A while ago (several years, in fact) I published a post on the speed of modems, i.e., digital communication. And since I am currently involved in two fiber-to-the-home projects I felt it was time to revisit the chart. I also notice that I could not find too many graphs that have been updated the last few years.
So, the story was to understand what kinds of speed that are offered to the user (and the backbone to distinguish them). In the previous post it was briefly pointed out why the speed increased, i.e., evolution accelerated, the last twenty years. If one plots the transmission speed for a subscriber line, it does not fit to an exponential trend line, nor any good polynomial interpolations either.
Let’s get (to) the picture first. [Click on the image to get to the google spreadsheet’s chart – there might be some more points added as time passes by. WordPress does not allow embedding google charts]. All transmission speeds are in bits per second. For the sake of clarity (?) I also added the population growth (blue line) where the numbers are in Millions. (Since it is a logarithmic scale it is just a matter of moving the blue curve up and down – it looks rather flat anyway).
- Red stars: telephone lines, teleprinters, wireline modems (POTS, ADSL, ISDN), as well as optofibres (the right-most, top-most red stars). One could argue about the use of optofibres here, but I go by the connected-via-a-cable concept.
- Green pyramids: Radio (and optical systems). We see the development of radio standards for subscribers: 1G through 5G, Wifi. Notice the two green dots on the left-hand side! These are the early telegraphs. One of them being the Chappe telegraph system and the other being the Swedish system developed by Edelcrantz. Fascinating stories themselves.
- Pink bullets: Backbone speeds. Here I have bundled radio, optical, fibres, telephone lines. I have also added some of the more experimental recent results – some claiming to have reached speeds at 1.5 Pbps, i.e., 1.500.000.000.000.000 bits per second. That’s ridiculously fast – about 30000 movies … per second.
- One could argue that some telegraph systems should be treated as backbone (pink) rather than subscriber (red). However, as long as there was no concept of frequency-division multiplexing or similar coding formats, only time-domain multiplexing, i.e., only one message sent at the time over the line, I count them as subscriber lines instead.
There is somewhat of a mess in terms of how speed is reported for all these various formats. The sources refer to measures such as: bits per second, words per minute, symbols per minute, symbols per second, baud rate, etc. For example, the Edelcrantz machinery: “A message could take 30 minutes to be sent from Stockholm to Gävle [some 200 km]”. This particular system, however, used 10 on/off shutters and we can thus translate that to 10 bits per symbol, etc.
For Morse code, there is often a words-per-minute count. In this case one has to assume an average symbol length and an average word length (in British it is 5.1 characters per word), the minute is also a bit vaguely defined since there is no standard delay between characters nor short/long beats on the key. In order to translate to bits per second, I have used what I have interpreted as best practice (various sources on the web).
Voice channels had a bandwidth of some 1 to 8 kHz in the early days and I have assumed 1 bps per Hz. Which is arguably
However, same holds for any system. Even if we say bps there might be additional overhead in terms of error correction coding, repetition, preambles, handshaking, etc. on the line. I have reported the “raw” numbers and not cared to how much information as such is transferred. (With that said, how much information is there in a “Hi, how are you?” message?)
Notice that distances are not considered in the graph! One could argue that it is a bit unfair to compare apples and pears. I have focused a bit on the transatlantic cable as a use case. It illustrates the real challenges. The experimental Pbps optical fibres do not face the same type of challenges. Gauss first telegraphs were a couple of km. The Chappe telegraphs spanned 300 km and more. A longer cable would distort and attenuate the signal more aggressively and transmission speed would be lower. The very first transatlantic cable in the late 1850s (the notable red star at the bottom of the graph) only obtained a “few words per hour”! Horribly bad – but what an exceptional achievement?!
Curve-fitting does not work
As mentioned above – it is hard to get a polynomial interpolation to fit to the data in order to see trends. Exponential approximation do not seem to fit very well either. The R2 values are not impressive. It is quite clear that we instead should look at the graph and think in terms of distinct, paradigm-shifting scenarios. We have
- the arrival of the telegraph
- the introduction of Internet for the “masses”
- the arrival of the optical fibre
- satellite communication
- mobile phone
The telegraph (and telephony systems) shows a rather flat line from the 1820s to the 1930s. After the war, things start to roll. The launch of the SCORE satellite (1958) demonstrated communication across the globe – without cables. One could argue that Sputnik did the same, a year earlier, but it did not relay signals. The first transatlantic fiber cables were laid down in the 1980s.
Notable systems not mentioned in the graph
I did not include smoke signals nor the innovative time-coded pneumatic system used by the Greek in the fourth century.
I did not include the transmission speeds reported in the Star Trek movies (it is too far in the future and would stretch the graph too much). For the same reason I omitted the Star Wars across-the-galaxy-without-latency communication as it happened “a long time ago in a galaxy far, far away”.