The fillgaps
function
uses autoregressive modeling to interpolate missing large contiguous
portions of signals. For an example that uses this function, see Reconstructing Missing Data.
The findchangepts
function
locates the points at which a signal significantly changes its mean,
its variance, or both. The cusum
function
detects small incremental changes in the mean of a process by monitoring
where the cumulative sum of the signal drifts beyond a target mean.
For an example that uses these functions, see Detecting Outbreaks and Significant Changes in Signals.
The dtw
function stretches
two nonlinearly related signals onto a common time axis such that
a global measure of distance between the signals is smallest. The
process highlights similarities between signals by making equivalent
features appear at the same location on the common axis. For an example
that uses this function, see Extracting Classification Features from Physiological Signals.
The periodogram
function
now computes reassigned spectra. The reassignment technique sharpens
the frequency localization of spectral estimates and produces periodograms
that are easier to read and interpret. For an example that uses this
feature, see Measuring the Power of Deterministic Periodic Signals.
The Signal Analyzer app enables plotting, alignment, and comparison of any signals in the MATLAB^{®} workspace.
xcorr
Function: Generate faster code for long input vectorsFor long input vectors, code generation for xcorr
now uses a frequency-domain calculation
instead of a time-domain calculation. The resulting code can be faster
than in previous releases.
chebwin
Function: Compute Dolph-Chebyshev windows fasterThe chebwin
function
has been rewritten and exhibits improved performance for windows of
any size.
The spectrogram
function
now computes reassigned spectra. The reassignment technique sharpens
the time and frequency localization of spectral estimates and produces
spectrograms that are easier to read and interpret. See the spectrogram
reference
page for examples. The enhanced function also enables you to specify
a threshold such that spectrum values with less power are set to zero.
The new hampel
function
detects and removes outliers from data. The function computes a moving
median and considers as outliers those data points that deviate from
the median by more than an adjustable number of standard deviations.
The medfilt1
function
now uses a faster algorithm and enables you to choose how to treat
signals with NaN
s and how to filter signal edges.
The new rpmordermap
and rpmfreqmap
functions
perform order analysis on rotational machinery. rpmordermap
computes
order maps of rotating systems, helping you visualize vibrations that
vary linearly with rotational speed. rpmfreqmap
computes
frequency maps. You can also use these functions to plot the maps
on interactive figures.
The new envelope
function
computes the upper and lower envelopes of a waveform. You can compute
the envelopes using an FFT-based analytic function, an FIR Hilbert
filter, or the function peaks or RMS values.
The new finddelay
and alignsignals
functions
compute the delay between two signals and align signals in time.
This release introduces two functions that use spectral estimates to measure signal frequency:
meanfreq
estimates
the mean frequency of the spectrum of a signal.
medfreq
estimates
the median frequency of the spectrum of a signal.
These functions accept a time-domain signal, a power spectral density, or a power spectrum as input. They have visualization functionality and support multichannel signals.
This release introduces two functions that measure signal bandwidth:
obw
returns
the occupied bandwidth of a signal at a specified percentage.
powerbw
returns
the power bandwidth of a signal at a specified level.
These functions accept a time-domain signal, a power spectral density, or a power spectrum as input. They have visualization functionality and support multichannel signals.
resample
function accepts signals with nonuniform sampling or missing dataThis release enhances the resample
function
to accept nonuniformly sampled signals and reconstruct them on a regular
grid. The function also accepts NaN
s in the input
signal and treats them as missing data.
pwelch
function computes maximum-hold and minimum-hold spectraThe pwelch
function
now computes the maximum-hold and minimum-hold power spectral density
(PSD) estimates of a signal.
spectrogram
function computes two-sided, centered, power, and PSD spectraThe spectrogram
function
now computes two-sided or centered spectra and can output the mean-square
spectrum instead of the power spectral density. The enhanced function
also plots a color bar with units.
All the spectral analysis functions offered by the Signal Processing Toolbox™ product let you input a vector with the frequencies at which to compute PSD estimates. The functions traditionally have used the Goertzel algorithm in those cases. Starting with this release, the functions might switch to a faster method based on the chirp Z-transform when the input frequency vector has many points and the points are uniformly spaced. The change applies to the following functions:
Nonparametric methods: cpsd
, mscohere
, periodogram
, pmtm
, pwelch
, spectrogram
, tfestimate
This release introduces a new spectral estimation function, plomb
.
The function can compute spectra of nonuniformly sampled signals or
signals with missing samples by using the Lomb-Scargle algorithm. plomb
also
provides visualization functionality and support for multichannel
data.
This release enhances several parametric and nonparametric spectral analysis functions by offering support for multichannel data. The functions continue to process vector data as single channels. For matrix input, the functions process the data column by column and return a matrix of spectra.
The function findpeaks
now
lets you estimate the width of any peak and limit the results of peak
searches by width or prominence. When called with no output arguments,
the function plots the signal and annotates the value, width, and
prominence of every peak.
This release introduces a new filter design function, designfilt
.
A single command allows you to design lowpass, highpass, bandpass,
bandstop, differentiator, and Hilbert filters, including minimum-order
designs, in one step. The designfilt
interface
leads to readable and self-documenting code that is easy to maintain.
Use filter
to
filter signals with the digitalFilter
objects
generated by designfilt
.
Use fvtool
to
visualize digital filters created using designfilt
.
