Hacker News Reader

Comment by gumby | Hacker News Reader

The fax requires a 56KHz channel which has not been used for mobile telephony since the 1980s (when AMPS and GSM were developed. Before then it was just a radio link through an operator). That’s why you can’t use a high speed modem either.

Some gsm phones had the ability to transmit a fax but they actually used a different protocol to talk to a fax gateway. I never saw one of these in the wild.

I haven’t been in that “biz” for a couple of decades but nowadays a different approach will be used anyway (pdf via IP to a fax gateway -> POTS)

View on HN

> The fax requires a 56KHz channel

This is quite wrong. I believe you are confusing Hz with bitrate, as the most advanced modems communicate at 56kbit/s using PCM. To clear it up, a regular phone line has a 4000Hz bandwidth. With digital lines, you require sampling at at least 8000Hz (Nyquist theorem). It's impossible to send a frequency above 4000Hz using this line. Each sample is considered to have 7 usable bits of data, giving it a capacity of 8000 * 7 = 56 kbit/s.

Most faxes are compatible with the T.30 standard, which doesn't acknowledge the usage of digital lines, and sends data using the older V.34 modem standard. This is either at 28kbit/s or 34kbit/s. This uses a method called Quadrature Amplitude Modulation (QAM), and a maximum baud rate of 3429 symbols per second. Since the centre frequency is set at 1959Hz, you won't have frequencies greater than the centre + bandwidth/2 = 1959 + 3429/2 = 3674Hz. You get 10 bits with each symbol, resulting in 3429 * 10 = 34 kbit/s.

Edit: Here's someone attempting to use the slowest supported T.30 standard, 300 bit/s, over his cell phone, and you can see it's not perfect: https://www.youtube.com/watch?v=uQqWHLZjOjA

I figured that audio compression on mobile networks prevented the sending of faxes as tones.

You're right - though it's more about the precise timing of the tones than the frequencies themselves.

The T.38 protocol exists to allow faxes to send over a packetized network. Already if you're sending a fax there's a high chance it's been converted to T.38 at some point, as SPs gradually move toward a fully packetized core.

Wish this wasn't the highest-rated comment, as it's not correct.

If it needed a "56kHz channel", it would not have worked in the phone-over-copper-wire days either.

If we assume the parent meant 56kbps, that's still incorrect. The fastest fax machines use V.34bis modems (ignoring ISDN), which give you 33.6kbps, though many (most?) still support the oldest standard, V.27, at 2400bps!

The real issue sending a (T.30) fax over a cell phone line would be jitter and packet loss, since most cell networks are IP-based nowadays. T.30 is not very resilient to noise or timing issues. There's the T.38 standard, which was designed for sending faxes over a packet-switched network, but it's still sensitive to timing issues (esp. since often you'd be talking to a real fax machine on the other end, after going through a T.38/T.30 gateway).

Cell phones certainly don't give you the bandwidth that a copper landline does, but it'd certainly be enough to transmit a fax... if you could keep the error rate low enough... which I doubt.

(Source: I built the initial production version of Twilio's fax product.)

You are confusing bitrates here, the fax emits an analog signal and doesn't interface the digital line directly- it's not a digital packet switch protocol.

Those 2400bps are not using 2400bps from the line, but are modulated to an analog signal that uses the lines full bitrate. From Nyquist–Shannon sampling theorem we know that a 56bps line can transfer something like half of it in analog form, so 33.6 is marginal.

T.38 is sensitive since it needs to end the analog side, practically receiving the fax.

Fax in general is sensitive since the standards have left too much freedom for implementers, especially like you said in the timing parts, causing incompatibilities between machines.

(Source: I have built fax relay over VoIP products in the late 90's)

CDMA also had a fax technology, under I think IS-95 data, but I didn't work on it myself. I've been out of the industry a couple years myself, but there were still a handful of customers using it that kept delaying the shutdown of the old tech, and there wasn't really an upgrade path.

IIRC from the water cooler talk, one of the remaining customers was a lighthouse, that had to send a daily report, things like that.

There were cellphones capable of sending/receiving a fax, in 2000 IIRC. Apparently it was used by about ~0 people worldwide, so the capability was discontinued.

Perhaps someone could start a relay service.

We made http://mailprincess.com/ a while ago which is $1 per fax

Also made http://maleprincess.com/ for fun

You should really get an SSL Cert. They are free these days

It exists, but will only deal with USA numbers = I live in Canada, maybe one day ....?

Like https://www.hellofax.com/ ?

These exist! They're not cheap and probably ripe for disruption.

The Twilio API supports fax, so I used it to make this: https://5dollarfax.com/

Cheapest service I know of if you infrequently need to send a fax and wait a few days for the possible reply.

Disrupting fax over internet? Why not just disrupt fax?

https://en.wikipedia.org/wiki/Email

More seriously though, the organizations still using faxes are not interested in "disruption", or anything else that is newer than the 80s.

They exist and cost $10 for a couple of pages, as the OP said.

