Main Content

fdesign

Filter design specification object

Description

Use the fdesign function to create a filter design specification object that contains the specifications for a filter, such as passband ripple, stopband attenuation, and filter order. Then, use the design function to design the filter from the filter design specifications object. For an example, see Design of Lowpass Decimator.

Here is the workflow diagram that shows the simple procedure to design, analyze, and finally apply the filter on streaming data.

For more control options, see Filter Design Procedure. For a complete workflow, see Design a Filter in Fdesign — Process Overview.

designSpecs = fdesign.response returns a design specification object for the filter with a given response.

example

designSpecs = fdesign.response(spec) specifies the variables to use that define your filter design. The filter design parameters are applied to the filter design method you choose for your filter. The specification option you choose determines which design methods apply to the fdesign object.

designSpecs = fdesign.response(___,Fs) specifies the sample rate in Hz to use in the filter specifications. The sample rate scalar must be the last input argument. If you specify a sample rate, all frequency specifications are in Hz.

designSpecs = fdesign.response(___,magunits) specifies the units for any magnitude specification you provide in the input arguments.

Examples

collapse all

Design a 100-tap FIR lowpass decimator filter that reduces the sample rate of a signal from 60 kHz to 20 kHz. The passband of the filter extends up to 6 kHz. Specify a passband ripple of 0.01 dB and a stopband attenuation of 100 dB.

Fs = 60e3;
N = 99;
Fpass = 6e3;
Apass = 0.01;
Astop = 100;
M = Fs/20e3;

Setup the filter design specifications object using the fdesign.decimator function.

filtSpecs = fdesign.decimator(M,'lowpass','N,Fp,Ap,Ast',N,Fpass,Apass,Astop,Fs);

Design the FIR lowpass decimator using the design function.

The resulting filter is a dsp.FIRDecimator System object™. For details on how to apply this filter to streaming data, refer to dsp.FIRDecimator.

decimFIR = design(filtSpecs,'SystemObject',true)
decimFIR = 
  dsp.FIRDecimator with properties:

     NumeratorSource: 'Property'
           Numerator: [1x100 double]
    DecimationFactor: 3
           Structure: 'Direct form'

  Show all properties

Use info to display information about the filter.

info(decimFIR)
ans = 10x56 char array
    'Discrete-Time FIR Multirate Filter (real)               '
    '-----------------------------------------               '
    'Filter Structure   : Direct-Form FIR Polyphase Decimator'
    'Decimation Factor  : 3                                  '
    'Polyphase Length   : 34                                 '
    'Filter Length      : 100                                '
    'Stable             : Yes                                '
    'Linear Phase       : Yes (Type 2)                       '
    '                                                        '
    'Arithmetic         : double                             '

Visualize the magnitude response of the filter using fvtool.

fvtool(decimFIR,'Fs',Fs)

Design a lowpass filter to use on a signal sampled at 96 kHz. The passband of the filter extends up to 20 kHz. The stopband of the filter starts at 24 kHz. Specify a passband ripple of 0.01 dB and a stopband attenuation of 80 dB. Determine automatically the order required to meet the specifications.

Set up the filter design specifications object using the fdesign.lowpass function.

Fs = 96e3;
Fpass = 20e3;
Fstop = 24e3;
Apass = 0.01;
Astop = 80;

filtSpecs = fdesign.lowpass(Fpass,Fstop,Apass,Astop,Fs);

Determine the available design algorithms using the designmethods function.

designmethods(filtSpecs,'SystemObject',true)
Design Methods that support System objects for class fdesign.lowpass (Fp,Fst,Ap,Ast):


butter
cheby1
cheby2
ellip
equiripple
ifir
kaiserwin
multistage

Using the design function, design an equiripple FIR filter and an elliptic IIR filter that meet the specifications.

lpFIR = design(filtSpecs,'equiripple','SystemObject',true)
lpFIR = 
  dsp.FIRFilter with properties:

            Structure: 'Direct form'
      NumeratorSource: 'Property'
            Numerator: [1x101 double]
    InitialConditions: 0

  Show all properties

lpIIR = design(filtSpecs,'ellip','SystemObject',true)
lpIIR = 
  dsp.BiquadFilter with properties:

                   Structure: 'Direct form II'
             SOSMatrixSource: 'Property'
                   SOSMatrix: [5x6 double]
                 ScaleValues: [6x1 double]
           InitialConditions: 0
    OptimizeUnityScaleValues: true

  Show all properties

You can also measure the designs to verify that the filters satisfy the constraints.

