What limits the speed at which information is sent/received over a fibre optic line?

Question

I was trying to find an answer everywhere, but all I can find is absorbtion and scattering, which doesn't limit the speed itself, just the quality of the signal.

Answer 1

The speed at which you turn on and off the light source (laser) and the speed at which you can reliably read it, are typically the major limiters. Fiber optics may allow you to use light as your data medium, and of course at the speed of light, it would seem like an almost limitless way to transmit data. But in reality, the way in which the signal is modulated and encoded are the limiting factors. On each end of a given fiber optic run, equipment has to encode and decode the pulses and frequencies of light. This is also the transition point from optics to electronics.

This is why the fiber optic cable itself is relatively inexpensive, and the communications equipment it connects to can be very costly.

Answer 2

I've got a really informative book on this, unfortunately it is not where I am right now. So, with probably a lot of omissions and incorrect terms (you've got the internet, you look it up!), off the top of my head:

Over a short distance in ideal conditions, the limit is the laser & receiver; - Modulation Speed: How fast you can pulse the laser / shortest pulse the Rx can detect - Wavelength: Not to be confused with modulation speed, this is the colour of the light and how accurate/stable it is. There are various effects linked to this, but the big one in modern systems is that you can use multiple colours down one fibre to shove more data down. The more colours you can reliably send & filter out at the other end, the more data you can get down it. This is called (Dense) Wave(length) Division Multiplexing. The "dense" got added when they went over about 8 colours, but that could've been Marconi's marketing department.

Over long distances, lots of factors come into play, already mentioned are absorption & scattering, obviously you have attenuation (especially across joints/splices/connectors), reflection (again, especially at joints etc.), dispersion - all of those have more than one mode (see Wikipedia) plus the problem of regeneration.

In fact, Wikipedia does a reasonable job of explaining the basics so I'm going to stop typing now.

[Edit] To add: I found the book but didn't have time to do more than skim the pretty pictures, basically there are multiple weird modes of dispersion/distortion that happen; colours shift, the polarisation of the wave can shift / get twisted, and in multi-mode fibres especially the waves that travel nearer the "edges" end up going slower due to the graduated refractive index, which distorts the pulses and whatnot.

On the plus side, it's still a damn sight better than copper, although ultimately neither are immune to the scourge of all telecommunications links - the guy in the JCB not paying full attention to where he's digging.

And any smart-arse who says radio links are immune to JCB-guy, trust me, the data has to get to & from the radio dish somehow... ;)

Answer 3

I am not an EE, but I do have an MS in CS and there is something I can provide here.

More generally, the speed of any communication medium is defined by Shannon's Theorem. This is the theoretical upper-limit that the medium itself can carry and helps explain why, for example, you could never get beyond 56k on a dial-up modem (analog voice lines operated at approximately 3Khz with SNR of 45db)

DSL lines allowed for the use of higher (inaudible) frequencies for transmission and as such had higher bandwidth potential. But, as you can see by market evidence, DSL has essentially lost to cable modems and other technologies because phone lines are disappearing and even if they weren't, the medium itself can only carry so much data on a copper line (electromagnetic interference drops the SNR, thus limiting bandwidth ala shannon) and distance is very limited at those frequencies.

So as far as fiber goes: Each light frequency has the potential to hold bandwidth. The higher the frequency the more bandwidth potential. But, higher frequencies require faster (de)modulating hardware (an EE can correct me here but I believe it requires sampling at 2x the carrier frequency to ensure no data loss). And, while lasers are robust against interference they can have frequency drift and dispersion.

So, as others have mentioned, to add to the subject a bit, the send/receive hardware really has its work cut out for it. It has to sort out colors, compensate for errors (if possible) and then (de)modulate all those signals at silly-high speeds.

Answer 4

I've got a bit interested in this because of a related question, so beware that I have no personal expertise in this, I'm just parroting a 2012 textbook/monograph: Design of Integrated Circuits for Optical Communications (2nd ed.) by Behzad Razavi.