>The fax requires a 56KHz channel

hmm so you're basically saying that fax machines only listen to data coming across on a 56KHz carrier wave? i guess that makes sense since the same device (a fax machine) can function both as a telephone and fax (and you'd need to distinguish in hardware which mode you were operating in upon "picking up the phone").

but what i don't understand is the telephony (i'm not a network engineer): when i make a phone call from either a fax or a phone isn't there, at this point in time, a hop through a digital connection? so doesn't some interchange between me and the person i'm calling ultimately decide what frequency carrier the receiver hears?

Yes, once your call gets to the central office it's moved onto a digital connection -- essentially an ISDN line.

Notably, the entire telephone network is standardized on G.711 as a codec, which is raw PCM audio at 64 kbps, from 300–3400 Hz sampled at 8,000 samples/sec. (There's some regional differences that determine whether the network uses G.711u or G.711a, but that's not really important here.)

Also relevant: The digital telephone network is exclusively a switched (not packet-based) network, so you're guaranteed near-zero latency. This is particularly important for fax modems, which don't tolerate any jitter.

The end result is a connection that might not sound the best (due to the narrow bandwidth), but has call quality characteristics that are standardized and guaranteed across the network.

(At some point in the future the telephone network will upgrade codecs to G.722 and AMR-WB, which will give higher voice quality. This is the "HD Voice" feature you see on VoLTE phones. But it'll be a while before that's universal.)

>The digital telephone network is exclusively a switched (not packet-based) network, so you're guaranteed near-zero latency.

Seems like this fact could be exploited more often in internet applications. High bandwidth channel for static assets, low latency channel for e.g. player movement data?

One of the professors when I was in a university was from ex-national telco R&D, and he was very vocal about those “real” bandwidth/latency/privacy assured telecom techs, “far superior” than “toys” like TCP/IP.

And that was for the sake of education. He’d hold his lab meetings in Skype, not over a multipoint ISDN video calls on circuit switched wireless ATM. Only when the show is on, speaking to us undergrads through the mic, he declined to admit that IP is more adopted.

Circuit switched data looks pretty on paper, that’s true, but at the end of the day no one pays for or bother with it.

Low jitter is a luxury for the document web, but if approximately all human social interaction is going to be through digital video for the foreseeable future, we might start to care...

It is already used with QoS [1]

[1] https://en.m.wikipedia.org/wiki/Quality_of_service

Which is traffic prioritization on the same channel. Having an extra link for high-priority traffic would be...exotic...but workable, sure.

It isn't a 56kHz channel. It is a 56k bps channel encoded as complex tones onto a 4kHz voice connection. The telephone system wiring does not have the bandwidth to carry a signal that is 2-3x the range of human hearing.

> so doesn't some interchange between me and the person i'm calling ultimately decide what frequency carrier the receiver hears?

Probably it's an interchange between your POTS handset and your local phone service provider's head-end, that determines the "media type" of the call; and then, in the case of fax, that call is forwarded (including media type) to the other land-line's service provider's head-end, where the service provider then negotiates+accepts the media type, rings the fax on the line until off-hook, and then sets the opened line to operate at that frequency.

I'm guessing that the thing picking up the carrier on the sender's head-end is the same thing that parses DTMF tones off the line; and therefore, not a component that exists in the signal chain of cellular service providers.

(A similar thing happens in cellular voice signalling—the negotiation of a voice codec—but it happens out-of-band, not in-band, AFAIK—on both ends (like a web browser talking to a web server) unless one end is POTS, in which case the cellular voice provider jumps to the POTS network with a DAC+bank of line couplers. In theory, this means that a cell phone might be able to send a fax to a fax machine, if there was a special voice codec option in GSM/CDMA/VoLTE/etc. for "lossless 56KHz PCM", that would be decoded at the POTS gateway into a raw line signal.)

Copper wire has a lot of bandwidth, but traditional phone lines had low-pass filter at 4KHz, so anything above is cut off. Between switching stations signal is modulated and multiplexed, but on the receiving end it is again restored to the original analog signal. Therefore fax-machines couldn't possibly expect anything on 56KHz for sure, it would never get to them. Also, faxes make that well-known noise when establishing connection, and if it were around 56KHz, we wouldn't be able to hear it - anything above 15KHz is very hard to hear for adults. So the OP probably meant 56kbps as a transmission rate, which again is not true because faxes existed much before 56K was possible and many worked on as low rates as 2400bps.

I seem to recall some occasions in the early 90's where there was exactly this problem and fax was not guaranteed to work with some land line due to the digital nature.

In the early days of running "analog" telephony over digital circuits, the edges of the audible hearing range would get clipped and compression would muddy the sharpness of the tones used. Humans couldn't tell much of a difference but devices cared a great deal. Nowadays you can run a fax over specific codecs designed to not do this (G.729, I think? Maybe G.711, it's been a few years).

We would also pull off a clever trick with the reverse. If the modem sensed a digital carrier and your ISP was digitally-connected to your telephone company, the connection could be put into a sort of digital-hybrid mode where the downlink of the connection would be 56kbps (the upstream would still be 28.8kbps maximum). This is where US Robotics "X2" modem names came from.

Reply on news.ycombinator.com