- The sound cuts through ALL background noise.
- Nobody else uses it, so I don't worry about false positives
- I geek a Nostalgia-dopamine hit when I hear it.
The best thing about this was having the infrastructure on public show, a bit like having a building's water pipes visible on the outside where everyone can see them. This exposure, combined with natural curiosity, meant many people had an idea of how it worked.
One takeaway from the diagram is how adaptation and channel estimation can increase channel capacity. We start off simple on the left hand side, with the dial tone and DTMF sending only a couple of frequencies at low speeds. As we move to the right each end adapts/learns and more frequencies come into play increasing the information density. Eventually the adaptation has reached some form of optimum and the information density (channel capacity) is maximised with the spectrogram being a solid mass of signals.
I heard this sound so many times that eventually I could tell if it wasn't going to negotiate properly, or negotiate at a lower speed, just by listening to it. If that was happening I could kill the attempt early and try again.
It was the time where autists were still in control :D No way today a business school product manager let engineers ship millions of next gen communication equipment sounding like loudly dying lamas :D
If I understand correctly there was absolutely 0 need to make it noisy to your ear by default except for the tester at the factory to test if it worked.
Having started with compuserve on a teletype-like device that used paper rolls and an acoustic-coupler 300baud modem, it's humorous to me to watch people learn about this like brand new history in 2021.
(but wow it's weird to be that old, just plain weird, I don't feel that old, just have these weird old experiences/memories)
I remember when 1200baud came out, it just blew our minds.
Then the battle for 2400bps and 9600bps "standards" hayes vs USR, etc.
Then thousands of jobs were made (and eventually lost two decades later) just having to support 33/56k modems at places like AOL. Just non-stop tech support calls, phone lines were just not made for that.
The slowest modem I ever had the pleasure of using was 9600 baud, but even at that rate, it was way too fast for me to actually comprehend that there was a series of high and low tones representing bits.
Back in college, we had acoustically-coupled TDD devices at most payphones on campus - you'd pick up the handset, dial a number, and put it down on the coupler. Or you could just type on the coupler, and hear tones come out the speaker, at a blistering 45 baud. 45 baud is slow enough for your brain to process the different tones.
If you run minimodem as `minimodem --tx tdd`, you can hear the same thing yourself today, without the coupler.
Minus some further nifty hacks above 2400 baud, it was very convenient to think of modem breath as "the TDD tones, but much faster".
> 9600 baud, but even at that rate, it was way too fast for me to actually comprehend that there was a series of high and low tones representing bits.
To be fair, 9600 bps modem already used fairly advanced modulation (QAM+trellis) and not just high and low tones, so it is not really comprehendible to humans the same way simpler modulations are. Even at much slower rates I don't think you could make much sense of QAM signal.
I really like how you can figure out the phone number dialed with the representation of the DTMF tones. This was basically an audio coordinate system (or just a 2D array): an X tone and a Y tone played at the same time.
The upper component has three positions representing the columns on a phone keypad.
The lower four positions represent the rows.
For example, zeros are represented by upper tone 2 for the middle column (2,5,8,0) and lower tone 4 for the bottom row (*,0,#).
There was a fourth column (A,B,C,D) meant for menu selection, and for the US military to indicate call precedence. Also used for various kinds of system automation and signaling but not accessible directly.
You have also just described how the old analog touch tone phones actually generated the tones. Each column of buttons generated a different tone per column, and each row also generated another tone per row. A given button when pressed activated a column and a row to generate the two tones associated with that button.
Where I'm from we used pulse dialing. I remember being so proud of myself for figuring out that the number of pulses the phone sent represented the digit you're dialing.
The Quadrature Amplitude Modulation used by the modem (the last part of the sequence, which sounds like white noise) really is a sequence of auditory vectors.
ADSL/VDSL uses phone lines but operate in the hundred kilohertz to several megahertz range, so all the sounds are ultrasonic. Still, it would be nice to see the recordings and plot them visually in a spectrogram. DOCSIS uses many megahertz of bandwidth too, and similarly you can't hear stuff but it would be interesting to see a spectrogram.
Things no longer work this way. Old skool modems took advantage of existing lines designed for voice, so the tones were somewhat audible. Now, it's pretty much an ethernet connection straight to your house (a gross over simplification), but modern connections to ISPs are no longer the same.
Underneath the Ethernet layer there is still an old skool layer of modulation over a copper wire (unless you have fibre to your premises or are on wireless).
It's not that far removed from the old 56k modems. The main difference is that VDSL uses multiple carriers at different frequencies, but each individual carrier bears some resemblance to a 56k modem. VDSL is basically 10000 old skool modems running in parallel. It's no coincidence that the VDSL tone spacing is about 4kHz, or the bandwidth of a phone line. Your "voice phone line" is just carrier #0.
Also VDSL uses some very sophisticated channel modelling, whereby it measures the coupling between your copper wire and all the other copper wires in a bundle of telephone wires and cancels the interference out. This allows it to send carriers down the wire at frequencies that were previously unusable. It's basically a beefed up version of the 56k training, taking advantage of today's computers running faster and being able to run a more complex channel model/estimator in real time.
As a whole the VDSL signal is mostly on ultrasonic carriers, but it would be possible to tune into one of the carriers and listen to it. (Interesting project anyone?)
But if they could make grandma marvel at the sound of two blackholes merging by transposing a faint gravitational wave to sound, I suppose we could amaze grandma by taking some kind of wavy view of the digital signal while someone load the picture of a cute girl's black hole.
Advantages:
Disadvantages: