I am new to VHDL, and I trying to verify UART receiver how is it works.
I synthesized the code below (quoted form the book) and its fine but if needs more let me know :).
The frequency for my board is 100 Mhz and the data I want receive is 8 bits, baud rate is 115200, how the clock and tick should be in the testbench or what is the right testbench here?
library IEEE;
use IEEE.STD_LOGIC_1164.ALL;
use IEEE.numeric_std.all;
entity uart_rx is
generic(
data_bits: integer := 8; -- # d a t a b i t s
stop_bit_ticks: integer := 16 -- # t i c k s f o r s t o p b i t s
);
Port ( rx : in STD_LOGIC;
clk : in STD_LOGIC;
reset: in STD_LOGIC;
tick : in STD_LOGIC;
rx_done : out STD_LOGIC;
data_out : out STD_LOGIC_VECTOR (7 downto 0));
end uart_rx;
architecture arch of uart_rx is
type state_type is (idle, start, data, stop);
SIGNAL state_reg, state_next: state_type;
SIGNAL s_reg, s_next: UNSIGNED(3 downto 0);
SIGNAL n_reg, n_next: UNSIGNED(2 downto 0);
SIGNAL b_reg, b_next: STD_LOGIC_VECTOR(7 downto 0);
begin
-- FSMD s t a t e & d a t a r e g i s t e r s
process(clk, reset) -- FSMD state and data regs.
begin
if (reset = '1') then
state_reg <= idle;
s_reg <= (others => '0');
n_reg <= (others => '0');
b_reg <= (others => '0');
--rx_done <= '0';
-- rx <= '1';
elsif (clk'event and clk='1') then
state_reg <= state_next;
s_reg <= s_next;
n_reg <= n_next;
b_reg <= b_next;
end if;
end process;
-- n e x t - s t a t e l o g i c & d a t a p a t h f u n c t i o n a l u n i t s / r o u t i n g
process (state_reg, s_reg, n_reg, b_reg, tick, rx)
begin
state_next <= state_reg;
s_next <= s_reg;
n_next <= n_reg;
b_next <= b_reg;
rx_done <= '0';
case state_reg is
when idle =>
if (rx = '0') then
state_next <= start;
s_next <= (others => '0');
end if;
when start =>
if (tick = '1') then
if (s_reg = 7) then
state_next <= data;
s_next <= (others => '0');
n_next <= (others => '0');
else
s_next <= s_reg + 1;
end if;
end if;
when data =>
if (tick = '1') then
if (s_reg = 15) then
s_next <= (others => '0');
b_next <= rx & b_reg(7 downto 1);
if (n_reg = (data_bits - 1)) then
state_next <= stop;
else
n_next <= n_reg + 1;
end if;
else
s_next <= s_reg + 1;
end if;
end if;
when stop =>
if (tick = '1') then
if (s_reg = (stop_bit_ticks - 1)) then
state_next <= idle;
rx_done <= '1';
else
s_next <= s_reg + 1;
end if;
end if;
end case;
end process;
data_out <= b_reg;
end arch;
Typically, UART receivers run at 8 times the bit rate. If your bit rate is 115200, this means a sample rate of 921600. If you are running at 100Mzh, you will need to create a clock divider to get you from 100 MHz to the desired sample rate. To go from 921600 to 100 MHz the following will work:
100 MHz = 100,000,000 Hz
921600 samples/sec = 921,600 Hz
divider = 100,000,000/921,600 = 108.51.
Thus, you will need a counter that will count up to 109 (we round up as we have to sample at an integer of the clock rate) on rising_edge(clock), then raise an enable signal that tells your component its time to sample the line and reset the counter. The example above assumed 8 samples/bit which is typical to my knowledge. Thus, if you set the period of your main clock in the simulation to be 1ns and set up the counter circuit I described above, you should get the test bench you are looking for.
EDIT: warning about uneven clock division
Almost forgot to mention this. Since your clock rate does not divide evenly into the bit rate for the UART, some extra care must be taken when coding up this circuit. Your sample rate will move later and later in the transmission with the scheme I have proposed. You will probably have to add a simple offset to change your counter to 108 on the even bits to keep you more aligned with the incoming data bits.
See here for some more info: https://electronics.stackexchange.com/questions/42236/uart-receiver-clock-speed
Related
I'm making a custom hardware ARINC 429 Core.
For now I have described the module in transmission (TX-FSM), according to the ARINC 429 standard and a FIFO in transmission from which it takes the data and sends them to the outside.
The FIFO works at a frequency of 2MHz (clk2M), while TX-FSM can generate a frequency of 100kb / s or 12.5kb / s (clk429) from 2MHz as per standard.
Since the FIFO works at a higher frequency (2 MHz), and the TX-FSM works at a lower frequency (100 kb/s), when the TX-FSM requests a data from the FIFO by raising the "TX_FIFO_rd" signal ("rd_en" on FIFO ), the FIFO supplies all the data contained within it, since in the FIFO clock domain the "rd_en" signal remains high for several cycles.
The FIFO should only provide one data at a time. Once the data has been transmitted, the TX-FSM will request the next data.
How can I make the FIFO and TX-FSM work in sync using a single clock?
FIFO VHDL code:
library ieee;
use ieee.std_logic_1164.all;
use ieee.numeric_std.all;
entity FIFO is
generic (
FIFO_WIDTH : natural := 32;
FIFO_DEPTH : integer := 10;
ALMOST_FULL_LEVEL : integer := 8;
ALMOST_EMPTY_LEVEL : integer := 2
);
port (
reset : in std_logic;
clk : in std_logic;
-- FIFO Write Interface
wr_en : in std_logic;
wr_data : in std_logic_vector(FIFO_WIDTH-1 downto 0);
ALMOST_FULL : out std_logic;
FULL : out std_logic;
-- FIFO Read Interface
rd_en : in std_logic;
rd_data : out std_logic_vector(FIFO_WIDTH-1 downto 0);
ALMOST_EMPTY : out std_logic;
EMPTY : out std_logic
);
end FIFO;
architecture rtl of FIFO is
type t_FIFO_DATA is array (0 to FIFO_DEPTH) of std_logic_vector(FIFO_WIDTH-1 downto 0);
signal r_FIFO_DATA : t_FIFO_DATA := (others => (others => '0'));
signal r_WR_INDEX : integer range 0 to FIFO_DEPTH -1 := 0;
signal r_RD_INDEX : integer range 0 to FIFO_DEPTH -1 := 0;
-- # Words in FIFO, has extra range to allow for assert conditions
signal r_FIFO_COUNT : integer range -1 to FIFO_DEPTH+1 := 0;
signal w_FULL : std_logic;
signal w_EMPTY : std_logic;
begin
-- FIFO process
-------------------------------------------------------------------
-------------------------------------------------------------------
WRITE_INDEX : process(clk)
begin
if rising_edge(clk) then
if reset = '1' then
r_WR_INDEX <= 1;
else
