Hacker News Reader

Comment by jedberg | Hacker News Reader

But now if I want three more full speed ports, I have to get a four port dock/hub that costs $179 and has to get plugged into the wall.

Am I not looking correctly or is there some technical limitation I'm not aware of or ...?

Edit: It seems like the consensus is basically "people wouldn't read the specs", which is a fair criticism. USB-C supports so many different types of devices that it would be impossible to make one that supports them all, and if you support a subset people would get mad that their device isn't in that subset.

View on HN

You might enjoy this story:

USB-C hubs and my slow descent into madness

https://overengineer.dev/blog/2021/04/25/usb-c-hub-madness.h...

It was recently discussed on HN https://news.ycombinator.com/item?id=30911598

One interesting takeaway for me was that there are actually some quite cheap hubs available from AliExpress and the "branded" ones are essentially the same with a logo printed on (and hopefully some additional QA).

> and hopefully some additional QA

Very unlikely. For this sort of thing it’s always been safe to assume that Amazon or EBay are AliExpress with faster domestic shipping and better customer service - that is what the premium is buying you.

USB hubs are only one of many items that are sold rebadged or not on both Amazon and Ali.

> one of many items that are sold rebadged or not

way to narrow it down, there

I can add this. A lot of specialized tools like watch tools, like high priced Bergeon/Horia, are found much cheaper on Alibaba with generic names.

Vector watch lathes are Sincere lathes but cleaned up machinist wize, and put in a fancy wood laser etched box. Basically you are paying 2-3x for packaging, and minor fine tuning.

I was once for tartifs.

I wish we could go back to products being made in certain European countries, and the USA, but that ship has sailed. (Meaning countries that paid a living wage, and offered social services to their poor. USA is sadily not even trying to help its poorest citizens.)

I still buy overpriced tooling/products on Amazon/ebay with jacked up prices because I just can't be sure on my intuition, and hate giving the Chinese my debt card. I should just buy prepaid cards, but they are another added expense, and very scammy.

(There are no laws stating Prepaid cards need to tell you how much you have left on their cards. In most cases you need to talk to a representative of the company. And yes--the amount of money consumers forget they have on their prepaid cards is staggering. (A simple fix would be a federal law stating that any money left on prepaid card is returned to buyer, or at least email them of their balance.)

If the Chinese government guaranteed they would go after any merchant that was fraudulent; I would buy most products from China.

Why? Because buying American does not seem to filter down to the employees, or in taxes. The CEO's just get richer?

Not sure, about the only items, my Lenovo soundbar is actually not a Lenovo one, it's printed on it, but they didn't assemble it, but a lot of producrs do this.

> only one of many items

I think i posted about it on that previous discussion, but my solution to this was to buy the plugable 3 port only-a-tb4/usb3-hub which let me be more selective about the other parts i used with it and I’ve been really happy with that.

It’s not cheap either but at least it’s flexible and will be useable for a long time as my needs change. I’ve been really happy with it.

I think these all in one things are always gonna cheap out on something in the end.

This one: https://plugable.com/products/tbt4-hub3c

Plugged into it is two monitors with usbc->hdmi cables (I think they use dp lanes internally) and one usb-a+ethernet hub. That's all I really need.

I'm pretty surprised that works for display passthru, especially for multiple displays. My impression had been that most TB/USB-C docks provide their display interfaces by internally being a PCIe graphics card.

I mean, the thing is it’s not really a “dock”. It’s just a tb4 controller in a box.

This one, apparently https://ark.intel.com/content/www/us/en/ark/products/189982/...

Many USB-C docks use DisplayLink to create a virtual graphics card over USB, it is implemented as software that renders using your computers/laptops graphics card that then sends the output over USB to the DisplayLink controller.

Thunderbolt docks (and some USB-C docks) instead will use Alternate Mode to pass-through the DisplayPort directly from the computer/laptop to the attached monitor. There is no additional graphics card or PCIe device involved in the matter, and no virtual graphics rendering that is sending frames across USB.

DisplayLink is not great, especially on macOS, even on Windows it would cause random slow downs and in general when using it with multiple monitors it would cause random tearing and other issues. It was fine for spreadsheets and the like, but for video it broke down quickly.

Whereas alternate mode just passes through your graphics card output, so there is none of that, and instead you get the full rendering power.

There are also non-Thunderbolt USB-C docks that support DP Alt mode. E.g. I had a Lenovo USB-C Dock Gen2, which uses 2 (superspeed) lanes for DisplayPort with HBR3. If a machine supports it, it can pass through 4k@60Hz. It uses the other two lanes for USB 3.1 Gen 2 10Gbit.

Though, I would always go for a Thunderbolt Dock. The extra bandwidth is nice (e.g. most Thunderbolt Docks support two 4k@60Hz screens). Also, Thunderbolt at least offers the option of using PCIe NICs, offering alternatives to the crappy Realtek USB ethernet controllers. Be cautious - even many expensive Thunderbolt hubs use USB Realtek. However, there are also models that have an Intel I210 or I225 on PCIe.

(The story changes a bit with USB4, since it can also optionally tunnel PCIe, but I wouldn't expect cheap USB-C adapters to do that.)

Right, there are some USB-C docks that are not Thunderbolt that do DP alt mode, but many of the cheaper ones seems to come with DisplayLink technology built-in.

DisplayLink is the only way to stop my MacBook from overheating when working with 2 external monitors

I have had multiple MacBook/MacBook Pro's, never had an issue with overheating while using external monitors. At one point I had 3 hooked up to my MacBook Pro.

The key thing is to not use it in clamshell mode.

Also DisplayLink wouldn't really help much with running cooler, since its drawing a virtual frame buffer using your existing graphics card anyway, although I don't know if macOS would recognize it and switch to the discrete graphics when in use vs continuing to use the iGPU.

