STD_LOGIC_VECTOR multipication

The multiplication returns and unsigned, not a std_logic_Vector. So with your code, you need to write:

EXPx63_LTCHD <= std_logic_Vector( unsigned(EXP_LTCHD) * unsigned(COEFF) );

But why have you used std_logic_vector. Surely its better just to write:

constant COEFF : unsigned(5 downto 0) := x"3F";
signal EXPx63_LTCHD : unsigned(27 downto 0);
signal EXP_LTCHD:	unsigned(22 downto 0);
EXPx63_LTCHD <= EXP_LTCHD * COEFF;

Remember, ports do NOT have to be std_logic_Vector.

I am trying to write the VHDL code to do the same job as a few of the modules! I used SLV's with these modules and it all worked OK. Then I came back to the code and started using Integers to do the calculation and found I soon went past 32bits and overflowed some of my variables (as i believe integers are limited to 32bit). So I went back and started using SLV's in my claculations as I suspect they cannot be limited in the same way as integers.

But you say that unsigned is the way to go? Can I get past 32bit arithmetic problems this way?

Many thanks again tricky for your quick response!

D

Hi,

EXPx63_LTCHD <= std_logic_vector(unsigned(EXP_LTCHD) * unsigned(COEFF));

Pay attention to x"3F" that might be considered as a negative number.

--- Quote Start ---

EXPx63_LTCHD <= std_logic_vector(unsigned(EXP_LTCHD) * unsigned(COEFF));

--- Quote End ---

Yes this solves the first problem involved with the code.

--- Quote Start ---

Pay attention to x"3F" that might be considered as a negative number.

--- Quote End ---

There's no reason to expect this. But x"3F" is 8 bit length where 6 bits are needed.

"111111" is the intended 6-Bit value.

Third problem, the product bit length must be increased by one.

--- Quote Start ---

I am trying to write the VHDL code to do the same job as a few of the modules! I used SLV's with these modules and it all worked OK. Then I came back to the code and started using Integers to do the calculation and found I soon went past 32bits and overflowed some of my variables (as i believe integers are limited to 32bit). So I went back and started using SLV's in my claculations as I suspect they cannot be limited in the same way as integers.

But you say that unsigned is the way to go? Can I get past 32bit arithmetic problems this way?

Many thanks again tricky for your quick response!

D

--- Quote End ---

Technically, integer has no bits, but it's maximum value is determined by the compiler. I think the VHDL LRM allows for any sized integer (8, 16, 32 or 64bit), but everyone sticks with 32 bits (well not quite - range is -2^-31+1 to 2^31-1, so not quite the full 32 bits range). Unisnged and signed are just arrays that allows the user to set the size using a natual, so an unsigned, signed or std_logic_vector can be up to 2^31 bits wide. SO yes, you get well around the integer limitations doing this (But never try to cast a number wider than 32 bits - 31 for unsigned - back to an integer type).

--- Quote Start ---

Pay attention to x"3F" that might be considered as a negative number.

--- Quote End ---

It wont be in unsigned arithmatic.

Thanks FvM,

I did spot the size issue with the product eventually. Thanks for your help.

D

Thanks all. Really helpful as always.

D

I do have one final quick question relating to this:

I have a variable called duty that is an unsigned(5 downto 0); and i want to slice off the (5 downto 0) bits of the following equation and assign it to DUTY;

DUTY <= (((WIDTH_LTCHD*PERCENT)/PERIOD_LTCHD);

AT the moment (in the example above) duty is unnecesarily 23 bits wide due to the variables in the equation, but I know for a fact (due to limits) that the result cannot be bigger than a 6bit result;

I tried a few variations and couldnt get the slice right; i tried somethings like:

DUTY <= (((WIDTH_LTCHD*PERCENT)/PERIOD_LTCHD),(5 downto 0));

Thanks guys!

[pedant]Its not a variable, with use of the <= assignment I see its a signal - there is a difference between the two [/pedant]

try:

DUTY <= ((WIDTH_LTCHD*PERCENT)/PERIOD_LTCHD)(5 downto 0);

otherwise just use a temporary 23 bit signal and slice that:

temp <= (WIDTH_LTCHD*PERCENT)/PERIOD_LTCHD;

DATA <= temp(5 downto 0);

Thanks Tricky, sometimes I cant get out of thinking like a 'C' programmer, hence variables everywhere! ;-)

D