Hello,

first you should be clear about the dimension of your byte variable.

For the word, it is clear: 0000 .. FFFF represents 0..100 % duty cycle with a resolution of 1/65535. But what about the byte ? Is it +- .125 %? Or +-12.5 % ?

+- .125 as a value from 0 to 1 

Got it.

In your word variable ('word[15..0]'), 0x1FFF is equal to 0.125.

To get something that 'computes', I'd suggest to convert your byte to 2's complement, adjust it to 16 Bits and add. As I'm not familiar with VHDL 'coding' - only with the algorithms behind that, I'll give you an algorithmic description what to do:

Be aware that you may have to check for over- resp. underflow of the result of the sum.

Hope the above is understandable

does this works with both positive and negative values?

That's the 'trick' with the 2's complement format: adding of negative values is easily done. Regarding subtraction, it is just about getting the 2's complement of the subtrahend and adding :)

Regards

i really appreciate your help 

but i need examples let us say that the duty is .5 which is in 16 bit format 

100000000000000

and the dutystep is 

.0335

which is

10100010

the results of the algorithm will be

1000100010010110 which is correct

the next step the duty step is -.1024 which is 00010100

and the duty is the new value 

1000100010010110

the results 

0100110111010110

while the results should be 

0110110110111111

Hi,

if 01024 is 00010100, -0.1024 is 1111111111101100 in 2's complement.

"'s complement for 16 bit integers is easiest built by inverting the code bit-wise and adding 1. (Or by subtracting the code from 10000000000000000. No error - this constant is a 17 bit value!)

Starting with your last correct value 1000100010010110 

-0.01024 is adding '-0.1024' and neglecting the overflow = 1000100010000010

Still your value.

But what does the algorithm do ? Is it in unsigned short or in signed short ?