I have two big monitors, I do not have space on my desk to keep laptop lid open (and that would be waste, because I would never even look at built-in screen anyway).

As for running cooler it really did change for me after buying DisplayLink docking station - before I would hear the fan noise constantly, now the laptop just stays quiet.

Hubs are trash in USB-C land.

I use USB-A hubs and an A to C adapter and that is FAR FAR better on the vaunted M1 macs for external keyboard+USB.

Fancy 200 dollar hubs? Lose my basic keyboard/mouse.

USB: a dumpster fire going on 25 years now.

If you don't need wired Ethernet, take a look at Kingston's Nucleum.

I picked this 3.2 Gen 2 hub for 10 bux, 10gbs, 3.2 usb, works great. I use it for my dual port keyboard + mouse, so I can just plug into my laptop or pc, so I only have to move 1 cable. Not sure why backwards compatible with usb 2 is a bad thing.

https://www.amazon.com/Inateck-Ports-Speed-Cable-HB2025A/dp/...

and hopefully some additional QA

You’d hope so, but I wouldn’t be too confident. The Satechi USB-C hub that my wife got (reboxed Chinese design) claimed to support 4k@60Hz, except that it didn’t work. Afterwards we found that other people had the same issue. Satechi support provided some undocumented Realtek firmware updater for Windows that didn’t work.

Can confirm, but not sure about better QA of the branded ones. Company issued macbook came with a "Satechi" hub, which died within 6 months, while two other noname AliExpress ones have worked with no issue for 2 years now

Tons of answers here. I think, if you want complex usb-pd offerings, you need to accept a pretty high $$$ price for all those power regulation subsystems you are demanding. I would be surprised to hear that indeed all the current offerings require wall power, that none can operate off laptop/uplink- I think you are wrong. But perhaps.

Overall I think the main current barrier to entry has been usb-c alt modes. If someone plugs a hdmi adapter into one of the hubs ports, ehat happens then? USB3+C's crazy lane splitting has been pretty suboptimal for figuring out what the role of a hub really is in the world today.

USB4 at least fixes that, makes usb a packet switched transport that tunnels other protocols. Now if someome plugs a displayport adapter into a hub, it's just going to use bandwidth. The whole tree under the usb-root doesnt have to totslly rebuild itself to make available a lane for some other purpose. The flexibility of usb-c, with it's two channels, is a complexity we've only finally specified a proper means to make use of. It's getting better. Looking forward to some other companies starting to offer hub chips, seeing this young new scene begin to evolve.

Does USB4 tunnel original USB (1.0-2.0), or is that still on its own separate pins? I was under the impression that electrical and timing requirements meant it was virtually impossible to tunnel it while remaining in spec.

Worse: at least thunderbolt provided USB downstream ports by including a PCIe-backed USB root controller. On USB4, all USB 2.0 downstream devices have to share the limited USB 2.0 bandwidth at the upstream port. They could have just specified a translator like they did for USB 1.1 -> USB 2.0, but they didn't and it's not possible to make one that works reliably/portably.

My understanding is that Displayport is already packetized, having USB4 packetize it again seems suboptimal? I think both Displayport and HDMI have official specs that work over the USB-C connector.

It's adding a small percent overhead in exchange for making it routable over a hub. Without the extra header it requires dedicated lanes.

USB-C has two physical channels, and the act of having to decide that a channel needs to be used for usb or hdmi or displayport is a complication that hubs were never well specified to handle. Hence there not being many such hubs.

USB4's tunneling means that the connector does not have to enter an alt-mode. Even when transmitting displayport or hdmi, the hub or hubs and host can remain using usb4. As well as making hubs suddenly make sense (after being an unspecified ball of confusion with alt modes), this allows bandwidth sharing that was previously impossible.

In short, that these other specs had packets is irrelevant, because those solutions werent really compatible with hub-like devices. There needed to be a common mode of interoperation, which hadnt existed before usb4.

>I think both Displayport and HDMI have official specs that work over the USB-C connector.

You've answered your own question here. "over the USB-C connector" meaning direct connections between two devices. The moment you put a hub inbetween it won't work.

USB C Port -HUB-> DP Port works, USB C Port -HUB-> 4x USB C port -USB-C-to-Display-Port-adapter-> DP Port does not.

> makes usb a packet switched transport

Guess that means the next logical step is to evolve USB into ethernet.

You kid, but USB PD reinvented ethernet among other things: https://nitter.net/whitequark/status/1035760558237204480?lan...

Eventually ethernet always wins.

and then infiniband.

> I would be surprised to hear that indeed all the current offerings require wall power, that none can operate off laptop/uplink- I think you are wrong. But perhaps.

Indeed, there are products that don't require wall power. I got a Chinese model [0] off Amazon that does Ethernet + 4K@60 over DP/HDMI (alt-mode, not DisplayLink) and can even drive an external spinning drive, all from the USB connection from the PC. The catch is that it comes with a dual-port cable, meaning it requires two (adjacent!) ports to work. I haven't tried it with a regular cable, so I don't know whether that's a hard requirement. It's also able to do PD pass-through if you connect it to a wall.

One downside is that I've sometimes had the screen go dark for a second, and other people have complained about this on the Amazon page. So it's not clear how reliable this will be in the long run.

---

[0] https://www.amazon.fr/gp/product/B09MTTZ36K

Isn't the problem that anyone making such a device would have to deal with an endless stream of returns from customers who assumed they could just plug anything with the right form factor in only to discover that no, the bargain hub won't handle Thunderbolt, or 100W PD, or...

Maybe! I know better and read the specs, but you're right, a lot of people might get confused that it isn't charging their 100W laptop.

We've trained people for decades that if it fits it should work. Now we've got a multitude of effectively different protocols all using the same connector shape.

