Worldwide
Search:

DSP

-

Features

The SystemVue DSP Library supports the IEEE and C4x fixed and floating-point arithmetic formats found in today's DSP chips. Its tokens allow you to specify the bit length of both the exponent and the mantissa for flexible DSP implementation.

Screenshot

Overview

The SystemVue DSP Library supports fixed and floating-point arithmetic formats found in today’s popular DSP chips. Both IEEE and C4x floating-point formats are supported by the library. While the IEEE format is nearly universal, the library includes support for several C4x formats, including standard and extended modes, as implemented on popular Texas Instruments floating-point DSPs. The DSP Library tokens provide the ability to specify the bit length of both the exponent and the mantissa, and to inspect the resulting condition codes when simulating floating-point operations.

Variable precision and condition codes are supported for integer operations as well. A set of elementary arithmetic operations are included, such as Negate, Add, Subtract, Multiply and Divide. Integer operations include AND, OR, NOT, XOR, SHIFT, rotate and bit reversal. Fixed-point integer operations allow values to be represented in integer and fractional form.

In addition to modeling algorithms for a particular DSP, the DSP Library can also be used in the design of custom DSP devices. For example, the designer can quickly try out different tokens and optimize a design based on the results of successive SystemVue simulations.

The DSP library also provides a wide selection of high-level processing functions. These include a mixed-radix FFT, a complete FIR and IIR filter graphical design environment, and a family of block transforms including the Hadamard, and Discrete Cosine and Sine transforms. Other DSP library tokens include an interpolator, a detrend operator, a multiply and accumulate operator and a FIFO. Design Example

The DSP system in this example uses inputs x and y to compute a bit-true arctan(y,x) output. The four quadrant arctangent function described below is used in many DSP applications for such procedures as Cartesian-to-Polar coordinate conversion.

The DSP system above uses inputs x and y to compute an arctan(y,x) output.

In this example file, two continuous signal sinusoids are digitized into 12-bit signed integers using the DSP Library Converter tokens. The input x and y components are reduced to a first- quadrant calculation of the arctangent by taking the absolute value of each input using Negate and Switch tokens (the Switch token takes its control from the Sign condition code output of the converter).

A further range reduction is taken by using the identity arctangent(abs(y/x)) is equal to Pi/2-arctangent(abs(x/y)). Thus, we can choose the smaller of y/x or x/y as input into a Chebychev polynomial evaluation block within the meta token. Comparison of the x and y is accomplished with the Subtract token. The Sign condition code output is then used to choose the smaller of x and y to be the numerator and the larger to be the denominator for input to a Divide token, after converting each to a fix point 12.11 quantity so that the quotient is expressed in [0,1]. Note that the Divide Overflow condition code that is a result of a divide-by-zero condition drives the arctangent output to zero.

Screenshot

1Hz sinusoid inputs drive the output from 0 radians to
+Pi to -Pi to 0, and so forth, in this SystemVue output window.

The MetaSystem token performs the polynomial calculation using Add and Multiply tokens after converting the 12.11 fix point format to a 12.9 format to cover the arctangent output range [-Pi,+Pi]. The MetaSystem also inputs the result of the input comparison to output either arctangent(abs(y/x)) or Pi-arctangent(abs(x/y)).

If the MetaSystem output is z, then z, Pi-z, z-Pi and -z are the arctangent results for quadrants 0,1,2 and 3 respectively. Quadrant information is fed from the Sign condition code output of the input Converter tokens and combined with the Overflow condition code output of the Divide token to create an input to the quadrant switch. The quadrant switch then selects which quadrant calculation, if any, is to be output in the fixed point 12.9 format

Library Tokens

Arithmetic Models

  • Adder
  • Adder (2 Port)
  • Constant Multiply
  • Divider
  • Multiplier
  • Multiplier (2 Port)
  • Negate
  • Reciprocal
  • Subtract

Bit Logic Models

  • Bit Extract
  • Bit To Symbol
  • Bitwise AND
  • Bitwise Exclusive OR
  • Bitwise NOT
  • Bitwise OR
  • Reverse
  • Rotate
  • Shift
  • Symbol To Bit

Input/Output Models

  • Converter
  • FIFO Buffer
  • Lookup Table

Signal Processing Models

  • Comb Filter
  • Convolution
  • Cross Correlator
  • Detrend
  • Discrete Cosine Transform
  • Discrete Hadamard Transform
  • Discrete Sine Transform
  • Fourier Transform
  • Interpolator
  • Real Input Fourier Transform

Operators

  • Integrator
  • Multiply Accumulate
  • Sine Cosine Table Lookup
  • Square Root
  • Filter (located in SystemVue Operator Library)
  • Bit-True System Design and Test
  • Variable-Precision Support for C4x and IEEE Floating-Point Formats
  • Variable-Precision Support for Unsigned or Signed Two's Complement Integer Formats
  • Fixed Binary-point Integer Operations
  • Primitive Integer Operations, such as Shift and Rotate
  • Condition Outputs such as Underflow/Overflow Resulting from DSP Token Operation
  • Various High-Level Functions Implemented with DSP Library Primitive Operations, such as FIR/IIR Filter Design, Detrend, Cross-Correlation, and FIFOs
  • Various Discrete Transform Operators Including Mixed-Radix FFT and Discrete Sine/Cosine/Hadamard Transforms

Additional Resources

Read the manual (PDF 840kb)

top of pagetop of page     printer-friendly versionprinter-friendly version     email this pageemail this page
* Contact Agilent EEsof
* Agilent EEsof EDA
   
* Product Information
* Support and Services
* Applications/Examples
* Education and Training
* Product Documentation
* Online Product Demos
* News and Events
   
* Subscribe to Newsletter
* Newsletter Archive
* Site Help

.

Click Here for RSS Feeds Subscribe now for instant product, support, and application news!