FPGA Intellectual Property
PCI Express*, Networking and Connectivity, Memory Interfaces, DSP IP, and Video IP
6510 Discussions

Using CIC Decimation Filter with Multi-channel Support

Altera_Forum
Honored Contributor II
1,943 Views

Hi All, 

 

I have questions about this designe example: 

http://www.altera.com/support/examples/dsp-builder/exm-cic-decimation-filter.html 

 

I have attached the design files. In the file fir_compiler_v7_2.vhd, it goes 

"number_of_input_channels" value="2" 

"input_bit_width" value="16" 

"output_bit_width" value="35" 

 

So the two input channels of fir conveys 8-bit 'I' and 8-bit 'Q' data respectively (correct me if it is wrong). Does that mean 'I' and 'Q' are merged into 35-bit packet on the output interface so the first half contains 'I' and the second half has 'Q'?  

 

Then the 35-bit output is sent to Avalon-ST Packet Format Converter. The converter seperates the 35-bit 'IQ' packet into two sources, 35-bit 'I' and 35-bit 'Q'. Does the converter perform the bit extension to add more bits to I and Q outputs, respectively, to have the bit width of 35? 

 

Many Thanks, 

Lee
0 Kudos
7 Replies
Altera_Forum
Honored Contributor II
510 Views

 

--- Quote Start ---  

Hi All, 

 

I have questions about this designe example: 

http://www.altera.com/support/examples/dsp-builder/exm-cic-decimation-filter.html 

 

I have attached the design files. In the file fir_compiler_v7_2.vhd, it goes 

"number_of_input_channels" value="2" 

"input_bit_width" value="16" 

"output_bit_width" value="35" 

 

So the two input channels of fir conveys 8-bit 'I' and 8-bit 'Q' data respectively (correct me if it is wrong). Does that mean 'I' and 'Q' are merged into 35-bit packet on the output interface so the first half contains 'I' and the second half has 'Q'?  

 

Then the 35-bit output is sent to Avalon-ST Packet Format Converter. The converter seperates the 35-bit 'IQ' packet into two sources, 35-bit 'I' and 35-bit 'Q'. Does the converter perform the bit extension to add more bits to I and Q outputs, respectively, to have the bit width of 35? 

 

Many Thanks, 

Lee 

--- Quote End ---  

 

 

I would think that each of I and Q is 16 bits wide. The final result of each is 35 bits wide. I and Q results are interleaved at FIR output and dinterleaved by the converter.
0 Kudos
Altera_Forum
Honored Contributor II
510 Views

Hi Kaz, 

 

You mentioned "I and Q are interleaved at FIR output". What does interleaved mean? Does that mean there are two output channels for 35-bit I and 35-bit Q, respectively? 

 

Thanks, 

Lee
0 Kudos
Altera_Forum
Honored Contributor II
510 Views

 

--- Quote Start ---  

Hi Kaz, 

 

You mentioned "I and Q are interleaved at FIR output". What does interleaved mean? Does that mean there are two output channels for 35-bit I and 35-bit Q, respectively? 

 

Thanks, 

Lee 

--- Quote End ---  

 

 

Yes that is what I imagine as I-Q-I-Q-... 

 

you can check that by setting I input to zeros and Q to nonzeros
0 Kudos
Altera_Forum
Honored Contributor II
510 Views

Hi Kaz, 

 

I'm trying to run the simulink model CICDownMISO.mdl but I re-created the model using 13.1 version of CIC and FIR compilers and re-named it as CICExample.mdl. I have attached the file. In the Step bolck, it has the parameter Sample time: clock1. I got an error saying undefined variable clock1. How do I define clock1 in the model? I'm very new to Simulink DSP Builder. 

 

 

Thanks, 

Lee
0 Kudos
Altera_Forum
Honored Contributor II
510 Views

The assignment of variable clock1 can be found here: 

In the Simulink model, 

File->Model Properties -> Callbacks (Tab) -> InitFcn* (Model Initialization Function)-> clock1 = 1; ....
0 Kudos
Altera_Forum
Honored Contributor II
510 Views

Hi All, 

 

How to decide the bit width for the fir filter coefficient? Is the bit width the same as the input data bit width? 

 

Many Thanks, 

Lee
0 Kudos
Altera_Forum
Honored Contributor II
510 Views

 

--- Quote Start ---  

Hi All, 

 

How to decide the bit width for the fir filter coefficient? Is the bit width the same as the input data bit width? 

 

Many Thanks, 

Lee 

--- Quote End ---  

 

 

quantise your coeffs and see response. If you and spec are happy with it so be it.
0 Kudos
Reply