dlock | Distributed lock manager | Architecture library

1. Your reset, as written in the VHDL, is active low. This means you're holding the circuit in reset most of the time. Your data pattern looks like you thought your reset was active high.

2. Your error signal, insofar as I can see in the image of the waveform posted, is behaving correctly. Every time you exit reset for a cycle, your data is 0, which sends you to the error state. Of course this only persists for one cycle since you immediately reset again.

3. These are just glitches, if you zoom in you'll see that the phantom unlocks are happening for 0 time (or very small time periods depending on your gate models). This is one reason why the output of combinational logic isn't used for clocking data. Passing the value through a flip-flop will remove glitches.

EDIT: Furthermore, your state assignment table and your state output table disagree with each other. One lists the Q values from Q2 downto Q0 and the other lists from Q0 to Q2, but both list the unlocked state as "110". This doesn't cause issues for the Error state since "111" reads the same forwards and backwards.

EDIT2: As far as avoiding glitches... glitches are the nature of combinational logic.

You could have locked sourced directly from a flop without adding latency by having the input to a "locked" flop be dictated by the same preconditions of the unlocked state (i.e. locked_d = not((state=s4 or state=locked) and data=1) and use locked_q.

You could just avoiding having locked be a function of multiple state bits by converting the state machine machine encoding to a one-hot or hybrid one-hot (where there is a dedicated bit for the locked/error states because they drive output bits , but the other 5 states use 3 shared state bits).

Think of a state table like this: