Zustandsautomaten können sehr elegant in VHDL beschrieben werden.
Hier ein Automat, der die Sequenz “101” erkennt und dann am Ausgang eine “1” ausgibt. Ein Moore Automat entsteht, wenn die Ausgangssignale nur vom State und nicht vom Eingangssignal abhängen.
library ieee; use ieee.std_logic_1164.all; entity automat is port ( clk_i: in std_ulogic; reset_ni: in std_ulogic; in_i: in std_ulogic; out_o: out std_ulogic ); end; architecture rtl of automat is type state_type is (start_s, erste_eins_s, null_in_der_mitte_s, fertig_s); signal current_state, next_state : state_type; begin next_state_and_output_p : process(current_state, in_i) begin out_o <= '0'; next_state <= current_state; case current_state is when start_s => if in_i = '1' then next_state <= erste_eins_s; end if; when erste_eins_s => if in_i = '0' then next_state <= null_in_der_mitte_s; else next_state <= erste_eins_s; end if; when null_in_der_mitte_s => if in_i = '1' then next_state <= fertig_s; else next_state <= start_s; end if; when fertig_s => out_o <= '1'; if in_i = '1' then next_state <= erste_eins_s; else next_state <= start_s; end if; end case; end process next_state_and_output_p; -- The sequential process for flipflop instantiation -- All signal assignments in this process will result in flipflops. state_reg_p : process (clk_i, reset_ni) begin if reset_ni = '0' then current_state <= start_s; elsif rising_edge(clk_i) then current_state <= next_state; end if; end process state_reg_p; end architecture;
Und hier ist eine geeignete Testbench:
library ieee; use ieee.std_logic_1164.all; use ieee.math_real.all; entity automat_tb is port ( clk_i: in std_ulogic; reset_ni: in std_ulogic; in_i: in std_ulogic; out_o: out std_ulogic ); end; architecture beh of automat_tb is component automat is port ( clk_i: in std_ulogic; reset_ni: in std_ulogic; in_i: in std_ulogic; out_o: out std_ulogic ); end component; signal clk, reset, stim, output : std_ulogic; begin -- This process generates the clock clk_p : process begin clk <= '0'; wait for 5 ns; clk <= '1'; wait for 5 ns; end process; -- Here the automat is instantiated as the device under test dut: automat port map ( clk_i => clk, reset_ni => reset, in_i => stim, out_o => output ); test : process begin reset <= '0'; -- Wait for some time wait for 100 ns; wait until falling_edge(clk); -- Reset Release reset <= '1'; -- Now the positive 101 Sequence stim <= '1'; wait until falling_edge(clk); stim <= '0'; wait until falling_edge(clk); stim <= '1'; wait until falling_edge(clk); stim <= '0'; wait until falling_edge(clk); wait until falling_edge(clk); wait; end process test; end architecture;
The previous testbench only produced a stimuli sequence which results in the output going high. The testbench did not check the output signal however. For checking the output, the assert command is used in VHDL. The following modified testbench checks the output.
test : process begin reset <= '0'; -- Wait for some time wait for 100 ns; report "Checking Reset Value"; assert output = '0' report "Resetvalue of output is not o.k." severity error; wait until falling_edge(clk); -- Reset Release reset <= '1'; -- Now the positive 101 Sequence stim <= '1'; wait until falling_edge(clk); assert output = '0' report "Output is not zero after first one" severity error; stim <= '0'; wait until falling_edge(clk); assert output = '0' report "Output is not zero after zero in the middle" severity error; stim <= '1'; wait until falling_edge(clk); assert output = '1' report "Output is not O.K." severity error; stim <= '0'; wait until falling_edge(clk); wait until falling_edge(clk); assert false report "End of Simulation" severity failure; end process test;
In addition the simulation is stopped with the final assert false statement. The previous stimuli sequence only shows one positive example for the good case. It is not tested that other sequences may also trigger the output going high.
The following code test process produces a random sequence for the input pattern. It is checked if the “101” sequence is in the random pattern. With “101” the output is checked for “1”, otherwise the output has to be “0”.
use ieee.math_real.all; --- architecture and begin follows test2_p : process variable zufall : real; variable seed1 : integer := 5; variable seed2 : integer := 17; variable inp_letzter, inp_vorletzter : std_ulogic; begin reset <= '0'; wait for 10 ns; reset <= '1'; for count in 0 to 1500 loop wait until falling_edge(clk); if (inp_vorletzter = '1' and inp_letzter = '0' and stim = '1') then assert output = '1' report "Fehler: Habe eine Eins erwartet" severity error; else assert output = '0' report "Fehler am Ausgang - Null erwartet" severity error; end if; inp_vorletzter := inp_letzter; inp_letzter := stim; uniform(seed1, seed2, zufall); if (zufall <= 0.5) then stim <= '0'; else stim <= '1'; end if; end loop; assert false report "End of Simulation" severity failure; end process test2_p;
This code simulates 1500 cycles. Also this code does not test all possible input sequences - which is not possible anyway. Without knowledge about the internal state it is not possible to do a 100% verification of the statemachine implementation.