All the filter analysis and data filtering functions offered
by the Signal Processing Toolbox product can be used with digital
filters designed with designfilt
.
designfilt
also
features a Filter Design Assistant—a smart run-time error-recovery
mechanism that helps you correct faulty syntax and generates MATLAB code
at the Command Window or in the Editor.
This release enhances SNR, distortion, intermodulation, and spurious free dynamic range measurement functions to provide visualization functionality.
The function sfdr
plots
the spectrum of the input signal and labels its fundamental component
and largest spur. The plot shades the spurious free dynamic range
and displays its value; it shows the fundamental, the DC value, and
the rest of the signal in different colors. sfdr
now
uses a Kaiser window by default.
The function sinad
plots
the spectrum of the input signal, labels its fundamental component,
and displays its signal to noise and distortion ratio. The plot shows
the fundamental component, the DC value, and the noise in different
colors.
The function snr
plots
the spectrum of the input signal, labels its fundamental component
and higher harmonics, and displays its signal-to-noise ratio. The
plot shows the fundamental, the noise, and the DC value and harmonics
using different colors.
The function thd
plots
the spectrum of the input signal, labels its fundamental component
and harmonics, and displays its total harmonic distortion. The plot
shows the fundamental, the harmonics, and the DC level and noise using
different colors.
The function toi
plots
the spectrum of the input signal, annotates its lower and upper fundamentals
and intermodulation products, and displays its third-order intercept
point.
The R2014a release changes the name-value pair arguments 'MidPct'
and 'PctRefLevels'
to 'MidPercentReferenceLevel'
and 'PercentReferenceLevels'
for
the following signal measurement functions: dutycycle
, falltime
, midcross
, pulseperiod
, pulsesep
, pulsewidth
, risetime
, settlingtime
, slewrate
, overshoot
,
and undershoot
.
Functionality | What Happens When You Use This Functionality? | Use This Instead | Compatibility Considerations |
---|---|---|---|
'MidPct' name-value pair of measurement
functions | Still runs | 'MidPercentReferenceLevel' | Replace all instances of 'MidPct' with 'MidPercentReferenceLevel' |
'PctRefLevels' name-value pair of measurement
functions | Still runs | 'PercentReferenceLevels' | Replace all instances of 'PctRefLevels' with 'PercentReferenceLevels' |
This release introduces four frequency-domain measurement functions that allow you to characterize the nonlinearity of a system. Quantitative measures of system linearity are important in a number of applications including audio system analysis, power electronics, and radio-frequency (RF) network analysis.
Use thd
to
measure the total harmonic distortion (THD) of a sinusoidal signal.
THD is appropriate for signals with discrete spectra consisting of
a fundamental frequency and one or more harmonics.
Use sinad
to
measure the signal to noise and distortion ratio (SINAD) of a signal.
SINAD is appropriate for signals with mixed spectra consisting of
a fundamental frequency with one or more harmonics and additive noise.
Use toi
to
measure the third-order intercept (TOI) point. TOI is used to quantify
the intermodulation distortion of a system in response to a two-tone
input.
Use snr
to
measure the signal-to-noise ratio (SNR) of a signal. SNR is appropriate
when you want to measure the power of the fundamental frequency in
comparison to the variance of additive noise.
This release introduces two new functions for raised cosine and Gaussian pulse-shaping filter design. The new functions provide a uniform interface for transmit and receive pulse-shaping filters that is designed specifically for communication engineers.
Raised cosine filters are used to minimize intersymbol interference
(ISI) by shaping pulses with a filter that satisfies the Nyquist ISI
criterion. Use rcosdesign
to
design a raised cosine or square-root raised cosine finite impulse
response pulse-shaping filter.
Gaussian pulse-shaping filters are used in modulation schemes
such as Gaussian minimum shift keying (GMSK) to limit the bandwidth
of pulse waveforms. Use gaussdesign
to
design a Gaussian finite impulse response pulse-shaping filter.
Functionality | What Happens When You Use This Functionality? | Use This Instead | Compatibility Considerations |
---|---|---|---|
firrcos | Still runs | rcosdesign | Replace all instances of firrcos with rcosdesign .
See Migrate from firrcos to rcosdesign . |
gaussfir | Still runs | gaussdesign | Replace all instances of gaussfir with gaussdesign .
See Migrate from gaussfir to gaussdesign . |
fdesign.pulseshaping | Still runs | rcosdesign or gaussdesign | Replace all instances of fdesign.pulseshaping with rcosdesign or gaussdesign ,
as appropriate. See Migrate from fdesign.pulseshaping to rcosdesign or gaussdesign . |
Migrate from firrcos
to rcosdesign
When you update legacy code using firrcos
to
use rcosdesign
instead, keep in mind that rcosdesign
designs
a square-root raised cosine filter by default. If you want a normal
raised cosine filter you must set the optional parameter, shape
,
to 'normal'
. firrcos
has the
opposite default behavior.
The number of samples per symbol must be an integer.
