DSP Blockset Release Notes

Enhanced Blocks

For summaries of individual block enhancements for Release 13 and links to their online reference pages, see the following sections:

• Audio Blocks -- New 24- and 32-Bit Support
• Autocorrelation, Correlation, Convolution -- New Optimization Options
• Cumulative Sum -- Intraframe Running Sums of Frame-Based Columns
• DCT and IDCT -- New Optimization Options
• Digital Filter Design -- Supports More Filter Structures
• Dyadic Analysis Filter Bank and Dyadic Synthesis Filter Bank Enhancements
• Filter Realization Wizard --  New Filter Design Options, Better Fixed-Point Support, More Structures
• Random Source -- Better Random Signals and New Option
• Sine Wave -- Accepts Zero and Negative Frequencies

Audio Blocks -- New 24- and 32-Bit Support

The To Wave Device and From Wave Device blocks now support 24-bit audio devices in addition to 8- and 16-bit audio devices.

The To Wave File and From Wave File blocks now support 24 bits per sample and 32 bits per sample in addition to the previous 8- and 16-bit support.

All four audio blocks are now located in a new library. For more information, see Audio Blocks Relocated.

Autocorrelation, Correlation, Convolution -- New Optimization Options

The Autocorrelation, Correlation, and Convolution blocks used to compute only in the time domain. Now you can set their computation domains to one of the following:

• Time domain -- Minimizes memory use
• Frequency domain -- Depending on the input length, may require fewer computations than computing in the time domain

The Correlation and Convolution blocks also have a third computation domain option, Fastest domain, which computes in the domain that minimizes the number of computations (time domain or frequency domain). The Autocorrelation block does not yet support this option.

Cumulative Sum -- Intraframe Running Sums of Frame-Based Columns

The Cumulative Sum block can now compute a running sum of each column of a frame-based input, independent of the running sum of columns of previous inputs.

DCT and IDCT -- New Optimization Options

The DCT and IDCT blocks use sine and cosine values to compute the discrete cosine transform and its inverse. You can now set the blocks to compute sines and cosines in one of the following ways:

• Table lookup -- Look up sine and cosine values in a speed-optimized table
• Trigonometric function calls -- Make sine and cosine function calls

Digital Filter Design -- Supports More Filter Structures

The Digital Filter Design block, which previously supported only the direct form II transposed structure, now supports all of the following structures:

• Direct form I
• Direct form II
• Direct form I transposed
• Direct form II transposed
• Direct form FIR
• Direct form FIR transposed

Dyadic Analysis Filter Bank and Dyadic Synthesis Filter Bank Enhancements

The Dyadic Analysis Filter Bank block and Dyadic Synthesis Filter Bank block (referred to as "analysis block" and "synthesis block") share several changes and enhancements described below.

New Wavelet Option.   Both blocks now allow you to specify wavelet-based highpass and lowpass filters in addition to the previously supported user-defined filters. Use of wavelets requires the Wavelet Toolbox.

New Option for Single-Port Inputs and Outputs.   Previously, the analysis block output each subband on a separate output port and the synthesis block accepted each subband through a separate input port. Now the analysis block provides an option to output a single vector or matrix of concatenated subbands, and the synthesis block provides an option to accept such inputs through a single input port.

Frame-Based Support Only.   The analysis block now accepts only frame-based input signals. The synthesis block now always outputs frame-based signals. To decompose or reconstruct a sample-based signal using filter banks, you can create your own filter banks using the new Two-Channel Analysis Subband Filter and Two-Channel Synthesis Subband Filter blocks in the Multirate Filters library (see Two-Channel Analysis Subband Filter and Two-Channel Synthesis Subband Filter).

Filter Realization Wizard -- 
New Filter Design Options, Better Fixed-Point Support, More Structures

The Filter Realization Wizard block has several significant enhancements described below.

New Interface Provides Filter Design and Analysis Options.   The Filter Realization Wizard block interface is now a part of the Filter Design and Analysis Tool (FDATool) GUI. Previously, the block required you to specify the filter by typing in its coefficients (you had to predesign the filter elsewhere). You now have the option to design your filter within the block; the extensive set of filter design and analysis tools in FDATool allow the block to automatically implement your filter design without you having to type its coefficients.

Enhanced Fixed-Point Support.   The Filter Realization Wizard block now better supports fixed-point filters due to the following enhancements:

• New fixed-point filter design capabilities -- The block now allows you to design fixed-point filters by using the Set Quantization Parameters panel in the new block interface. Use of this panel requires the Filter Design Toolbox.
• Better fixed-point filter implementation -- The block now implements a much better fixed-point filter when you install the Fixed-Point Blockset and Filter Design Toolbox. To implement a good fixed-point filter, you should first design a fixed-point filter as described above and then set the block to implement the filter using Fixed-Point Blockset blocks.

Corresponding dfilt and qfilt Methods.   dfilt (dfilt.calattice and dfilt.calatticepc) and qfilt objects now have a new realizemdl method. This method allows you to access the filter realization capabilities of the Filter Realization Wizard block from the command line.

Supports More Filter Structures.   The block can now realize filters using any of the following filter structures:

Direct form I Direct form II Direct form I transposed Direct form II transposed Second order sections for direct form I and II, and their transposes Direct form FIR Direct form FIR transposed Direct form asymmetric FIR Direct form symmetric FIR

Lattice ARMA Lattice AR Lattice MA (same as lattice minimum phase) Lattice all-pass Lattice maximum phase Cascade Parallel

You may not be able to directly access some of the above structures through the block interface, but you can access all of them by creating a qfilt or dfilt object with the desired structure, and then importing the filter into the block.

Random Source -- Better Random Signals and New Option

The Random Source block now outputs much better random signals due to the following algorithmic enhancements:

• Uses improved random generators -- The block now uses the same random generator algorithms as the current rand and randn functions. These generators yield much better random signals than the MATLAB Version 4 generators that were previously used by the Random Source block.
• Better automatic seed generation -- When you opt to specify a single initial seed for a multichannel output, the block automatically generates an initial seed for each channel (using the seed you provide). The block now generates these initial seeds using an improved method that results in better random signals.
• Better complex random signal generation -- The block now generates complex random values using an improved method that results in better complex random signals.

The Random Source block also now offers a second method of computing Gaussian (normally distributed) random signals based on the central limit theorem. When set to use this new method, the block computes Gaussian random values by adding and scaling uniformly distributed random signals. The block allows you to specify the number of uniform values to sum.

Sine Wave -- Accepts Zero and Negative Frequencies

The Sine Wave block now accepts zero and negative frequency values in addition to positive frequency values.

New Blocks

New and Enhanced Demos