if (wr_en = '1' and w_FULL = '0') then
if r_WR_INDEX = FIFO_DEPTH-1 then
r_WR_INDEX <= 1;
else
r_WR_INDEX <= r_WR_INDEX + 1;
end if;
end if;
end if;
end if;
end process;
READ_INDEX : process(clk)
begin
if rising_edge(clk) then
if reset = '1' then
r_RD_INDEX <= 0;
else
if (rd_en = '1' and w_EMPTY = '0') then
if r_RD_INDEX = FIFO_DEPTH-1 then
r_RD_INDEX <= 0;
else
r_RD_INDEX <= r_RD_INDEX + 1;
end if;
end if;
end if;
end if;
end process;
COUNT_INDEX : process(clk)
begin
if rising_edge(clk) then
if reset = '1' then
r_FIFO_COUNT <= 0;
else
if (wr_en = '1' and rd_en = '0') then
r_FIFO_COUNT <= r_FIFO_COUNT + 1;
elsif (wr_en = '0' and rd_en = '1') then
if r_FIFO_COUNT > 0 then
r_FIFO_COUNT <= r_FIFO_COUNT - 1;
end if;
end if;
end if;
end if;
end process;
Write_Data : process (clk) is
begin
if rising_edge(clk) then
if wr_en = '1' then
r_FIFO_DATA(r_WR_INDEX) <= wr_data;
end if;
end if;
end process;
rd_data <= r_FIFO_DATA(r_RD_INDEX);
w_FULL <= '1' when r_FIFO_COUNT = FIFO_DEPTH else '0';
w_EMPTY <= '1' when r_FIFO_COUNT = 0 else '0';
ALMOST_FULL <= '1' when r_FIFO_COUNT > ALMOST_FULL_LEVEL else '0';
ALMOST_EMPTY <= '1' when r_FIFO_COUNT < ALMOST_EMPTY_LEVEL else '0';
FULL <= w_FULL;
EMPTY <= w_EMPTY;
end rtl;
TX-FSM code
-- Arinc 429 trasmitter interface
library IEEE;
use IEEE.STD_LOGIC_1164.ALL;
use IEEE.STD_LOGIC_ARITH.ALL;
use IEEE.STD_LOGIC_UNSIGNED.ALL;
entity Tx is
port
(
--INPUT
clk2M : in std_logic; -- clock signal
reset : in std_logic; -- reset signal
enable : in std_logic; -- enable signal
en_parity : in std_logic; -- enable parity bit
parity : in std_logic; -- odd/even parity
speed : in std_logic; -- speed 100kbps or 12.5kbps
gap : in std_logic; -- gap between two messages: 4 or 64 bit of gap
TX_FIFO_ep : in std_logic; -- TX FIFO EMPTY
a429TX_in : in std_logic_vector (31 downto 0); -- data in
--OUTPUT
a429TX_outA : out std_logic; -- positive out
a429TX_outB : out std_logic; -- negative out
TX_FIFO_rd : out std_logic -- TX FIFO READ
);
end entity;
architecture RTL_A429TX of Tx is
-- FSM state name
type state_type is (IDLE,START, PAR,TRANSMITTING,WAITING);
signal state : state_type;
-- FSM register
signal shift_reg : std_logic_vector (31 downto 0);
signal shift_counter : std_logic_vector (4 downto 0);
signal gap_counter : std_logic_vector (6 downto 0);
-- speed clock register
signal clk429 : std_logic;
signal clk429_counter : integer;
signal clk429_max_count : integer;
signal clk429_half_count : integer;
begin
-- speed clock process
-------------------------------------------------------------------
-------------------------------------------------------------------
-- select speed process
process (speed)
begin
if (speed = '1') then
clk429_max_count <= 19; -- 100kbs/s
clk429_half_count <= 10;
else
clk429_max_count <= 159; -- 12.5kbs/s
clk429_half_count <= 80;
end if;
end process;
-- clock429 generate speed process
process (clk2M, reset)
begin
if (reset = '1') then
clk429 <= '0';
elsif rising_edge(clk2M) then
if (clk429_counter <= clk429_half_count ) then
clk429 <= '1';
else
clk429 <= '0';
end if;
end if;
end process;
-- counter activity process
process (clk2M, reset)
begin
if (reset = '1') then
clk429_counter <= 0;
elsif rising_edge(clk2M) then
if (clk429_counter >= clk429_max_count) then
clk429_counter <= 0;
else
clk429_counter <= clk429_counter + 1;
end if;
end if;
end process;
-------------------------------------------------------------------
-------------------------------------------------------------------
-- a429TX interface process
process (clk429, reset)
variable p : std_logic;
begin
if reset = '1' then
state <= IDLE;
shift_reg <= (others => '0');
shift_counter <= (others => '0');
gap_counter <= (others => '0');
a429TX_outA <= '0';
a429TX_outB <= '0';
TX_FIFO_rd <= '0';
elsif rising_edge(clk429) then
case state is
when IDLE => -- idle state
if (enable = '1') then
if (gap = '1') then
gap_counter <= "0000100"; -- 4
else
gap_counter <= "1000000"; -- 64
end if;
if TX_FIFO_ep = '0' then
TX_FIFO_rd <= '1';
state <= START;
else
state <= IDLE;
end if;
else
state <= IDLE;
end if;
when START =>
-- data formatting
TX_FIFO_rd <= '0';
shift_reg <= a429TX_in(31 downto 8)& a429TX_in(0) & a429TX_in(1) & a429TX_in(2) & a429TX_in(3) & a429TX_in(4) & a429TX_in(5) & a429TX_in(6) & a429TX_in(7);
shift_counter <= "11111";
if ( en_parity = '1') then
state <= PAR;
else
state <= TRANSMITTING;
end if;
when PAR => -- parity state
--TX_FIFO_rd <= '0';
p := '0';
for I in 31 downto 0 loop
p := p xor shift_reg(I);
end loop;
if (parity = '1') then
shift_reg(31) <= p; -- odd
else
shift_reg(31) <= not p; -- even
end if;
state <= TRANSMITTING;
when TRANSMITTING => -- transmission state
--TX_FIFO_rd <= '0';
a429TX_outA <= shift_reg(0);
a429TX_outB <= not shift_reg(0);
shift_reg <= shift_reg(0) & shift_reg(31 downto 1);
if (shift_counter = "00000") then
state <= WAITING;
else
shift_counter <= shift_counter -1;
state <= TRANSMITTING;
end if;
when WAITING => -- wait state. generate gap
a429TX_outA <= '0';
a429TX_outB <= '0';
if (gap_counter > 0) then
gap_counter <= gap_counter - 1;
state <= WAITING;
else
state <= IDLE;
end if;
when others => -- default
state <= IDLE;
end case;
elsif falling_edge (clk429) then
a429TX_outA <= '0';
a429TX_outB <= '0';
end if;
end process;
clk429 <= clk429;
end architecture;
Thanks for your help.
Run both FIFOs at the 2 MHz clk2M, and then generate a single cycle enable indication on TX_FIFO_rd when FIFO read data transfer is required.
Thereby you can get the benefit from synchronous design, without the hazzle of handling multiple clock domains.
Also, it is not good (but actually very bad :-) synchronous design practice to generate internal clock like the clk429, since it results in error prune design and more complex timing closure with Static Timing Analysis (STA). Instead make an enable signal that is asserted a single cycle, run the design on the clk2M, and the only update the relevant state when the enable signal is high.
Now i make a circuit to measure temperature and humidity, then display on LCD. This is my code for DHT22, i use Elbert V2.
After genarating my project, it did not go right.
