3G/4G modems - Alternatives?


Not saying that they will put anything in there.

If we need a small example perhaps relaying to a server the IEMI number with time date and connected cellular tower identification.


I think the emphasis was on “M.2”. That link is for the module with LGA connectivity.


The information is good but doesn’t tell us what is active on the m.2.


Good one, but if you really care, you neither trust the module nor the tower. IOW, assume that both spy on you. So, even if you had a blob-free modem, you would turn it off if you don’t want your location to be known. And that’s why I ask: Is it really an additional concern?
I still can’t think of something the modem (in the Librem 5) can do/know that the tower can’t.


I know that and I am not getting you wrong. It is just something that I wanted to make straight and (even it does not belong here) it has had just something to do with understanding of the broader picture/complexity of this campaign.


Anything sent by the modem to a cell tower can be collected by the cell tower and related elsewhere. The same is true of the carriers other hardware and software in the system. The question would be who wants the information and which path is best to obtain it. Best could mean most discreet or least difficult.

Eg. A carrier would be less likely to notice if the routing was done from the phone than their own cell tower hardware.

Eg. 2 Australian metadata laws and anti encryption laws could have them request the Thales Group to send the previously mentioned data to a server for storage. As knowing where the device is wouldn’t be determined until location information is available it would be likely all modems would have the firmware updates and information deleted or not sent if not located in Australia. I use the Thales Group as a strong example as they have strong ties to the Australian defence forces.


There is no need to work on the antenna itself.
All you need is your own antenna to monitor a given area or a virtual mobile operator to monitor the whole world.

Injecting infected firmware into Gemalto’s production line to monitor a few thousand Purism customers is absurd when you can collect data on tens of millions of users without too much effort.


The Purism customers could be unintended targets. The modems will not be exclusive to the Librem 5 and Purism. Other companies will also use it. Possibly some less then friendly people towards your chosen ideals/countries ideals could use it triggering the request to monitor the information instead.

This doesn’t mean it will happen.

And we’ve all heard “If have nothing to hide…”


Not for smartphone, in “other companies’ smartphone” modem is integrated in the main SoC.


@36w4r6 Personally I think use of the cellular network is very problematic to begin with. Having the cellular modem on is perfect for tracking/spying on people. And with telecoms selling our location data, it is a huge concern (to me at least)

As a dad I have to be reachable in case something happens at home. But I’m hoping to have the baseband kill switch engaged as much as possible. Hopefully that will minimize anything problematic in the modem firmware.


For modems that aren’t available in M.2 form factor, presumably it might be possible to design an M.2 card and solder the module onto it, breaking out the appropriate electrical interfaces onto the card-edge connector. I cannot design one, but it wouldn’t surprise me if that’s effectively what the manufacturers are doing when they offer an M.2 version.


The Librem 5 could be the start of a new trend of smartphones with easily replaceable and upgradable components. I doubt that Samsung or Apple would follow this trend. It could remain niche. But we can not predict where the Librem 5 will lead. The puzzle phone concept could make a return.


They are. We can not mitigate every concern and risk without controlling every part of the system and that just isn’t realistic at the moment. All we can do is reduced the risks we can and trust in the components of the system we can’t.


That is possible and exactly how we get the modules we have.

The Quectel EM06, EG06 and EP06 all use the same modem just in different form factors.

I did come past a company that has modems that where designed to be swapped in a slot like changing your CPU but they didn’t have an M.2 card for them.

Biggest problem would become costs. We would need someone to design the card. Maybe have it modular to accept modems from different companies. Then we need to make them. Production scale would reduce costs. Those of us that can make our own cards would be less affected.


I would love to see more cell phones with a replaceable cellular modem, but it is very unlikely to become a trend for the following reasons:

  1. An SoC with an integrated modem costs less than two chips and takes up less space, which is critical on phones. Qualcomm and Mediatek offer their integrated modems for so cheap that only companies with special reasons like Apple, Pine64 and Purism are not going to use them. Adding an M.2 or PCIe slot, plus a separate card makes the modem very expensive. The cheapest cellular modems on an M.2 card cost $30, so I would guesstimate that Purism can get them in large quantities for around $17 a piece, plus $3 for the M.2 connector, so around $20, plus $20 for the i.MX 8M Quad SoC, so $40 in total. In comparison, the Snapdragon 427 with an integrated modem probably costs around $10 in large quantities.
  2. Because the cellular modem can get hot if used for very long, you either need to cool it like the SoC (so you might as well include it in the SoC since it makes designing the cooling system easier) or you need to use a huge chip package like the Gemalto SPL8 which is 29x32x2mm in size to distribute the heat. It is very hard to provide cooling for a chip on an M.2 card, so you are stuck with giant chip packages on 3042 M.2 cards. Nobody makes cellular modems in smaller 1630 or 2230 M.2 cards because of the need for cooling. No phone maker (except for an oddball company like Purism) is willing to dedicate 30 x 42 x 2.3 mm of space to the cellular modem because it kills the aesthetics.
  3. Almost all the industry works on planned obsolescence to increase sales which is why batteries can’t be replaced and the headphone jack is being removed and companies like Apple make it hard to get original replacement parts, so they aren’t going to be at all enthusiastic about replaceable cellular modems and the cellular providers don’t want to promote phones which allow people to easily switch their networks.