I learnt "fits but won't work" running 5v(worked for the 486DX2) through a 3.3v CPU(DX4 100). The nymphs decided it was too hot and left.

Since then, building your own PC has gotten easier and easier. As you say, mostly if it fits it works. Which I think is great!

But how did plugging stuff in from the outside go the other way...

I think we're talking more about end user external connections. I'm sure I don't have to tell you that back in the day a bajillion things ran through a DB-9s and DB-25s in electrically incompatible ways and how much easier it got to tech support our families with USB 1 and 2.

On the very plus side, with ATX, you can't plug in the motherboard power incorrectly, unlike AT.

Depends on how strong you are.

Or smart

I absolutely agree with you that USB 1 and 2 did great things for family IT support.

But now we're at the USB 3 (and seemingly 4) stage of things - hence my "plugging stuff in from the outside go the other way..." statement.

Between varying capabilities offered through USB 3 and cables/hubs/docks that support those varying capabilities it is a bit hit and miss whether plugging stuff in will work or not.

I now know (not yet personally verified, from other comments on this article) Lightning 4 branded cables should support all connectivity possibilities and power delivery. It doesn't solve patchy device support but at least it is one step better. Connect two USB3 devices with a Lightning 4 cable is a bit silly marketing wise but at least it is easy to support.

Building a PC should be expected to require a certain level of RTFM that I don't you can responsibly expect from people just trying to charge their phone or plug in peripherals.

And display outputs too. What happens if the user plugs in multiple displays to the hub could be an important scenario to support...

(and for example, macOS doesn't support DisplayPort MST, requiring an encapsulation over Thunderbolt to support that scenario...)

tldr: it'd be possible to make such a hub, but it's so complexity fraught that a _lot_ of companies would avoid doing it, USB Type-C is just too complex.

This was the story moving from USB 1 to 2, and this is still the story for people trying to charge their laptop (USB-A to C cable) from a random phone charger or an underpowered USB-A hub.

Coupled with the nightmare of guessing which USB-C cable supports what, I have the feeling preventing us from further confusion has never been a decisive factor in product making.

I have a fantasy where everyone who worked on recent USB standards must personally provide tech support when there's a problem.

They probably do. It's an industry-driven standard; they added all that stuff exactly because they intended to sell it to consumers.

But the people who worked on the standard directly aren't doing the tech support

How is that different from any other technical standard or product design?

It behooves one to read the conversation before replying. The implication is this: perhaps if the people who wrote the spec personally had to offer tech support for the spec they wrote, they might have an incentive to write less insane specs.

I get that. But spec writers never have to do that, yet there are many other specs that are presumably less insane. So it's unlikely that not-requiring tech support is a causal factor in USB's supposed insanity.

Reading the thread would also explain that you're discussing somebody's 'fantasy', not expectation or belief of causality.

This is classic HN stuff, I find it actually valuable to see an explicit enumeration of all possible interpretations of some statement, also somewhat humorous.

To be honest, I wanted to essentially say that it's a thoroughly stupid fantasy. But in more polite terms.

“We could sell this feature to consumers even though it would confuse them”? That’s not what tech support would do.

I’d pay to watch them connect up a monitor, then try fix the issue when it mysteriously stops working. Then have the replacement also fail. Then find another that works but looks identical.

Support guy: "M'am, can you tell me what is on the plug of your hub? Is it a sort of pitch fork with three numbers, or is it more like a pitch fork with a lightning bolt and two numbers?"

Customer: "Son, I'm 75, I don't see so well anymore!"

Support guy: "Well, unless we can find out if you have a USB 3.1 Gen 2b compliant with Power Delivery 20W hub or a USB 3.2 Gen 1 DP+ SuperSpeed 10GB hub, we won't know what alternate modes your cable must support in able to connect that mouse to your laptop m'am"

But isn't that already the case with the ports directly on computers?

I have a laptop that can do USB 3.2 (or something). I can charge it through those ports, but it doesn't do Thunderbolt of any kind.

I have a small desktop in front of me that's power through an external adaptor, like a laptop. It has a USB-C port that doesn't do video output of any kind and which can't be used to power the computer. It has another USB-C port with a bolt on top of it. It does thunderbolt, it can output video (I've driven an Apple TB display with it), but still can't be powered through it. It also doesn't output video through the USB-C dock, which works on the laptop.

I have another laptop with a USB-C port with the bolt on it, and it can be powered through that port.

These are all ports directly on the computer, no hubs required.

> Relatedly, why can't I get a simple USB-C hub? I have four USB-C ports on my laptop and I need six. Back in the USB-A days, I'd just get a $30 hub and four new full speed ports at the cost of one on the laptop.

The problem is that, I believe, a USB-C port is effectively required to provide a minimum of 5V at 3A on every port. I believe that USB 3 only requires 5V at 0.9A and USB 2.0 only requires 500mA (you nominally have to negotiate up to 500mA from 100mA--but nobody ever does that they just suck down 500mA without paying attention).

At that point, you can see how much more power USB-C is shoveling around. A 4-port hub is shoveling 60W minimum for USB-C, roughly 20W maximum for USB 3 and roughly 10W maximum for USB 2.

And, to be fair, life wasn't magically better in the USB 2 hub days. I have had loads of USB 2.0 hubs that do screwball things once they get a little overdrawn on their power. I eventually discovered that a good way to predict reliability of my USB 2 hubs was to look at the amperage of the wall wart that they shipped with it (the bigger the better).

And USB 3 isn't automatically better. I literally just had to replace a whole bunch of old, very nice, very stable, aluminum wedge USB 3 hubs because Microsoft updated something and simply plugging them in would blue screen the computer. Once I bought all new plastic chintztastic USB 3 hubs, Windoze is now happy. Microsoft can go die in a fire.

What’s a simple USB hub?