I tested and my program did not to come to "end_sl"( last state). And i dont know why?. Any suggestions for me? thank you.
my code
----------------------------------------------------------------------------------------------------------------------------------------------------------------
entity DHT11 is
generic (
CLK_PERIOD_NS : positive := 83; -- 12MHz
N: positive:= 40);
port(
clk,rst : in std_logic ;
singer_bus: inout std_logic;
dataout: out std_logic_vector (N-1 downto 0);
tick_done: out std_logic
);
end DHT11;
architecture Behavioral of DHT11 is
constant DELAY_1_MS: positive := 1*10**6/CLK_PERIOD_NS+1;
constant DELAY_40_US: positive := 40*10**3/CLK_PERIOD_NS+1;
constant DELAY_80_US: positive := 80*10**3/CLK_PERIOD_NS+1;
constant DELAY_50_US: positive := 50*10**3/CLK_PERIOD_NS+1; --
constant TIME_70_US: positive := 80*10**3/CLK_PERIOD_NS+1; --bit > 70 us
constant TIME_28_uS: positive := 30*10**3/CLK_PERIOD_NS+1; -- bit 0 > 28 us
constant MAX_DELAY : positive := 5*10**6/CLK_PERIOD_NS+1; -- 5 ms
type state_type is (reset,start_m,wait_res_sl,response_sl,delay_sl,start_sl,consider_logic,end_sl);
signal index, next_index : natural range 0 to MAX_DELAY;
signal state, next_state : state_type;
signal data_out,next_data_out: std_logic_vector (N-1 downto 0);
signal bit_in, next_bit_in: std_logic;
signal number_bit,next_number_bit: natural range 0 to 40;
signal oe: std_logic; -- help to set input and output port.
begin
--register
regis_state:process (clk,rst) begin
if rst = '1' then
state <= reset;
index <= MAX_DELAY;
number_bit <= 0;
bit_in <= '1';
data_out <= (others => '0');
elsif rising_edge(clk) then
state <= next_state;
index <= next_index;
number_bit <= next_number_bit;
bit_in <= next_bit_in;
data_out <= next_data_out;
end if;
end process regis_state;
proces_state: process (singer_bus,index,state,bit_in,number_bit,data_out) begin
tick_done <= '0';
next_data_out <= data_out;
next_number_bit <= number_bit;
next_state <= state;
next_data_out <= data_out;
next_index <= index;
dataout <= (others => '0');
oe <= '0';
next_bit_in <= bit_in;
case(state) is
when reset => -- initial
if index = 0 then
next_state <= start_m;
next_index <= DELAY_1_MS;
next_number_bit <= N-1;
else
next_state <= reset;
next_index <= index - 1;
end if;
when start_m => -- master send '1' in 1ms
if index = 0 then
next_state <= wait_res_sl;
next_index <= DELAY_40_US;
else
oe <= '1';
next_state <= start_m;
next_index <= index -1;
end if ;
when wait_res_sl => -- wait for slave response in 40us --
next_bit_in <= singer_bus;
if bit_in ='1' and next_bit_in = '0' then --
next_state <= response_sl;
else
next_state <= wait_res_sl;
end if;
when response_sl => -- slave response in 80us
next_bit_in <= singer_bus;
if bit_in ='0' and next_bit_in = '1' then
next_state <= delay_sl;
else
next_state <= response_sl;
end if;
when delay_sl => -- wait for slave delay in 80us
if bit_in = '1' and next_bit_in ='0' then
next_state <= start_sl;
else
next_state <= delay_sl;
end if;
when start_sl => -- start to prepare in 50us
if (bit_in = '0') and (next_bit_in = '1') then
next_state <= consider_logic;
next_index <= 0;
elsif number_bit = 0 then
next_state <= end_sl;
next_index <= DELAY_50_US;
else
next_state <= start_sl;
end if;
when consider_logic => -- determine 1 bit-data of slave
next_index <= index + 1;
next_bit_in <= singer_bus;
if bit_in = '1' and next_bit_in = '0' then -- the end of logic state
next_number_bit <= number_bit -1;
if (index < TIME_28_uS) then -- time ~ 28 us - logic = '0'
next_data_out <= data_out(N-2 downto 0) & '0';
elsif (index < TIME_70_US) then -- time ~70 us - logic ='1'
next_data_out <= data_out(N-2 downto 0) & '1';
end if;
next_state <= start_sl;
next_index <= DELAY_50_US;
elsif bit_in ='1' and next_bit_in ='1' then
next_state <= consider_logic;
end if;
when end_sl => -- tick_done = '1' then dataout has full 40 bit.
if index = 0 then
next_index <= MAX_DELAY;
next_state <= reset;
else
tick_done <= '1';
dataout <= data_out;
next_index <= index -1;
next_state <= end_sl;
end if;
end case;
end process proces_state;
--tristate IOBUFFER
singer_bus <= '0' when oe ='1' else 'Z';
end Behavioral;
There are many errors in your code. How did you debug exactly? Because it seems like you did not.
Why wait for 60 ms after the reset? you waste (valuable) simulation time. 6 ms is more then enough.
Looking at the simulation output, you can see the state does not advance at all: it's stuck ini wait_res_sl. The problem is that you have not added all the signals read in the process to the sensitivity list. I.e.
bit_in ='1' and next_bit_in = '0'
Will not detect a change if next_bit_in is not in the sensitivity list.
A problem -a common mistake made- is that your 'test bench' only provides input stimuli.... But it does not actually test anything.
And then the counters. Why is the delay counter called index? It doesn't index anything.
Why do your time delays not match their label? 70us -> 80 us. 28us -> 30 us.
Small thing don't call a RTL architecture behavioral
I tried to clean your code, seems to work now.
library ieee;
use ieee.std_logic_1164.all;
entity dht2 is
generic (
clk_period_ns : positive := 83; -- 12mhz
data_width: positive:= 40);
port(
clk,rst : in std_logic ;
singer_bus: inout std_logic;
dataout: out std_logic_vector(data_width-1 downto 0);
tick_done: out std_logic
);
end entity;
architecture rtl of dht2 is
constant delay_1_ms: positive := 1*10**6/clk_period_ns+1;
constant delay_40_us: positive := 40*10**3/clk_period_ns+1;
constant delay_80_us: positive := 80*10**3/clk_period_ns+1;
constant delay_50_us: positive := 50*10**3/clk_period_ns+1; --
constant time_70_us: positive := 70*10**3/clk_period_ns+1; --bit > 70 us
constant time_28_us: positive := 28*10**3/clk_period_ns+1; -- bit 0 > 28 us
constant max_delay : positive := 5*10**6/clk_period_ns+1; -- 5 ms
signal input_sync : std_logic_vector(0 to 2);
type state_type is (reset,start_m,wait_res_sl,response_sl,delay_sl,start_sl,consider_logic,end_sl);
signal state : state_type;
signal delay_counter : natural range 0 to max_delay;
signal data_out : std_logic_vector (data_width-1 downto 0);
signal bus_rising_edge, bus_falling_edge : boolean;
signal number_bit : natural range 0 to data_width;
signal oe: std_logic; -- help to set input and output port.
