Given the " hacker " or technical nature of these laptops. I could see a large audience finding an external nvme connector useful.
Given the slim nature of the hardware, i think it would also just about fit.
Sure it wont be fast, but under many use cases having additional storage that isnt proprietary modules, limited capacity or most importantly,
seperate hardware you carry around, it could be quite handy.
May be possible, would be usb2, if you make some I’ll buy one or two, just to have. The nvme casings come with usb connections, maybe small enough to fit into a spacer. The matrix module is a great starter.
I also know a good way to utilize it. Backup in the background, like while charging, the system would slowly copy changes from the main ssd.
Are you looking at creating one yourself or suggesting it for others?
There is 3.7mm of height is available for the input modules. And I see M.2 connectors available that are 2.75mm high, such as MDT275M02001, so potentially doable.
That’s really interesting. A M.2 2280 would fit tidily in the width of a keyboard spacer, and seems to indicate that the thickness ought to work out as well. You only have USB 2.0 available, though, so you’d be looking at an A+E keyed module that didn’t provide any PCIe at all. I wonder what compatibility would be like for those, or if you’d need a chipset to provide PCIe?
Well there are chipsets available for USB 3.0, the question Is have is whether they’d downgrade properly if you plugged them into a USB2.0-only port. Probably?
I think a great way to get started would be to get something like this:
M.2 to USB Adapter, RIITOP NVMe to USB 3.1 Reader Card Compatible with Both NVMe (PCI-e) M Key SSD & (B+M Key SATA Based) NGFF SSD https://a.co/d/01wjwb7S
which uses a RTL9210 chipset (unfortunately I can’t easily find a datasheet, although JLCPCB has them listed). The desolder the USB3 connector and fit it with a adapter for the framework pogo connector (I ought to spin that up as a separate little dev board) and see how things go. Then 3d print a bit of a frame to tidy up the bodge.
EDIT: I took apart the M.2 adapter I had, and it turns out it has that same RTL9210B chipset. The size looks pretty reasonable:
The main hiccup is the chunky SOIC-8 all the way at the end, which is too far out to fit into the little recess provided for thick components in the input module area, like you can see here:
(See also https://www.seeedstudio.com/blog/2020/03/30/two-pcie-to-usb-bridge-chip-solution-to-use-nvme-ssd-on-your-raspberry-pi-4/ and A USB 3 to M.2 Reference Design Schematic - Electrical Engineering Stack Exchange which lists some common chipsets. It’s surprisingly hard to find reference designs on the english-speaking internet, though.)
Pretty sure they will all work fine with USB 2.0 only ports. The external nvme ssd enclosures I have all do.
I have a rtl9210 enclosure, haven’t noticed any issue with USB 2.0 only ports.
All of the ones I have including a chonky usb4 one downgrade to usb2 just fine, apart from the usb2 speeds.
I think one of the more serious problems with this is going to be heat. NVMEs run hot, and sometimes REALLY hot. Putting those all within reach of the hand, wrist, and arm, might not be the wisest of choices.
When bottlenecked by a usb2 connection that should not be a huge issue. many usb enclosures are quite good at turning down the power anyway. Other option is you’ll learn pretty quickly where not to touch if it actually is a problem.
It potentially damaging the screen may be a bigger issue if you actually find an ssd that goes all toasty on usb2.
Most LCDs have very high temperature tolerances, but i’d probably mount the ssd reversed or with an airgap to reduce heat transfer into the display.
Additionally, LCDs arent that sensitivie to heat ( 50c) and without direct contact, its unlikely to hit that.
Probably not a big issue, hell on the 13 the display gets blasted with the cooler exhaust and I haven’t seen any reported issues caused by that.
Not when running at usb2.
That’s actually a really good idea. The midplate in the FW16 is tied into the heat sinks anyway, so if you can transfer the heat down into the midplate it has the most chance of being efficiently conducted away.
That also eliminates one of the other issues I was having, which is that most of the converter PCBs I was looking at tend to have the components mounted on the “underside” which is where we don’t have a lot of thickness to play with. If you turn the whole thing upside down maybe things get simpler.
