Added g_big_endian_wb_in/out. Improved index order description.

libraries/dsp/filter/src/vhdl/fil_ppf_wide.vhd

@@ -18,41 +18,77 @@
 -- along with this program.  If not, see <http://www.gnu.org/licenses/>.
 --
 -------------------------------------------------------------------------------
--- Purpose: Performing a poly phase prefilter  (PPF) function on one or multiple wideband data stream.
+-- Purpose: Performing a poly phase prefilter (PPF) function on one or more
+--          wideband data stream.
 --
 -- Description:
---   The poly phase prefilter function is applied on multiple inputs. The
---   incoming data must be divided over the inputs as shown in the
---   following example for nof_streams=1 and wb_factor is 4:
+--   The poly phase prefilter function is applied on multiple inputs. In
+--   array notation:
 --
---     array   stream   wb      time index
---     index   index    index
---         3       0        0 : t0, t4,  t8, t12, t16...
---         2       0        1 : t1, t5,  t9, t13, t17...
---         1       0        2 : t2, t6, t10, t14, t18...
---         0       0        3 : t3, t7, t11, t15, t19...
+--               parallel                 serial             type
+--   in_dat_arr  [wb_factor][nof_streams] [t][nof_channels]  int
+--   out_dat_arr [wb_factor][nof_streams] [t][nof_channels]  int
+-- 
+--   If g_big_endian_wb_*=true then the time t to wb_factor P mapping for the
+--   fil_ppf_wide is t[0,1,2,3] = P[3,2,1,0], else when it is false then the
+--   mapping is t[3,2,1,0] = P[3,2,1,0]. The mapping can be selected
+--   independently for the in_dat_arr and the out_dat_arr.
+--
+--   The incoming data must be divided over the inputs as shown in the
+--   following example for nof_streams=1 and wb_factor is 4. The array
+--   index I runs from [wb_factor*nof_streams-1:0].
+--
+--     array   wb     stream   time index when g_big_endian_wb_*=true)
+--     index   index  index 
+--       I       P      S      t
+--       3       3      0    : 0, 4,  8, 12, 16, ...
+--       2       2      0    : 1, 5,  9, 13, 17, ...
+--       1       1      0    : 2, 6, 10, 14, 18, ...
+--       0       0      0    : 3, 7, 11, 15, 19, ...
+--                             ^
+--                             big endian
 --              
 --   Every array input will be filtered by a fil_ppf_single instance. It is
 --   also possible to offer multiple wideband input streams. Those wide
 --   band input streams will share the filter coefficients. For a system with
---   nof_streams=2 and wb_factor=4 the array inputs become (s0t0 means wide
---   band stream 0, timestamp 0):
+--   nof_streams=2 and wb_factor=4 the array inputs become:
+--
+--     array   wb     stream   time index when g_big_endian_wb_*=true)
+--     index   index  index  
+--       I       P      S      t
+--       7       3      1    : 0, 4, 8,  12, 16, ...
+--       6       3      0    : 0, 4, 8,  12, 16, ...
+--       5       2      1    : 1, 5, 9,  13, 17, ...
+--       4       2      0    : 1, 5, 9,  13, 17, ...
+--       3       1      1    : 2, 6, 10, 14, 18, ...
+--       2       1      0    : 2, 6, 10, 14, 18, ...
+--       1       0      1    : 3, 7, 11, 15, 19, ...
+--       0       0      0    : 3, 7, 11, 15, 19, ...
+--                             ^
+--                             big endian
 --
---     array   stream   wb      time index
---     index   index    index
---         7       0        0 : s0t0, s0t4, s0t8,  s0t12, s0t16...
---         6       1        0 : s1t0, s1t4, s1t8,  s1t12, s1t16...
---         5       0        1 : s0t1, s0t5, s0t9,  s0t13, s0t17...
---         4       1        1 : s1t1, s1t5, s1t9,  s1t13, s1t17...
---         3       0        2 : s0t2, s0t6, s0t10, s0t14, s0t18...
---         2       1        2 : s1t2, s1t6, s1t10, s1t14, s1t18...
---         1       0        3 : s0t3, s0t7, s0t11, s0t15, s0t19...
---         0       1        3 : s1t3, s1t7, s1t11, s1t15, s1t19...
+--   Note that I, P and S all increment in the same direction and t increments 
+--   in the opposite direction of P. This is the g_big_endian_wb_in=true and
+--   g_big_endian_wb_out=true format for the in_dat_arr and out_dat_arr.
 --
---   For fil_ppf_wide with wb_factor > 1 the array index runs from
---   [wb_factor*nof_streams-1:0]. However per stream the wb_factor index
---   needs to run in big endian order from [0:wb_factor-1], so this needs to
---   be accounted for when connecting wide band streams to fil_ppf_wide.
+--   If g_big_endian_wb_in=false and g_big_endian_wb_out=false for little endian
+--   format, then the time t increments in the same direction as P, for both
+--   in_dat_arr and out_dat_arr, so then I, P and S all increment in the same
+--   direction:
+--
+--     array   wb     stream   time index when g_big_endian_wb_*=false
+--     index   index  index  
+--       I       P      S      t
+--       7       3      1    : 3, 7, 11, 15, 19, ...
+--       6       3      0    : 3, 7, 11, 15, 19, ...
+--       5       2      1    : 2, 6, 10, 14, 18, ...
+--       4       2      0    : 2, 6, 10, 14, 18, ...
+--       3       1      1    : 1, 5, 9,  13, 17, ...
+--       2       1      0    : 1, 5, 9,  13, 17, ...
+--       1       0      1    : 0, 4, 8,  12, 16, ...
+--       0       0      0    : 0, 4, 8,  12, 16, ...
+--                             ^
+--                             little endian
 --
 --   With wb_factor > 1 and nof_streams > 1 the streams in in_dat_arr and
 --   out_dat_arr are looped first and then the wb_factor is looped. This
@@ -77,6 +113,8 @@ use work.fil_pkg.ALL;
 