begin
input_syncronizer : process(clk) begin
if rising_edge(clk) then
input_sync <= to_x01(singer_bus)&input_sync(0 to 1);
end if;
end process;
bus_rising_edge <= input_sync(1 to 2) = "10";
bus_falling_edge <= input_sync(1 to 2) = "01";
--register
regis_state:process (clk) begin
if rising_edge(clk) then
case(state) is
when reset => -- initial
if delay_counter = 0 then
number_bit <= data_width;
oe <= '1';
delay_counter <= delay_1_ms;
state <= start_m;
else
delay_counter <= delay_counter - 1;
end if;
when start_m => -- master send '1' in 1ms
if delay_counter = 0 then
oe <= '0';
delay_counter <= delay_40_us;
state <= wait_res_sl;
else
delay_counter <= delay_counter -1;
end if ;
when wait_res_sl => -- wait for slave response in 40us --
if bus_falling_edge then --
state <= response_sl;
end if;
when response_sl => -- slave response in 80us
if bus_rising_edge then
state <= delay_sl;
end if;
when delay_sl => -- wait for slave delay in 80us
if bus_falling_edge then
state <= start_sl;
end if;
when start_sl => -- start to prepare in 50us
if bus_rising_edge then
delay_counter <= 0;
state <= consider_logic;
elsif number_bit = 0 then
delay_counter <= delay_50_us;
state <= end_sl;
end if;
when consider_logic => -- determine 1 bit-data of slave
if bus_falling_edge then -- the end of logic state
number_bit <= number_bit - 1;
if (delay_counter < time_28_us) then -- time ~ 28 us - logic = '0'
data_out <= data_out(data_width-2 downto 0) & '0';
elsif (delay_counter < time_70_us) then -- time ~70 us - logic ='1'
data_out <= data_out(data_width-2 downto 0) & '1';
end if;
delay_counter <= delay_50_us;
state <= start_sl;
end if;
delay_counter <= delay_counter + 1;
when end_sl => -- tick_done = '1' then dataout has full 40 bit.
if delay_counter = 0 then
delay_counter <= max_delay;
state <= reset;
else
tick_done <= '1';
dataout <= data_out;
delay_counter <= delay_counter - 1;
end if;
end case;
if rst = '1' then
number_bit <= 0;
data_out <= (others => '0');
delay_counter <= max_delay;
state <= reset;
end if;
end if;
end process regis_state;
--tristate iobuffer
singer_bus <= '0' when oe ='1' else 'Z';
end architecture;
And test bench: I added one check, but you should make more checks: every time you do something, it should have an effect. You should test if that effect actually happens.
entity dht2_tb is end dht2_tb;
library ieee;
architecture behavior of dht2_tb is
use ieee.std_logic_1164.all;
--inputs
signal clk : std_logic := '0';
signal rst : std_logic := '0';
--bidirs
signal singer_bus : std_logic := 'H';
--outputs
signal tick_done : std_logic;
-- clock period definitions
constant clk_period : time := 83.33 ns; -- 12mhz
use ieee.math_real.all;
-- This function generates a 'slv_length'-bit std_logic_vector with
-- random values.
function random_slv(slv_length : positive) return std_logic_vector is
variable output : std_logic_vector(slv_length-1 downto 0);
variable seed1, seed2 : positive := 65; -- required for the uniform function
variable rand : real;
-- Assume mantissa of 23, according to IEEE-754:
-- as UNIFORM returns a 32-bit floating point value between 0 and 1
-- only 23 bits will be random: the rest has no value to us.
constant rand_bits : positive := 23;
-- for simplicity, calculate remaining number of bits here
constant end_bits : natural := slv_length rem rand_bits;
use ieee.numeric_std.all;
begin
-- fill sets of 23-bit of the output with the random values.
for i in 0 to slv_length/rand_bits-1 loop
uniform(seed1, seed2, rand); -- create random float
-- convert float to int and fill output
output((i+1)*rand_bits-1 downto i*rand_bits) :=
std_logic_vector(to_unsigned(integer(rand*(2.0**rand_bits)), rand_bits));
end loop;
-- fill final bits (< 23, so above loop will not work.
uniform(seed1, seed2, rand);
if end_bits /= 0 then
output(slv_length-1 downto slv_length-end_bits) :=
std_logic_vector(to_unsigned(integer(rand*(2.0**end_bits)), end_bits));
end if;
return output;
end function;
-- input + output definitions
constant test_data_length : positive := 32;
constant test_data : std_logic_vector(test_data_length-1 downto 0) := random_slv(test_data_length);
signal data_out : std_logic_vector(test_data_length-1 downto 0);
begin
-- instantiate the unit under test (uut)
uut: entity work.dht2 -- use entity instantiation: no component declaration needed
generic map (
clk_period_ns => clk_period / 1 ns,
data_width => test_data_length)
port map (
clk => clk,
rst => rst,
singer_bus => singer_bus,
dataout => data_out,
tick_done => tick_done
);
-- clock stimuli
clk_process: process begin
clk <= '0', '1' after clk_period/2;
wait for clk_period;
end process;
-- reset stimuli
rst_proc : process begin
rst <= '1', '0' after 100 us;
wait;
end process;
-- bidir bus pull-up
-- as you drive the bus from the uut and this test bench, it is a bidir
-- you need to simulate a pull-up ('H' = weak '1'). slv will resolve this.
singer_bus <= 'H';
-- stimulus process
bus_proc: process
-- we use procedures for stimuli. Increases maintainability of test bench
-- procedure bus_init initializes the slave device. (copied this from your code)
procedure bus_init is begin
-- singer_bus <= 'Z'; -- initial
wait for 6 ms;
-- singer_bus <= '0'; -- master send
-- wait for 1 ms;
singer_bus <= 'Z'; -- wait response for slave
wait for 40 us;
singer_bus <= '0'; -- slave pull low
wait for 80 us;
singer_bus <= 'Z'; -- slave pull up
wait for 80 us;
end procedure;
function to_string(input : std_logic_vector) return string is
variable output : string(1 to input'length);
variable j : positive := 1;
begin
for i in input'range loop
output(j) := std_logic'image(input(i))(2);
j := j + 1;
end loop;
return output;
end function;
-- procedure send_data
procedure send_data(data : std_logic_vector) is begin
-- we can now send a vector of data,length detected automatically
for i in data'range loop
singer_bus <= '0'; -- slave start data transmission
wait for 50 us;
singer_bus <= 'Z'; -- slave send bit;
-- I found the only difference between sending bit '0'
-- and '1' is the length of the delay after a '0' was send.
