[v0.1] Form Factor
gregdavill opened this issue · 20 comments
The current design features an upstream Type-C plug and downstream Type-C socket.
This is designed to plug directly into a host, but the Type-C paddle mounted connectors mean we're forced to use a 0.8mm PCB, and the overall design feels a bit flimsy.
Potentially a 3d printed case might help?
Would it be possible to use two Type-C sockets, and then maybe house the device in a small extruded aluminum case as shown?
This is designed to plug directly into a host, but the Type-C paddle mounted connectors mean we're forced to use a 0.8mm PCB, and the overall design feels a bit flimsy.
Potentially a 3d printed case might help?
These are exactly my thoughts--I think I was going to file a bug but forgot.
Would it be possible to use two Type-C sockets, and then maybe house the device in a small extruded aluminum case as shown?
Unfortunately, I don't think two sockets will work. If you are using something that requires e-marked cables, I think (based on my understanding of the PD spec) that will just garble the communication since there's only SOP' and SOP'' to address the markers. Also, if you are using something like Thunderbolt, two cables plus the PD sniffer PCB might exhaust the entire link budget. I am very interested in using it for both of these applications.
The latter reason also excludes e.g. having two connectors with a mux picking between them, since you cannot pick a mux that handles every conceivable altmode. The board has to stay passive and functional with only a single cable involved, or it will lose a lot of its usefulness.
I think a sturdy case, 3d printed or not, is the way to go. I expect some boards will break during use, but... it's not a complex or expensive board (now that we aren't using Google's absurd design rules) so it's not a huge loss. In any case, plugging suspicious type-C stuff into a type-C debugger directly electrically interfacing the lines is inherently hazardous. We don't even have protection from having 20V on the CC lines; anyone using this sniffer ought to buy them in pairs!
If you wanted to move away from a PCB mount plug then the other option is to use a captive cable.
It appears that this might be the correct cable, but I cannot tell as their data sheets seem to not match well with their product's pages.
https://www.wit-area.com/20v-5A-100w-USB-C-cable-63.html
If these are not exactly what you need they might have a different version that I couldn't find exact details from their data sheets. It seems pretty hard to find these online and listed.
If you wanted two receptacles then you would want to get a cable like this one from Acroname.
https://acroname.com/store/c38-usbc-uoc
This is the problem that Acroname had to solve for their passive USB-C switch.
https://acroname.com/store/s85-rdvr-usbcsw?sku=S85-PASS-USBCSW
With all this I agree with @whitequark that the plug with an enclosure would work reasonably well. Especially with a connector that has some mechanical stability designed in. e.g. the JAE DX07P024MJ1R1500
this connector can get some metal attached to the shell and get firmer mounting to the PCB or enclosure.
Note: it is super easy to break un-reinforced type-c plugs off of boards. The pads alone don't have enough adhesion to retain the connectors.
If you wanted to move away from a PCB mount plug then the other option is to use a captive cable.
This seems like it would be very hostile to DIY. Is that microcoax on the cable you linked? I think that would be a problem for many people, myself included.
If you wanted two receptacles then you would want to get a cable like this one from Acroname.
This is more reasonable but I think it's still kinda hostile to DIY; not only you need to get a custom device (which we can maybe get to be easily orderable through e.g. JLCPCB) but now you also need a custom cable with very few suppliers which is also not easy to find a second source for.
(Thanks for bringing it to my attention though!)
Note: it is super easy to break un-reinforced type-c plugs off of boards. The pads alone don't have enough adhesion to retain the connectors.
That's how I felt about it, too. Guess I'm going to be even more careful now that you've confirmed it!
There are different options for board edge midplane USB-C plugs, The DX07P024MJ1 from JAE has extra mechanical strength by having the shell extend back to get better leverage and mechanical durability as the connectors pins are now not the mounting point.
Generated with a SnapEDA footprint modified to work with the USB-Plug footprint.
Steps to reproduce:
- Download DX07P024MJ1
- Add to library
- Edit all shield pads to signal S1
- Add step file to footprint with z offset of -0.05mm
- Replace existing plug footprint with JAE footprint.
- A touch of layout and outline work.
I'm on Nightly and don't know KiCad 5, so not sure about submitting a branch with the modifications.
If you wanted to move away from a PCB mount plug then the other option is to use a captive cable.
This seems like it would be very hostile to DIY. Is that microcoax on the cable you linked? I think that would be a problem for many people, myself included.
I would expect the cable to be shielded twisted pair (STP). I agree that terminating shielded twisted pair is not DIY friendly, it would require good magnification. Terminating micro-coax is at the high end of difficult DIY.
To be explicit about my goals, I want to have a device that would reduce USB-related frustration, so I'm strongly biased towards solutions that are, perhaps, suboptimal, but do not require specialized skills, or even skills most people could probably acquire quickly but are afraid of (like BGA soldering for example). I think there is a space for a more complex PD debugger that is also more capable, but I'd really like to fill a baseline need first.
... plugging suspicious type-C stuff into a type-C debugger directly electrically interfacing the lines is inherently hazardous. We don't even have protection from having 20V on the CC lines; anyone using this sniffer ought to buy them in pairs!
Is there an easy way to add 20V protection? I notice this is a common feature on some of the integrated PD controllers.
I guess at a minimum we should try to ensure a 20V fault can't propagate back through the USB 2.0 interface.
Is there an easy way to add 20V protection?
Maybe pass FETs? Paging @electroniceel
A very quick search shows some integrated solutions, that at-least cover CC/SBU pins.
NX20P0407, or TPD8S300
Oh neat, I knew about TPD4 but not TPD8
24 pin Type-C connectors sorted for my opinion of easiest soldering:
Plugs:
- Vertical PCB mount: DX07VN24WA2
- 👍 All 24 pins exposed, has through hole pins to assist in alignment.