The standard is bonkers, and encapsulates other complex standards for kicks. Implementation requires a bunch of different chips and using the correct, unlabeled wire that is identical. My company is the proud owners of about 50,000 USB-C and USB—C/Thunderbolt docks.

They are both amazing and awful. One vendors thunderbolt dock is better than the plain USB-C, another is the opposite. Minor, undocumented differences in firmware or chip revisions may have a major impact on quality. One OEM’s driver bricked laptops.

When stuff works, it’s amazing to the point that it seems impossible. Some higher end workstations are pushing 5x more IO than a (admittedly aging) midrange SAN that’s on the floor can.

> What’s a simple USB hub?

The quintessential "simple USB hub": https://www.amazon.ca/UGREEN-Adapter-MacBook-Chromebook-Pixe...

Replace the USB-A ports in the above with USB-C ports, and it's what the GP poster is imagining.

But, for some reason, nobody makes that.

Even though it'd be USB-3.1 either way; even though you can plug USB-A-to-C cables into this hub, and it'd be exactly the same as plugging USB-C-to-C cables into a hypothetical USB-C-replicating hub.

I think the point is that the built-in ports on your computer might support any combination of a large number of possible features, and that it is difficult/expensive to make a hub that supports all of them, and people would hate a hub that happens to not support one particular feature that their computer’s built-in ports support.

But this is what I meant by:

> you can plug USB-A-to-C cables into this hub, and it'd be exactly the same as plugging USB-C-to-C cables into a hypothetical USB-C-replicating hub

I.e., anything you can plug into a USB-C port using a C-to-C cable, you can plug into a USB-A port using an A-to-C cable. So if the problem is idiots thinking "if it connects, it should work" — then wouldn't these same idiots be complaining that their Thunderbolt peripherals, external USB-C monitors, etc. don't work when plugged into one of these simple C-to-A hubs via an A-to-C cable?

But that doesn't seem to stop the C-to-A hubs from existing.

It’s not at all unreasonable for consumers to want a USB hub to function like “just give my computer more ports.” So no, it wouldn’t be unreasonable or idiotic to expect something that works plugged directly into your computer to also work when plugged into your computer via a hub.

It is less reasonable for a consumer to expect that any chain of adapters or cables that uses multiple totally different types/generations of connector would still support every feature of the computer’s built-in port.

> idiots thinking "if it connects, it should work"

You mean thinking things will work in the reasonable way that they have for decades prior to USB-C?

"If it connects, it should work" used to be largely true (modulo OS compatibility, which was easy to check by users, and wasn't even an issue anyway if you had a reasonably recent Windows version). It's hard to blame the users for not understanding why things have been suddenly made much more complicated by a standard that supposedly simplifies things.

100% agreed, and this is a royal problem with USB-C.

1. Pre-USB: Several different cords, with different shape ports to distinguish them.

2. USB-A: Several identical cords, with identical ports that function the same.

3. USB-C: Several different cords, with visually identical ports and no clear way to distinguish them.

It's like the designers of USB-C saw the success of USB-A and thought it was due to minimizing the number of shapes, rather than minimizing the number of factors to be tracked. When devices come with pleas to only use the provided charger and cord, rather than any cord that meets the standard, you know that something has gone horribly wrong.

Exactly. The fact this thread exists and everyone has a pretty informed, yet divergent point of view speaks to the problem.

Imagine being a consumer product marketing person trying to sell a device that meets the needs of a MacBook Pro user, a Samsung Galaxy user and a 6th grader with a Chromebook.

Egads, yeah. There's so many use cases that no standard can accomplish all of them. USB-C tries, and even it failed to cover some pretty big use cases. If I were to try to list out qualities that a USB-C port could have, I can come up with several different features.

0. Base level. No significant power draw in either direction. No significant data flow in either direction. I'll label this as the USB-A solution.

1. Significant power draw, recipient.

2. Significant power draw, provider.

4. Significant data throughput, recipient.

8. Significant data throughput, provider.

Between these, there's 16 different combinations, and pretty much all of them exist. As an exercise, I tried to see if I could come up with reasonable examples for every one of them.

0. Every single USB-A device.

1. Phone charger, typical use.

2. Wall wart for a phone charger.

3. Rechargeable battery.

4. Monitor with dedicated power supply

5. Monitor without dedicated power supply

6. Charge a phone while playing video from it.

7. External tablet, currently in use streaming video from a Raspberry Pi, Raspberry Pi powered via the tablet.

8. Desktop graphics card.

9. External graphics card.

10. Desktop Graphics card powering a single-cord monitor.

11. External graphics card powering a single-cord monitor.

12. High-speed external storage.

13. High-speed external storage without a dedicated power supply.

14. First in a daisy-chained series of monitors, with a dedicated power supply.

15. USB-C hub

Which comes back to why there are no decent USB-C hubs, because every single port would be expected to be effective at every single task. Which is ludicrous, and strikes me as why there are no good USB-C hubs, because it's a Herculean task.

USB-A did a lot of things incredibly well (non-streaming data transfer, peripheral management), and a lot of things relatively poorly (device charging, video display). USB-C tried to improve on the aspects that USB-A was bad at, and made an absolute mess of the things that USB-A was good at.

The advantage of USB-C, if you don't realize it, is that there are many devices that want to be clients sometimes, and hosts other times. You know, like being able to plug your phone into your computer, or your gamepad into your phone.

Previous to USB-C, USB ports were either "host ports" or "client ports" (thus all the different shapes); and devices were required to assume that if something was plugged into a client port of theirs, it was a host; and if something was plugged into a host port of theirs, it was a client. No dynamic host/client negotiation — there wasn't even a protocol for it.