If other companies decide to copy the Librem 5, they are very unlikely to copy its M.2 cellular modem. Sadly, the Librem 5 is probably going to be one of a kind, and it only happened because every cellular baseband in existence requires binary blobs in the Linux kernel, so the only way to have a 100% free software phone was to control the modem via a USB interface over M.2.

Sadly, there is no way to make an economical 100% free software phone.


If that is the case Purism might as well give up now and walk away. The impossible becomes possible all the time.

I am willing to pay a little bit more or sacrifice something unnecessary for something I consider important. Except for one phone all of mine have had user replaceable batteries. I’ve just never had the highest specifications in them except for 2.

A trend doesn’t have to be the most popular thing in the world. It just has to be a recurring event. If the niche phone makers take it up for most or all of their phones that would make it a trend in my opinion.[quote=“amosbatto, post:96, topic:6067”]
Adding an M.2 or PCIe slot, plus a separate card makes the modem very expensive.

True that it adds cost. It also adds complexity to the design. The question is are we the consumers willing to accept the increase in exchange for the feature. M.2 and mPCIe are not the only options.


I’m certainly willing to pay $100 more for a phone that has a replaceable cellular modem, because I care about the environmental impact of electronics and planned obsolescence, but it is important to understand why no company did it before Purism. Todd Weaver is truly unique in the hardware industry in what he is willing to do to run a Linux kernel without binary blobs. Purism is going to extraordinary measures by running the Wi-Fi/Bluetooth over SDIO and the cellular modem over USB+M.2, and putting the binary blob to initialize the LPDDR4 RAM in a separate Flash memory chip and running U-Boot on a separate Cortex-M4 core in order to execute that blob.

Frankly, I stand in awe of the company, but I can’t see any of the other Linux companies doing the same, because you need a company that is really committed to the ideals of Free Software Foundation to do it. If the Librem 5 becomes a huge commercial success, we might see copycats, which would be fantastic, but the economics are not in our favor.

Sorry, but they are the only options. As I explained in my previous post, nobody makes a cellular baseband chip package that is smaller than 29x32x2 mm because you need a giant heat spreader to redistribute the heat, so you are either stuck with a 30x42 mm M.2 card or a 30×50.95 mm mPCIe card. You can’t use the smaller sizes because the chip package won’t fit on them. If you have better cooling, then you can use a smaller chip package, but how are you going to implement better cooling on a plugin card? Even if you create a custom chip socket to put on the motherboard, how are you going to attach the cooler? Remember that Purism is a tiny company, so it can’t demand custom form factors from suppliers.

Purism is in a bind here. I assume that Purism didn’t select the Quectel EG25-G for the Librem 5 which supports 30 bands, because it required some binary blob, whereas the PinePhone is using it because Pine64 doesn’t care about free software. However, the reason might be because this chip is only available on mPCIe cards, and Purism couldn’t find another 8mm of space.

Librem 5 concern

Can you post a link or the name of the company? I would love to see how they cooled it.


This is probably what Purism people are curently doing/proofing if already done. Furthermore, Gemalto PLS-8 LGA chip is 2,9 mm thick. By adding M.2 connector PCB plate there is to expect to be more in thickness. And as stated, Gemalto’s proprietary LGA footprint (156 pad LGA mount) leads me to conclusion that it will be swappable on M.2 plate or hopefully offered as an upgradable M.2 modem card version (for replacement). It is already mentioned that: “unique LGA form factor compatible with past and next generation wireless modules ensures easy migration between wireless standards from a single design as technology needs evolve.”

Another eventual (expensive) and proprietary (again) LGA form factor option in 2,9 mm thickness may be Telit LE922A6-E2 LTE CAT-6 + required (developed) M.2 connector board. Additionally, to ensure good long term heat dissipation for such M.2 PCB plate requirement like the electroless nicker/immersion gold (ENIG) may be relatively expensive as well. And for example, finalized product versions of the Quectel EM06-E and EM06-A LTE-A Cat 6 M.2 modules are 2,3mm thick in total. I have great respect for Purism people effort.


Underrated tweet :wink:

The problem is, I think very few people know this, understand it, or even care.
Many don’t understand why the pine phone is so much cheaper. (Although I could imagine the price going up a bit).

Purism should make sure the reviewers of the press know of all those design decisions.