Hacker News Reader

Comment by zdw | Hacker News Reader

In watching the evolution of USB-C over the last few years, it seems like it's extremely hard to implement correctly with the huge number of modes, alternate modes, and power delivery in the spec.

When you connect two USB-C devices today, you have almost no idea what is actually going to happen, which device is the master, and which way power will flow.

While having one connector and cable type for everything seems like it would be a good idea, in practice it's turning out to be a giant mess. Maybe it'll clear up in a few years, but given the race to the bottom in price and quality in the accessory market, this seems doubtful.

View on HN

I partly agree that it's a bit harder than before. But a lot of the faulty devices we're seeing are just straight up using an incorrect resistor value or similar, and apparently released to the public without proper testing.

Maybe that's on USB spec people for not having good material, but on the other hand maybe that's on the manufacturers for not hiring EEs who can actually read a damn spec sheet properly...

It's cost cutting to make cheaper cables and ports.

Or intentional trivial incompatibility to sell _more expensive_ cables and ports :)

That's a scary possibility. Flood the market with barely-functional garbage in the low range so that consumers can't figure out which cheap ones are actually usable and just pay more to avoid the headache.

It's not dissimilar to the box store HDMI/audio/etc cable strategy: Cheap, absolutely garbage cables at a mid-range price, or upper-mid-range cables at grossly over-inflated prices. Consumers tricked because of course the $200 HDMI cable looks noticeable better than the $20 one. Reality is the real retail prices of cables they're comparing should be more like $2 and $20, and the difference between that $20 cable and one that actually would cost $200 is only noticeable with high-end test equipment (or by 'audiophiles', who have the super-power of being able to see/hear a difference in cables as long as they know the price).

This was the case when the original USB came out first too. It took a good couple of years to mature. I think these are normal standardization issues...

Difference is that when the original USB spec came out it wasn't delivering enough power to kill somebody, contrast that with today where a USB-C error could be fatal.

See: https://www.theverge.com/2016/2/4/10916264/usb-c-russian-rou...

This is incorrect. The high power charge is 20 volts, below the safe threshold of ~50V set by such agencies as: OSHA, NFPA, ANSI, IEEE, UL, IEC and others.

The real difference is USB was a completely new technology with zero market penetration so growing pains were inevitable. USB-C should not have these growing pains as they should have learned from 20+ years of developing the standard.

There is relatively little risk of electrocution but 100 watts can still be more dangerous than 10 watts. Overheating and fire are a much bigger risk.

Is it really large, established manufacturers who are making these mistakes? Or is it random fist-time hardware entrepreneurs in Shenzhen?

Analogy: we don’t blame Javascript-the-language-standard for the Node community’s constant (and usually half-assed) reinvention of build tools. Javascript itself didn’t cause that. More than likely, it was a glut of newbie developers joining the community—and building tools for one-another, rather than leaving it to the experienced people—that caused that.

> Is it really large, established manufacturers who are making these mistakes?

Yes. Just read the NathanK or Benson google+ pages. Very few USB-C accessories are compliant - almost every one has some bug in its implementation, some worse than others. Apple cables are good, but even they took a couple iterations to get it right.

Considering this article is about the switch not being compliant, yes. Nintendo has quite a lot of experience making hardware, and they didn't get it right. Right now USB-c is a mess, and I'm not aware of any controllers on the market that are compliant and easy to use.

Yes, Anker had to issue a recall for a USB C cable that could fry devices:

https://www.theregister.co.uk/2016/08/30/anker_usbc_cables_r...

5V to 20V can be fatal?

I can kill you with a 12V car battery at 8 amps clamped to your fingers, no problem.

No you can't. It has the capacity to deliver 8 amps, but your body's resistance is too high to draw that current.

https://www.seattlepi.com/news/article/Car-Talk-Car-battery-...

EDIT - here's some math:

The U.S. Navy FIRE CONTROLMAN Volumes 01 - 06 & FIREMAN gives 1500 ohms as a common resistance approximation between extremities, either hand to hand or hand to foot, for an average human body.

