What is inter symbol interference?
In: 0
So we like to send data through various means and much of these are what we classify as signals, continuous waveforms, I’m assuming you know this knowing this question.
Much of the data we send is digital, meaning at the end we need to decode this data into discrete parts. Now we send these along channels that are “continuous” in the real world, be it cables, internet, etc. There are a few different means of doing this, but more or less all involve sending what we call “symbols” (most commonly in binary these are simply 1s and 0s but it could be more). We send these symbols at a certain rate, and on the other end the receiver would pick it up by reading the signal at an instant for every period of that rate…
In an ideal world. Problem is we don’t live in an ideal world, and the universe hates us. We send a nice square wave like signal to signal high and low for 1s and 0s and the physics of the wire or wifi that carries that signal turns it into a mushy square wave. Also we don’t in the real world have the ability to read a signal’s strength at an instant, usually its a rough local average at best. And depending on other properties, resonance of the channel at certain frequencies and what not, the effects of sending one symbol may reverberate later on.
We call this inter symbol interference. And we have a few ways to try to deal with it.
First some naiive ways that don’t work. In most cases, any interference tends to diminish with time, as a result we can decrease the symbol rate to a level where the ISI doesn’t harm our measurements. However the problem is that this interference is still most often with neighboring symbols and most signals don’t guarantee that the measurement on the receiving end will be “in the middle of the symbol period,” so this doesn’t work amazingly but is a strategy.
Increasing the signal to noise ratio doesn’t work either, because this isn’t static, this is interference from other symbols. Here the noise is always proportional to the signal, so it will cause problems even if you say crank up the voltage.
Realistically heres what we can do.
First thing we can do is we can try to deal with the channel. The channel itself has certain properties causing this distortion, as you may know we can summarize this in a little piece of data called an impulse response (assuming you know what this is as well since you’re asking this). We can take this impulse response and often find a rough “opposite” to it, so that when it hits our data, they cancel out, fixing out line on the other end. Usually this is known as “impulse shaping” although has certain limitations, its hard to shape impulses and make exact filters like this, but it doesn’t have to be perfect. This also assumes a consistent channel, this works for example well with a wire, which we usually can assume has constant properties and thus a constant impulse response, this doesn’t work well for things like wifi however, which depend a lot on environment.
Beside that we can introduce a higher level of error correction at the cost of some bandwidth, errors will happen but can be caught and corrected before we output them as the final output.
First, the words, “inter” and “symbol”. “Inter” means “between two of…” (as opposed to “intra” which means “within one”). “symbol” is a generic way for the individual elements of some communication. Those may be letters in written communication or words in spoken communication.
So “intra symbol interference” means that there is interference between elements of some communication that should be distinct. Something like writing on a napkin where the ink of one letter runs into the ink of another letter, or when talking in a room with an echo, where the words run into each other.
The latter is a pretty good example, as echo is one of the major issues in electronic communication. No matter if it’s a copper wire or an optical fibre, there will be echo from the sides, the ends, every connection and so on.
That’s the “layman level”, for a deeper dive, see yalloc’s answer.
Latest Answers