Equivalently, the ratio Fs/(2*Fc)
, where Fs
is
the sampling frequency and Fc
is the cutoff frequency
of the filter, must be an integer.
firrcos
and rcosdesign
use
different normalizations for the filter coefficients. firrcos
normalizes
the coefficient so that the nominal passband gain is 1. rcosdesign
sets
the filter energy to 1.
firrcos | rcosdesign |
---|---|
Fs = 7000; N = 16; Fc = 500; df = 375; ag = firrcos(N,Fc,df,Fs) | R = df/2/Fc; sps = Fs/2/Fc; ga = rcosdesign(R,N/sps,sps,'normal'); ga = ga/max(ga)/sps |
ug = firrcos(N,Fc,df,Fs,'sqrt') | gu = rcosdesign(R,N/sps,sps); gu = gu/max(gu)*(1+R*(4/pi-1))/sps |
N = 16; Fc = 1000; R = 0.25; Fs = 8000; b1 = firrcos(N,Fc,R,Fs, ... 'rolloff','normal') | beta = R; sps = Fs/(2*Fc); span = N / sps; b1n = rcosdesign(beta,span,sps,'normal'); b1n = b1n / max(b1n) / sps |
N = 16; Fc = 1000; R = 0.25; Fs = 8000; b2 = firrcos(N,Fc,R,Fs, ... 'rolloff','sqrt') | beta = R; sps = Fs/(2*Fc); span = N / sps; b2n = rcosdesign(beta,span,sps,'sqrt'); b2n = b2n / max(b2n) /(pi*sps) ... * (pi*(1-beta) + 4*beta) |
Migrate from gaussfir
to gaussdesign
When you update legacy code using gaussfir
to
use gaussdesign
instead, keep in mind the differences
between the arguments in each function. In gaussfir
,
the second argument, nt
, is the number of symbol
periods between the start of the impulse response of the filter and
its peak. It is thus equal to half the second
argument, span
, of gaussdesign
.
Moreover, gaussdesign
has no default values for
its second and third arguments. gaussfir
respectively
has 3 and 2.
gaussfir | gaussdesign |
---|---|
bt = 0.3; h = gaussfir(bt) | hn = gaussdesign(bt,6,2) |
nt = 5; h = gaussfir(bt,nt) | span = 2*nt; hn = gaussdesign(bt,span,2) |
of = 3; h = gaussfir(bt,nt,of) | sps = of; hn = gaussdesign(bt,span,sps) |
Migrate from fdesign.pulseshaping
to rcosdesign
or gaussdesign
The use of fdesign.pulseshaping
is
not recommended. Use rcosdesign
or gaussdesign
,
as appropriate.
fdesign.pulseshaping | rcosdesign, gaussdesign |
---|---|
sps = 6; span = 4; Beta = 0.25; f1 = fdesign.pulseshaping(sps, ... 'Square Root Raised Cosine', ... 'Nsym,Beta',span,Beta); d1 = design(f1); n1 = d1.Numerator | n1n = rcosdesign(Beta,span,sps); n1n = n1n / max(n1n) * (-1/(pi*sps) ... * (pi*(Beta-1) - 4*Beta)) |
g1 = fdesign.pulseshaping(sps, ... 'Square Root Raised Cosine', ... 'N,Beta',sps*span,Beta); h1 = design(g1); k1 = h1.Numerator | k1n = rcosdesign(Beta,span,sps); k1n = k1n / max(k1n) * (-1/(pi*sps) ... * (pi*(Beta-1) - 4*Beta)) |
f2 = fdesign.pulseshaping(sps, ... 'Raised Cosine', ... 'Nsym,Beta',span,Beta); d2 = design(f2); n2 = d2.Numerator | n2n = rcosdesign(Beta,span,sps,'normal'); n2n = n2n/max(abs(n2n))/sps |
g2 = fdesign.pulseshaping(sps, ... 'Raised Cosine', ... 'N,Beta',sps*span,Beta); h2 = design(g2); k2 = h2.Numerator | k2n = rcosdesign(Beta,span,sps,'normal'); k2n = k2n/max(abs(k2n))/sps |
BT = 0.3; f3 = fdesign.pulseshaping(sps, ... 'Gaussian', ... 'Nsym,BT',span,BT); d3 = design(f3); n3 = d3.Numerator | n3n = gaussdesign(BT,span,sps) |
This release introduces three new functions for power measurements
in the frequency domain. To measure the average power contained in
a specified frequency interval, use bandpower
.
You can measure the spurious-free dynamic range using sfdr
.
The spurious-free dynamic range is the ratio of powers for the fundamental,
or carrier frequency, to the next largest component. The function enbw
measures
the equivalent noise bandwidth of a window. The equivalent noise bandwidth
is the width of an ideal rectangular filter with peak power equal
to the peak power of the window. The product of the peak power and
the equivalent noise bandwidth is equal to the power obtained by integrating
under the magnitude-squared Fourier transform of the window function.
This release enhances functions for nonparametric and parametric
power spectral density (PSD) function estimation. In R2013a, nonparametric
PSD estimators provide confidence intervals based on the chi-square
probability distribution. Parametric PSD estimators provide confidence
intervals based on the Gaussian probability distribution. All nonparametric
and parametric PSD estimators include the ability to obtain a two-sided
PSD estimate with 0 frequency (DC) in the center. See periodogram
, pwelch
,
and pmtm
for
information on obtaining confidence intervals and DC centered spectra
for nonparametric PSD estimation. See pburg
, pcov
, pmcov
,
and pyulear
for
information on obtaining confidence intervals and DC centered spectra
for parametric PSD estimation.
