Error when using math_real library

Hi
I am trying to use the acos function in the math_real library, but this gives the following error in the quartus 9.1 compiler:

Error (10482): VHDL error at Akselerometer.vhd(29): object "ACOS" is used but not declared

I know floating point numbers and fpga is not a good match, so maybe I should use look up tables instead(what is your opinion)?
My code is as follows:

Library ieee;
USE ieee.std_logic_1164.all;
use ieee.numeric_std.all;
library ieee_proposed;
use ieee_proposed.fixed_pkg.all;
use ieee.math_real.all;

ENTITY Akselerometer IS
PORT(
y : IN sfixed(7 downto -4);--Readed accelerometer value
Clock : IN STD_LOGIC;
alphaEstimated : OUT sfixed(7 downto -4));

END Akselerometer;

ARCHITECTURE Behavior OF Akselerometer IS
SIGNAL bias : sfixed(7 downto -4);
SIGNAL argCosREAL,alphaREAL : REAL;
SIGNAL argCos : sfixed(12 downto -10);

BEGIN
bias<=to_sfixed(1.586,bias);--fixed bias =1.586V
PROCESS--Start processing when a new value enters y-pins
BEGIN
IF (y>=bias)THEN
argCos<=bias/y;
argCosREAL<=to_REAL(argCos);
alphaREAL<=ACOS(argCosREAL);
alphaEstimated<=to_sfixed(alphaREAL,alphaEstimated);
ELSIF (y<bias) THEN
argCos<=y/bias;
argCosREAL<=to_REAL(argCos);
alphaREAL<=ACOS(argCosREAL);
alphaEstimated<=to_sfixed((alphaREAL),alphaEstimated);
END IF;
END PROCESS;
END Behavior;

I know floating point numbers and fpga is not a good match.

Click to expand...

The stament is missing the point. As a simple fact, IEEE.math_real isn't supported for synthesis at all. It's intended for simulation purposes, and you can use it with a design compiler like Quartus to calculate constants at compile time, and e.g. calculate a look-up-table. This applies to the real type in general as well to the functions provided by the library. Beside tables, CORDIC is a common way to generate trigonometric functions in FPGAs.

FvM said:

The stament is missing the point. As a simple fact, IEEE.math_real isn't supported for synthesis at all. It's intended for simulation purposes, and you can use it with a design compiler like Quartus to calculate constants at compile time, and e.g. calculate a look-up-table. This applies to the real type in general as well to the functions provided by the library. Beside tables, CORDIC is a common way to generate trigonometric functions in FPGAs.

Click to expand...

Hi
Tnx for replay, then I have to find another solution for the problem.

I think you can use fixed_pkg library alone for the calculation of cos_inverse. You just need to implement the equation given in this link:
Taylor series - Wikipedia, the free encyclopedia
for cos inverse.

The equation ideally contains infinite terms but since you just want only 5 bits fractional you can implement the function using first 4 or 5 terms.

vipinlal said:

I think you can use fixed_pkg library alone for the calculation of cos_inverse. You just need to implement the equation given in this link:
Taylor series - Wikipedia, the free encyclopedia
for cos inverse.

The equation ideally contains infinite terms but since you just want only 5 bits fractional you can implement the function using first 4 or 5 terms.

Click to expand...

Hi
Good idea


I wrote a conversion program using Taylor series, but when I tried to simulate it, I get the following warnings for every signal I tried to track. (only In and out signals is shown ok in the quartus 9.1 simulation report)

"Warning: Ignored node in vector source file. Can't find corresponding node name "n[0]" in design."

I obviously doing something wrong because it will not work on the evaluation board either.
Could it be the way I am treating y in the Process sensitivity list? I am using it like an interrupt, but is this the way to do it VHDL?

The code is as follows:

Library ieee;
USE ieee.std_logic_1164.all;
use ieee.numeric_std.all;
library ieee_proposed;
use ieee_proposed.fixed_pkg.all;


ENTITY Akselerometer IS
PORT(
y : IN sfixed(7 downto -4); --Readed accelerometer value
Clock : IN STD_LOGIC;
alphaEstOut : OUT sfixed(7 downto -4)); --Outputed accelerometer value (rad)


END Akselerometer;

ARCHITECTURE Behavior OF Akselerometer IS
SIGNAL Pi : sfixed(3 downto -5);
SIGNAL Factorial : sfixed(7 downto 0);
SIGNAL n : unsigned(2 downto 0);
SIGNAL nSfixed : sfixed(2 downto 0);
SIGNAL x : sfixed(7 downto -4);
SIGNAL bias : sfixed(7 downto -4);
SIGNAL binit,b,c,d,e,f : sfixed(7 downto -4);
SIGNAL alphaEst : sfixed(7 downto -4);
--For simulation purpose
SIGNAL FactorialSim :sfixed(7 downto 0);

BEGIN
bias<=to_sfixed(1.586,bias);--fixed bias =1.586V
Pi<=to_sfixed(3.1415,Pi); --Pi = 3.1415....
PROCESS--Start processing when a new value enters y-pins
BEGIN
Factorial<=to_sfixed(1,Factorial); --Initial value;
FactorialSim<=to_sfixed(1,FactorialSim);--Initial value;
binit<=to_sfixed(4,binit); --Initial value;
b<=to_sfixed(1,b); --Initial value;
c<=to_sfixed(0,c); --Initial value;
d<=to_sfixed(0,d); --Initial value;
E<=to_sfixed(0,E); --Initial value;
f<=to_sfixed(0,f); --Initial value;
x<=to_sfixed(0,x); --Initial value;
n<="000"; --Initial value;
nSfixed<=to_sfixed(0,nSfixed); --Initial value;
alphaEst<=to_sfixed(0,alphaEst); --Initial value;
IF (y>=bias)THEN
x<=resize(bias/y,x); --Calculate the argument for the arccos function
ELSIF (y<bias) THEN
x<=resize(y/bias,x); --Calculate the argument for the arccos function
END IF;
alphaEst<=resize((Pi/2)-x,alphaEst); --for term n=0 in Taylor series (Pi/2-n0)
FOR i IN 1 TO 5 LOOP
n<=n+1;
nSfixed<=sfixed;
Factorial<=resize(Factorial*nSfixed,Factorial);--Factorial,n=1=>1,n=2=>2,n=2=>6 etc
FactorialSim<=Factorial; --Just for Sim. purpose
b<=resize(b*binit,b); --For use in the Taylor series calculation
c<=resize(2*nSfixed+1,c); --For use in the Taylor series calculation
d<=resize(b*Factorial*Factorial*c,E); --For use in the Taylor series calculation
Case n IS --For use in the Taylor series calculation
WHEN "000"=>
e<=e;
WHEN "001"=>
e<=resize(x*x*x,e);
WHEN "010"=>
e<=resize(x*x*x*x*x,e);
WHEN "011"=>
e<=resize(x*x*x*x*x*x*x,e);
WHEN "100"=>
e<=resize(x*x*x*x*x*x*x*x*x,e);
WHEN "101"=>
e<=resize(x*x*x*x*x*x*x*x*x*x*x,e);
WHEN "110"=>
e<=e;
WHEN "111"=>
e<=e;
END CASE;
f<=resize(2*Factorial*e,f); --For use in the Taylor series calculation
alphaEst<=resize(alphaEst-(f/d),alphaEst); --Prev Taylor calculation(incl.Pi/2) - current Taylor calculation
END LOOP;
alphaEstOut<=alphaEst;
END PROCESS;

END Behavior;

I think you should clock the process.
process(clock)
...
..
end process

Synchronize everything and remove y from the sensitivity list. You may have to make some changes in the body of the process too.