# How does bipolar line coding such as AMI work with medium like fiber optics

We know that in bipolar encoding (sometimes called multilevel binary), there are three voltage levels: positive, negative, and zero. The voltage level for one data element is at zero, while the voltage level for the other element alternates between positive and negative as picture below shows:

but I always think for medium like optical fiber only support two signals, just like the classic approach high voltage means 1 and low voltage means 0. so for AMI, optical fiber has to use 3 different light signals to represent +, - and zero. And considering we have other mutilevel schemes like 2B1Q as the picture below shows:

now it seems like optical fiber has to use 4 different light signals to support it. so my questions are:

Q1. how many different light signals can be produced to support bipolar encoding?is it a limited number?

Q2. we know that fiber optics has a lot of tiny optical fibers as the picture below shows:

so does each tiny fiber represent one light signal?

Q3. Does optical fiber transmit data by parallel transmission or serial transmission.

since fiber optics uses a lot of tiny fibers, so I guess it is parallel transmission?

how many different light signals can be produced to support bipolar encoding?is it a limited number?

Several different light signal levels can be used but the number is limited by a couple of things: -

The transmitting laser diode will have a range of light level outputs ranging from a low level to a high level as shown in this picture: -

As can be seen, you have to avoid letting the laser current drop below $$\I_{TH}\$$ because coherence is lost and there can be a significant delay (loss of data) in restoring the device into proper laser action. And, as the device warms up, the light level will change for a given data modulation of the bias level. $$\I_{TH}\$$ also rises with higher temperatures and the "operational" slope decreases.

Taking all these things into account and assuming that the laser drive levels can be corrected on-the-fly means there are only a handful of distinct light output levels that can be used. I can't put a number on it but I'd have thought 8 might be too many whereas 4 is probably achievable (taking into account the whole system). The extinction ratio (maximum usable light output divided by minimum usable light output) for a laser might only be 10 dB - that's the whole dynamic range the laser can produce so any intermediate levels have to fit into that range.

Some laser diodes will be better; some will be worse.

The 2nd thing is basic signal-to noise at the receiving end. The light levels get attenuated by the fibre and therefore the signal levels at the receiver can be quite small and the self-induced noise at the receiving photodiode has to be significantly below the signal level or you will get bit errors. This brings into action the Shannon–Hartley theorem: -

This theorem supports the idea of using multiple levels to represent binary numbers but, the larger the noise, the fewer the number of different levels are allowed and data throughput drops. It's a big subject.

Finally (and I'm sure there's plenty of other things to consider) is the demodulation process and the data formatting. If each symbol transmitted has (say) 4 levels (i.e. can encode a two-bit binary number), given that there are drifts on the system (as well as noise), you have to ensure that each of those four levels is sent regularly so that the receiver can "calibrate" itself should things drift off in one direction or the other.

so does each tiny fiber represent one light signal?

Each fibre is likely to be unrelated to the other fibres and hence each fibre may send pure 1 and 0 digital data or use a multi-level transmission as described above.

since fiber optics uses a lot of tiny fibers, so I guess it is parallel transmission?

The concept of the arrival of simultaneous parallel data bits is fraught with practical problems because if one fibre is longer than the others or one optical interface is a bit slower than the others (propagation delay), it will mean you have significant bit misalignment problems. If all the bits are shipped out serially these problems cannot happen and, in the long run, at high data rates, serial transmission proves to be faster than parallel transmission AND uses far fewer resources.

• Thanks for your great answer. So just want to confirm that, if I use AMI line encoding, the each tiny fiber should be able to send 3 different light signals to represent one bit?is my understanding correct? Jun 13, 2020 at 11:08
• Yes, each tiny fibre can "transport" multiple level light signals. The transmitter sends it, the fibre transports it. Each tiny fibre is unrelated to each other except they may be paired to allow a more convenient setup for transmit (1 fibre) and receive (the other fibre). Date is normally sent serially so 110 is sent 1 then 1 then 0. Jun 13, 2020 at 12:06
• Parallel transmission of data is not done, the chief reason being that you cannot control the arrival of all the parallel data to be simultaneous and therefore sending serial data (everything in one bag) is the preferred method because the sequence timing is barely affected. Basically serial data can be faster than parallel data. Jun 13, 2020 at 12:59
• Ansolutely they do and, if the data requirements for two or more users is within the speed capabilities of one fibre then I don't see why they can't share one fibre but don't ask how they might disentangle each users data! Fibre is cheap but installing is expensive so it makes total sense to future proof an installation with a surplus of data carrying capability. Jun 13, 2020 at 13:29
• @amjad For long distance fiber, they may bury 5 or 10 fibers at once since it would be expensive to dig up the cable and install more later. Usually each individual fiber is shared across a huge number of different channels, each using a tiny range of wavelengths. Jun 13, 2020 at 15:17