how to call decimal values in verilog

[emerson_11]

hy
i am working in a image processing based project using FPGA. i want to call my features into addition block which are decimals in the range of 0.1 to 0.9. Can anyone tell me some techniques to convert these numbers into fixed point binary?

 

[KlausST]

Hi,

simply use "1...9" as integers.
And imagine as there was a "0." before it.

Or use them as integers and divide them by 10...if you really need this.
****

Maybe you need to give a bit more information, why you especially need to know that it is a value between 0 and 1.

Klaus

 

[dipin]

hi,

sometimes i will use this for conversion

https://www.exploringbinary.com/binary-converter/

then what is the fixed integer & fractional length ?

regards

 

[emerson_11]

can anyone explain about the roundoff to the nearest power of 2 concept? can i use this for converting my decimals into rounded integer and then into binary

 

[ads-ee]

Nearest power of two is finding the 2^N that is nearest to your number.

e.g.

2 => 2^1 (2)
5 => 2^2 (4), or 2^3 (8, for round up)
13 => 2^4 (16)

As you can see this won't help at all for converting between decimals and binary.

BTW in hardware all decimal numbers are stored as binary values, there are no 10 state FFs. So if you look at the following code:

wire [3:0] a;
assign a = 10;
// 10 in binary is 1010
// a[0] = 1'b0
// a[1] = 1'b1
// a[2] = 1'b0
// a[3] = 1'b1

For your original question the easiest thing to do is scale all your numbers by *10. Using any kind of 2^N scaling will never give you an exact 0.9 value. What you need to do is determine your required integer and fractional bit widths to obtain the accuracy you desire.

 

[emerson_11]

can you tell me how to choose an effective scaling factor? whether it should be only *10 s or any other number?

 

[ads-ee]

You scale based on the number of bit of precision you require.

 

[Altaero]

Fixed point arithmetics is absoultely the same as integer arithmetics, in fact it IS integer arithmetics. It's up to you where to put decimal point.