sfixed to ufixed conversition

[fanwel]

Hi all;

I have learn about ufixed and sfixed operation. I want to convert -0.129 to sfixed using the code below and its has no problem;

signal n1: ufixed (4 downto -4);
...
out_A <= to_sfixed(-0.129,n1);

Then, I want to convert this out_A(sfixed value) to ufixed..How can I do like that?
Need helps..thanks

 

[TrickyDicky]

ufixed means unsigned fixed
sfixed means signed fixed.

Therefore assigning an sfixed to a ufixed is inappropriate. You need to either use an abs function or declare out_a as an sfixed.

 

[fanwel]

Sorry for the miss typing; signal n1: sfixed (4 downto -4);
How can I convert the sfixed to ufixed using abs function? Can you give any links or note on abs function to study..thanks for reply

---------- Post added at 03:25 ---------- Previous post was at 02:06 ----------

I write the code below and its work;
-------------------------------------------------------------
Library ieee;
USE ieee.std_logic_1164.all;
use ieee.numeric_std.all;
library ieee_proposed;
use ieee_proposed.fixed_pkg.all;

ENTITY try IS
PORT(
y : IN sfixed(4 downto -4);
Clock : IN STD_LOGIC;
alphaEstOut : OUT ufixed(5 downto -4));
END try;

ARCHITECTURE Behavior OF try IS
SIGNAL aa : sfixed(y'high +1 downto y'low);
BEGIN
PROCESS(Clock)
BEGIN
IF rising_edge(Clock) THEN

aa<=abs;
alphaEstOut<=ufixed(aa);

END IF;
END PROCESS;
END Behavior;

 

[fanwel]

Now, I proceed to next stage which is multiply the conversion output with ufixed values (input).
library ieee;
library ieee_proposed;
use ieee_proposed.fixed_pkg.all;

package my_data_types is
type vector is array (natural range <>) of integer;
type ufixed_array_t is array (0 to 3) of ufixed (9 downto -10);
end my_data_types;

library ieee;
library ieee_proposed;
use ieee_proposed.fixed_pkg.all;
use work.my_data_types.all;

entity mul_four is
port (clk: in bit;
A: in vector (0 to 3);
out_A: out ufixed_array_t);
end mul_four;

architecture mul_four of mul_four is
signal n1: sfixed (4 downto -4);
signal n2: sfixed (n1'high + 1 downto n1'low);
signal n3: ufixed (5 downto -4);
signal u: ufixed (5 downto -4);
begin
process(clk)
begin
if (clk'event and clk='1') then
n1 <= to_sfixed(-0.129,n1);
n2 <= abs(n1);
n3 <= ufixed(n2);
end if;

for i in 0 to 3 loop
out_A(i) <= (to_ufixed (A(i),u)) * n3;
end loop;
end process;
end mul_four;

There are no error when I compile in quartus, but then an error occur in ModelSim:
# Cannot continue because of fatal error. ( at line 36).

Can anyone check my code?..do I write in wrong way?..need helps

 

[TrickyDicky]

move the for loop inside the clocked part of the process.

Plus, did you realise you had the sizing of the arrays wrong? the assignment out_A(i) is basically doing a /4.

 

[fanwel]

Oh, I not realise the size of the array is wrong. Thanks for remind me..

---------- Post added at 09:39 ---------- Previous post was at 09:15 ----------

This is the architecture code:
...
architecture mul_four of mul_four is
signal n1: sfixed (4 downto -4);
signal n2: sfixed (n1'high + 1 downto n1'low);
signal n3: ufixed (5 downto -4);
signal u: ufixed (5 downto -4);

begin
n1 <= to_sfixed(-0.129,n1);
n2 <= abs(n1);
n3 <= ufixed(n2);

process(clk)
begin
if (clk'event and clk='1') then
for i in 0 to 3 loop
out_A(i) <= (to_ufixed (A(i),u)) * n3;
end loop;
end if;
end process;
end mul_four;

I have simulate in ModelSim and the output seem like correct. I just want to confirm,do I convert the sfixed to ufixed in the right way?..thanks for helps

 

[TrickyDicky]

Yes you did, but I dont understand the point of what you're doing. why bother using -0.129? you're just converting it to 0.129, so all results will be A(i) * 0.129.

why not just use ufixed all the time and use 0.129?

---------- Post added at 10:32 ---------- Previous post was at 10:31 ----------

this code is pointless:

n1 <= to_sfixed(-0.129,n1);
n2 <= abs(n1);
n3 <= ufixed(n2);

you might aswell just have:

out_A(i) <= to_ufixed(A(i), u) * to_ufixed(0.129, u);

 

[fanwel]

Sorry Tricky, actually I have one fixed value (-0.129) that will be multiply by an input of integer.
I follows the rule where the signed number need to convert to sfixed. Then convert it to ufixed for multiply with the integer (ufixed).
I not realise that without doing 2's complement I can directly multiply -0.129 with the integer using only ufixed. You have show me the simple way to do this multiply operation..thanks for help

 

[TrickyDicky]

if you want -0.129, you need sfixed. Out_A needs to be an array of sfixed.

then:

out_A(i) <= to_sfixed(A(i), u) * to_sfixed(-0.129, u);

 

[fanwel]

Both equation will give same final output right? I mean after convert it back to integer type..
1) out_A(i) <= to_ufixed(A(i), u) * to_ufixed(0.129, u);
2) out_A(i) <= to_sfixed(A(i), u) * to_sfixed(-0.129, u);

 

[TrickyDicky]

No.
1 will always be positive, because ufixed is treated as unsigned (ie. always positive). When you convert the integer to a ufixed, if A is negative, it will be converted to 0 (I think) or may just give you an error.
2 will be positive or negative, depending on the integer.

 

[fanwel]

For example I have (13 x -0.129) + (11 x 0.32);
What the better way to do this operation? Using sfixed or ufixed? Can you give suggestion..thanks for reply

 

[TrickyDicky]

if you need signed, you need sfixed.
if you're always going to have positive, use ufixed.

with your equation - everything needs to be sfixed.

 

[fanwel]

Thanks for your helps..I really appreciate.