Hacker News new | comments | ask | show | jobs | submit login

> I was looking forward for Ken Shirriff-style charger reviews

Ha ha, thanks! I agree with you that it looks okay from a safety perspective. It looks like they built the charger with reasonable quality, not cutting corners, but it's not at the Apple level of (over-)engineering. The one sketchy thing is the charger panel that just pops off (instead of being glued/welded), potentially exposing the user to high voltage.

One interesting thing is the amount of complexity that USB-C adds. The charger has a separate daughter board for the Cypress USB-C controller chip. This chip contains a 32-bit Arm Cortex-M0 CPU running at 48 MHz. I believe that works out to about 8 Cray 1 supercomputers using the Dhrystone benchmark.

The switching power supply is a quasi-resonant flyback topology. To oversimplify, the incoming AC is rectified to DC, chopped up into pulses that are fed through the flyback transformer. The output from the transformer is rectified, yielding the low-voltage, high-current DC output.

One somewhat advanced feature is that the output is not rectified by a diode, but by a MOSFET controlled by the controller chip. This is called synchronous rectification. This improves efficiency because you don't have the voltage drop you get across a diode.

The SMPS controller chip is interesting. Most switching power supplies have an optoisolator to provide feedback between the output and the control chip. But this control chip connects to both the input side and output side; it contains an inductive isolator internally. The control chip also contains the MOSFET that chops up the input voltage. So the big controller chip replaces multiple components in a typical charger.

The LED indicator is a bit puzzling. There's a TL431 voltage reference chip next to it. The TL431 is extremely common in chargers to provide the feedback for voltage regulation, but apparently it's being used here to drive the LED.

You're saying that our phone chargers have the equivalent of EIGHT of the fastest computers in the world in 1976? Wow.

A usb c controller is quite complicated. The protocol is not simple at all.

It's still mind-blowing that if you wanted a USB-C controller in 1976, you'd need $64 million dollars worth of computing power.

You would need more than that, since a significant cost is already built into the economies of scale of making non-leading-edge chips.

Exponential growth is a crazy thing.

If it's doing just power charging it should be much simpler, no? Maybe there just aren't any USB-C charger-only microcontrollers out there? Or perhaps the protocol negotiation happens at such fast speed that you have no choice but to use a fairly fast microcontroller?

I guess it's important to remove the charger from the socket when you're not using it, or that computer will keep eating power.

On the other hand, would be fun if you could run complicated calculations on just your charger.

The USB chip (like many microcontrollers) has clock gating and sleep modes so it uses almost no power when idle. When running it uses 10 mA (50 mW) which is pretty low, but in deep sleep mode it uses just 100 µA (0.5 mW). In comparison, a desktop processor can use 95 W or more and the Cray 1 used over 100 kW.

Need constant current for steady LED brightness, which the TL431 can certainly do. Maybe they just wanted to use a chip they were already sourcing in large quantities.

Constant current for a single blue LED is a little nuts to me, but you’re probably right and it made financial sense somehow.

> The one sketchy thing is the charger panel that just pops off (instead of being glued/welded), potentially exposing the user to high voltage.

That sounds pretty sketchy indeed. Isn't that terribly dangerous? Could a child pop it off? Does that meet legal safety requirements for this kind of product?

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact