Thread by @joemmac

**puts on EE hat**

AirPower's "problems" were well known in our circle since that early autumn day years ago but we just assumed Apple had "cracked" something as they often do.

Well, they didn't.

The obvious problem is that adjacent and overlapping coils are NOT insulated from each other just because they're not touching because (obviously) we're talking about contactless systems here.

Of course, multi coil chargers DO exist. And I don't mean "two devices" I mean two overlapping coils to embiggen the "Sweet spot." Most upright charging pads use a pair of slightly overlapping vertically stacked.

The logic for choosing between two coils can be implemented in hardware.

Where Apple fucked up was needing an OS

The idea behind the product was (reportedly, we'll never get to tear down a prototype) to use software to which up to 3 of dozens to coils to transmit from.

Two coils is easy. There's crap in the overlap but it doesn't add up.

A honeycomb of dozens of coils is not "merely" a scaling problem. Your extra coils pick up your own signal you're trying to handshake with. 2 or 3 coils to one device you can filter this pretty easily - just "unplug" all but one.

Even 2 devices for 5 coils is simple enough - align your coils in 1 dimension(-ish) and don't even think about talking to adjacent coils.

Making sure you don't burn the thing down with a 1D array of coils is still in MATLAB homework assignment territory.

The Apple Watch and AirPods case are VERY small. You need to get two axis right to get a standard pad to talk to them - even a multi coil pad - so Apple probably started putting coils in the other dimension.

(they're also small enough to put too close together)

What they probably found was that with something like "only" an 8x3 coil matrix charging an 8Plus, Watch, and AirPods required just as much if not more placement precision than 3 discrete coils.

At this point we still have two choices for mediocre products though

As others have snarked, that didn't stop the MacBook Pro keyboards 😂😭

I really think they thought they could bring that AirPod pairing magic to inductive charging. "Since we need an OS to run the algorithm on the sparse pad, let's just dial up the coil count"

A while back chip design hit an interesting wall. A the relevant voltages related to size adjacent features started having capacitance.

It's a first order differential equation. It could be designed around. It took extra hardware, diminishing Moore's law gains, but worth it.

Smaller still, inductance between adjacent features became a thing. Simple circuits have a solvable second order ODE but these were not. The solution came from materials science. Higher dielectric constants got us back to "ignoring" the inductance.

A clusterfuck of coils is a clusterfuck of differential equations, but since you WANT them to talk through each other, there's no cheating with a dielectric constant. This is their "laws of physics" wall.

(And we're probably going to hit it again in silicon but theoretically we can keep trying random doping until something works - RISC machines are easier to add error correction to as well in case anyone had any x86 optimism left after Intel's last few years)

I just hope the lesson Apple takes from this isn't
"We could've had it if we weren't bound to working with Qi too"