case data(i) is
when '0' => wait for 24 us;
when '1' => wait for 68 us;
when others =>
report "metavalues not supported for bus_proc send_data"
severity failure;
end case;
singer_bus <= '0';
end loop;
-- next is VHDL-2008 (else use ieee.std_logic_textio.all;)
report "transmitted: "&to_string(data);
end procedure;
begin
wait until rst = '0';
bus_init; -- call procedure
send_data(test_data); -- call procedure
wait for 100 us; -- final delay
singer_bus <= 'Z'; -- release bus
report "received: "&to_string(data_out);
-- test correctness of output
assert data_out = test_data
report "data output does not match send data"
severity error;
report "end of simulation" severity failure;
end process;
end architecture;
I have to do UART with vhdl on the Xilinx which will send 16 chars string. I wrote such code
library IEEE;
use IEEE.STD_LOGIC_1164.ALL;
use IEEE.STD_LOGIC_UNSIGNED.ALL;
use ieee.numeric_std.ALL;
entity uartByJackob is
Port ( CLK, A, B, C : in STD_LOGIC;
RESET : in STD_LOGIC;
TxD, TxDOSC : out STD_LOGIC);
end uartByJackob;
architecture Behavioral of uartByJackob is
signal K: std_logic_vector(14 downto 0);
signal Q: std_logic_vector(3 downto 0);
signal CLK_Txd: std_logic;
signal ENABLE: std_logic;
signal QTxD: std_logic_vector(9 downto 0);
signal DATA : STD_LOGIC_VECTOR(7 downto 0);
-- freq of clock
begin
process(CLK, RESET)
begin
if rising_edge(CLK) then
if(A = '1' and K < 10416) then
K <= K + 1;
CLK_Txd <= K(13);
elsif(B = '1' and K < 5208) then
K <= K + 1;
CLK_Txd <= K(12);
elsif(C = '1' and K < 20832) then
K <= K + 1;
CLK_Txd <= K(14);
else
K <= (others => '0');
end if;
end if;
end process;
--counter
process(CLK_Txd, RESET, ENABLE)
begin
if(RESET = '1' and ENABLE = '0') then
Q <= "0000";
elsif (rising_edge(CLK_Txd)) then
Q <= Q + 1;
end if;
end process;
--comparator
ENABLE <= '1' when (Q > 4) else '0';
--transcoder
process(Q, CLK_Txd)
begin
if (rising_edge(CLK_Txd)) then
case Q is
when "0001" => DATA <= x"40";
when "0010" => DATA <= x"41";
when "0011" => DATA <= x"42";
when "0100" => DATA <= x"43";
when "0101" => DATA <= x"44";
when "0110" => DATA <= x"45";
when "0111" => DATA <= x"46";
when "1000" => DATA <= x"47";
when "1001" => DATA <= x"48";
when "1010" => DATA <= x"49";
when "1011" => DATA <= x"50";
when "1100" => DATA <= x"51";
when "1101" => DATA <= x"52";
when "1110" => DATA <= x"53";
when "1111" => DATA <= x"54";
when others => DATA <= x"55";
end case;
end if;
end process;
--uart
process(CLK_Txd, ENABLE, DATA)
begin
if(ENABLE = '0') then
QTxD <= DATA & "01";
elsif rising_edge(CLK_Txd) then
QTxD <= '1'&QTxD(9 downto 1);
end if;
end process;
TxD <= QTxD(0);
TxDOSC <= QTxD(0);
end Behavioral;
It's send data completely not connected with that what i have in transcoder and realy dont know why. Do you have any ideas what is wrong with my code, or do you have any diffrent examples of it how to send your own 16 chars with uart? I suppose that something is wrong with my counter or comparator.
--EDIT
Thans for your effort, i can't try your code at the Xilinx right now couse I am workin on it at my university. I see that you made a lot of changes in my code. Of course first i try to do it like you show and i hope this will be acceptable, but I propably have to do it with transcoder according to this picture.
From last time i made such changes i my code
library IEEE;
use IEEE.STD_LOGIC_1164.ALL;
use IEEE.STD_LOGIC_UNSIGNED.ALL;
use ieee.numeric_std.ALL;
entity uartByJackob is
Port ( CLK, A, B, C : in STD_LOGIC;
RESET : in STD_LOGIC;
TxD, TxDOSC : out STD_LOGIC);
end uartByJackob;
architecture Behavioral of uartByJackob is
signal K: std_logic_vector(14 downto 0);
signal Q: std_logic_vector(7 downto 0);
signal CLK_Txd: std_logic;
signal ENABLE: std_logic;
signal QTxD: std_logic_vector(7 downto 0);
signal DATA : STD_LOGIC_VECTOR(7 downto 0);
signal QPrim: std_logic_vector(3 downto 0);
begin
process(CLK, RESET)
begin
CLK_Txd <= CLK;
end process;
process(CLK_Txd, RESET, ENABLE)
begin
if(ENABLE = '0') then
Q <= "00000000";
elsif (rising_edge(CLK_Txd)) then
Q <= Q + 1;
end if;
end process;
ENABLE <= '1' when (Q <= 255) else '0';
process(Q(7 downto 4))
begin
case Q(7 downto 4) is
when "0000" => DATA <= x"40";
when "0001" => DATA <= x"41";
when "0010" => DATA <= x"42";
when "0011" => DATA <= x"43";
when "0100" => DATA <= x"44";
when "0101" => DATA <= x"45";
when "0110" => DATA <= x"46";
when "0111" => DATA <= x"47";
when "1000" => DATA <= x"48";
when "1001" => DATA <= x"49";
when "1010" => DATA <= x"50";
when "1011" => DATA <= x"51";
when "1100" => DATA <= x"52";
when "1101" => DATA <= x"53";
when "1110" => DATA <= x"54";
when "1111" => DATA <= x"55";
when others => DATA <= x"56";
end case;
end process;
process(CLK_Txd, ENABLE, DATA)
begin
if(ENABLE = '1') then
QTxD <= DATA;
elsif rising_edge(CLK_Txd) then
QTxD <= '1'&QTxD(7 downto 1);
end if;
end process;
TxD <= QTxD(0);
TxDOSC <= QTxD(0);
end Behavioral;
According to that i send MSB to transcoder and LSB to comparator but my program all the time still send x"40" to DATA and it is propably connected with this counter which you were talking about.
There is my simulation efect. I becoming upset with that couse i don't have enough skills in vhdl to do it by my self. I hope that you will help me to do rebuild my project. On simulation it looks good i dont know how it looks on Xilinx.
Can you show me a piece of code? - Stefan
The entire purpose to providing the link to Adrian Adamcyzk's code (Altera FPGA hardware (has an issue) vs ModelSim simulation (ok) - self implemented UART) was to provide an example with a bit (baud) counter and flip flop used to control sending the message once.
Here's Jackob's modified:
library ieee;
use ieee.std_logic_1164.all;
-- use ieee.std_logic_unsigned.all;
use ieee.numeric_std.all;
entity uartbyjackob is
port (
clk, a, b, c: in std_logic;
reset: in std_logic;
txd, txdosc: out std_logic
);
end entity uartbyjackob;
architecture foo of uartbyjackob is
-- signal k: unsigned(14 downto 0); -- FOR simulation
-- note if k were used in simulation it would require initialization
signal q: unsigned (3 downto 0); -- WAS std_logic_vector
signal clk_txd: std_logic;
signal enable: std_logic;
signal qtxd: std_logic_vector(9 downto 0);
-- signal data: std_logic_vector(7 downto 0);
-- added:
signal bdcnt: unsigned (3 downto 0);
signal ldqtxd: std_logic;
signal davl: std_logic;
type data_lut is array (0 to 15) of std_logic_vector (7 downto 0);
constant data: data_lut := (
x"40", x"41", x"42", x"43", x"44", x"45", x"46", x"47",
x"48", x"49", x"50", X"51", x"52", X"53", x"54", x"55"
);
signal datalut: std_logic_vector (7 downto 0); -- FOR SIMULATION visibility
begin
-- -- freq of clock -- NOTE k never in known binary state for simulation
-- process (clk, reset)
-- begin
-- if rising_edge(clk) then
-- if a = '1' and k < 10416 then
-- k <= k + 1;
-- clk_txd <= k(13);
-- elsif b = '1' and k < 5208 then
-- k <= k + 1;
-- clk_txd <= k(12);
-- elsif c = '1' and k < 20832 then
-- k <= k + 1;
-- clk_txd <= k(14);
-- else
-- k <= (others => '0');
-- end if;
-- end if;
-- end process;
clk_txd <= clk; -- SHORTENS SIMULATION
DAVL_FF: -- DATA_AVAILABLE to send
process (clk_txd, reset)
begin
if reset = '1' then
davl <= '0';
elsif rising_edge (clk_txd) then
if q = 15 and bdcnt = 9 then -- a JK FF equivalent
davl <= '0';
elsif q = 0 then
davl <= '1'; -- one clock holderover from reset
-- else
-- davl <= davl;
end if;
end if;
end process;
-- process(clk_txd, reset, enable)
-- begin
