Fixed point format FIR compiler

[phoe_137]

Hi all,

I am using FIR compiler v7.2 for my system. I input the data to FIR with format 16Q15 (or Q.15) and the output of FIR is 32Q15 (or Q17.15). I want to convert this format to Q.15 like input. So, is there anyway to do this ?

Thank you,

[drjohnsmith]

Thinking of base 10 numbers,

    if you multiply 99 by 99, you get 9801,

           to put that back into the sam eformat , yo move the decimal point twoto the left, to give 98.01 , as an integer 98

'All' the Q format is doing it saying where the binary point is,

[drjohnsmith]

https://en.wikipedia.org/wiki/Fixed-point_arithmetic

https://en.wikipedia.org/wiki/Q_(number_format)

http://www.ee.ic.ac.uk/pcheung/teaching/ee3_Study_Project/lecture%205(4).pdf

[drjohnsmith]

I have just quickly checked,

you can also set the output size in the FIR compiler, and the method of truncation

[meherp · Accepted solution]

I am using FIR compiler v7.2 for my system. I input the data to FIR with format 16Q15 (or Q.15) and the output of FIR is 32Q15 (or Q17.15). I want to convert this format to Q.15 like input. So, is there anyway to do this ?

As mentioned by @drjohnsmith, we can use the truncate option here. But it will not work here due to below reason, 

Reason: The output Q17.15 i.e. 17 interger bits are required for full scaled output. If we use Q.15 format, then we have 1 bit for integer output. Hence the lower 16 bit will be discarded by the compiler. In the other words, the output will be attenuated by a factor of 2^16. 

 Another thing to notice is, your input is 16Q15, but output is 32Q15 i.e. we have large value of coefficients (as the output is multiplication and add operation of coefficent and input data). May be you can play with coefficient values to reduce the output size and then use the truncate option.