fir_filter: wip V2fir_filter

Added a rough state machine to control the parts of the filter

1 files changed, 73 insertions, 7 deletions

diff --git a/fir_filter/fir_filter.vhd b/fir_filter/fir_filter.vhd
index 429b09a..63a39b7 100644
--- a/fir_filter/fir_filter.vhd
+++ b/fir_filter/fir_filter.vhd
@@ -42,6 +42,8 @@ architecture behav of fir_filter is
 	signal mul0_sig_in:  signed(data_width-1 downto 0);
 	signal mul0_coef_in: signed(coef_width-1 downto 0);
 	signal mul0_out:     signed(data_width+coef_width-1 downto 0);
+	signal mul0_sel:     integer (range 0 to coef_num);
+	signal mul0_en:      std_logic;
 
 	signal acc_sum: signed(data_width+overflow_width+fraction_width-1
 	                       downto 0);
@@ -50,7 +52,70 @@ architecture behav of fir_filter is
 
 	signal data_delay: array (0 to coef_num-1) of
 	                   signed (data_width-1 downto 0);
+
+	signal fir_shift: std_logic;
+
+	signal FSM_states is (clear, waiting, processing)
+	signal state_cur, state_next: FSM_states;
 begin
+	-- state machine here
+	-- inputs: en,
+	-- outputs: mul0_sel, mul0_en
+	fsm_logic process (en, clk)
+		variable mul0_sel_prev  -- TODO past value so we know
+		                        -- what's next
+	begin
+		-- TODO shift the FIR regs
+		case state_cur is
+		when clear =>
+			state_next <= waiting;
+			mul0_sel   <= 0;
+			mul0_en    <= '0';
+			ready      <= '0';
+			fir_shift  <= '0';
+		when waiting =>
+		when proc_coef =>
+			if mul0_sel='1' then
+				state_next <= ?;
+			else
+				state_next <= processing;
+			end if;
+			mul0_sel   <=  mul0_sel_prev - 1;
+			mul0_en    <= '1';
+			ready      <= '0';  --TODO confirm
+			fir_shift  <= '0';
+		when proc_last =>
+			if (en='1') then
+				state_next <= waiting;
+				mul0_sel   <=  0;
+				mul0_en    <= '1';
+				ready      <= '0';
+				fir_shift  <= '1';
+			else
+				state_next <= proc_last;
+				mul0_sel   <=  0;
+				mul0_en    <= '0';
+				ready      <= '1';
+				fir_shift  <= '0';
+			end if;
+		when others =>
+			state_next <= clear;
+			mul0_sel   <=  0;
+			mul0_en    <= '0';
+			ready      <= '0';
+			fir_shift  <= '0';
+		end case;
+	end process fsm_logic;
+
+	fsm_reg: process (clk, clr)
+	begin
+		if (clr='1') then
+			state_cur <= clear;
+		elsif (clk'event and clk='1') then
+			state_cur <= state_next;
+		end if;
+	end process fsm_reg;
+
 	-- our one multipler
 	-- output will be data_width + coef_width - coefs_offset . coefs_offset
 	mul0_out <= mul0_sig_in * mul0_coef_in;
@@ -68,7 +133,6 @@ begin
 		end if;
 	end process;
 
-
 	-- data_width + coef_width bit from multipler
 	-- data_width = N
 	-- coef_width = M
@@ -78,14 +142,16 @@ begin
 	-- fraction_width = K
 	-- multi out = N + M - Moff . Moff
 
-	process ( TODO )
+	process (mul0_sel, mul0_en)
+		variable N: integer (range 0 to coef_num);
 	begin
-		if (save  coef(N)) then
+		N := mul0_sel;
+		if (mul0_en) then
 			mul0_sign_in <= signal_delay(N);
 			mul0_coef_in <= coefs(N);
 			acc_in <= resize(mul0_out(data_width+data_width-1
 			                          downto
-			                          coefs_offset(N)-fraction_width)
+			                          coefs_offset(N)-fraction_width),
 			                 running_sum'length);
 			acc_en <= '1';
 		else
@@ -96,14 +162,14 @@ begin
 		end if;
 	end process;
 
-	-- cycle here just means increment the saved registers
+	-- fir_shift here just means increment the saved registers
 	data_delay(0) <= input;  -- no delay
-	process (clr, clk, cycle)
+	process (clr, clk)
 	begin
 		if (clr='1') then
 			data_delay(1 to coef_num-1)  <= (others (others => '0'));
 		elsif (clk'event and clk='1') then
-			if (cycle='1') then
+			if (fir_shift='1') then
 				data_delay(1 to coef_num-1)
 				    <= data_delay(0 to coef_num-2);
 			end if;