Instead, there was "USB On The Go" (USB-OTG), which would signal a "USB-OTG" compatible device on its normally-a-client port, to switch that port+USB controller into host mode, based on the physical wiring of the host side of the special "client but actually a host" connector on a special USB-OTG cable. And, of course, users would never consider this problem ahead of time — why would they? — so there was no demand for host/client hybrid devices to ship with USB-OTG cables; instead, they were something you'd realize you'd need only too late, once you already had the problem; and now you're running to an electronics store at 8PM so you can plug your Android tablet into a scanner.

With USB-C, all that is gone. USB-C devices are required to signal whether they're a host, client, or hybrid device, so that hybrid devices can auto-discover whether it's a host, client, or another hybrid device on the other end of the line, and do the sensible thing in each case. No need for special "override" cables. And no need for discrete host-side vs. client-side connectors. Host-ness vs. client-ness is now a logical part of the USB protocol, rather than an electrical part of the protocol.

This also means that if you have, say, a laptop, you only need to bring one (good, beefy) USB-C cable with you — probably the one that came with your laptop's USB-C AC adapter; and this cable will both charge your laptop from the wall, and allow you to plug in any USB-C devices you might need to connect to. Or, in the case of a display or hub with power-delivery — both! (IMHO the point of the original Macbook's single-port USB-C design was to serve as a forcing function to make people realize that the combination of battery life + USB-C "switch-around-ability" means they can — in most casual use-cases — just alternate between charging and connectivity through a single port+cable, rather than needing both at once and thus needing two ports + two cables.)

> “If it connects, it should work" used to be largely true

Have you ever worked with DB-9 or DB-25 serial cables? If you had a cable that was wired correctly for the connection you wanted to make, it should indeed work, but you still would have to figure out baud rate, number of stop bits, and parity. Detecting what kind of null-modem cable you had basically came down to getting out your multimeter.

On PCs, you also had the traditional parallel port (mostly unidirectional, with IIRC two bits of information that a printer could send back: “don’t send more data; I’m busy handling the earlier data”, and “out of paper”) versus the much faster, bidirectional iteration that was used to connect such things as hard drives and CD players.

> "If it connects, it should work" used to be largely true

Gods no! There are innumerable devices with D-sub, mini-DIN, RJ-11, RCA, or TRS connectors, that put those connectors to entirely different and incompatible purposes. With entirely-incompatible electrical standards, even, such that plugging client-devices intended for one purpose into host-devices intended for a different purpose will electrically short one or both.

Remember the iPod Shuffle's "USB-A to TRRS" cable? Guess what happens when you try to use that cable to, say, plug your computer into your hi-fi? Or — less absurdly — into your TI-84 calculator?

The "Universal" in USB is in contrast to the DB-9 serial port, which was decidedly not universal. Yes, there were actual "serial ports" and "serial port devices" — the spec is called EIA/TIA 232, if you're curious — but there were tons of other things that used DB-9 connectors (male or female) but weren't EIA/TIA 232 devices. Plug one of those into a serial port, and poof! — there goes the magic smoke.

You don't have to worry about that with USB (including USB-C); USB ports and cables all follow a single unified electrical pin-out specification, whatever they're carrying. Even USB-C devices that mis-implement features like USB-PD (e.g. the Nintendo Switch) won't ever start an electrical fire, because — whatever logic-level protocol support they might have mis-implemented — they're still "USB-standard devices" on the electrical-pinout level.

Now, these are all fairly-old connectors I'm mentioning, so you'd think we'd have learned better by now; but special mention goes to the Atari joystick port (technically a type of D-sub connector), which is still to this day slapped onto arbitrary generic Shenzhen-special "TV plug-and-play consoles" — with no two devices being wired the same way, and so only actually supporting the gamepads the console comes with.

When you think about it, relative to this disaster, USB-C does guarantee you that 1. plugging any two USB-C ports together with a USB-C cable won't throw your circuit breaker; and 2. as long as both devices are data devices, you'll get at least USB2 data-transfer out of the arrangement (if you're plugging in a legacy USB2 device), and much more likely USB3 speeds. Everything is actually compatible with everything, as long as you don't care about getting the fastest line-rates possible.

It seems we are pretty well settled on the fact that on a good laptop, all usb C ports should be capable of all features. And then on a hub, you can expect the usb-c port to be charging pass through and nothing more. All other ports on the hub work as you would expect.

I wouldn't ever expect to see a hub where the usb-c ports pass on the full thunderbolt capabilities.

> I wouldn't ever expect to see a hub where the usb-c ports pass on the full thunderbolt capabilities.

Thunderbolt is specifically designed to permit daisy chaining (unlike USB). So it's actually quite normal for a device to pass Thunderbolt through to another, as well themselves being a Thunderbolt device.

The problem is the cost of building a little hub that has a backplane capable of handling 40Gb is not cheap. Should alt mode work? Does anyone actually understand wtf that means?

Combine that with the market expectation that a hub is a feature of a $12 keyboard and the baffling combination of cables, etc, and you get an expensive, limited utility product that will have low sales and high return rate.

USB-C is a comically poorly designed human interface.

I would expect a basic hub to do 5Gbps or 10Gbps. No need for 40.

> USB-C is a comically poorly designed human interface.

You can blame alt modes, but the port itself is fine and when used for pure USB the issues are quite minor.

Yeah, I just want the classic dumb USB Squid, but for USB-C. No power delivery, just a port multiplier. I did find one that was way too expensive that gives me 3 ports instead of 1.

$40? https://www.amazon.com/Satechi-4-Port-USB-C-Hub-Compatible/d...

Found it bouncing through https://www.amazon.com/USB-C-Gen-Hub-Adapter-7-Ports/dp/B09M...

Left hub out, googled "USB-C 4 port" to get there.

That's data transfer only. I want it for my iPad and phone, so I want charging too.