I = V/R = 12V / 1500 Ohms = 0.008 A

8 mA is not nearly enough to kill you.

On the other hand, if skin is breached, the resistance is quite a bit smaller, and it won't take too much voltage to cause a deadly current.

There's a story floating around the internet of someone having done by stabbing through their thumbs with a 9V multimeter, and maybe it's possible, but I'm skeptical of the story's veracity. I can see those wacky stories for a lot of situations, but you'd think with somebody dying that there'd be some kind of incident report or other documentation.

But sure, it's probably best to not touch the car battery terminals when your hands are bleeding.

I'm genuinely curious how you'd do that... I can touch both battery terminals with my hands and feel nothing. Even if my hands are damp, nothing happens. My battery is capable of outputting over 700amps.

People have latched on to the mantra that it’s not the volts that kill you, it’s the amps, but they never seem to remember that you need a certain voltage to get those amps in the first place.

Exactly. And also the mantra "electricity takes the path of least resistance". While that is true, some people seem to interpret it as "electricity only takes the path of least resistance" which is very wrong. Electricity takes all paths!

I don't know squat about electricity

If electricity takes all paths, then whats the point of the mantra? It includes the highest and least resistance, and everything in between. The mantra is almost deliberately meant to mislead?

> If electricity takes all paths, then whats the point of the mantra? It includes the highest and least resistance, and everything in between. The mantra is almost deliberately meant to mislead?

Maybe mantra wasn't the right word, Wikipedia calls it a heuristic. Regardless, I agree it is very misleading and I would never say anything like that when teaching or explaining electricity. Why does it keep getting repeated? Who knows. Inertia of the masses without a much understanding of electricity would be my guess...

https://en.wikipedia.org/wiki/Path_of_least_resistance

Most of the power goes along the path of least resistance. It's an oversimplification, but you'll make much more dangerous mistakes if you let intuition lead you into assuming a pretty spread-out flow.

It's technically correct, but the path of least resistance is the path of least resistance at the instant of measurement. Voltage and current are abstractions that represent the aggregate motion of charge carriers in a medium. Like charge carriers repel each other. At any instant, an individual charge carrier will take the path of least resistance. Calculus is required to model the current in a non-trivial system. It frequently leads to counter-intuitive results.

In short, it's complicated.

A better mantra would be: electricity takes all paths in inverse proportion to their resistance.

So with a single low resistance path, give or take, all the electricity will go that way. But with two paths of equal resistance, half the electricity will go each way.

Electricity takes the path of least resistance...and all other paths, in amounts depending on their resistance.

You don't die instantly. You are able to take exactly 5 steps before you die.

Well, to be fair he said he could. Yes, it would be possible using a boost converter to jump the voltage, you can certainly draw enough current from an auto battery. But without using external components the only way a car battery would kill you is by creating a spark and igniting hydrogen that the battery produced.

If you pierce the skin with the "teeth" of the clamp you can do it, the lower resistance will be enough to send lethal amounts of current.

No way. If that were the case, people would be dying all the time when trying to jump cars. You would need to wear lineman gloves to do it safely and working on cars in general would be quite hazardous. I've touched both car battery terminals before. You don't feel a thing.

Indirectly, it could start a fire...

A 9V battery in the wrong place can kill you.

As can a AAA battery, if accelerated sufficiently beforehand.

Well, I mean from electrocution specifically. 100-200mA through your heart is probably going to be fatal, and the low resistance required for that to happen with a simple 9V battery source has at least once been achieved by inserting multimeter probes in the thumbs such that the current passed via the blood stream through the heart, causing fibrillation.

Interestingly, there exists a sort of upper threshold where the amperage is less likely to be fatal due to severe cramps preventing fibrillation IIRC.

Would you care to give us a scenario? The best source I could find places the minimum resistance of already damaged human skin at 500 ohms. With 9v of potential, that translates to .018 amps of current, which is just barely above the threshold for sensation, let alone death.