Firstly, the power (which dictates distance that you can transmit to) and switching speed are negatively correlated because a powerful laser requires a physically bigger semiconductor device, which in turn has higher capacitance, so it's harder/slower to switch. So absorbtion & scattering in medium (fiber) does indirectly affect switching speed too.

As far as switching of laser diodes is concerned there are some additional factors limiting speed:

• There's a turn-on delay when a laser diode is (ahem) turned on before the carrier density reaches a threshold level. During this time/delay you only get spontaneous emission of photons instead of stimulated emission (aka real laser). Although it isn't really clear to me how it happens, the books says (on p. 43) that there's substantial randomness in when the stimulated emission threshold is reached because the spontaneous emission itself has randomness that apparently affects/impedes a more deterministic threshold crossover. Anyway, the net result is jitter in the transmission of actual data transmitted (on stimulated emission).
• The oscillation frequency of a laser (and thus the frequency of the carrier wave in the fiber) depends on the refractive index of the medium (in which it is emitted). But this refractive index is actually not constant in laser diodes because it varies with the carrier concentration (which one must vary to emit pulses). So, a side effect of varying the carrier concentration is a variation in the frequency of the carrier wave itself, which is called frequency chirping. This broadens the spectrum of the carrier by as much as one order of magnitude according to the book, which in turns causes substantial dispersion of signals in the fiber. (And recall the first [non-bullet] point in this answer: worse transmission parameters require more transmission power to counteract, which in turn diminishes switching speed.)
• Step signals also cause a variation of optical output power of a laser diode. This is basically the well-known ringing phenomenon in electrical circuits. In the jargon of the laser guys, this optical ringing is called relaxation oscillation. This optical ringing happens because a high photon density causes the density of the electrons on the upper energy level(s) to fall, which in turn lowers the photon density. The electrons then accumulate and initiate a stronger light emission, etc. This ringing also limits transmission speed because it also causes jitter and potentially even inter-symbol interference (meaning overlap so you might not even be able to tell two adjacent symbols apart). Furthermore, the variation in electron density due to ringing accentuates frequency chirping as well. A common method for reducing ringing is to only partially turn off the laser diode, which creates (due to spontaneous emission) a sort of background glow for the low level symbol (typical the zero symbol); this method essentially reduces the signal-to-noise ratio (called extinction ration [ER] in this case) so it's harder to tell apart the high an low logical levels. So there's a trade off between ER (and the power needed to transmit clearly given low ER, thus switching speed) and ringing, which itself affects switching speed directly. So the engineering choice here is optimizing the spot between a rock and a hard place... The usual figure chosen for ER is between 10 and 15dB, according to the book.
• Finally, the threshold (at which stimulated emission occurs) in laser diodes drifts as they age and also drifts with the operating temperature. (Anyone who has worked with optocouplers should be familiar with a similar phenomenon.) The drift with temperature is about 1-2% per degree Celsius. In order to maintain a constant ER (from the previous bullet), a servo loop needs to control the high and low ends of the current pump driving the diode. The optical output of the laser diode is measured [i.e. converted to an electrical signal] using a photodiode (which is typically physically packaged with the laser diode); the servo then compares the electrical output of the photodiode with a reference value and uses the error to adjust the current driving the laser diode. Obviously this servo loop adds its own time delay and thus presumably some jitter.

Because of these issue, above 10Gb/s the transmission technique is a bit different, employing a continuous wave laser to emit light with constant intensity whose phase shift is then varied (in order to encode data) by a Mach-Zehnder modulator. For engineering reasons that aren't totally clear to me, phase modulation is not sent over the fiber, but rather two Mach-Zehnder modulators are used interferentially to create amplitude modulation. The input/output characteristic of a Mach-Zehnder modulator also drifts with age and temperature, so a control/servo loop must still be employed, although the Mach-Zehnder modulator is a voltage-controlled device, so this control circuit differs from the one used for laser diodes. Additionally at 10Gb/s there are apparently breakdown issues with getting transistors that both switch fast enough and have enough voltage swing to drive the Mach-Zehnder modulators, even though these require only 4 to 6 volts to drive...

For more details you should read the aforementioned book by Razavi, which is rather fascinating for a curious neophyte like myself.