- 👎 This causes the board to mount funny to ports.
- Mid plane board edge mount, with physical reinforcement: DX07P024MJ1
- 👍 has mechanical side supports to help the connector from breaking free of board.
- 👍 Side supports can be soldered in place to help align the fine pitch SMT pads.
- 👎 low connector break off resistance, should have a case or other support bracket.
- (Note: no vendor linkable photo.)
- Existing connector style
- 👍 most common style
- 👎 very fragile, depends on external mechanical hardware to prevent ripping off SMT pads.
Receptacles.
- Through hole style: usb4085
- 👍 Solderable with fine tip and small solder.
- 👍 in KiCad library
- SMT & through hole mid-plane with mechanical retention: CX70M-24P1
- 👍 very robust mounting
- 👍 can hand solder
- 👎 harder to reflow solder (solder in pin tuning)
- Current Dual row SMT pads:
- 👎 impossible to hand solder. (I feel that this alone should disqualify the part for DIY)
So thinking a bit more about this, it shouldn't matter about the SS and USB2 pins being unprotected. Since we're just passively connecting them through. It will be upto the downstream device to have adequate protection.
Unless we really want to add protection.
I think we'll stick with a style of mid-plane connector, whether that has the extra tabs or not. And also develop an enclosure for the device, for adding rigidity. It could be cool to do a milled case, but we'll also need to have a design for FDM printing for DIY cases.
Is it possible to get 100% THT full featured 24pin USB-Receptacles? The one you linked is only USB-2.0 pins.
The dual row SMT means you are required to use solder-paste/reflow. IMO, they're the best option automated assembly, as it removes the need for THT soldering.
I'll look into SMT+THT as an option.
Whoops, I though it was 24 contacts to the board. Appears that when I searched and it showed up as a 24 pin part, I misinterpreted as 24 contacts wired in.
The enclosure would make the mounting strength issues less important. Also I bet liberal amounts of glue would work as well for those without printers and the like.
Is there an easy way to add 20V protection?
Maybe pass FETs? Paging @electroniceel
For the CC lines you could do it like I have planned for my Glasgow protection addon: https://github.com/electroniceel/io-protection-notes
Unfortunately I haven't continued to work on this since I started doing other Glasgow-related work. So this is more in the state of a design concept and not a fully tried and tested schematics you can just copy and it works.
If something more fancy is done for the USB 2.0 data lines requiring protection. Then 20V protection could be integrated with a TS5USBC41 Unfortunately it is a DSBGA package. :-/
The NXP part above also has a sibling part for USB 2.0 protection NX20P0408. Unfortunately it is also WLCSP12.
Protecting D+/D- from >20 V is the really tricky part and these seem to do it. Nice. I didn't know there was an IC that could do it.
Since these are really small, we could consider adding them to the regular Glasgow ports in a future rev. They don't protect from negative voltages and don't cover Vio, but they'd still make regular Glasgow much more resilient.
Only real downside are the packages.
I will order some samples and play with them a bit.
I bought some Type-C plug breakout boards with the JAE DX07P024MJ1, they are much more stable than other SMT midmount options.
I would use this in my lab without a case.
Some thoughts:
- Twinkie is already 20V tolerant on CC as long as you are only sniffing (Q1 provides that protection). If you start transmitting into 20V, you'll probably blow the low side of Q2/Q3.
- The Google design is stupid expensive because we only had access to really weak prototype plug connectors and wanted to minimize the torque on that connector, so we basically crammed everything as tight as we could (Twinkie was made during early development of the USB-C spec). This is actually well beyond the tech most of our Chromebooks are built to.
- Keeping one side of the Type-C plug narrow is useful to ensure you can fit two next to each other on a hub or other minimally-spaced multi-port machine. This makes the panel mount connector less attractive, as you've noted.
- The mounting side lobes on some plug connectors are good for pull resistance, but still not great for the up-and-down torque you get from having essentially a 2-inch long connector. The only way to make that better other than doing radical things is to find a connector that works with a thicker board...
- Twinkie had a molded plastic case in two halves, glued together -- this did not provide adequate strain relief on the plug. 3D printed might work if you can keep the plug end as a single piece, with tight tolerances on the plug hole. Maybe a bucket-style case you can slide the freshness into before sealing it off at the end...just like a real Twinkie :)
- If you want to go really robust without needing a case, Antenk has some vertical-mount SMT connectors, both plug and receptacle, that are low-profile. I've played with them a little for low-speed stuff; they seem alright and let you make a much thicker board. I haven't tested them for SI. You can put the connectors just above each other vertically, on opposite sides of the board, and have very short traces, with the sniffing sections flagging up vertically. It'll be a bizarre form-factor but might work well. Main downside is that those connectors are pretty much impossible to hand-solder. U3.1MS04-S-L-P-G0-D for the plug and U3.1FM02-S-L-P-G0-D for the receptacle.
Thanks for these insights @dnschneid!
FWIW, I solved this in an unconventional way in the Twonkie: I'm using vertical-mount connectors and using them as straddle mounts (or rather, touch-the-board-edge mounts). This requires some minor physical modification to the connectors (removing some mounting features that are in the way and slightly bending the pins for better contact) but it worked like a charm on two specimens so far, much easier to get right than I expected, and since I solder the shell to the PCB surrounding the connector it's also rock solid mechanically. Also I can use an 1.6mm PCB which is good for stability (and availability).
I was a little afraid that the little stubs created by the pins would cause problems on the high speed links but both boards I built so far support TB3 20Gb/s links just fine.
Of course this is not suitable for automated assembly but it works pretty well for DIY.