(At my house, my only options are VoIP and cellular. POTS service has been shut down.)
I wonder how well modems work over VoIP though? My understanding is that the voice-oriented compression causes problems (although you can turn that off by using G.711), but the bigger issue is the unreliable timing of most IP connections upsets the modem and it keeps on disconnecting.
ITU V.150.1 (aka V.MOIP) is supposed to fix that but I don't know think many VoIP providers implement it.
Didn't V.90 (56k) require a special setup with ISPs where you had fewer that one analog-to-digital hop? And IIRC, you could go from a 56k modem to another 56k modem. ISPs had different hardware.
Nope, the theoretical max between two 56k modems via a normal CO connection is 38k, in reality it usually meant 33.6k at best.
As you sorta alluded to, a 56k modem increases that due to the removal of the ADC on the ISP side, with the ISP having a digital connection to the CO.
As to the asymmetrical part of your comment, definitely true in my experience there. I could go into detail on the technical reasons why that was the case, but don’t see the point unless someone wants me too.
I visit a cabin that has a POTS line that is active but rarely used, and would love basic connectivity.
LTE is not available in that area.
I wonder if telephone providers have some sort of recognition to detect people using (abusing?) a free call plan for transfering data. Certainly, if your "call" takes several days, that may violate some TOS.
(The use case I have in mind is to save money by using cheap prepaid SIM cards without an internet plan. It is purely for interest/proof of concept, not for scamming...)
Uhm, you are off on your numbers. GSM channel size is 200khz, but TDMA is used on top of that only leaving ~20khz for voice traffic (remember analog modems are “voice” traffic). So naively one might assume since 14,400 is under 20,000, it would be theoretically possible to achieve, but that’s not true in reality due to losses in the digital to analog to digital conversion due to noise. In the case of POTS, which has a max channel size of 64k, the theoretical maximum is roughly 38k (the reason 56k modems worked is due to the removal of the ADC conversion on the ISP side). Using a similar conversion for GSM, that gives you a theoretical maximum of 12k. If nothing else was a factor, then 9600 would be doable, but due to other issues with GSM, the real world experience usually was closer to 2400, sometimes getting 9.6k when everything was working great.
I owned a dialup ISP in the 90s and routinely logged in from remote to manage user accounts from my 386 Thinkpad connected to a GSM phone. After selling the ISP, I worked for multiple years in the telco field, on the RBOC side (Southwestern Bell and later SBC), the cell carrier side (Voicestream and later T-mobile), and LDDS side (MCI and later MCI Worldcom).
Forgot to mention, GPRS uses the same time slots as voice calls do over GSM. Each time slot is still limited to about 9.6k useable data rate as with using a modem, but GPRS allows bonding to use more than one time slot. As such, using three time slots would get you 28.8k, using six gets you 57.6k, etc.
I suspect it won't be a very user-friendly experience, but as a POC? Sure!
1) per minute charge on the phone (typically 1-10p)
2) per monthly charge to phone company for line rental
3) per monthly charge to isp for access
4) per minute charge to isp while we were dialed in
There were some Linux projects to do this in a generic way - basically generate the required tones over the audio output and interpret tones over an audio input - and implement the various ITU V-series standards (though some of them were patented and could not be implemented, like V.92 for 56k).
There's another one other than iaxmodem (https://stackoverflow.com/questions/3069945/open-source-soft...) but I can't remember what it is.
This lists a real serial modem as a prerequisite, and those are kinda getting to be collectors' items. Is it feasible to build a modem out of a Raspberry Pi Zero or an Arduino? You'd need to hook the phone line to the GPIO pins, do the ADC and DAC stuff in software, implement the V.22 protocol in software, and also implement the Hayes Smartmodem stuff in software, meaning this is one of the few times a Hacker News reader would have occasion to implement the Hayes Code. Is there any show-stopping problem I'm not seeing? Massive electrical incompatibility?
I have my trusty old USR Courier and USR Sportster 14400 in a box. But I'm a bit more tempted to try and get pcboard running again using some of this setup. I don't have a landline in my house, only fiber so the asterisk option would be the way to go.
I also have disk images of my ms dos 6.22 floppies and PCBoard 15.22. I also have desqview for multinodes but I remember it was a PITA to get it running.
Now I just need an old 486 that is still running, my oldest running machine has dual pentium pro cpus and it doesn't behave all that nice with ms dos.