Before we can buy modems from Librem, as versus companies such as 4G LTE (https://4gltemall.com), there needs to be some supply of B-slot M.2 card modem choices out there. While there are many SATA cards for data transmission, there are nearly none with voice compatibility that I can find on the market. Perhaps Quectel EM06, though I am not even sure of that!
I would hope for some independent development and some need for chip development and software writing for firmware to give voice compatibility is needed. A clear problem has been lag time necessary for development of ARM chip manufacture.
Some prospect of risc-v (“risc-five”) developed chips as an alternative to ARM allows for some potential for a rapid development free from licensing. Qualcomm as well as Western Digital is supporting this hardware alternative and modem chips have been developed such as Espressif (https://docs.espressif.com/projects/esp-idf/en/latest/get-started/). The chip noted here is easily small enough to use for an M.2 card modem with Linux compatibility present and Bluetooth 4.0 and modem functionality, but applications need to be made for both hardware utilization for this as well as firmware to allow it to interact properly with CPU designs for voice capability if not voice capacity.
Whether some present hardware already on M.2 card can be given modification, some chip already in Qualcomm or other company’s domain can be made to function with M.2 placement as Librem has done with the existing Gemalto chips, there need to be options. Perhaps some rapid development may come to the market with risc-v chips as I have noted. A problem here is that choice generally is based on demand and even Gemalto probably has no demand for M.2 card modems beyond Librem 5.
In essence, what I am arguing is that until the market becomes large enough to make it worthwhile, we may fantasize about alternative modems to purchase, whether from Librem or elsewhere, but we will hardly have any selection worth a tinker’s darn. I am not sure that much of a potential for upgrade can be had even if we are vocal with Librem of Qualcomm or other manufacturers.
With these efforts being supported by the open source community, there is one big advantage. Any one who can read a datasheet and who has a surfacemount soldering station and some basic PCB design skills can adapt a modem chip IC to an M.2 card. It’s a little more complicated than that, but not by much. Then all you need is to find someone out there who can write the drivers (the more difficult part, but more people have these skills). For someone like me, the hardware part would be an easy hobby project. The hardware part should be very easy for a lot of people. With everyone sharing their hardware designs and their code, the best modems will rise to the top as the most popular. If a PCB designer wants to, he or she can give their PCB fabrication company permission to sell unpopulated boards of their design to anyone who wants to place the order from the board manufacturer. Small custom designed unpopulated PCBs (no components soldered in yet) are very inexpensive (a few dollars per square inch with minimum quantities as low as ten). With enough people working on this issue, it won’t take long to have the best solution out there in everyone’s hands.
This area is not my forte, such IC hobbying not done by me at all, but this is quite nice to know.
Apparently, Purism uses Kicad. It’s a free program if you register to download it. There are some very expensive schematic/pcb programs. But you don’t need to pay $5K for such a program, anymore than you need to pay Microsoft for an Office suite program. You first design your schematic in a schematic program, or in a program that does both schematic and pcb, but in a schematic view where you see only schematic symbols. Sometimes, this only requires copying-in to the program, the manufacturer’s reference design. Then you open your design in a pcb view or using a compatible pcb program. From there, you see only a 2D grid and components that look like the real components that will be soldered to the board. You position all of your components on the board grid where you want them, and then route the traces (wires). As you hold your cursor over a ‘stretchey line’ and hold the left-click down and move the cursor, the rubber-band like lines between components flatten out and stick to the board grid. These are the traces (wires). After setting up Design rules, the program won’t let you accidentally cause a short circuit or leave a circuit open (unrouted). If you need a design change at this point, you should go back to the schematic view and make those changes there. After you finish your pcb layout, the program will generate ‘Gerber files’ that you email to a pcb fabrication company. They send you a price quotation. Then you pay them by credit card. Between one to four weeks later, your boards arrive in the mail. The longer you’re willing to wait, the less expensive the boards will be. Circuit design and PCB layout is a learned skill. For me, it’s easier to learn than C programming. Like programming, you never learn everything. You learn what you need to know as you go and you get better at it over time. Connecting a system-on-chip to another interface using manufacturers’ reference design documents is about as easy as it ever gets.