 entity fil_ppf_wide is
   generic (
+    g_big_endian_wb_in  : boolean            := true;
+    g_big_endian_wb_out : boolean            := true;
     g_fil_ppf           : t_fil_ppf          := c_fil_ppf;    
     g_fil_ppf_pipeline  : t_fil_ppf_pipeline := c_fil_ppf_pipeline; 
     g_file_index_arr    : t_nat_natural_arr  := array_init(0, 128, 1);  -- default use the instance index as file index 0, 1, 2, 3, 4 ...
@@ -100,8 +138,9 @@ architecture rtl of fil_ppf_wide is
 
   type t_fil_ppf_arr      is array(integer range <> ) of t_fil_ppf;                       -- An array of t_fil_ppf's generics.                                                                              
   type t_nat_natural_arr2 is array(integer range <> ) of t_nat_natural_arr(127 downto 0); -- An array of arrays, used to point to the right .mif files for the coefficients
-  type t_out_arr          is array(integer range <> ) of std_logic_vector(g_fil_ppf.nof_streams*g_fil_ppf.out_dat_w -1 downto 0);
-  type t_in_arr           is array(integer range <> ) of std_logic_vector(g_fil_ppf.nof_streams*g_fil_ppf.in_dat_w  -1 downto 0);
+
+  type t_streams_in_arr   is array(integer range <> ) of std_logic_vector(g_fil_ppf.nof_streams*g_fil_ppf.in_dat_w  -1 downto 0);
+  type t_streams_out_arr  is array(integer range <> ) of std_logic_vector(g_fil_ppf.nof_streams*g_fil_ppf.out_dat_w -1 downto 0);
   