FIRmeas = measure(lpFIR)
FIRmeas = 
Sample Rate      : 96 kHz      
Passband Edge    : 20 kHz      
3-dB Point       : 21.4297 kHz 
6-dB Point       : 21.8447 kHz 
Stopband Edge    : 24 kHz      
Passband Ripple  : 0.0092309 dB
Stopband Atten.  : 80.6014 dB  
Transition Width : 4 kHz       
 
IIRmeas = measure(lpIIR)
IIRmeas = 
Sample Rate      : 96 kHz     
Passband Edge    : 20 kHz     
3-dB Point       : 20.5524 kHz
6-dB Point       : 20.7138 kHz
Stopband Edge    : 24 kHz     
Passband Ripple  : 0.01 dB    
Stopband Atten.  : 80 dB      
Transition Width : 4 kHz      
 

Estimate and display the computational cost of each filter. The equiripple FIR filter requires many more coefficients than the elliptic IIR filter.

FIRcost = cost(lpFIR)
FIRcost = struct with fields:
                  NumCoefficients: 101
                        NumStates: 100
    MultiplicationsPerInputSample: 101
          AdditionsPerInputSample: 100

IIRcost = cost(lpIIR)
IIRcost = struct with fields:
                  NumCoefficients: 20
                        NumStates: 10
    MultiplicationsPerInputSample: 20
          AdditionsPerInputSample: 20

Use fvtool function to visualize the resulting designs and compare their properties.

fvtool(lpFIR,lpIIR,'Fs',Fs);
legend('FIR Equiripple','Elliptic IIR')

Design a lowpass Butterworth filter that has a passband edge frequency of 0.4π rad/sample, a stopband frequency of 0.5π rad/sample, a passband ripple of 1 dB, and a stopband attenuation of 80 dB.

Create a lowpass filter design specification object using the fdesign.lowpass function. Specify the design parameters.

lowpassSpecs = fdesign.lowpass(0.4,0.5,1,80);

To view a list of design methods available for the specification object, use the designmethods function. If multiple methods are available, pick one that best meets the design criteria. For this example, pick 'butter'.

designmethods(lowpassSpecs,'SystemObject',true)
Design Methods that support System objects for class fdesign.lowpass (Fp,Fst,Ap,Ast):


butter
cheby1
cheby2
ellip
equiripple
ifir
kaiserwin
multistage

Furthermore, you can specify the design options used in designing the filter. To see a list of available options, run the designoptions function on lowpassSpecs. The design options are dependent on the design method you pick. The design method, in this case, 'butter', must be specified as an argument to the designoptions function.

designoptions(lowpassSpecs,'butter','Systemobject',true)
ans = struct with fields:
           FilterStructure: {1x6 cell}
              SOSScaleNorm: 'ustring'
              SOSScaleOpts: 'fdopts.sosscaling'
              MatchExactly: {'passband'  'stopband'}
    DefaultFilterStructure: 'df2sos'
       DefaultMatchExactly: 'stopband'
       DefaultSOSScaleNorm: ''
       DefaultSOSScaleOpts: [1x1 fdopts.sosscaling]

The filter order necessary to meet a set of design constraints must also be rounded up to an integer value. This loosens some of the constraints, and as a consequence, some design specifications are met while others are exceeded. The 'MatchExactly' option allows you to match the passband or stopband exactly while exceeding the specification for the other band. Design the filter so that it matches the passband exactly.

