“…In Figure 1, consider for each user a binary rate k/n convolutional encoder, where k and n are positive integers and k < n. For simplicity we assume each encoder has overall constraint length kν and can be implemented by k shift registers, each of length ν. Let u = u ) denote the output subblocks associated with information blocks u t = (u In order to use error-correcting codes in a noisy 2-BAC and avoid the well known ambiguity resulting from the input pairs (u t = 0, d t = 1) and (u t = 1, d t = 0), a construction was proposed in [4] which employs the same error-correcting code in systematic form, for both users in Figure 1, in a serial concatenation with noiseless 2-BAC codes. However, the construction described in [4] still leads to a form of ambiguity expressed as the conditional probability equality P{u t = 0, d t = 1|y} = P{u t = 1, d t = 0|y}, which forbids the decoder of separating the symbols sent by each user in the 2-BAC at time instant t, except for the trivial cases, i.e., where u t = d t .…”