The Implementation Options screen defines the coefficient structure to use and to configure the various datapath and coefficient options.
Coefficient Options
- Coefficient Type
- The Implementation Options screen is used to define the coefficient structure to use and to configure the various datapath and coefficient options.
Coefficient Options
- Type
- Specify the coefficient data as either signed or unsigned. When the signed option is selected, conventional two’s complement representation is assumed.
- Quantization
- Specifies the quantization method. Available options are
Integer Coefficients, Quantize Only, or Maximize Dynamic Range.
- The Integer Coefficients option is only available when the filter coefficients are specified using only integer values.
- The Quantize Only option rounds the provided values to the nearest quantum using a simple rounding towards zero algorithm.
- The Maximize Dynamic Range option scales all coefficients such that the maximum coefficient is equal to the maximum representable number in the specified bit width, thus maximizing the dynamic range of the filter (with the current implementation, overflow is not possible, as the accumulator width is automatically set to accommodate maximum bit growth within the filter). See Coefficient Quantization for more information.
- The Normalize to Center Coefficient option scales the coefficients such that the center coefficient can be represented as a power of 2. This option is only available for a single coefficients set with an odd number of taps.
- Width
- The bit precision of the filter coefficients. Use with the filter response graph to explore the possibilities for more efficient implementation by limiting coefficient bit width to the minimum required to meet your target specification for the filter.
- Best Precision Fraction Length
- When selected, the coefficient fractional width is automatically set to maximize the precision of the specified filter coefficients. See Best Precision Fractional Length for further information.
- Fractional Bits
- Specifies the number of coefficient bits to represent the fractional portion of the provided filter coefficients. The maximum value it supports is the Coefficient Width value minus the required integer bit width. The integer bit width value is static and is automatically determined by calculating the integer bit width required to represent the maximum value contained in the provided coefficient sets. When the coefficient width is less than the required integer bit width, this field reports zero. When the required integer bit width is zero, this parameter can take a value greater than the Coefficient Width. See Coefficient Quantization for more information.
Input Data Options
- Type
- Specify as either signed or unsigned. The signed option employs conventional two’s complement arithmetic. This parameter is automatically set in the IP integrator but can also be overridden.
- Width
- The precision (in bits) of the filter input data samples. This parameter is automatically set in IP integrator but can also be overridden.
- Fractional Bits
- The number of Input Data Width bits to represent the fractional portion of the filter input data samples. This field is for information only. It is used with Coefficient Fractional Bits to calculate the filter Output Fractional Bits value. This parameter is automatically set in the IP integrator but can also be overridden.
Output Data Options
- Rounding Mode
- Specifies the type of rounding to be applied to the output of the filter.
- Width
- When using Full Precision, this field is disabled and indicates the output precision (in bits) of the filter output data samples, including bit growth. When using any other Rounding Mode, this field allows you to specify the desired output sample width.
- Fractional Bits
- Reports the number Output Width bits used to represent the fractional portion of the filter output samples.
Filter Structure
- Filter Structure
- Supports five coefficient structures: Non-symmetric, Symmetric, Negative Symmetric, Half-band, and Hilbert. The structure can also be inferred from the coefficient file directly (default setting), or specified directly. The inference algorithm only analyzes the first 2048 coefficients. Only valid structure options, based on analysis of the provided coefficient file, are available for you to specify directly. If Hilbert has been specified as the Filter Type, then Hilbert is forced for the coefficient structure.