cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Highlighted
Observer
Observer
4,463 Views
Registered: ‎02-09-2010

Simulation error in Xilinx FIR compiler

Jump to solution

Hello all,

 

Im simulating an FIR filter which does interpolation by a factor of two, and I keep receiving this simulation error:

 

interpolator_2_inst/U0/g_mac/mac/ : Warning: NUMERIC_STD.TO_SIGNED: vector truncated

 

What can cause this? XCO file of filter is attached

0 Kudos
1 Solution

Accepted Solutions
Highlighted
Teacher
Teacher
5,172 Views
Registered: ‎08-14-2007

Hi,

first of all, it's no error, but a warning.

 

Warnings highlight situations that can lead to malfunction but they can be ignored if the design can deal with such situations.

 

So you have to decide.

Truncation in MAC structures is one solution to keep the word width in a reasonable size.

Another method would be rounding. Both cause some loss of information, but have different influence on the stability of your filter.

So, if your filter is stable with truncated values everything is just fine.

Use Matlab for a comprehensive test of your filters numerical stability.

 

Have a nice synthesis

  Eilert

View solution in original post

2 Replies
Highlighted
Teacher
Teacher
5,173 Views
Registered: ‎08-14-2007

Hi,

first of all, it's no error, but a warning.

 

Warnings highlight situations that can lead to malfunction but they can be ignored if the design can deal with such situations.

 

So you have to decide.

Truncation in MAC structures is one solution to keep the word width in a reasonable size.

Another method would be rounding. Both cause some loss of information, but have different influence on the stability of your filter.

So, if your filter is stable with truncated values everything is just fine.

Use Matlab for a comprehensive test of your filters numerical stability.

 

Have a nice synthesis

  Eilert

View solution in original post

Highlighted
Observer
Observer
4,449 Views
Registered: ‎02-09-2010

Yes of course, you are correct. It is no error, my mistake! :smileywink:

 

I have simulated the filter chain and analyzed the outputs of respective filters using octave.

The filter precision (and noise..) is within our tolerance, both in simulation and synthesis.

 

I am using the non-truncated output of the filters. I round the output outside the filter core itself.

 

In conclusion, everything works ok, I was just wondering if this something I should consider investigating more.

 

Regards

0 Kudos