Gaussian filter digital implementation

[elec-eng]

Hi all,

I have a question regarding the digital implementation of a Gaussian filter for GFSK modulation used in bluetooth

I know that the filter can be implemented directly as an FIR filter, but when I searched the internet I found that there are other methods used to implement the filter like:
-recursive implementation
-optimized FFT implementation
-repeated convolution implementation

So what is the method used in reality to implement Gaussian filters?

Thanks in advance

 

[RBB]

I used a state machine that calculated intermediate points (about 16) between the mark and the space values.
A state machine would cycle through these 16 points as the input data was changed. So the state machine ran at 16x the baud rate.
The intermediate points where calculated off of the peak deviation.

That's how I did it. I'm sure there are multiple ways.

 

[elec-eng]

Thanks Rbb very much

But could you provide more details about the state machine implementation )resourses to read, complexity, speed,...)

Thanks very much

 

[RBB]

Basically at the start of transmission a state machine calculates the peak deviation step steps. The steps are: 32/32, 31/32, 29/32, 25/32, 21/32, 15/32, 9/32, 3/32, 3/32, 9/32, 15/32, 21/32, 25/32, 29/32, 31/32, 32/32.

So if the peak deviation was 32kHz & we are transitioning from a space to a mark then the intermediate peak deviation value would be:
-32kHz, -31kHz, -29kHz, -25kHz, -21kHz, -15kHz, -9kHz, -3kHz, 3kHz, 9kHz, 15kHz, 21kHz, 25kHz, 29kHz, 31kHz, 32kHz.
The intermediate peak deviation is then added to the LO.
Just to reiterate the intermediate steps are only calculated once, upon initialization. Then the state machine running at 16x the baud rate adds the intermediate peak deviation values to the LO.

 

[elec-eng]

Is there any paper or tutorial to read more about this kind of implementation

Thanks

 

[RBB]

Not that I know of.