Marco wrote:
> what i did wrong?

1

** Error: pl.vhd(30): Type error resolving infix expression "-" as type ieee.NUMERIC_STD.SIGNED.

Your types didn't match:

1

    error  <= to_integer(ref) - to_integer(feed);

2

    signed <= integer         - integer;

See the defined functions in numeric_std:
https://tams-www.informatik.uni-hamburg.de/vhdl/doc/faq/FAQ1.html#4.8.1

If you change the code to

1

  process(reference_signal, encoder_signal)is 

2

  begin

3

    -- Calcolo dell'errore tra velocità di riferimento e velocità effettiva

4

    error <= ref - to_integer(feed);

5

6

    -- Calcolo del termine integrativo

7

    integral_term <= integral_term + error;

8

9

    -- Calcolo del segnale di controllo

10

    pwm_output <= resize( encoder_signal + natural(Kp) * error + natural(Ki) * integral_term, pwm_output'length); -- è l'equivalenente di (7 downto 0) ma più skillato

11

  end process;

It will compile, but you need to check with a testbench if it will work.

BTW: you don't need no process here. These are simply 3 concurrent 
statements.

First o all, I want to thanks everyone for the answers, but i have 
another question about this. If i don't need a process here, where do i 
need a process? It's only where there are more statements or what?
Thank you.
P.S. i will post more as  update about the work

IT COMPILES!!! but why?

Marco wrote:
> If i don't need a process here, where do i need a process?
You don't need the process here because inside it are only 3 statements 
which are calculated without any relation to each other.

Or in other words: a concurrent statement has an implicit process with 
an implicit sensitivity list.

This concurrent statement here:

1

a <= b + c;

Is the very same like that chatty process:

1

process (b,c) is

2

begin

3

  a <= b + c;

4

end process;

Marco wrote:
> IT COMPILES!!! but why?
Hint: look as said for the defined functions of the operators.

E.g. for the - operator there is no overloaded functions for 
integer-integer with a signed as a result. This would work also:

error  <= to_signed(to_integer(ref) - to_integer(feed)),8);