Álvaro PrietoDo you have a J-Link compact and wonder why it only works with _some_ USB-C cables? I looked into why that is 🤬
Krutonium://@alvaro Not gonna lie, loving the write up, but that I can't scroll if my cursor is off on the left or right side of the text is irritating as hell
@krutonium probably the damn Hugo template I’m using. 🤦♂️
I hate UIs
I hate UIs
@alvaro I've been using Paper, maybe give that a peek? (Or just mod what you're using)
@krutonium not bad. Thanks for the recommendation! Now I’ll have to find the time to play with it and make sure nothing breaks 😅
Michael Cook@alvaro Thanks for the explanation. I deal with a few devices that act like this and it’s always a huge pain.
Worik@alvaro fascinating
> get the maximum power available(over 3A and up to 48V)
I had no idea that nearly 150 watts can be delivered by USB C
That is huge....
@worik the new standard supports 48V @ 5A!
Jernej Simončič �
Wilfried Klaebe@jernej__s The @frameworkcomputer Laptop 16 comes with a 180W USB-PD PSU. I have not yet seen anything doing more than that.
@wonka @frameworkcomputer @worik @alvaro I thought HP had some laptops with 240W power supplies, but it looks like I misread, and they're "only" 140W. The new G6 docks support up to 180W over USB-C (but the 280W version still uses a companion barrel plug).
Florian 
Seems they already reached the 240W now ⚡
https://www.theverge.com/news/765542/framework-240w-usb-c-pd-charger-first-framework-16
Bye-bye barrel jack: Framework brings 240W USB-C charging to laptops
Framework is becoming the first manufacturer to finally fulfill the promise of 240 watts through a USB plug. Today, it’s opening preorders for the first standards-compliant 240W USB-C PD power adapter ever sold by a computer company.
Jan Rychter@alvaro oh no... I can't believe even Segger messed this up. TBH, they should offer a replacement program.
Z̈oé
Gus
Luke Wren@alvaro I'd guess the 2 x 10k vs 1 x 5k is a BoM optimisation? But yeah it's funny to see this in the wild on a board presumably designed after the first production rev of Pi 4
Timon 🛠@wren6991 @alvaro yea smells like 2x10K parallel being on purpose to save the extra reel and they only ever intended to short the CC lines together.
Wild to still see this error being made 10y into USB-C adoption, really wonder where that comes from, there gotta be some wrong guidance around this somewhere.
Wild to still see this error being made 10y into USB-C adoption, really wonder where that comes from, there gotta be some wrong guidance around this somewhere.
@timonsku @alvaro Because an engineer sees this circuit in the spec and says "oh so I only need one of them". It's almost irresistible.
Also the same shortcut is correct in other areas, like in theory the sink detects orientation of CC1/CC2 to select the correct D-/D+ pair, but every USB 2.0 board I see just shorts the two pairs together. As far as I know this is valid.
@timonsku @alvaro The USB C spec should probably lead with "use exactly this circuit with no deviations if you just want 5V" because those are the people who will dedicate the least time to reading the spec. Committees are too excited about their favourite pet feature to think about how people use the docs they write.
That schematic does exist later on (pic), and I understand the point of view that this is a normative circuit and it's your fault if you implement anything different. On the other hand the way the shorted CC lines fail with an e-marked cable is only obvious with the benefit of hindsight, so I sympathise with people who make this mistake.
vriska, thief of light 
@leo @wren6991 @timonsku yeah, it’s very easy to make the assumption “oh, no cable has both CC lines, so this will always work!”
And in the past, most people would use the cheapest cables they could get away with. These days there’s a lot more eMarked cables.
I’m surprised that in all the testing it never came up, and if it did, that they just left it unfixed when the solution is so simple and inexpensive. Sure you have to spin the board, but cmon…
@leo @wren6991 @timonsku I don’t think “always just use the included cable and only that cable” is a good solution. If that was the case it should be built in.
I’m surprised they don’t have this documented on their website at the very least. There must be some amount of support calls this generates 🤷♂️
@wren6991 @alvaro I feel like the type of person too lazy to read a spec in detail (most people really) would just google how to implement USB-C and use the first vaguely reliable looking circuit.
You need to be a weird type of half assing person to want to not do the lazy google route, read the spec but then not fully read the spec to actually implement it correctly :D
You need to be a weird type of half assing person to want to not do the lazy google route, read the spec but then not fully read the spec to actually implement it correctly :D
twosky2000@wren6991 @timonsku @alvaro [help me understand]
So the problem is that the CC is measured/forwarded to the sink chip after pulldown resistor and not before?(I don't get the 2x 10k, but I think it boom optimisation is ok, but it has been done wrong.)
