Mh, just had a quick look at the schematics but noticed something substantial: You’re shorted the USB datalines on the ESD protection IC (U102, U301, U401). Pin 1 and 6 as well as pin 3 and 4 are internally connected (as basically also shown in the schematic symbol).
But I like the structured design of the schematic.
For the PCB:
I like the decoupling for the hub-IC .
0.2mm vias are usually more expensive and don’t seem really necessary here.
I’d remove the copper pour on the top layer, you have an internal ground plane that you can easily access and it currently disturbs the controlled impedance of the USB traces.
You can get away with way less power vias than you used, one of those can carry around 1.5A, the traces that connect to them can withstand far less.
With the saved space you can try to bring all components to one side (for cheaper assembling costs).
I probably won’t use 0201 passives, depending on if you want to have it assembled the usual cheap Chinese manufacturers charge extra for that size and if you want to assemble it by yourself I recommend checking if you really want to deal with such small things. Also I’m not sure if e.g. 10uF caps are available in 0201 packages.
I have a prototype version of my design on the way, I’m curious if it actually works as intended. I have to assemble it by myself tho because assembly for one might cost around an arm and a leg…
I have an idea, but I am not sure of its viability.
Here is my proposal:
Would it be possible to have 2 USB-C ports to connect the modules to the system itself? Such that the electrical constraints that exist would be avoided as a dual USB-C expansion card just becomes a pass through solution. Would it also be possible to use the two USB-C inputs to increase total data throughput? For example, if there is a new DP standard released that currently exceeds the speed of a single USB 4 port, then two connectors could combine their total bandwidth to provide sufficient throughput in a single expansion card.
I believe there might be a hardware constraint at play here, each Thunderbolt controller can only handle 2 USB-C Ports iirc.
The chassis would also have to be redesigned as well as most expansion cards, so probably not very possible in the foreseeable future.
Combining 2 ports’ bandwith into a single data-stream isn’t really supported by current technologies as well. For instance, you don’t see monitors that accept two seperate HDMI inputs.
For PCBA assembly, we’ve had some quick prototypes done at places like PCBWay and found that the quality and speed are acceptable and costs are shockingly low.
If you are trying to decide between PCBWay and JLCPCB, I’d recommend PCBWay, I’ve found PCBway to be generally a very similar price but sales reps seem to be nicer and almost always reply within an hour during their business hours. In case you haven’t seen it previously you can request reimbursement through this link https://forms.gle/uaP9kWxwcK5VZNuA8
But for that, cannot use 6 layers… 6 layers is just exponentially more expensive, unfortunately. However, when doing in large quantities the price per item with 6 layers is actually really low. 500 pc for ~281$ (.56$ each).
But thanks for the tip. If there’s a way these people can fit it into 4 layers, it will be very cost effective. Much more than I anticipated. Maybe the concept of board and daughter-board isn’t that much of a far fetch.
Doesn’t matter. The prototypes are the really expensive part. After the prototypes, you can make extra PCBA for very cheap each. I’m sure it will sell 200 easily (even if need to wait some months). For that, the prices are very reasonable, even with 6 layers.
So framework could probably contribute in a prototype production run?
I did find an old blog post about that but it seemed it is long gone.
I have a proof of concept PCB design on the way, but I’ll hand assemble it (which hopefully works, some parts are smaller than expected , and the connectors could be weird to assemble) since I couldn’t spend that much on assembly. But since that’s the first high speed design I did I’m not sure how well it works at all. Also there’s probably quite some room for improvement on the BOM cost and (especially) the DFM side.
I’ll report back here .
Why not start a geographical pool to see how many people are interested in such dual expansion ? This way we could estimate how the bulk price + shipment would be and maybe run a batch of it ?
That’s the thing to discover. Maybe a topic can be made sticky with a clear title and the poll in it to let everyone interested vote for their needs. Look at the recent topic the CEO has done to ask for inputs for the 16" laptop…
For what it’s worth, I would be in favor of a dual USB-C module. I would greatly benefit from that sort of setup and would be glad to help crowd fund or whatever it would take to make it happen. It would free up one of my slots to have an SD Card reader as well. Right now I keep two USB-C modules in place, USB-A and swap out the HDMI and SD Cards back and forth since those are use less often.
Truly, the Framework is such a cool product and the fact that this discussion is even taking place, along with so many others is absolutely thrilling!
Seems like dual usb-c would be more helpful for the 13-inch framework model because of only having 4 usb-c ports. However, with the 16-inch having 6 usb-c ports, I wonder how many people will need more than that
Sounds like you got the displeasure of using a displaylink based one or something.
Actual Displayport+usb or thunderbolt based ones are neither laggy nor particularly more resource intensive than plugging it “directly” into the ports on the laptop.
.
, and the connectors could be weird to assemble) since I couldn’t spend that much on assembly. But since that’s the first high speed design I did I’m not sure how well it works at all. Also there’s probably quite some room for improvement on the BOM cost and (especially) the DFM side.