13

I recently read about CRC, a checksum to detect package corruption and drop corrupted packages. But how does a packet in an IP-based network get corrupted in the first place?

1
  • 6
    Your question basically boils down to "Why are systems imperfect?" There's implementation mistakes, environmental conditions... Systems don't run in perfect isolation. No matter how good a system you make, it will still be affected by random noise from the environment - cosmic rays, EM interference, "that faulty device on the network/power supply"... Even your computer's RAM/HDD has error-detection and correction mechanisms, and that's right there in your computer's box - would you expect communication over thousands of kilometers to be less faulty?
    – Luaan
    May 16, 2017 at 13:49

1 Answer 1

21

Packets are long streams of binary numbers (zeros and ones).

The zeros and ones are usually changes of an electric signal, specifically voltage changes.

To simplify, suppose that 0 is 0 volts and 1 is represented alternatively as 5 volts and -5 volts.

Then a long string of zeros and ones will look like this in terms of voltage variations in the cable:

enter image description here

As you can see, the first 1 is 5 volts, but the next 1is -5 volts. This is neccesary to avoid that a long string of 1s energizes the communication line.

If the cable is exposed to electromagnetic interference, then the voltage levels can be altered. If a -5 volts value becomes -2 volts then the receiver can't identify it precisely as a zero or a one.

If your data cable runs for a long stretch parallel to power cables, then the power cable can cause interference to the data cable and induce errors.

2
  • 3
    This gets worse for multi level PAM encodings, like 5 level 1000BASE-T or 10GBASE-T with 16 levels. Just look at one of their eye diagrams: 0397dc3.netsolhost.com/images/wp_ChooseCAT5/wp_ChooseCAT5_2.jpg
    – PlasmaHH
    May 16, 2017 at 14:25
  • 1
    WiFi interference is maybe more obvious/likely. Most wireless systems include some level of FEC because there is often some interference and the parity overhead is worth it to avoid the massive time hit that giving a bad packet to higher layers might make as they need to retry.
    – Nick T
    May 16, 2017 at 22:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.