from Hacker News

The sad state of voice support in cellular modems

by bahjoite on 9/27/17, 9:00 AM with 20 comments

• by mixedbit on 9/28/17, 2:15 PM

  I have recently bought Huawei USB modem and discovered that these things run Linux with Android. This is pretty off-putting. With such a modem it is no longer enough to keep your system & drivers up to date, because the device has it's own full-blown operating system. An attacker no longer needs to hack your machine to capture your traffic or make you a part of a botnet, it can hack your modem. And the modem, like Android phones, is probably not going to receive software updates for long.

• by derefr on 9/28/17, 10:53 PM

  Do cellular modems really "need" to do audio any more? I don't mean because everyone could just be using data-only SIMs with OTT SIP software. I mean, instead: sure, audio is part of the cellular standards—but why is it part of the cellular baseband hardware? Phones aren't made in a way where the baseband chip "is" the phone any more. The modem is now a tiny part of the processing power of even the dumbest of dumbphones.

  So, given this, why can't the standards require that the modem expect there to be some other chip (e.g. the SoC in a smartphone; a specialized DSP on a dumbphone; a ground resistor on a mobile data stick) to feed the modem audio packets which have been pre-encoded and serialized to the format it would have done itself, so it can just take them, wrap them in baseband frames, and directly put them on the wire?

  In other words, why is the (DAC + GSM/G.722/etc. DSP + ser/des) logic a feature of cellular modems, rather than a feature of phones?

  Think about it by analogy: you wouldn't expect a PSTN modem to come with a feature where a computer can make analogue telephone calls using it by hooking it up directly to your microphone and speakers. You'd expect that, if it did any such thing, it'd be by DMAing pre-encoded-for-the-wire packets out of a realtime ring buffer, the same way any network card consumes any other packet. Maybe your sound-card would provide hardware support for some codecs (by consuming one DMA ring and producing another. Or maybe you'd do the audio encoding on your GPU with a GPGPU shader, allowing your phone to speak new voice codecs with just a firmware update.

• by mseebach on 9/28/17, 1:36 PM

  I suspect that the reason for this is the behaviour of most OS's around USB audio devices: switch over the default inputs and outputs to the new device -- which would mean that when you plug in your modem, suddenly your Skype would configure itself to "use" that device for it's audio IO.

  Also, I'm not sure how easy it would be to even configure the OS to route modem output to default sound card input (your headphones), and the same soundcard's output (your microphone) to the modem input. The assumption of most OS's seems to be that you have one active input and one active output at any given time. It's certainly not something I'd immediately know how to wire up using the default mixer in Windows.

• by mmjaa on 10/1/17, 6:57 PM

  The 'reason' for this distinction between breaking out the audio of the modem to a standard interface, versus not breaking it out - but rather requiring a secondary processor to submit proprietary packets - is simply this: cell equipment manufacturers have to comply with the requirement that it be relatively difficult to record calls made with their equipment.

  Putting the DAC elsewhere absolves them of this requirement and drastically simplifies their time-to-market statistics. If you make the OS and other layers of the stack responsible for the ADAC, then you don't need to worry about it - as long as there is a usable interface.

  Around the world, different markets have regulations governing the recording of calls. Its never supposed to be easy to do this, at a hardware level, because its supposed to be very hard even for the players in various markets to do this - without a third party (i.e. government) involved.

  tl;dr the reason this is so, is because hardware vendors have to comply with legislation, and doing squirrely ADAC instead of out of the box Audio routing absolves them of the responsibility.

• by TorKlingberg on 9/28/17, 2:15 PM

  One correction: Modems in phones are not normally LGA modules, but rather components soldered directly to the main board. Sometimes the baseband processor and protocol stack are even integrated in the main processor SoC.

  Systems that use the LGA or m.2 modules often don't use the audio at all, so nobody has cared to make a nice interface.

• by Nexxxeh on 9/28/17, 12:42 PM

  How annoying. Especially as several of the Huawei USB dongles support voice over USB.

  https://stackoverflow.com/questions/8367864/how-make-use-of-...

• by cesarb on 9/28/17, 12:56 PM

  Would VoLTE make the situation better, since voice is sent as data?