This release enhances a number of filter analysis and implementation
functions to accept single-precision floating point inputs. In R2013a,
you can use single-precision filter coefficients with filter analysis
functions, such as freqz
, fvtool
, grpdelay
, impz
, phasez
,
and zerophase
.
If you input single-precision filter coefficients, the analysis function
outputs are single precision.
In R2013a, the filtering function, sosfilt
,
accepts single-precision inputs for both the filter coefficients and
data. If either the second-order section matrix input or the data
is single-precision, sosfilt
outputs
single-precision data.
The filtering functions fftfilt
, filtfilt
,
and latcfilt
still require double-precision
inputs.
This release removes the requirement for DSP System Toolbox™ to generate standalone C and C++ code for supported Signal Processing Toolbox functions. You must have MATLAB Coder™ and Signal Processing Toolbox software to generate standalone C and C++ code for supported functions. See Functions Supported for Code Generation for a list of Signal Processing Toolbox functions with code generation support if you have MATLAB Coder software.
Functionality | What Happens When You Use This Functionality? | Use This Instead | Compatibility Considerations |
---|---|---|---|
dspdata.avgpower | Still runs | bandpower | Replace all instances of dspdata.avgpower with bandpower |
dspdata.msspectrum | Still runs | periodogram , pwelch | Replace all instances of dspdata.msspectrum with
the appropriate function interface |
dspdata.psd | Still runs | pburg , pcov , periodogram , pmcov , pmtm , pwelch , pyulear | Replace all instances of dspdata.psd with
the appropriate function interface |
dspdata.pseudospectrum | Still runs | peig , pmusic | Replace all instances of dspdata.pseudospectrum with
the appropriate function interface |
dspdata.sfdr | Still runs | sfdr | Replace all instances of dspdata.sfdr with sfdr |
sigwin.barthannwin | Still runs | barthannwin | Replace all instances of sigwin.barthannwin with barthannwin |
sigwin.bartlett | Still runs | bartlett | Replace all instances of sigwin.bartlett with bartlett |
sigwin.blackman | Still runs | blackman | Replace all instances of sigwin.blackman with blackman |
sigwin.blackmanharris | Still runs | blackmanharris | Replace all instances of sigwin.blackmanharris with blackmanharris |
sigwin.bohmanwin | Still runs | bohmanwin | Replace all instances of sigwin.bohmanwin with bohmanwin |
sigwin.chebwin | Still runs | chebwin | Replace all instances of sigwin.chebwin with chebwin |
sigwin.flattopwin | Still runs | flattopwin | Replace all instances of sigwin.flattopwin with flattopwin |
sigwin.gausswin | Still runs | gausswin | Replace all instances of sigwin.gausswin with gausswin |
sigwin.hamming | Still runs | hamming | Replace all instances of sigwin.hamming with hamming |
sigwin.hann | Still runs | hann | Replace all instances of sigwin.hann with hann |
sigwin.kaiser | Still runs | kaiser | Replace all instances of sigwin.kaiser with kaiser |
sigwin.nuttallwin | Still runs | nuttallwin | Replace all instances of sigwin.nuttallwin with nuttallwin |
sigwin.parzenwin | Still runs | parzenwin | Replace all instances of sigwin.parzenwin with parzenwin |
sigwin.rectwin | Still runs | rectwin | Replace all instances of sigwin.rectwin with rectwin |
sigwin.taylorwin | Still runs | taylorwin | Replace all instances of sigwin.taylorwin with taylorwin |
sigwin.triang | Still runs | triang | Replace all instances of sigwin.triang with triang |
sigwin.tukeywin | Still runs | tukeywin | Replace all instances of sigwin.tukeywin with tukeywin |
spectrum.burg | Still runs | pburg | Replace all instances of spectrum.burg with pburg |
spectrum.cov | Still runs | pcov | Replace all instances of spectrum.cov with pcov |
spectrum.eig | Still runs | peig | Replace all instances of spectrum.eig with peig |
spectrum.mcov | Still runs | pmcov | Replace all instances of spectrum.mcov with pmcov |
spectrum.mtm | Still runs | pmtm | Replace all instances of spectrum.mtm with pmtm |
spectrum.music | Still runs | pmusic | Replace all instances of spectrum.music with pmusic |
spectrum.periodogram | Still runs | periodogram | Replace all instances of spectrum.periodogram with periodogram |
spectrum.welch | Still runs | pwelch | Replace all instances of spectrum.welch with pwelch |
The R2012b release introduces a revamped Signal Browser GUI. Whenever you select one or more signals in the Signals list box and click the View button, The Signal Browser GUI launches from the SPTool GUI. The Signal Browser allows you to perform the following additional operations:
Time Domain Measurements Panels — The Signal Browser includes new side panels labeled Trace Selection, Cursor Measurements, Signal Statistics, Bilevel Measurements, and Peak Finder.
The Trace Selection panel allows you to choose which signal to make the active signal when you select multiple input signals in the Signals list box in SPTool. This panel replaces the Select Trace toolbar button ( ) and Select a trace dialog box that were used in previous releases to choose an active signal. In the Signal Browser menu, select Tools > Measurements > Trace Selection.
The Cursor Measurements panel displays screen cursors. This panel replaces the Markers menu and all the corresponding toolbar buttons that were used in previous releases to show and configure markers. In the Signal Browser toolbar, click the Cursor Measurements button ( ). Alternatively, in the Signal Browser menu, select Tools > Measurements > Cursor Measurements.