0.018 is way above the threshold of sensation, well into the territory of very painful shocks (18 times the threshold of sensation at ~1mA). Of course this will usually be local with a 9V battery, but with some deliberation you can pass the current through your heart via the blood stream, where the resistance is sometimes much lower which can get you very close to what'll likely be fatal.

Edit: the above holds true for AC but not necessarily for DC from a battery, where the thresholds are typically higher. But the heart is very sensitive and fibrillation occurs at very low currents passing directly through it.

Sadly it's tiny watch batteries in the wrong place doing most of the killing and disfiguring.

False. USB-C cannot be fatal.

Yeah, agreed - those things never worked. It took something like a decade before it got properly reliable.

Is it just a resistor, or is it one of hundreds of possible resistor depending on factor X? USB or any connection should be simple - you know, like code. The more exceptions and options you add, the harder it becomes to properly implement it. This is exponential, too, not linear.

>EEs who can actually read a damn spec sheet properly...

I think this is the hardware equivalent of expecting programmers to write bug free code on the first try.

Most of the time when a programmer wants to implement something a little complex and a little outside their expertise, they use an external library. Likewise, EEs will often buy a chip produced by a third party manufacturer and use that to handle it. But the abstractions available to EEs are often a bit more leaky than those available to us programmers, since they are more constrained by the laws of physics. and as a result that third party chip is still harder to use than a software library.

That is still an avoidable error. You shouldn't be adding parts without understanding how they work or how they will affect your circuit as a whole.

On the first try? I would hope that hardware manufacturers don't take the engineers' first attempt and immediately go into mass production.

That isn't the way EE's are trained in my experience. When new hardware is designed that is outside the EE's experience there is usually an evaluation board available from the manufacturer. They get one of those in house, read the application notes and data sheets, and use their test equipment to observe the 'proper' signal levels and waveforms on a correctly implemented system.

Then they add this part to their design, and lay out the schematic and the printed circuit board according to best practices by the manufacturer. Once the first boards return they put measure the signal levels and waveforms in their circuit to verify they are with the specifications and that they match the ones on the eval board. Then they will 'corner' test the circuit (corners are low/hi temperature range, low and high voltage range) and verify it continues to work according to specification at all the 'corners' (if it does then you are generally ok with assuming it will work at all points "inside" those for corners.)

There are people who are either in a hurry or don't care who wire something up according to the application note, power it up once and call it 'good'. I've seen a number of cost reduced 'clone' equivalents that meet that description. @kens has done a number of blog posts that show this sort of mentality in detail.

Your process outlined here takes time and is likely part of why the quality ones aren't to market yet. :(

It must be more than "engineers can't read a spec" because decices using previous versions of USB tend to be fine. So why can engineers read those specs? (Why didn't they "choose an incorrect resistor value" for previous generations of USB, then?)

This spec by contrast must be far more complicated[1]; also since it is so much faster, can't quality components supporting it all be more expensive - leading to cutting corners as a cost cutting feature?

Also if as a consumer you can't predict what "should" be about to happen (which way power will flow, what will be host) that is down to a bad spec re connector types. For example couldn't a universal visual indicator on a port (including the bottom of a phone) indicate its capabilities and what will happen very clearly? (Through symbols.) They do not. Couldn't a software popup make you choose? They do not.

Not to mention that it is not clear what should happen. If a phone can power a USB peripheral (like usb on the go) that means it could charge a second phone.

So if you connect two phones which one should charge the other? With old versions of USB phones the answer is simple: whichever one you put an on the go converter into, making it a master instead of a phone's usual role as a slave.

The fact that the spec does not make these as clear for USB C is down to the spec and design.

Can't USV C simply be badly designed? (overdesigned, underdesigned, badly designed.)

[1] Consider that a 1-page spec that can be read and implemented quickly in 15 minutes, can be implemented by anyone. A five-thousand page spec that takes 1,000 hours to read and understand can be implemented by no one (only teams). In between we have both previous versions of USB and USB C - but is it possible that USB C is too far in the latter direction?

> Couldn't a software popup make you choose? They do not.

They do when I e.g. connect two android phones together.

Got it (never tried it).

A sane USB-C spec would not have required a resistor for an A-to-C cable.