The resulting filter is a dsp.BiquadFiter System object™. For details on how to apply this filter on streaming data, refer to dsp.BiquadFilter.

IIRbutter = design(lowpassSpecs,'butter','MatchExactly','passband', ...
    'SystemObject',true)
IIRbutter = 
  dsp.BiquadFilter with properties:

                   Structure: 'Direct form II'
             SOSMatrixSource: 'Property'
                   SOSMatrix: [16x6 double]
                 ScaleValues: [17x1 double]
           InitialConditions: 0
    OptimizeUnityScaleValues: true

  Show all properties

Use fvtool to visualize the magnitude response of the filter.

fvtool(IIRbutter)

Input Arguments

collapse all

The table specifies the possible filter responses.

fdesign Response Method

Description

arbgrpdelay

fdesign.arbgrpdelay creates an object to specify allpass arbitrary group delay filters.

arbmag

fdesign.arbmag creates an object to specify IIR filters that have arbitrary magnitude responses defined by the input arguments.

arbmagnphase

fdesign.arbmagnphase creates an object to specify IIR filters that have arbitrary magnitude and phase responses defined by the input arguments.

audioweighting

fdesign.audioweighting creates a filter design specification object for audio weighting filters. The supported audio weighting types are: A, C, C-message, ITU-T 0.41, and ITU-R 468-4 weighting.

bandpass

fdesign.bandpass creates an object to specify bandpass filters.

bandstop

fdesign.bandstop creates an object to specify bandstop filters.

ciccomp

fdesign.ciccomp creates an object to specify filters that compensate for the CIC decimator or interpolator response curves.

comb

fdesign.comb creates an object to specify a notching or peaking comb filter.

decimator

fdesign.decimator creates an object to specify decimators.

differentiator

fdesign.differentiator creates an object to specify an FIR differentiator filter.

fracdelay

fdesign.fracdelay creates an object to specify fractional delay filters.

halfband

fdesign.halfband creates an object to specify halfband filters.

highpass

fdesign.highpass creates an object to specify highpass filters.

hilbert

fdesign.hilbert creates an object to specify an FIR Hilbert transformer.

interpolator

fdesign.interpolator creates an object to specify interpolators.

isinchp

fdesign.isinchp creates an object to specify an inverse sinc highpass filter.

isinclp

fdesign.isinclp creates an object to specify an inverse sinc lowpass filters.

lowpass

fdesign.lowpass creates an object to specify lowpass filters.

notch

fdesign.notch creates an object to specify notch filters.

nyquist

fdesign.nyquist creates an object to specify Nyquist filters.

octave

fdesign.octave creates an object to specify octave and fractional octave filters.

parameq

fdesign.parameq creates an object to specify parametric equalizer filters.

peak

fdesign.peak creates an object to specify peak filters.

polysrc

fdesign.polysrc creates an object to specify polynomial sample-rate converter filters.

rsrc

fdesign.rsrc creates an object to specify rational-factor sample-rate convertors.

Use the doc fdesign.response syntax at the MATLAB prompt to get help on a specific structure. For example, this command provides more information about the lowpass specification object:

doc fdesign.lowpass

Each response has a Specification property that defines the specifications to use to design your filter. You can use defaults or specify the Specification property when you construct the specifications object.

Using the Specification property, you can provide filter constraints such as the filter order or the passband attenuation to use when you construct your filter from the specification object.

Filter design specifications, specified as a character vector. The set of available specification options depends on the fdesign.response function. For more information, refer to the individual fdesign.response pages.

The filter design is based on the specifications provided by the fdesign.response object. For example, when you create a default lowpass filter design specification object, fdesign.lowpass, the specification expression is set to 'Fp,Fst,Ap,Ast'. The filter design parameters — Fp (passband frequency), Fst (stopband frequency), Ap (passband ripple), and Ast (stopband attenuation) — are set to default values. The design function designs the filter based on these parameters.