Okay, well that doesn't seem surprising that it wants to be plugged into the wall then (because people will ignore the real limitation on power input from the supply device and complain).

Yup, that's always the answer. The customer doesn't give you the full specs, complains about why this isn't available already. When you deliver your proposed solution it turns out they had some extra physically impossible to fulfill requirements. Reminds me of "The Expert" skit.

The customer is used to being able to plug all of their stuff into a lot standard wall outlet and their headphones in any standard 3.5mm jack so it's not at all surprising that they don't automatically provide all the specs. And frankly customers shouldn't have to care. USB engineers and PMs completely botched the design of USB-C.

I think it’s interesting how a lot of knowledgeable people will chime in on here on certain feature compromises and limitations with USB-C and it always sounds reasonable in isolation. But you only have to take a step back and look at the specification in its totality to see that it’s a shit show. It’s like this thing was designed by a bunch of subcommittees and no one had a clear understanding of or oversight over the big picture.

I am ok with various devices having various capabilities. Not too surprising given they are made from various bits.

But I want one damn cable to bring them all and in the darkness bind them. And a way to know what cable that is.

Sure, that'll be $129 for a 1.8m cable. Will you be buying one to power your wireless charger? https://www.apple.com/shop/product/MN713AM/A/thunderbolt-4-p...

That's a bit dramatic. It's a longer active cable. You can get a certified passive Thunderbolt 4 cable for 25 Euro including VAT:

http://shop.caldigit.com/eu/Adapters%20And%20Cables/Cables?p...

Short answer to the grandparent: buy a certified Thunderbolt 4 cable. They have clear markings (a Thunderbolt logo, along with the digit 4). They are backwards compatible with Thunderbolt 4, Thunderbolt 3, USB 4, USB 3.2/3.1/3.0, USB-C Gen 1/2, USB 2.0 and can charge 100W.

Well, that exists, it's called Thunderbolt 3 (or 4, since PC makers decided to try to get away with a shitty half-version of TB3).

And the cable exists too, but you have to pay through the nose for it, and people expect all cables to be $5 at Monoprice.

Turns out that 40Gbps and supporting any and all monitors at the same time and 100W of power isn't an easy spec.

Everyone mocked Apple for going with Thunderbolt so early, but it turns out they had it exactly right, just like they did with FireWire, another standard that was super-fast for its time and also included robust power delivery.

The customer assuredly cannot do that for a standard wall wallet.

The original poster said they wanted to plug in an iPad and phone for charging. The right comparison would be a customer buying a power strip, plugging in two space heaters at the same time, and wondering why the circuit breaker triggers.

The closer comparison would be a customer wanting a power strip that they can plug multiple phone chargers in to, I think.

A fast-charging phone is to a standard USB port as a space heater is to a standard wall socket.

The promise of usb (any version) is that you could have one port that everything plugs into. Usb-a isn't perfect in that regard, but it's mostly there. Usb-a backslid considerably on that promise.

Okay, but old powered USB hubs were still under $30 and the power cable was optional.

They didn’t provide dozens of watts per port.

I don't need dozens of watts per port. I'm not trying to charge my laptop. I just want to power e.g. a spinning-rust 2.5" external USB hard drive. Which USB ports are able to do at their default line voltage, pre USB-PD negotiation; but which these "data transfer only" USB-C hubs cannot do for some reason.

It wasn't intentional on my part, but it appears the hub I link will provide 3.5 watts of power to each port:

https://support.satechi.com/hc/en-us/articles/4405450897435-...

And I'm again not familiar with power requirements for external drives, but it appears that is sufficient for at least some 2.5" USB drives?

Not being able to power spinning external drives from USB A ports on laptops is a known issue for a while. Some manufacturers even provided you with a double USB A ports on the laptop side to make sure there is enough power for that.

You're thinking of 3.5" drives. 2.5" drives are a bit easier to power.

So you want to get 4 × 10W out of a single 2.5W (at most 10W) upstream port?

From what I've read the MacBook can output 15W on a port, so I'd expect a hub to be able to output 15W when one device is plugged in, 10W+5W with two, etc.

The macbook is using thunderbolt, not pure usb-c.

This is also a good point. Even I use the terms interchangeably when I shouldn't, and I should know better.

Note that these terms are not strictly comparable. USB-C is just a connector specification. There are many protocols that utilize USB-C: USB 3.2, USB4 and Thunderbolt being some of them.

Not having ports with constant capabilities is a source of customer dissatisfaction when devices work when plugged in independently but not together.

Consider for instance the new 14" and 16" MacBook pros: They got rid of the 4th Thunderbolt port in favor of an HDMI port, likely because of internal bandwith constraints. They could have kept Thunderbolt and HDMI, and only have one Thunderbolt port on the right side work when HDMI is plugged in, but they didn't, because having your external drive work only when your monitor is unplugged is unintuitive.

They were right originally when then went usb-c only. You could output hdmi on any port or do anything else with the port. We had gotten to the point where most offices had a usb-c adapter on the end of every loose hdmi cable so it wasn't an issue.

You mean Thunderbolt only. Not USB-C. Let's credit Apple for what they actually did, which was to go with the non-shitty port from the beginning even though it was expensive and everyone mocked them for doing so.

doesnt §4.6 of the USB-C spec show the precedence of power source usage? I.E. there is no 10W option, just 3A, 1.5A, 0.9A, 0.5A at 5V or 15W, 7.5W, 4.5W, or 2.5W. so then your 15W port turns into 2x 7.5W ports or 3x 5W ports. I dont know if you can count on the upstream port being 15W. if its 7.5W do you do 4.5+2.5W and 3x 2.5W?

also, I've found that no one obeys the current limit for inrush so now your downstream ports have to allow 7.5+W until they enumerate and you can lower the current limit.

additionally, people are complaining about not being able to tell cables apart. can they tell the power requirements and calculate how much power a combination of devices requires?