-- if reset = '1' and enable = '0' then
-- q <= "0000";
-- elsif rising_edge(clk_txd) then
-- q <= q + 1;
-- end if;
-- end process;
QCNT:
process (clk_txd, reset)
begin
if reset = '1' then
q <= (others => '0');
elsif rising_edge (clk_txd) then
if enable = '1' then
q <= q + 1;
end if;
end if;
end process;
BAUD_COUNTER:
process (clk_txd, reset)
begin
if reset = '1' then
bdcnt <= (others => '0');
elsif rising_edge (clk_txd) then
if davl = '0' or bdcnt = 9 then
bdcnt <= (others => '0');
else
bdcnt <= bdcnt + 1;
end if;
end if;
end process;
-- comparator
-- enable <= '1' when (q > 4) else '0';
enable <= '1' when bdcnt = 9 and davl = '1' and q /= 15 else
'0';
-- q latches at 15;
ldqtxd <= '1' when bdcnt = 9 and davl = '1' else
'0';
datalut <= data(to_integer(q)); -- FOR SIMULATION VISIBILITIY
--transcoder
-- process(q, clk_txd)
-- begin
-- if rising_edge(clk_txd) then
-- case q is
-- when "0001" => data <= x"40";
-- when "0010" => data <= x"41";
-- when "0011" => data <= x"42";
-- when "0100" => data <= x"43";
-- when "0101" => data <= x"44";
-- when "0110" => data <= x"45";
-- when "0111" => data <= x"46";
-- when "1000" => data <= x"47";
-- when "1001" => data <= x"48";
-- when "1010" => data <= x"49";
-- when "1011" => data <= x"50";
-- when "1100" => data <= x"51";
-- when "1101" => data <= x"52";
-- when "1110" => data <= x"53";
-- when "1111" => data <= x"54";
-- when others => data <= x"55";
-- end case;
-- end if;
-- end process;
-- uart
-- process (clk_txd, enable, data)
-- begin
-- if enable = '0' then
-- qtxd <= data & "01";
-- elsif rising_edge(clk_txd) then
-- qtxd <= '1' & qtxd(9 downto 1);
-- end if;
-- end process;
TX_SHIFT_REG:
process (clk_txd, reset) -- shift regiseter Tx UART
begin
if reset = '1' then
qtxd <= (others => '1'); -- output mark by default
elsif rising_edge (clk_txd) then
if ldqtxd = '1' then
qtxd <= '1' & data(to_integer(q)) & '0';
-- STOP & Data(q) 7 downto 0 & START , a MUX and expansion
else
qtxd <= '1' & qtxd(9 downto 1); -- shift out;
end if;
end if;
end process;
txd <= qtxd(0);
txdosc <= qtxd(0);
end architecture foo;
library ieee;
use ieee.std_logic_1164.all;
entity uartbyjackob_tb is
end entity;
architecture foo of uartbyjackob_tb is
signal clk: std_logic := '0';
signal reset: std_logic := '0';
signal txd: std_logic;
begin
DUT:
entity work.uartbyjackob
port map (
clk => clk, -- clk_txd driven by clk
a => 'X',
b => 'X',
c => 'X', -- a, b, c aren't used
reset => reset,
txd => txd,
txdosc => open
);
CLOCK:
process
begin
wait for 52.35 us;
clk <= not clk;
if now > 20000 us then
wait;
end if;
end process;
STIMULUS:
process
begin
wait for 104.7 us;
reset <= '1';
wait for 104.7 us;
reset <= '0';
wait;
end process;
end architecture;
The model has been modified for faster simulation, ignoring the baud rate clock generator.
There's an added flip flop (davl) for enabling the UART to run. There's an added baud (bit) counter bdcnt.
I changed the order of the start, stop and data values loaded into QTxD so the start bit came out first, followed by 8 data bits and the stop bit.
You can read off TxD from left to right start bit, data(q)(0) ... data(q(7), stop bit. The enable or ldqtxd will occur at the same time as a stop bit.
There's only one observable draw back to this implementation, if you reset while a value in the shift register hasn't finished loading you'll cause a framing error for the receiver. Don't reset it for 10 baud times after davl goes false.
The simulation is shown with a 9600 baud clk_txd, the characters go out back to back.
It has fewer flip flops than the original (disregarding k). There is no data register separate from QTxD ( - 8 FFs) plus bdcnt (+ 4) plus davl (+ 1). There are two comparisons (optimized to two) bdcnt = 9, q =, /= 9. Those could be expressed separately so it doesn't require optimization during synthesis.
I changed the look up table style, a matter of personal preference also the excuse for changing counters to type unsigned and using only package numeric_std for arithmetic.
The little testbench likewise doesn't expect the k counter to generate the baud clock.
Running the testbench gives:
Where there's an added signal datalut to show the value being shifted out after ldqtxd.
After your change making the q counter (7 downto 0)
We still see from your waveform that it doesn't work.
This is due to the enable and the shift register.
If you use a single counter with the upper four bits indexing the output character your character is transmitted in 10 out of the 16 clk_txd times indexed by the lower four bits of the counter. The remaining clock times TxD should be '1' (idle line marks in RS-232 parlance).
The order for data to be transmitted will be a space (the start bit), data(0) through data(7) and a mark (the stop bit). (Shown left to right on TxD).
For simulation the k counter is not used. I included it commented out below.
I made several changes for proper simulation. These include synchronously loading the shift register containing QTxD, synchronously clearing the rightmost bit of QTxD to provide a full width and moving enable to occur once every sixteen clocks (clk_txd). The enable is preceded by a new clear for the start bit and both been offset to prevent it from occurring during reset which has the effect of causing a framing error on the first character for any receiver.
Simulation is done with the same testbench I provide above.
The changes to your new code are shown by comments:
architecture behavioral of uartbyjackob is
-- signal k: std_logic_vector(14 downto 0);
signal q: unsigned (7 downto 0); -- std_logic_vector(7 downto 0);
signal clk_txd: std_logic;
signal enable: std_logic;
signal qtxd: std_logic_vector(7 downto 0);
-- using an 8 bit shift register requires a method of outputting a
-- synchronous start bit (the width is important for receive framing)
-- and synchronous stop bit
signal data: std_logic_vector(7 downto 0);
signal qprim: std_logic_vector(3 downto 0);
signal clear: std_logic; -- synchronous clear for start bit
begin
-- let's keep this here for when you put it the FPGA
-- -- freq of clock -- NOTE k never in known binary state for simulation
-- process (clk, reset)
-- begin
-- if rising_edge(clk then
-- if a = '1' and k < 10416 then
-- k <= k + 1;
-- clk_txd <= k(13);
-- elsif b = '1' and k < 5208 then
-- k <= k + 1;
-- clk_txd <= k(12);
-- elsif c = '1' and k < 20832 then
-- k <= k + 1;
-- clk_txd <= k(14);
-- else
-- k <= (others => '0');
-- end if;
-- end if;
-- end process;
process (clk) -- , reset)
begin
clk_txd <= clk; -- if simply a concurrent assignment statement this
end process; -- would look similar to the elaborated equivalent
-- process. The difference, no sensitivity list and
-- an explict wait on clk statement at the end.
-- This process wants to be removed and replaced by
-- the above commented out process for synthesis
process (clk_txd, reset) -- , reset, enable) -- enable a reset?
begin
-- if enable = '0' then
if reset = '1' then -- puts q counter in known state for simulation
q <= "00000000";
elsif rising_edge(clk_txd) then
if q /= 255 then -- stop after sending once
q <= q + 1;
end if;
end if;
end process;
-- enable <= '1' when q <= 255 else '0'; -- this appears incorrect
enable <= '1' when q(3 downto 0) = "0010" else
'0';
clear <= '1' when q(3 downto 0) = "0001" else
'0';
-- USING ONE COUNTER requires some clocks output MARKS
-- (idle bits) each 16 clocks. It requires the load (enable)
-- occur once every 16 clocks.