A single resistor is seriously not challenging to get right. If they can't do that they sure as hell shouldn't be making a cable.

Your average cable is some wire, some connectors, and plastic. Adding a resistor adds a whole extra element to the process, which is a big deal for a part that sells for probably well under $1 wholesale. Cable shops probably have no experience with circuits. The USB-C spec for some reason made it so that cables with no resistor or with the wrong resistor still seem to work but are unsafe. Of course the result is a mess.

Just about every video game console made between the late 1980s and mid-2000s had an official video-out cable with one or more resistors [1]. It's not an unheard-of thing.

[1]: http://members.optusnet.com.au/eviltim/gamescart/gamescart.h...

Is it really that simple? Just changing a resistor value? I got an XPS 13 for work, it’s all USB-C, and the versatility of the port is fantastic (doesn’t matter which port I plug into, it still charges, so I don’t have to wrap the power cable around the back of the machine or sit in a weird position to keep some slack in the wire). However, I took it to the office and naturally I plugged my power adapter into one port. I then plugged in the monitor (also USB-C, delivering power), and the two ports cancelled themselves out and confused the OS. Windows thought I was charging the laptop yet it also said the battery was depleting. They were negating each other and the battery loss continued as it would with no power connected.

It makes sense for a monitor to power the laptop because it’s mains connected. But then you have one of those classic programming problems: if there are multiple inputs providing the same thing which do you consider the source of truth?

They could always ask the user if they are that confused then remember the answer so they don't have to ask again.

Can you imagine a typical laptop user who assumes everything 'just works' (like me, by virtue of thinking I need to power my laptop while attaching a powered monitor) seeing a prompt in the OS about which thing should provide power and which should either reject it or use it to charge a battery?

Can you imagine someone not understanding a clearly worded UI that they use to do the things they have to do to function in society?

There are some USB c complaince issues which are yes literally just the wrong resistor.

What you're describing might go beyond that though, in particular if there's multiple ports and hubs involved, that could be not just a usb problem.

Hmm, this have anything to do with why trying to hook up my macbook to external monitor with usb-c to display port cable blew my motherboard today?

USB PD requires communication over the CC pins via setting different resistor values and checking them at specific times in the hand shaking process. The communication in the latest spec is supposed to happen at about 300kbs using a specific protocol on a specific CC line. There have been multiple specs since about 2011, some of which used communication over the VBUS line using BFSK, which has since been deprecated. This adds to the confusion.

On Macs, it will choose the power supply that advertises the highest available wattage. This has been working flawlessly on my MacBook Pro.

There needs to be a way to decide which device is the master, like the prompt on Android devices asking if you wish to power or be powered.

Wonder if you could do it on the cable with little embedded LED lights. One end turns red for "charging" the other turns blue for "supplying power". Would add to the BOM cost on the cable for sure, but it would not only be cool looking but rather useful.

oh yeah, because my bedroom doesn't have enough glowing devices when i'm trying to go to sleep. let's add LEDs to every cable too!

Hah. Good point. Yes, too many devices want to put big giant blue LED's on their product.... even ones that are specifically targeted for the bedroom and specifically designed for sleeping...

Barring the comments about adding more sources of light in a dark room (yep, I agree with this), I still appreciate the MagSafe adapter for my MBP lighting a tiny LED to say charging/full. The new USB-C MBPs we have at work don't have any indication whatsoever that the adapter is working unless the machine is booted and logged in (to pre-empt comments about the chime, what if it's muted, as is often true of workplace laptops?). MagSafe tells me no matter what state the computer is in.

The LEDs don't have to be bright, just dim little indicators that are only really noticeable if you're looking for them. Then you get an instant idea of what the devices are doing.

This. I’m always completely bemused when I first connect a USB-C charger to a USB-C battery pack before plugging into the wall - the LED on the charger lights up like the battery is powering the charger.

There’s also tons of times where the device you want to power from the battery starts charging the battery instead. With both ends of the connector being identical I’m not sure there even is a _good_ way to fix this that ordinary consumers will understand, the better solution might have been to still keep different cable ends. I hope this mess eventually gets better.