Those are only for dumb power supplies using resistors to signal max amount. Negotiated power is very granular.

> additionally, people are complaining about not being able to tell cables apart. can they tell the power requirements and calculate how much power a combination of devices requires?

If they don't know device power levels they can just learn what can simultaneously go on a single port when they try, like with USB 2.

At most 10W?

2A at 5V is the maximum you can get without USB-PD. 500mA 5V is the maximum that’s guaranteed.

Why are we assuming no PD?

That's the issue really. There are so many features that you will never get a device that does exactly what you want and no more. So you either get the basic usb3 only or you get the fully featured thunderbolt dock which costs a lot.

I’m pretty sure for my 7 port usb-a hub (usb 3) I have to plug it into the wall power too. My max pro doesn’t send enough current to to hub so if I want to to be power stable I have to supply extra power…

Spendy but capable-looking: https://plugable.com/products/usb4-hub3a (I haven't tried it, but I have had good luck with some of their other products, they take Linux support a little more seriously than most... though not on this hub, apparently...)

Alternatively, any old existing usb3 hub plus usb-a to usb-c adapters will give you a lot more flexibility.

I finally went ahead & bought a decent quantity of adapters- they're tiny, cheap, and add lots of flexibility. They have helped me consolidate my cable collection significantly while retaining a bunch of charging devices I already had installed.

Once you start trying to layer in usb-pd things get much much more complex. But, as a bus for peripherals, longstanding usb hubs do more than fine.

The problem is the plethora of features USB-C supports, and the bandwidth required to pass them all downstream. In other words, you need to passthru PCIe to the hub, and break out to USB-C and other ports from there,

Which is simply expensive.

On the other hand, Kingston's Nucleum is a nice hub. It provides two additional USB-C ports with PD support in one of them, provides two USB 3.0 USB-A ports, a proper card reader and an HDMI port.

I take photos and my camera produces 50MB images, and carrying them over a nice card reader which can use the full bandwidth of the card is nice, and Nucleum delivers.

I looked into how USB 2 unpowered hubs work and it sounds like "not very well."

For example, Anker sells a USB 3 hub and it's up to the user to make sure that whatever is plugged into it doesn't exceed 900 milliamps.

I'm thinking that a good way to go would be to get a hub that connects to a USB-C port and lets you plug in USB 2/3 devices? That way there's plenty of power for them without needing a power adapter. Plus USB 2 devices are cheaper.

These hubs need some sort of power indicator. Red flashing = running out of power.

"Simple USB-C hub" is very underspecified. Exactly what do you want it to do?

Speaking with my usability hat on (my degree is human machine interaction) that isn't a useful question as there is one correct answer: plug whatever devices in, in whatever order and combination.

OK, you plug a power supply, two laptops, and three monitors into a USB-C hub. What happens?

Any kind of bus where peers can negotiate protocols is going to have severe usability issues. Just look at the MacBook that's plugged into itself: https://photos.app.goo.gl/G2PivEo1JrKKtA9W6

Given the fundamental usability problems of USB-C, I don't think there is anything anyone can do toward making a simple USB-C hub.

> Given the fundamental usability problems of USB-C

That is the whole point.

Okay, so you want a hub that provides 240W per port and supports 40Gbps per port, right?

Are you willing to pay the hundreds of dollars that’s going to cost?

Cost to create a hub should have been a consideration of the spec designers.

I'm involved (on the very edge) of C++ standards, and we often reject proposals because it would be too hard to implement, or too hard to teach. The proposal itself would be a useful, but the costs are too high.

I use a USB-C to USB-A converter, plug in an old USB-A hub, and plug USB-A to USB-C converters into each port.

Works great. You'll be stuck with USB 2.0 speeds, but it connects to, passes data to, and charges every single dumb device I have with no problems. Mice, keyboard, USB-C headphones, my phone, other laptops that charge over USB-C... and that's all I ever really wanted.

Apple's 2-USB-C-port laptop was literal insanity.

> Apple's 2-USB-C-port laptop was literal insanity.

Actually... I do get the idea. 99.9% of the time I only use one of my four USB-C/TB ports, and that's for power. The other 0.1% is when I connect my Android phone for a firmware update (it's rooted, so have to go the hard way) or an SD card reader.

I routinely have keyboard + monitor + power :| I've done a lot of "unplug power to do something briefly" because of it. And tbh the vast majority of monitors-with-usb-hubs ports I've used have been one of: not worked at all, absurdly slow like <1Mb/s, or provide only like 0.5A of power shared across all ports. They're so, so bad.

I know good ones exist, but offices don't tend to buy those. And I don't really want to drop $1k on a new monitor. Thankfully this is a secondary device, for fixing stuff that breaks from the M1 migration.

This is a fantastic reflection on what I consider to be a pretty big failure of the while One Port to Rule Them All idea.

Relatedly, why can't I get a simple USB-C hub? I have four USB-C ports on my laptop and I need six. Back in the USB-A days, I'd just get a $30 hub and four new full speed ports at the cost of one on the laptop. But now if I want three more full speed ports, I have to get a four port dock/hub that costs $179 and has to get plugged into the wall.

I'd be happy paying the $179, as long as I was sure the stupid thing did everything the port on my laptop does. And didn't require external power.

Right now, we're paying the $179, unsure if it actually does what we think (because the specs are so insane), and it's less usable. USB-C, in practice, is a giant step backwards. In theory, it's amazing, and when it works, it really is amazing, but holy crap, what a bowl of spaghetti.

> as long as I was sure the stupid thing did everything the port on my laptop does. And didn't require external power.

As I understand if you want it to do everything your laptop port does and expect to be able to use more than one port at a time there is absolutely no way around external power. And if you can only use one port at a time it's an adapter not a hub. You're asking for the impossible.