   ----------------------------------------------------------
   -- This function creates an array of t_fil_ppf generics 
@@ -115,22 +154,22 @@ architecture rtl of fil_ppf_wide is
     variable v_nof_bands : natural := input.nof_bands/input.wb_factor;                     -- The nof_bands for the single channel poly phase filters
     variable v_return    : t_fil_ppf_arr(input.wb_factor-1 downto 0) := (others => input); -- Variable that holds the return values
   begin
-    for I in 0 to input.wb_factor-1 loop
-      v_return(I).nof_bands := v_nof_bands;     -- The new number of bands
+    for P in 0 to input.wb_factor-1 loop
+      v_return(P).nof_bands := v_nof_bands;     -- The new number of bands
     end loop;                               
     return v_return; 
   end;
   
   ----------------------------------------------------------
-  -- Function that divids the input file index array into 
+  -- Function that divides the input file index array into 
   -- "wb_factor" new file index arrays. 
   ----------------------------------------------------------
   function func_create_file_index_array(input: t_nat_natural_arr; wb_factor: natural; nof_taps: natural) return t_nat_natural_arr2 is
     variable v_return : t_nat_natural_arr2(wb_factor-1 downto 0) := (others => input);    -- Variable that holds the return values
   begin
-    for I in 0 to wb_factor-1 loop
-      for J in 0 to nof_taps-1 loop
-        v_return(I)(J) := input(I*nof_taps+J);
+    for P in 0 to wb_factor-1 loop
+      for T in 0 to nof_taps-1 loop
+        v_return(P)(T) := input(P*nof_taps+T);
       end loop;
     end loop;                               
     return v_return; 
@@ -141,11 +180,12 @@ architecture rtl of fil_ppf_wide is
   
   constant c_mem_addr_w      : natural := ceil_log2(g_fil_ppf.nof_bands * g_fil_ppf.nof_taps / g_fil_ppf.wb_factor); 
 
-  signal ram_coefs_mosi_arr  : t_mem_mosi_arr(g_fil_ppf.wb_factor-1   downto 0);                               
-  signal ram_coefs_miso_arr  : t_mem_miso_arr(g_fil_ppf.wb_factor-1   downto 0) := (others => c_mem_miso_rst); 
-  signal out_val_i           : std_logic_vector(g_fil_ppf.wb_factor-1 downto 0);
-  signal out_arr_i           : t_out_arr(g_fil_ppf.wb_factor-1        downto 0);
-  signal in_dat_vect_arr     : t_in_arr(g_fil_ppf.wb_factor-1         downto 0);
+  signal ram_coefs_mosi_arr  : t_mem_mosi_arr(g_fil_ppf.wb_factor-1 downto 0);                               
+  signal ram_coefs_miso_arr  : t_mem_miso_arr(g_fil_ppf.wb_factor-1 downto 0) := (others => c_mem_miso_rst); 
+
+  signal streams_in_arr      : t_streams_in_arr( g_fil_ppf.wb_factor-1 downto 0);
+  signal streams_out_arr     : t_streams_out_arr(g_fil_ppf.wb_factor-1 downto 0);
+  signal streams_out_val_arr : std_logic_vector( g_fil_ppf.wb_factor-1 downto 0);
   
 begin
   ---------------------------------------------------------------
@@ -165,22 +205,30 @@ begin
     miso_arr => ram_coefs_miso_arr
   );
 
-  gen_in_vect_wb : for I in 0 to g_fil_ppf.wb_factor-1 generate
-    gen_in_vect_streams : for J in 0 to g_fil_ppf.nof_streams-1 generate
-       --in_dat_vect_arr(I)((J+1)*g_fil_ppf.in_dat_w-1 downto J*g_fil_ppf.in_dat_w) <= in_dat_arr(J*g_fil_ppf.wb_factor+I)(g_fil_ppf.in_dat_w-1 downto 0);
-       in_dat_vect_arr(I)((J+1)*g_fil_ppf.in_dat_w-1 downto J*g_fil_ppf.in_dat_w) <= in_dat_arr(I*g_fil_ppf.nof_streams+J)(g_fil_ppf.in_dat_w-1 downto 0);
-    end generate;
-  end generate;
+  p_wire_input : process(in_dat_arr)
+    variable vP : natural;
+  begin
+    for P in 0 to g_fil_ppf.wb_factor-1 loop
+      if g_big_endian_wb_in=true then
+        vP := P;                        -- big endian time [0,1,2,3] to P [3,2,1,0] index mapping
+      else
+        vP := g_fil_ppf.wb_factor-1-P;  -- little endian time [0,1,2,3] to P [0,1,2,3] index mapping 
+      end if;
+      for S in 0 to g_fil_ppf.nof_streams-1 loop
+        streams_in_arr(vP)((S+1)*g_fil_ppf.in_dat_w-1 downto S*g_fil_ppf.in_dat_w) <= in_dat_arr(P*g_fil_ppf.nof_streams+S)(g_fil_ppf.in_dat_w-1 downto 0);
+      end loop;
+    end loop;
+  end process;
 