The Signal Statistics panel displays the maximum, minimum, peak-to-peak difference, mean, median, and RMS values of a selected signal. It also displays the times at which the maximum and minimum values occur. In the Signal Browser toolbar, click the Signal Statistics button ( ). Alternatively, in the Signal Browser menu, select Tools > Measurements > Signal Statistics.
The Bilevel Measurements panel displays information about a selected signal's transitions, overshoots or undershoots, and cycles. In the Signal Browser toolbar, click the Bilevel Measurements button ( ). Alternatively, in the Signal Browser menu, select Tools > Measurements > Bilevel Measurements.
The Peak Finder panel displays maxima and the times at which they occur. These displays allow you to modify the settings for peak threshold, maximum number of peaks, and peak excursion. In the Signal Browser toolbar, click the Peak Finder button ( ). Alternatively, in the Signal Browser menu, select Tools > Measurements > Peak Finder.
Multiple Display Support — R2012b enhances the Signal Browser by allowing you to choose to have multiple displays. This feature allows you to tile your screen into a number of separate displays, up to a grid of 4 rows and 4 columns. You may find multiple displays useful when you select multiple input signals in the Signals list box in SPTool. To set the number of displays, in the Signal Browser toolbar, click the Layout button ( ). Alternatively, in the Signal Browser menu, select View > Layout.
Style dialog box — R2012b enhances the Signal Browser by allowing you to customize the style of displays using a Style dialog box. You can change the color of the figure containing the displays, the background and foreground colors of display axes, and properties of lines in a display. The Style dialog box replaces the Line Properties toolbar button and Edit Line dialog box that were used in previous releases for customizing line properties. To view or modify the line style of the active signal, in the Signal Browser menu, select View > Style. You can also right-click on the display and select Style. The Signal Browser — Style dialog box opens.
Sampled data as stairs — The Signal Browser
can display a signal as a stairstep graph.
A stairstep graph includes only horizontal lines and vertical lines.
Each horizontal line represents the signal value for a discrete sample
period and is connected to two vertical lines. Each vertical line
represents a change in values occurring at a sample. Using this approach
is equivalent to using the MATLAB stairs
function.
Stairstep graphs are useful for drawing time history graphs of digitally
sampled data. To display a sampled signal as a Stairstep graph, in
the Signal Browser — Style dialog box, set the Plot
type parameter to Stairs
.
Properties dialog box — R2012b enhances the Signal Browser by providing a central location where you can modify the properties of a display. To change options for a display, in the Signal Browser menu, select View > Properties. You can also right-click on the display and select Properties. The Signal Browser — Visuals:Time Domain Options dialog box opens.
Complex data support — The Signal Browser accepts complex-valued input signals and can visualize them in two distinct fashions. By default, the complex data is displayed in real and imaginary form as different-colored lines on the same axes. Alternately, you can display the magnitude and phase of the signal on separate axes in the same display. To change the complex data options, in the Signal Browser — Visuals:Time Domain Options dialog box, select the Display tab. Then, select or clear the Plot signal(s) as magnitude and phase check box.
Ability to change the time units of the display —
The Signal Browser allows you to label the time-axis
in three different ways. First, you can confirm the default operation,
in which the Signal Browser displays time in metric units. The Signal
Browser chooses the appropriate metric units, based on the minimum time-axis
limit and the maximum time-axis limit of the scope
window. Second, you can verify that the time-axis
is always labeled as Time (seconds)
and that the
appropriate power of 10 appears in the bottom-right corner of the
Time Scope display. Finally, you remove the units in the time-axis
label entirely. To change the manner in which the time units are displayed,
in the Signal Browser — Visuals:Time Domain Options dialog
box, select the Main tab. Then, set the Time
Units parameter to either Metric (based on
Time Span)
, Seconds
, or None
,
respectively.
See the Signal Browser reference topic for more information.
The R2012b Signal Browser GUI replaces the tool available in previous releases that was also named Signal Browser. In R2012b, you can still use the Signal Browser from previous releases, which is hereafter referred to as the Legacy Browser. To do so, execute the following steps.
In the SPTool menu, select File > Preferences.
In the list at the left side, select Signal Browser.
Under Signal Browser, select the Use Legacy Browser (to be removed) check box.
Click OK.
Now when you select one or more signals from the Signals list and click the View button, the Legacy Browser opens.
xcorr
, xcorr2
, fftfilt
, xcov
, and cconv
functionsThe R2012b release introduces GPU acceleration for xcorr
, xcorr2
, fftfilt
, xcov
,
and cconv
.
GPU acceleration for these functions requires Parallel Computing Toolbox™ software
and meeting the GPU computing requirements detailed here:
.
If you have the Parallel Computing Toolbox, you can use gpuArray
to
create a GPUArray object. The supported Signal Processing Toolbox functions
accept GPUArray objects as inputs.
In R2012a, you can perform a number of basic measurements on bilevel pulse waveforms. These measurements include:
State levels — You can estimate the state levels
of a bilevel waveform using the histogram method with statelevels
.
Transition metrics — You can measure the rise
time, fall time, and mid-reference level instants of waveform transitions.
See the help for midcross
, risetime
, falltime
,
and slewrate
for
details. Additionally, R2012a introduces pulse metrics to measure
bilevel waveform behavior in pretransition and posttransition regions
including, overshoot
, undershoot
,
and settlingtime
.
