Use the Filter Analysis Express VI to observe the response of the fixed-point filter. Use the calculated results to optimize the fixed-point filter.

The Filter Analysis Express VI performs fixed-point analysis in the frequency domain. You might need to analyze and adjust the filter design iteratively until the calculated results are satisfactory.

Ensure that the fixed-point filter is stable by verifying that all poles are within the unit circle and that the filter maintains a satisfactory frequency response. If the fixed-point characteristics do not satisfy the requirements, try one or more of the following options:

  • Return to the quantization step and change the quantizer settings.
  • Change the implementation structure.
  • Change the floating-point reference filter specifications to allow more headroom for finite-precision effects.
  • For Infinite Impulse Response (IIR) filters, reduce the pole radius constraint of the reference floating-point filter.
If you notice large distortions in the response, look for underflow or overflow conditions. Use the DFD FXP Coef Report VI to determine if overflows or underflows exist in the quantized coefficients. If you find occurrences of overflows or underflows in the coefficients report, repeat the quantization step by using an increased iwl value for the coefficients quantizer. Increasing the iwl value helps eliminate overflows and underflows and improve the frequency response of the filter. If you do not find any occurrences of overflows or underflows, try reducing the value of iwl to allow more digits of precision for the quantized coefficients. If the minimum allowable iwl value still returns large distortions in the frequency response, you need to increase the value of wl to minimize the distortion.
Note Underflows and overflows that happen during quantization do not always affect filter responses. Therefore, if the filter response after coefficients quantization is satisfactory, you do not need to make adjustments to avoid overflows or underflows.