Does the onboard (to the laptop) USB-C port/controller not pass enough power for more than one port? Or just not enough power to provide "full power" to extra ports?

I'd be happy to compromise that aspect as long as the various other data specs are met. I'm not trying to charge several big devices here, just AirPods and a phone, plus run two external monitors. And since I need to use one of the USB-C on the laptop to charge the laptop, I'm one port short sometimes.

From my experience it never was simple in USB-A days either. The cheap hubs or the one mounted on the desk just didn't provide enough voltage. Sure, it might work to run your mouse (if it's not a fancy one with LEDs), but good luck connecting an external HDD to one.

The limitation is that by wanting a USB-C hub, you've painted a dollar sign on your head. USB-A hubs are cheap and plentiful, and for many applications work fine with C devices with an A->C cable. If that's not sufficient, you must be a big spender.

This doesn't seem accurate at all. There are cheap usb-c hubs which only pass down the usb3 capabilities. The expense comes down when you want to be driving 4k monitors, 1gbit ethernet, and charging your laptop through the same hub and cable which requires both extremely high speed data exchange combined with heavy duty power transfer and management.

The expense is not the problem, inability to look at hubs and cables and reliably understand which ones can actually drive 4k monitors, 1Gbit Ethernet and charge the laptop at the same time without issues is the problem.

The Apple one does that, I have tested that specific combo. Everything else I wouldn't be confident of.

this is precisely it. there's just so much confusion around USB C and the fact that not every Type-C plug functions identically.

I miss USB 2.0.

Usb 2.0 and 3.0 still works exactly as it always did. If you only care about basic usb function, they all work. It only gets complex when you are trying to run a 4K video and 80w over the cable.

> I have four USB-C ports on my laptop and I need six.

I have four USB ports on my laptop and I need three. Unfortunately two are C and all my cables are A.

I really wouldn't mind if USB-C just went away. I've got a single device with a C plug (which came with an A to C charging cable, of course) and some other brand new charging cables won't charge it if I don't hold the cable pushed in manually. Not sure if the cables are bad or the hole. I also find it harder to plug in than micro (maybe because there's no narrow point on the plug which you can put in the middle of the hole and then move it to the edge to automatically align it perfectly) and hate the metal on metal sound/feeling this gives. The plug seems to be badly designed compared to the other USB ports (never had this issue, from A to B to mini to micro to micro B) and very often has incompatibilities where you're just guessing whether a display is going to work at all.

And then there's the ports problem. I've got a number of old cheap USB hubs laying around, like every techie I guess, but C hubs? Well... So this can now all go in the trash. We should stop this madness and come up with a better plug rather than a worse one, if we need to change all hardware anyway we might as well do it right the first time.

I have a Platinet 4 port USB 3.x hub/docker (USB-C uplink) that doesn't need a external power, however I didn't benchmark the speed, but I don't have the disconnection problems that I had with unpowered 2.0 Hubs.

USB-C is a pretty complex spec with many options. There aren’t a lot of chipsets for it. Most hubs make a lot of trade offs to get the cost down.

Indeed, it's the chipsets that are the limit. The IP is complex, so there's no one who really wants to design whole new cut-down silicon to implement a "simple" USB 3 dock with fewer features to sell for less money. And since the chipset supports it, and technically you can't certify it as USB 3 without it (not that all these devices are fully certified), you might as well just use what the chipset gives you. In fact, in terms of implementation, that's probably the easiest as you just follow the reference design. So that's what they do.

I have also been looking for such a device for a long time. It's baffling that it doesn't seem to exist.

how much operating capital we need for some stuff like this designed to spec out of shenzhen?

USB 1 and 2 were relatively simple protocols, and had very low energy profile (I think max power on the original USB 2.0 spec is 2.5W (5V, 0.5A) [0]. Same with bandwidth.

Back in USB 2.0 days, a single port could consume at most 480 Mbps in bandwidth, so let's say that your typical USB 2.0 hub consumed at most 2Gbps. That's very easy to accomplish in terms of I/O [1] (1 PCIe Gen 1.0 x1 port).

Now imagine 4 USB 3.2 or eventually 4.0 full speed ports: that is 20 Gbps per port, and Thunderbolt 3 has 40 Gbps per port. That means a 4-port HUB would need to have I/O capabilities of 80 or 160 Gbps. That is even beyond PCI express 4.0 w/ 4 lanes (8GB/s). A single Thunderbolt 3 port can consume theoretically half as much as the best and fastest NVMe SSDs (around 7 GB/s).

Basically, what I am trying to say is that it is getting harder and harder to achieve with USB 3.2 (and eventually 4.0) what was easy to do with USB 2.0. Our computers are just not getting fast enough to support all these I/O speeds. That's why most PCIe extension cards only offer a single USB-C port. A standard PCIe Gen 4.0 1x port cannot even support that at top speed!.

Take a look at this card [2]: a single USB port doesn't even support 20Gbps speeds, and half the ports share the same 10Gbps bandwidth. And this card requires a PCIe 3.0 x4 port.

And this is just bandwith. In terms of power, it is pretty much the same issue, Power Delivery can consume 100W at most. That's more power than the normal draw of a typical "ultrabook" laptop (around 60-65W). Now imagine 4 of these, 400W. That's a standard desktop PC PSU requirement.

tl;dr it is just not possible to support this much I/O speed and power draw with cheap consumer technology. That's why good USB 3.2 and Thunderbolt 3 hubs cost much, much more than USB 2.0 ever did. Also why they are much more complicated devices.

[0] https://www.tripplite.com/products/usb-charging [1] https://www.crucial.com/support/articles-faq-ssd/pcie-speeds... [2] https://www.newegg.com/inateck-ku8211-red-pci-express-to-usb...

Reply on news.ycombinator.com