I guess the CC shorting is problematic, especially for the ones that just think USB 2.0 and ignore CC/keep a false minimum. So likely devices that were developed for USB 2.0 micro.
And here I'm unsure, the reference has passive sink (apple).
@twosky2000 @timonsku @alvaro The problem is the short across the CC lines at the sink causes the source to see the sink-side 1k resistor on the opposite CC line on an e-marked cable
@twosky2000 @timonsku @alvaro You can't see the problem in this diagram because this type of cable works fine. You need to look at the e-marked cable diagram in Alvaro's blog post
Pim de Groot@timonsku @wren6991 @alvaro as far as I understand the standard, for usb 2.0 it is mandatory to short the d+, and d- pins together. I couldn’t find a way to make this conform to the usb c standard: https://hackaday.com/2021/03/22/cursed-usb-c-when-plug-orientation-matters/
@mifune @wren6991 @alvaro my theory is that they are separate pins because they are easiest/best to short on the PCB. You need the pads in the connector for rotational symmetry and having them be individual pads happens by default due to how the construction works. Shorting them in the connector is rather complex manufacturing wise/an additional step and you can't do it in the cable without extra cables or creating an excessive stub via the plug+connector
Ignas Kielalike in theory the sink detects orientation of CC1/CC2 to select the correct D-/D+ pair, but every USB 2.0 board I see just shorts the two pairs togetherThat's incorrect, shorting is exactly to spec and there's supposed to be only a single D-/D+ pair on the cable, orientation detection is only for high speed signals.
Graham Sutherland / Polynomial@gsuberland @wren6991 @alvaro I wish the industry took more care into adopting cmsis-dap+pack. J-link still the least headache for many parts unfortunately. Maybe there needs to be an org vendors can throw money at to add support for their part like they do with Segger.
Quite a few shoddy half done cmsis packs/svd files out there
Quite a few shoddy half done cmsis packs/svd files out there
Jana
@gsuberland @wren6991 @alvaro In recent releases they started to detect and reject them unfortunately
@timonsku I'd suggest the Glasgow Interface Explorer designed by @whitequark .
janusfox 🍅@alvaro Wow, I've learned so much about usb-c, thanks!
Frost, wolf of winter 🐺🎄
@alvaro (not that it's really relevant, I just have no idea what a J-Link compact is and think it's pretty funny that it doesn't tell you)
UpLateGeek@alvaro I would void that thing’s warranty in no time flat. It’d be way quicker to just bodge the fix on the device than looking everywhere to find the magic cable that makes it work. (Also, F those nonstandard cables.)
Linus (@@alvaro
This reminds me of the first-generation Raspberry Pi 4B, which had the same problem. (Now that I think about it, that might prevent it from working with the 5's power brick.)
Also: You need to escape
This reminds me of the first-generation Raspberry Pi 4B, which had the same problem. (Now that I think about it, that might prevent it from working with the 5's power brick.)
Also: You need to escape
$ in CommonMark, because it encloses LaTeX Math formula code.
Vincent Sparks@alvaro any spec that is possible to misimplement, will be misimplemented
Piotr Esden-Tempski@alvaro Great article. (finally came around to read it ;) ) It took me a moment to parse what they did from the PCB photo. Maybe you can add a schematic of the JLink abomination to the article? ;) ... that said, it is a decent idea to optimize BOM lines using 10k resistors. If you already have 1% 10k resistors in the BOM and you have board space, this might be a decent optimization to do.
rhempel@alvaro I don't know for sure, but I doubt this was a cost saving measure - the more likely explanation is that the circuit worked with the cable they supply with the device, and there are so many corner cases and cable combinations and complexity to the USB standard that it's a simple design error.
I assume you have let them know it's an issue?
🇺🇦 haxadecimal 🚫👑@alvaro
Nice work identifying the problem, and great writeup!
It amazes me how many companies think they have invented a clever way to save the negligible cost of a 1/20W SMD resistor, despite the standard actually being quite clear that two are required.
And then even crazier that Segger used two resistors in parallel, to actually save not even the negligible cost of one resistor.
There's never time to do it right, but there's always time to do it over.
Nice work identifying the problem, and great writeup!
It amazes me how many companies think they have invented a clever way to save the negligible cost of a 1/20W SMD resistor, despite the standard actually being quite clear that two are required.
And then even crazier that Segger used two resistors in parallel, to actually save not even the negligible cost of one resistor.
There's never time to do it right, but there's always time to do it over.
I love this, so I@alvaro oh dear $DEITY not again. USB-C isn’t even that hard folks! And in a device that expensive I’d expect not just passive CC1/2 resistors but active current capability measuring…