How a filter with coefficient of 3.234 is represented in register of Fixed Point DSP?

[cmos babe]

Let's say we have a filter that has coefficient = 3.234 . How is this represented in a register of a fixed point Digital Signal Processor.

 

[Circuit_seller]

Re: Fixed Point DSP

Hi

you can display it in Q2.13.

 

[dsp_forall]

Re: Fixed Point DSP

if you want to convert in fixed point multiply required number with 2 raised power n where n is number of bits

 

[vinay.b]

Re: Fixed Point DSP

hi,

Go thru the document attached u may find some answers on dealing with floating point numbers in fixed point processors.



rgds,
vinay.b

 

[cmos babe]

Re: Fixed Point DSP

I'm reading a section on fixed point numbers in DSP computer-based approach by Mitra. It mentions that in practice fixed point numbers are stored as either fractions or integers. How can this be done?

 

[chiron_17]

Re: Fixed Point DSP

hi
in case of fixed point implementation of the floating pt values...the number u want to convert is first normalised i.e. if u remember the elementry maths on logarithms is converting into a 1 followed by the decimal point and then followed up by the fraction but here we further reduce it to below 1 i.e. -1<N<1(in case of unsigned 0<N<1) ... in this case the dividing factor is 4 hence we need 3 bits before the pt to represent the division factor and then the next number of bits depends on the precision or bit-length of the data type.. hence in fixed pt dsp we use Q format here for 2 byte precision the Q format is Q3.12

after this the fraction i.e. 3.1415/4 is multiplied with the precision value ie number of bits representing the fraction part... here 12 bits so
fixed pt value = 3.1415/4 *(2^12)

 

[cmos babe]

Re: Fixed Point DSP

Hi

you can display it in Q2.13.

Hi,
What's confusing me is that DSP books (Proakis, Mitra,..) suggest the use of fractional fixed point numbers (I.e Q0.15) and they don't mention anything about these various formats or if using some format is better than the other. I'd appreciate it if someone would help me end this confusion.

 

[iTdl]

Re: Fixed Point DSP

You can reffer to the post:

 

[amit_8561]

Fixed Point DSP

you can represent it using binary representation of decimal value as in any standard text book....