Specifications that do not contain the filter order result in minimum-order designs when you invoke the design function:

d = fdesign.lowpass;      % Specification is 'Fp,Fst,Ap,Ast'
FIReq = design(d,'equiripple','SystemObject',true);
length(FIReq.Numerator)           % Returns 43. The filter order is 42

The specification option you choose determines which design methods are applicable. You can use the setspecs function to set all of the specifications simultaneously.

You can set filter specification values by passing them after the Specification argument, or by passing the values without the Specification.

Filter object constructors take the input arguments in the same order as setspecs and Specification.

When the first input to fdesign.response is not a valid Specification option, fdesign assumes that the input argument is a filter specification and applies it using the default Specification option. For example, 'Fp,Fst,Ap,Ast' is the default for a lowpass object.

Sample rate to use in filter specifications, specified in Hz. The sample rate scalar must be the last input argument. If you specify a sample rate, all frequency specifications are in Hz.

Data Types: single | double | int8 | int16 | int32 | int64 | uint8 | uint16 | uint32 | uint64

Units for magnitude specification, specified as:

  • 'dB' –– decibels

  • 'linear' –– linear units

  • 'squared' –– power units

When you omit the magunits argument, fdesign assumes that all magnitudes are in dB. Note that fdesign stores all magnitude specifications in dB. If you set magunits to an option other than 'dB', the function converts the unit to 'dB'.

Output Arguments

collapse all

fdesign returns a filter design specification object. Every filter design specification object has these properties:

Property Name

Default Value

Description

Response

Depends on the chosen type

Defines the type of filter to design, such as an interpolator or bandpass filter. This is a read-only value.

Specification

Depends on the chosen type

Defines the filter characteristics used to define the desired filter performance, such as the cutoff frequency Fc or the filter order N.

Description

Depends on the filter type you choose

Contains descriptions of the filter specifications used to define the object, and the filter specifications you use when you create a filter from the object. This is a read-only value.

NormalizedFrequency

Logical true

Determines whether the filter calculation uses a normalized frequency from 0 to 1, or the frequency band from 0 to Fs/2, the sampling frequency. Accepts either true or false without single quotation marks. Audio weighting filters do not support normalized frequency.

In addition to these properties, filter design specification objects may have other properties as well, depending on whether they design single-rate filters or multirate filters.

Added Properties for Multirate Filters

Description

DecimationFactor

Specifies the amount to decrease the sampling rate. Always a positive integer.

InterpolationFactor

Specifies the amount to increase the sampling rate. Always a positive integer.

PolyphaseLength

Polyphase length is the length of each polyphase subfilter that composes the decimator or interpolator or rate-change factor filters. Total filter length is the product of pl and the rate change factors. pl must be an even integer.

More About

collapse all

Filter Design Procedure

Here is the workflow diagram of the overall procedure for designing and analyzing the filter.

Here are the steps in detail:

  1. Create an fdesign.response specification object to specify the design parameters.

  2. Use designmethods to determine the filter design methods that work for your new filter specification object. If you choose to use the default design method, then this step is optional.

  3. If you prefer to change the design options and would like to see a list of available options, run the designoptions function on the specification object. The output also shows the design options the filter uses by default.

  4. Use design to design the filter from the filter specification object. Specify the design method (determined from step 2) as an input. If the design options must change from the default values, specify them as name-value pairs following the design method.

    If you call the design function without any output arguments, FVTool is launched and shows the magnitude response of the designed filter.

    Alternatively, use the fvtool function.

  5. Further analysis, such as viewing the frequency response of the filter, computing the cost of implementing the filter, and measuring the filter response characteristics, can be done using one of the supported Analysis Methods for Filter System Objects.

  6. Once you analyze the filter and determine that the filter satisfies the design constraints, you can apply the filter object to streaming input data. For details on how to pass data to the filter object, refer to the corresponding filter System object™ reference pages.

For a detailed example on the design and analysis, see Lowpass Butterworth Filter Specification and Design.

Introduced in R2009a