Duration metrics — See the help for pulsewidth
, pulseperiod
, pulsesep
,
and dutycycle
for
details.
In R2011b, you can specify passband and stopband weights for
a fixed-order unconstrained partial band differentiator filter design.
You can access this capability through filterbuilder
and fdesign.differentiator
.
With fdesign.differentiator
, use the
specification string 'N,Fp,Fst'
, and
set the design method to 'equiripple'
.
The following example shows you how to see the passband weight, Wpass
,
and stopband weight, Wstop
, design options.
d = fdesign.differentiator('N,Fp,Fst',30,0.25,0.5); designopts(d,'equiripple')
Specify the passband or stopband weight values when you design your equiripple filter. For example:
Hd = design(d,'equiripple','Wstop',4);
In R2011b, you can specify different numerator and denominator
orders in filterbuilder
for
lowpass and highpass Butterworth (maxflat) designs.
Access this option in filterbuilder
by
setting Impulse response under Filter
specifications to IIR
and Order
mode to Specify
.
The default is equal order for both the numerator and denominator.
Check Denominator order to specify
a different denominator order.
For R2011b, Signal Processing Toolbox error and warning message identifiers have changed.
If you have scripts or functions that use these changed message
identifiers, you must update the code to use the new identifiers.
Typically, message identifiers are used to turn off specific warning
messages. You can also use them in code that uses a try/catch
statement
and performs an action based on a specific error identifier.
If your code checks for a message identifier in a warning or error, you must update it to check for the new warning or error instead. To determine the identifier for a warning, run the following command just after you see the warning:
[MSG,MSGID] = lastwarn;
This command saves the message identifier to the
variable MSGID
.
To determine the identifier for an error, run the following command just after you see the error:
exception = MException.last; MSGID = exception.identifier;
Note: Warning messages indicate a potential issue with your code. While you can turn off a warning, a suggested alternative is to change your code so that your code does not generate warnings. |
In R2011a, there are two major enhancements to filtfilt
:
filtfilt
has
been completely rewritten to improve performance. Actual performance
improvement depends on your hardware, filter length, signal length,
and number of channels.
filtfilt
now
accepts IIR filters in second-order section (biquad) form.
In R2011a, blackmanharris
and nuttallwin
have
a symmetric window design option. The input argument SFLAG
controls
the window option and defaults to 'symmetric'
. 'periodic'
returns
a N-periodic window. The default symmetric option is preferred in
FIR filter design because it results in linear phase. In spectral
analysis applications, the periodic option is preferred.
In releases previous to R2011a, blackmanharris
and nuttallwin
only
return N-periodic windows. To reproduce behavior in R2011a consistent
with behavior in previous releases, use:
win = blackman(N,'periodic'); % N is the window length
win = nuttallwin(N,'periodic'); % N is the window length
In R2011a, rectpuls
returns
a double-precision vector instead of a logical vector.
In previous releases, rectpuls
returns
a logical vector. To produce behavior in R2011a consistent with previous
releases, cast the output of rectpuls
to a logical
vector.
t = linspace(0,1,0.01); y = logical(rectpuls(t));
In R2011a, MathWorks^{®} is no longer using the term Embedded MATLAB to refer to the language subset that supports code generation from MATLAB algorithms. This nomenclature incorrectly implies that the generated code is used in embedded systems only.
The new term is code generation from MATLAB. This terminology better reflects the full extent of the capability for translating MATLAB algorithms into readable, efficient, and compact MEX and C/C++ code for deployment to both desktop and embedded systems.
Signal Processing Toolbox users who have the DSP System Toolbox and MATLAB Coder software can generate deployable C/C++ code and MEX files using supported functions in the Signal Processing Toolbox.
You can find material on using Code Generation from MATLAB with the Signal Processing Toolbox software in Code Generation from MATLAB Support in Signal Processing Toolbox.
Users who have the DSP System Toolbox and Fixed-Point Toolbox™ can
accelerate MEX-files for fixed-point applications using fiaccel
.
The functionality associated with C/C++ and MEX code generation from MATLAB has changed in R2011a. These changes include:
The compiler flag #codegen
replaces #eml
.
codegen
replaces emlc
and emlmex
for
generating deployable C/C++ code and MEX-files from MATLAB algorithms.
You must have the DSP System Toolbox and the MATLAB Coder software
to use codegen
.
fiaccel
replaces emlmex
for
generating fixed-point MEX code from MATLAB algorithms. To use fiaccel
,
you must have the DSP System Toolbox and Fixed-Point Toolbox software
and your MATLAB code must satisfy the conditions described on
the fiaccel
reference
page.
In R2010b, Embedded MATLAB^{®} supports upsample
and downsample
in
the Signal Processing Toolbox.
In R2010a, users can construct dfilt
objects
with single-precision floating point arithmetic. Set the Arithmetic
property
to 'single'
to obtain a single-precision
floating point representation of the filter coefficients.
In R2010a, Embedded MATLAB supports additional functions in the Signal Processing Toolbox. You can find a comprehensive list of supported functions in the Function Library Reference. You can find examples of using supported functions with the Signal Processing Toolbox software at Code Generation from MATLAB Support in Signal Processing Toolbox.