-- q(3 downto 0) is selected for enable to prevent outputting spaces
-- TxD during reset (q is reset to all '0's). This would cause a receive
-- framing error.
process (q(7 downto 4))
begin
case q(7 downto 4) is
when "0000" => data <= x"40";
when "0001" => data <= x"41";
when "0010" => data <= x"42";
when "0011" => data <= x"43";
when "0100" => data <= x"44";
when "0101" => data <= x"45";
when "0110" => data <= x"46";
when "0111" => data <= x"47";
when "1000" => data <= x"48";
when "1001" => data <= x"49";
when "1010" => data <= x"50";
when "1011" => data <= x"51";
when "1100" => data <= x"52";
when "1101" => data <= x"53";
when "1110" => data <= x"54";
when "1111" => data <= x"55";
when others => data <= x"56";
end case;
end process;
process (clk_txd) -- , enable, data) -- synchronous enable and clear
begin
-- if enable = '1' then -- this appears incorrect
-- qtxd <= data;
if reset = '1' then
qtxd <= (others => '1'); -- outputs mark after reset
elsif rising_edge(clk_txd) then
if clear = '1' then -- synchronous clear for start bit
qtxd(0) <= '0';
elsif enable = '1' then -- synchronous load
qtxd <= data;
else
qtxd <= '1' & qtxd(7 downto 1); -- shift right
end if;
end if;
end process;
-- the synchronous load prevents the first start bit from being stretched
-- q(3 downto 0) the following in hex notation
-- q(3 downto 0) = 2 is the start bit
-- = 3 is data(0)
-- ...
-- = A is data(7)
-- = B is the stop bit
-- = C - 1 are mark (idle) bits (q(3 downto 0) rolls over)
-- = 1 enable occurs loading qtxd
--
-- The offset is caused by synchronous load (1 clk_txd) and the load point
-- (q(3 downto 0) = 1 in enable term).
--
-- The load point wants to occur in the first 6 counts of q(3 downto 0) to
-- insure a trailing mark when q is stopped.
--
-- q(3 downto 0) = 1 is selected for enable to prevent spurious spaces
-- during reset from causing a receive framing error.
txd <= qtxd(0);
txdosc <= qtxd(0);
end architecture behavioral;
The comment table:
-- the synchronous load prevents the first start bit from being stretched
-- q(3 downto 0) the following in hex notation
-- q(3 downto 0) = 2 is the start bit
-- = 3 is data(0)
-- ...
-- = A is data(7)
-- = B is the stop bit
-- = C - 1 are mark (idle) bits (q(3 downto 0) rolls over)
-- = 1 enable occurs loading qtxd
--
-- The offset is caused by synchronous load (1 clk_txd) and the load point
-- (q(3 downto 0) = 1 in enable term).
--
-- The load point wants to occur in the first 6 counts of q(3 downto 0) to
-- insure a trailing mark when q is stopped.
--
-- q(3 downto 0) = 1 is selected for enable to prevent spurious spaces
-- during reset from causing a receive framing error.
tells you where to find bits of the data(q(7 downto 0)) selected character. In the following waveform q is shown as hex to match:
You'll find with the fixes the first character transmitted is 0x40, the second 0x41,...
I have an sensor it has an unlocked byte sequence which needs to be sent to it to unlock it and then it can receive the other command data.
The sensor receive data at a baudrate of 115200 bps, 8 data bits, even parity, 2 stop bits.
and before receiving any command data( used to set parameters) It needs to recieve d4 (hexadecimal number, byte) 10 times at an interval of 1ms.
I send the d4 converted into bits 11010100 added with parity and stop bits becomes 11010100011 to the TX port of uart at the baud rate of 115200 but how to create a delay between two d4 data byte sent ? I am writing the code if not clear please let me know I would put more details.
entity Uart_tx is
port (
TX : out std_logic;
clk_in : in std_logic;
but_div_clk : out std_logic;
clk_in_2 : in std_logic
);
end Uart_tx;
architecture Behavioral of Uart_tx is
signal tx_clk : std_logic := '0';
signal clk_1Khz : std_logic := '0';
signal q : unsigned(8 downto 0) := (others => '0');
signal p : unsigned(8 downto 0) := (others => '0');
type state_type is (idle, start);
signal state : state_type;
signal tick_in : std_logic := '0';
subtype byte is std_logic_Vector(7 downto 0);
type byte_array is array(natural range <>) of byte;
signal data_byte_array : byte_array(1 to 8);
-- signal curr_byte : std_logic_vector(7 downto 0);
signal byte_index : unsigned(2 downto 0) := "000";
subtype reg is std_logic_Vector(10 downto 0);
type reg_array is array(natural range <>) of reg;
signal TxDataReg_array : reg_array(1 to 8);
signal cur_Tx_reg : std_logic_vector(10 downto 0);
signal current_reg : unsigned(3 downto 0) := "0001";
signal count : unsigned (4 downto 0) := (others => '0');
signal count_d : unsigned (4 downto 0) := (others => '0');
signal sent_d4 : unsigned (3 downto 0) := (others => '0');
signal send_d4 : std_logic := '1';
signal D_4 : std_logic_vector(10 downto 0) :="11000101011";
begin
-- below are random entry ..actual data will come from slv_reg registers.
data_byte_array(1) <= "10101010"; -- slv_reg0(7 downto 0);
data_byte_array(2) <= "10101011"; -- slv_reg0(15 downto 8);
data_byte_array(3) <= "10101010"; -- slv_reg0(23 downto 16);
data_byte_array(4) <= "10101011"; -- slv_reg0(31 downto 24);
data_byte_array(5) <= "10101010"; -- slv_reg1(39 downto 32);
data_byte_array(6) <= "10101011"; -- slv_reg1(47 downto 40);
data_byte_array(7) <= "10101010"; -- slv_reg1(55 downto 48);
data_byte_array(8) <= "10101011"; -- slv_reg1(63 downto 56);
tick_in <= '1';
---------------------------------------Clk_div-----------------------------------------
process ( clk_in ) is
begin
if clk_in'event and clk_in = '1' then
q <= q + 1;
tx_clk <= q(8); --- 58.gdfg/2^8 =~ 230Khz baud rate = 115200
but_div_clk <= tx_clk;
end if;
end process;
---------------------------------------Clk_div------------------------------------------
---------------------------------------Clk_div------------------------------------------
process( clk_in_2 ) is
begin
if clk_in_2'event and clk_in_2 = '1' then
p <= p + 1;
clk_1Khz <= p(7);
end if;
end process;
---------------------------------------------------------------------------------------
--------------------------------------TX_Process----------------------------------------
process( state, tx_clk , tick_in) is
variable parity : std_logic := '0';
variable curr_byte : std_logic_vector(7 downto 0) := (others => '0');
begin
case state is
when idle => TX <= '1';
if tick_in = '1' then
state <= start;
else
TX <= '1';
end if;
when start =>
if send_d4 = '1' then
if (rising_edge(clk_1Khz)) then
case count_d is
when "00000" => TX <= D_4(0);
when "00001" => TX <= D_4(1);
when "00010" => TX <= D_4(2);
when "00011" => TX <= D_4(3);
when "00100" => TX <= D_4(4);
when "00101" => TX <= D_4(5);
when "00110" => TX <= D_4(6);
when "00111" => TX <= D_4(7);
when "01000" => TX <= D_4(8);
when "01001" => TX <= D_4(9);
when "01010" => TX <= D_4(10);
when others => TX <= '1';
end case;
count_d <= count_d +1;
sent_d4 <= sent_d4 + 1;
if to_integer(count_d) = 11 then
count_d <= "00000";
end if;
if to_integer(sent_d4) = 10 then
send_d4 <= '0' ;
end if;
end if;
else
for i in 1 to 8 loop
curr_byte := data_byte_array(i);
parity := '0';
for j in curr_byte'range loop
parity := parity xor curr_byte(j);
end loop;
if parity = '0' then
TxDataReg_array(i) <= "110" & curr_byte ;
else
TxDataReg_array(i) <= "111" & curr_byte ;
end if;
end loop;
cur_Tx_reg <= TxDataReg_array(to_integer(byte_index)+1);
byte_index <= byte_index + 1;
if rising_edge(tx_clk) then
case count is
when "00000" => TX <= cur_Tx_reg(0);
when "00001" => TX <= cur_Tx_reg(1);
when "00010" => TX <= cur_Tx_reg(2);
when "00011" => TX <= cur_Tx_reg(3);
when "00100" => TX <= cur_Tx_reg(4);
when "00101" => TX <= cur_Tx_reg(5);
when "00110" => TX <= cur_Tx_reg(6);
when "00111" => TX <= cur_Tx_reg(7);
when "01000" => TX <= cur_Tx_reg(8);
when "01001" => TX <= cur_Tx_reg(9);
when "01010" => TX <= cur_Tx_reg(10);
when others => TX <= '1';
end case;
count <= count+1;
if to_integer(count) = 11 then
count <= "00000";
state <= idle;
-- TX <= '1';
end if;
end if;
end if;
when others => TX <= '1';
end case;
end process;
end Behavioral;
To get a timed delay you have to implement a counter that ticks off a computed number of clock cycles equal to 1 ms. You then need to insert states into your FSM that activate the counter and wait for it to complete when needed. It is possible to manually calculate the counter value but you can make the tools do the work for you and avoid having magic numbers in your code.
library ieee;
use ieee.std_logic_1164.all;
use ieee.numeric_std.all;
...
constant CLOCK_FREQ : real := 50.0e6; -- 50 MHz system clock
constant SENSOR_DELAY : real := 1.0e-3; -- 1 ms delay
constant DELAY_COUNT : natural := integer(CLOCK_FREQ * SENSOR_DELAY);
-- This could be auto calculated with a ceil_log2() function
constant TIMER_SIZE : natural := 16;
signal timer : unsigned(TIMER_SIZE-1 downto 0);
constant DELAY_INIT : unsigned(timer'range)
:= to_unsigned(DELAY_COUNT, timer'length);
...
-- Initialize the timer sometime before you want the delay
timer <= DELAY_INIT;
...
-- Somewhere in your FSM
when WAIT_1MS =>
timer <= timer - 1;
if timer = 0 then
state <= WHATEVER_YOU_WANT_NEXT;
end if;
This method of using real constants to compute integer values is subject to rounding errors and general floating-point inaccuracies. For these sort of long delays, the small error (typically off-by-one) that could happen isn't usually of concern.
Note that you need to rework your state machine to follow more conventional patterns. You have created a process that mixes pure combinational logic with synchronous. You should not mix the two. You should not have the rising_edge() tests inside your FSM case statement but rather there should be a single if-block evaluating rising_edge() that contains your FSM.
If you need a clock sensitive process then it should only have the clock and an (optional) asynchronous reset in its sensitivity list. Any other pure combinational code should be put in a separate process. In a design such as this that shouldn't be necessary though.
VHDL allows what you have now but synthesis tools expect the use of a more limited style when describing the hardware. You risk having unexpected results if they can handle your code at all.
I've written code for a binary divider that takes in an 8 bit dividend, 3 bit divisor, and gives a 5 bit quotient (3 bit remainder). I've literally spent hours trying to fix a bug that gives incorrect results but I haven't been able to identify it. Any help would be GREATLY appreciated! I basically get wrong answers for my inputs but I can't figure out why. There is a bus that takes in values and on the first clock cycle where st is 1, the dividend register is loaded. On the second clock cycle after, the divisor register is loaded and the calculation is made for the next three clock cycles.
The V signal is the output to signify that an overflow has occured (the result can't be fit into the five bits of the quotient), my st is the start signal to start the process, sh is the shift signal for the shift register, su is the subtract signal for the subtractor.
library IEEE;
use IEEE.STD_LOGIC_1164.all;
use IEEE.STD_LOGIC_ARITH.all;
use IEEE.STD_LOGIC_UNSIGNED.all;
entity Divider is
Port (bus_in: in std_logic_vector(8 downto 0);
St, Clk, reset: in std_logic;
Quotient: out std_logic_vector(4 downto 0);
Remainder: out std_logic_vector(2 downto 0);
v: out std_logic);
end Divider;
architecture Behavioral of Divider is
signal State, NextState: integer range 0 to 5;
signal C, Ld1, Ld2, Su, Sh: std_logic;
signal Divisor: std_logic_vector(2 downto 0);
signal Subout: std_logic_vector(3 downto 0);
signal Dividend: std_logic_vector(8 downto 0);
begin
Subout <= Dividend(8 downto 5) - ('0' & divisor);
C <= not Subout (3);
Remainder <= Dividend(7 downto 5);
Quotient <= Dividend(4 downto 0);
State_Graph: process (State, St, C)
begin
Ld1 <= '0';
Ld2<='0';
v <= '0';
Sh <= '0';
Su <= '0';
case State is
when 0 =>
if (St = '1') then
Ld1 <= '1';
NextState <= 1;
else
NextState <= 0;
end if;
when 1 =>
if (St = '1') then
Ld2 <= '1';
NextState <= 2;
else
Ld2<='1';
NextState <= 2;
end if;
when 2 =>
if (C = '1') then
v <= '1';
NextState <= 0;
else
Sh <= '1';
NextState <= 3;
end if;
when 3 | 4 =>
if (C = '1') then
Su <= '1';
NextState <= State;
else
Sh <= '1';
NextState <= State + 1;
end if;
when 5 =>
if (C = '1') then
Su <= '1';
end if;
NextState <= 0;
end case;
end process State_Graph;
Update: process (Clk)
begin
if Clk'event and Clk = '1' then
State <= NextState;
--if Load = '1' then
-- Dividend <= '0' & bus_in;
--end if;
if Ld1 = '1' then
Dividend <= '0'&Bus_in(7 downto 0);
end if;
if Ld2 = '1' then
Divisor <= Bus_in(2 downto 0);
end if;
if Su = '1' then
Dividend(8 downto 5) <= Subout;
Dividend(0) <= '1';
end if;
if Sh = '1' then --94
Dividend <= Dividend(7 downto 0) & '0';
end if;
end if;
end process update;
end Behavioral;
Here's my input and outputs:
[Signals]: http://imgur.com/fqfiYJZ 1
The picture shows that my registers for the divisor and dividend is loading correctly. So I think the issue is with the actual division code. The state machine also seems to be working correctly.
Don't write this yourself. You are re-inventing the wheel.
Either write q <= a / b;
or use an IP core from your FPGA vendor.