   ---------------------------------------------------------------
   -- INSTANTIATE MULTIPLE SINGLE CHANNEL POLY PHASE FILTERS
   ---------------------------------------------------------------
-  gen_ppf_singles : for I in 0 to g_fil_ppf.wb_factor-1 generate
-    u_single_filter : entity work.fil_ppf_single 
+  gen_fil_ppf_singles : for P in 0 to g_fil_ppf.wb_factor-1 generate
+    u_fil_ppf_single : entity work.fil_ppf_single 
     generic map (
-      g_fil_ppf           => c_fil_ppf_arr(I),
+      g_fil_ppf           => c_fil_ppf_arr(P),
       g_fil_ppf_pipeline  => g_fil_ppf_pipeline,
-      g_file_index_arr    => c_file_index_arr2(I),
+      g_file_index_arr    => c_file_index_arr2(P),
       g_coefs_file_prefix => g_coefs_file_prefix 
     )
     port map (
@@ -188,18 +236,30 @@ begin
       dp_rst         => dp_rst,
       mm_clk         => mm_clk,
       mm_rst         => mm_rst,
-      ram_coefs_mosi => ram_coefs_mosi_arr(I), 
-      ram_coefs_miso => ram_coefs_miso_arr(I), 
-      in_dat         => in_dat_vect_arr(I),
+      ram_coefs_mosi => ram_coefs_mosi_arr(P), 
+      ram_coefs_miso => ram_coefs_miso_arr(P), 
+      in_dat         => streams_in_arr(P),
       in_val         => in_val,
-      out_dat        => out_arr_i(I),
-      out_val        => out_val_i(I)
+      out_dat        => streams_out_arr(P),
+      out_val        => streams_out_val_arr(P)
     ); 
-    gen_in_vect_streams : for J in 0 to g_fil_ppf.nof_streams-1 generate   
-      out_dat_arr(I*g_fil_ppf.nof_streams+J) <= RESIZE_SVEC(out_arr_i(I)((J+1)*g_fil_ppf.out_dat_w-1 downto J*g_fil_ppf.out_dat_w), out_dat_arr(I)'LENGTH);
-    end generate;
   end generate;  
   
-  out_val <= out_val_i(0);                             
+  p_wire_output : process(streams_out_arr)
+    variable vP : natural;
+  begin
+    for P in 0 to g_fil_ppf.wb_factor-1 loop
+      if g_big_endian_wb_in=true then
+        vP := P;                        -- big endian time [0,1,2,3] to P [3,2,1,0] index mapping
+      else
+        vP := g_fil_ppf.wb_factor-1-P;  -- little endian time [0,1,2,3] to P [0,1,2,3] index mapping 
+      end if;
+      for S in 0 to g_fil_ppf.nof_streams-1 loop
+        out_dat_arr(vP*g_fil_ppf.nof_streams+S) <= RESIZE_SVEC_32(streams_out_arr(P)((S+1)*g_fil_ppf.out_dat_w-1 downto S*g_fil_ppf.out_dat_w));
+      end loop;
+    end loop;
+  end process;
+    
+  out_val <= streams_out_val_arr(0);                             
   
 end rtl;