Name | What Happens When you use the Function, Object, Object Method, or Object Property | Use Instead | Compatibility Considerations |
---|---|---|---|
addsection (Method for dfilt.cascade and dfilt.parallel objects) | Errors | addstage method | Replace all instances of the addsection method
with the addstage method for dfilt.cascade and dfilt.parallel objects. |
FFTLength (Property of spectrum objects) | Errors | 'NFFT' parameter in psd or msspectrum methods | Replace all instances of the FFTLength property
in spectrum objects
by the 'NFFT' parameter in the psd or msspectrum methods. |
freqzplot | Warns | fvtool | Replace all instances of freqzplot with fvtool . |
removesection (Method for dfilt.cascade and dfilt.parallel objects) | Errors | removestage method | Replace all instances of the removesection method
with the removestage method for dfilt.cascade and dfilt.parallel objects. |
Section (Property of dfilt.cascade and dfilt.parallel objects) | Errors | Stage property | Replace all instances of the Section property
with the Stage property for dfilt.cascade and dfilt.parallel objects. |
Sidelobe_atten (Property of sigwin.chebwin object) | Errors | SidelobeAtten property | Replace all instances of the Sidelobe_atten property
with the SidelobeAtten property for sigwin.chebwin objects. |
specplot | Warns | plot method with DSPDATA object | Replace all instances of specplot with
the plot method for DSPDATA objects. |
In R2010a, the use of Filter Designer in sptool
is
not recommended. Use fdatool
instead.
Under File —> Preferences-> Filter
Designer
in SPTool, you may still select to use Filter
Designer, but you will receive a warning that Filter Designer will
be removed in a future release. If you select Filter Designer, you
will be prompted to change your preferences in the sigprefs.mat
and startup.spt
files
when you exit SPTool. Changing your preferences to use Filter Designer
results in a warning each time SPTool starts. See Setting
Preferences and Saving
and Loading Sessions for details.
Because a future release will remove Filter Designer , use fdatool
instead.
Filters created in Filter Designer are not compatible with FDATool.
Under File —> Preferences-> Filter Designer
in sptool
,
you can select to use FDATool. You receive a prompt to convert filters
created in Filter Designer to a format compatible with FDATool. When
you exit sptool
after
changing your preferences, you receive another prompt, instructing
you to update your preferences to use FDATool.
In R2009b, Embedded MATLAB supports the generation of embeddable C code for a subset of Signal Processing Toolbox filter design and window generation functions. You must install both the Signal Processing Toolbox and Signal Processing Blockset™ software to use this feature. Depending on which Embedded MATLAB feature you wish to use, additional products are required. The generated C code meets the strict memory and data type requirements of embedded target environments. See Code Generation from MATLAB Support in Signal Processing Toolbox for a list of supported Signal Processing Toolbox functions and examples.
If you use Simulink^{®}, you can now use the new MapCoeffstoPorts
property
with realizemdl
to
map filter coefficients from dfilt
objects
to constant blocks. The coefficients also appear in the MATLAB workspace
providing tunability to the realized Simulink model. See dfilt
for
a list of supported filter structures and any restrictions.
A new and more robust way to design filters has been added to
the toolbox —fdesign
objects
and the filterbuilder
GUI.
The following filter responses are supported: lowpass, highpass, bandpass,
bandstop, Hilbert, differentiator, pulse-shaping (including FIR Gaussian)
and arbitrary magnitude. Advanced design methods and additional responses
are available in Filter Design
Toolbox™.
A new option has been added to the dfilt block method to specify filter coefficients via Simulink block ports.
A new marcumq
function,
which implements the generalized Marcum Q function, has been added
to the toolbox.
You can now specify whether to keep or drop the last taper for
calculating the pmtm
,
which returns the PSD using the Thomson multitaper method. By default,
the last taper is dropped. If you set the DropLastTaper property false,
the last taper is included.
The spectrum
object
has been enhanced with a new method for calculating confidence intervals
for PSDs and mean-squared spectra.
The dspdata
object
has been enhanced with a new method to measure spurious-free dynamic
range (SFDR) for mean-squared spectra.
A new function (cconv
)
that computes circular convolution has been added to the toolbox.
spectrum
objects
now support computing the spectrum and pseudospectrum on a user-specified
vector of frequencies. This vector identifies the frequencies at which
the spectrum or pseudospectrum is calculated.
A more detailed explanation of the factorize algorithm and an
example have been added to the cceps
reference
page.
You can now create a Simulink block from dfilt.statespace
objects
with the dfilt realizemdl
method.
The spectral analysis command line functions (pburg
, pcov
, peig
, periodogram
, pmcov
, pmtm
, pmusic
, pwelch
,
and pyulear
)
now accept a frequency vector as an input parameter. This vector identifies
the frequencies at which the spectral analysis function returns an
estimate.
For functions that use the Goertzel algorithm (periodogram
, pmtm
, pwelch
,
and spectrogram
),
the frequency inputs are rounded to match the nearest bin value used
by the algorithm.
The FFTLength
parameter has been
removed from all spectrum
objects
and you now specify the number of FFT points (NFFT
)
via the psd
, msspectrum
,
or pseudospectrum
estimation method.
The NFFT
value can be an integer or
a string (either 'Nextpow2'
or 'Auto'
). 'Nextpow2'
is
the default and sets the number of FFT points to the next power of
2 greater than the input signal length (or the segment length for spectrum.welch
objects). 'Auto
'
sets the number of FFT points to be equal to the input signal or segment
length.