What's more annoying is when you connect a wacom tablet to your laptop and your laptop says it's charging. Only when your laptop is dead do you realize it was low battery. I had this happen during a demo and it was terrible.

The simplest solutions are directional cables (remember directional HDMI cables?) or plug-in order (supplier first, receiver last).

Not an EE, possibly a dumb question, but... do USB-C devices know that a cable has been plugged into them before the other end of the cable is plugged in to something else?

The plug can have (has?) resistors that let the socket "feel" it, even if the cable is not attached to anything.

Just imagine this: Device A <--> Device B

Device A receives the cable first, upon receiving the cable, Device A notices that the cable is not "hot" ("hot" in the sense that there is voltage/activity on the wire), so then Device A decides to "turn on" the cable, make it "hot".

Now Device B receives the cable, it notices that the cable is already "hot" and decides that it must be the second device.

I would imagine it could work that way,

Im also pretty sure that the usb cables are "dumb" cables and not active cables like QSFP (these have chips at each end).

That assumes a device can tell the difference between not hot and not there, which was the parent's question.

This is not how USB Type C works. The cables are active (chipped), and the end devices have sense resistors. Each device can detect the presence and type of the cable, as well as the presence and type of an end device and total cable orientation (swapped or not).

This is all done before power is applied, as USB Type C is very explicit about not going hot (apart from a weak 5V V_conn for powering the cable).

Also, QSFP is not a cable, but a pluggable spec. A QSFP module is active, but the fiber optic cable you connect to it to is dumb. Link detection there works by sensing beam power.

I recently got a new workstation at work that is a MacBook Pro that connects to 2 monitors via USB-C.

I've learned somewhat superstitiously that if I don't connect and disconnect the USB-C cables in the right order the laptop may stop responding and I have to do a hard reboot. I thought this solved the problem but even now it still occasionally happens.

Your question makes me wonder if something like this isn't the cause (Mac misreading USB-C signal).

Based on my experience with a pre-USB-C Macbook connected to multiple monitors, USB-C is not at all required to get them to do strange things, up to and including hard crashing, depending on the order and speed of connecting monitors and sleeping the system.

If all devices actually follow the standard, then yes, a device should be able to detect a cable was plugged in even if the other end is disconnected. The trouble is finding compliant hardware, I fully expect it will be at least 2 more years before we start seeing common hardware that follows the spec.

Yes, the cables are active and can ID themselves when connected. Detection of another connected device (and orientation) is separate.

So yes, it is possible to detect which end gets connected first, but this absolutely shouldn't have any side-effects. It is unintuitive and fragile (if the 'master' device restarts, it will think the other end connected first, suddenly becoming 'slave').

Plug in order is a terrible idea. With potentially every restart/power cycle, the cable would have to be unplugged and re-plugged into the receiver.

Yeah i don't get how they went from a reversible USB plug (you can already find those in the A format, that use a small lip with contact pads on both sides) to a 20(?) wire monstrosity carrying anything from several amps of 20V electricity to thunderbolt wrapped PCI traffic.

Well, the problem is that you can't patent Ethernet anymore so they have to complexify it. I'm being a bit snarky, but if you look at practically all the modern standards (HDMI, Thunderbolt, UCB-C), they are basically 2-TX and 2-RX pairs like Ethernet.

The core problem with power is that you can't really pump laptop charge wattage down a USB cable without raising the voltage above 5V. Once you accept that, the engineering gets quite a bit more complicated and you have to negotiate so one side doesn't blow the other up.

It isn't extremely hard; lots of IC vendors make USB-PD chips that are fully compliant with USB-PD.

And, they're not expensive. EEs are just cheap mother fuckers, often, and they think they can do it "well enough" themselves and save a cent or two on the bill of materials.

This is not the fault of the connector.

It is the fault of the protocol.

USB is a clusterfuck.

More and more devices are implementing it. Whats your tource its on its way to the bottom?

Murphy's Law: If it can go wrong, it will go wrong.

Reply on news.ycombinator.com