In convolutional codes each output not only depends on the current input but also on previous inputs. The constraint length (symbolized by the letter K) is an integer that specifies the "memory" of the code. For instance, each output of a K=3 encoder actually depends on 3 bits, the current incoming bit and the 2 previous ones.
The finite state machine underlying the encoder has exactly 2^(K-1) states because the "history" of the FSM can be represented using the K-1 previous bits.
Bigger constraint lengths yield better codes, but at the cost of an exponential increase in complexity.
The online version of the generator supports K ranging from 7 to 20. To generate codes outside this range, please contact us.