You should update any existing code that specifies the FFTLength
parameter
and instead use the NFFT
parameter associated with
an estimation method.
You should also verify that any spectrum.welch
objects
are using the desired FFT length, since the FFT length is now based
on the segment length instead of the input signal length.
The dfilt
block
method
now allows you to specify a target subsystem in your Simulink model
where you want to place the block. Two new parameters implement this: 'Destination'
and 'Link2Obj'
.
The 'Destination'
specifies where to
place the block and 'Link2Obj'
creates
a link between the block in your model and inputs from the command
line.
The alpha parameter in the formula used in gaussfir
has
been updated to match the formula in Rappaport T.S., Wireless
Communications Principles and Practice, 2nd Edition, Prentice
Hall, 2001.
FDATool has replaced the SPTool Filter Designer as the preferred method for designing filters for use in SPTool. For details, see FDATool in the Signal Processing Toolbox documentation.
The format in which filters are saved differs between SPTool Filter Designer and FDATool. When you load an SPTool session with saved filters, you are prompted to upgrade your filters to use FDATool format.
A new parameter zeroflag
has been
added to zp2sos
.
This parameter is used for real zeros that are the negatives of each
other. For these zeros it specifies whether to keep them together
instead of ordering them according to proximity to poles.
A new delay structure (dfilt.delay
)
has been added to dfilt
objects. This
structure adds latency to any signal filtered with it.
A Normalize magnitude
option has
been added to the Analysis Parameters of
WinTool and WVTool magnitude plots.
You can draw lines on your filter response in FDATool to indicate rejection areas.
FDATool now supports generating C header files for complex filters.
A new Tip of the Day dialog displays when you start FDATool. It contains tips and hints for using FDATool.
FDATool no longer supports state space filters.
If you load a saved FDATool session that contains a state space filter, it is converted to a direct-form II transposed filter.
Three new options have been added to the Analysis Parameters for magnitude response displays.
Normalize Magnitude to 1 (0 dB) — displays the magnitude so that the maximum magnitude value occurs at 0 dB
Autoscale axes — automatically scales the response data y-axis
dB Display Range — If you are not using autoscale and the magnitude display is in dB, this allows you to specify the y-axis limits, .
The coefficient view in FVTool now displays each section of a second-order section filter as a separate filter with its own numerator, denominator, and gain.
The default units for the phase response in FVTool
have been changed to radians. This is consistent with the phasez
function.
You should verify that the FVTool phase plots generated by existing code display the desired units.
The dfilt
coefficients
method
has been changed to the coeffs
method, which returns
a structure. See the Methods section of dfilt
for
information.
You should update any code that uses the coefficients
method
to use the new coeffs
method and its returned structure.
You cannot pass filter states (initial and final
conditions) via the dfilt
filter
method.
You must use the states
property. See dfilt
for
more information.
You should update any code that passed filter states via the dfilt
filter
method
to use the new states
property.
spectrogram
has
been added to replace the grandfathered specgram
function.
If you use this function with no outputs, a surface plot is displayed,
instead of an image.
You should update any code that references specgram
to
use the new spectrogram
function. spectrogram
uses
different default values than specgram
and the
order of the inputs has changed.
pwelch
(and
the other spectrum analysis functions) no longer accept 'half'
or 'whole'
.
You must use 'onesided'
or 'twosided'
to
indicate the type of analysis you want.
You should update any code that uses 'half'
or 'whole'
with
spectral analysis functions and instead use 'onesided'
or 'twosided'
,
respectively.
The following functions and methods now generate standard MATLAB plots, instead of launching an interactive plot. Refer to the MATLAB documentation for information on plots.
gaussfir
has
been added to replace the grandfathered firgauss
function. gaussfir
uses
parameters that are common to communications systems.
You should update any code that references firgauss
to
use the new gaussfir
.
The firpm
and cfirpm
functions
now take function handles as inputs instead of strings.
You should update any code that uses firpm
or cfirpm
so
that it will work correctly with function handle inputs instead of
string input.
Signal Processing Toolbox demos have been reorganized and a new demo on the analysis of a numerically controlled oscillator (NCO) has been added.
The Filter Wizard no longer requires Filter Design Toolbox software. You can
use the Filter Wizard if you have Signal Processing Toolbox software
and Simulink installed. If you have the Filter
Design Toolbox™ software installed, more options are available.
See dspfwiz
for
more information.
Release | Features or Changes with Compatibility Considerations |
---|---|
R2016a | None |
R2015b | None |
R2015a | None |
R2014b | None |
R2014a | Changes in name-value pair arguments of measurement functions |
R2013b | Functionality Being Removed or Changed |
R2013a | Functionality Being Removed or Changed |
R2012b | Signal Browser in SPTool |
R2012a | None |
R2011b | Conversion of Error and Warning Message Identifiers |
R2011a | |
R2010b | None |
R2010a | |
R2009b | None |
R2009a | None |
R2008b | None |
R2008a | None |
R2007b | None |
R2007a | None |
R2006b | FFT Length in Spectral Analyses Changed |
R2006a | SPTool Filter Designer Replaced by FDATool |
R14SP3 | None |
R14SP2 |
You can also select a location from the following list: