the contacts of the Magsafe connector displays the ID code on the bottom line of the display. The top line indicates the components of the code: CRC check, customer id, wattage, revision, serial number, and family.

A 1-Wire ID reader with LCD display. Touching the wires to the contacts of the Magsafe connector displays the ID code on the bottom line of the display. The top line indicates the components of the code: CRC check, customer id, wattage, revision, serial number, and family.

Controlling the Magsafe status light

The Mac controls the status light in the Magsafe connector by sending commands through the adapter sense pin to the 1-Wire DS2413 switch IC to turn the two pairs of LEDs on or off. By sending the appropriate commands to the IC through the adapter sense pin, an Arduino can control the LEDs as desired.

The picture below demonstrates the setup. The same simple resistor circuit as before is used to communicate with the chip, along with a simple Arduino program that sends commands via the 1-Wire protocol. These commands are described in the DS2413 datasheet but should be obvious from the program code.

I used a cable removed from a dead charger for simplicity. The LEDs are normally powered by the charger's voltage, which I simulated with two 9-volt batteries. To hook the Arduino to the connector, this time I used a Mac DC input board that I got on eBay; this is the board in a Mac that the Magsafe connector plugs into. The only purpose of the board here is to give me a safer way to attach the wires than poking at the pins.

The connector contains a pair of orange/red LEDs and a pair of green LEDs, which can be switched on and off independently. When both pairs are lit, the resulting color is yellow. Thus, the connector can display three colors. The Arduino program cycles through the three colors and off, as you can see from the pictures above.

The charger startup process

When the Magsafe connector is plugged into a Mac, a lot more happens than you might expect. I believe the following steps take place:
  1. The charger provides a very low current (about 100 µA) 6 volt signal on the power pins (3 volts for Magsafe 2).
  2. When the Magsafe connector is plugged into the Mac, the Mac applies a resistive load (e.g. 39.41KΩ), pulling the power input low to about 1.7 volts.
  3. The charger detects the power input has been pulled low, but not too low. (A short or a significant load will not enable the charger.) After exactly one second, the charger switches to full voltage (14.85 to 20 volts depending on model and wattage). There's a 16-bit microprocessor inside the charger to control this and other charger functions.
  4. The Mac detects the full voltage on the power input and reads the charger ID using the 1-Wire protocol.
  5. If the Mac is happy with the charger ID, it switches the power input to the internal power conversion circuit and starts using the input power. The Mac switches on the appropriate LED on the connector using the 1-Wire protocol.

This process explains why there is a delay of a second after you connect the charger before the light turns on and the computer indicates the battery is charging. It also explains why if you measure the charger output with a voltmeter, you don't find much voltage.

The complex sequence of steps provides more safety than a typical charger. Because the charger is providing extremeley low current at first, there is less risk of shorting something out while attaching the connector. Since the charger waits a full second before powering up, the Magsafe connector is likely to be firmly attached by the time full power is applied. The safety feature are not foolproof, though, as the burnt-up connector I tore apart shows.

Don't try this at home

Warning: I recommend you don't try any of these experiments. 85 watts is enough to do lots of damage: blow out your Mac's DC input board, send flames out of a component, blow fuses, or vaporize PC traces, and that's just the things I've had happen to me. The Mac and charger both have various protection mechanisms, but they won't take care of everything. Poking at your charger while it's plugged in is a high-risk activity.

Reading your charger's ID by probing the pins while it's not plugged in is considerably safer, but I can't guarantee it. If you mess up your charger, computer or Arduino you're on your own.

Conclusions

There's more to the Magsafe charger connector than you might expect. The center pin of the connector - the adapter sense pin - controls a tiny chip that both identifies the charger and controls the status LED. It is part of a complex interaction between the charger and the Mac. Using an Arduino microcontroller, this chip can be accessed and controlled using the 1-Wire protocol. Is this useful? Not really, but hopefully you found it interesting.