Cast and Quantize Data

Quantization to fixed-point and floating-point data types, casting between data types

When designing fixed-point algorithms, use cast, zeros, ones, eye, and subsasgn to separate the core algorithm from data type definitions. These functions allow you to use double-precision, single-precision, and fixed-point data types in the same code. Manual Fixed-Point Conversion Best Practices describes how to get from generic MATLAB^® code to an efficient fixed-point implementation.

To simulate full-precision arithmetic using doubles and quantize only at the output of the algorithm, use quantizenumeric.

To simulate floating-point data with arbitrary word length and exponent length, use the quantizer object and associated object functions.

Functions

expand all

Cast Between Types

`eye`	Create identity matrix with fixed-point properties
`ones`	Create array of all ones with fixed-point properties
`zeros`	Create array of all zeros with fixed-point properties
`cast`	Cast variable to different data type
`cast64BitFiToInt`	Cast `fi` object types that can be exactly represented to a 64-bit integer data type (Since R2020a)
`cast64BitIntToFi`	Cast 64-bit integer types to an equivalent `fi` object type (Since R2020a)
`castFiToInt`	Cast fi object to equivalent integer data type (Since R2020a)
`castFiToMATLAB`	Cast `fi` object type to an equivalent built-in MATLAB data type (Since R2020a)
`castIntToFi`	Cast an integer data type to equivalent `fi` type (Since R2020a)
`subsasgn`	Subscripted assignment
`subsref`	Subscripted reference

Quantize Numeric Data

quantizenumeric Quantize numeric data

Quantize Doubles Using `quantizer` Objects

`quantizer`	Create `quantizer` object
`quantize`	Quantize numeric data using `quantizer` object
`assignmentquantizer`	Create `quantizer` object with `fi` object attributes
`copyobj`	Make independent copy of `quantizer` object
`randquant`	Generate uniformly distributed, quantized random number using `quantizer` object
`reset`	Reset objects to initial conditions
`unitquantize`	Quantize numeric data using `quantizer` object except numbers within `eps` of `+1`
`unitquantizer`	Create `unitquantizer` object

Properties of `quantizer` Objects

`denormalmax`	Largest denormalized quantized number for `quantizer` object
`denormalmin`	Smallest denormalized quantized number for `quantizer` object
`eps`	Quantized relative accuracy for `fi` or `quantizer` objects
`exponentbias`	Exponent bias for `quantizer` object
`exponentlength`	Exponent length of `quantizer` object
`exponentmax`	Maximum exponent for `quantizer` object
`exponentmin`	Minimum exponent for `quantizer` object
`fractionlength`	Fraction length of `quantizer` object
`get`	Property values of object
`lsb`	Scaling of least significant bit of `fi` object, or value of least significant bit of `quantizer` object
`range`	Numerical range of `fi` or `quantizer` object
`realmax`	Largest positive fixed-point value or quantized number
`realmin`	Smallest positive normalized fixed-point value or quantized number
`set`	Set or display property values for `quantizer` objects
`tostring`	Convert `fi`, `fimath`, `numerictype`, or `quantizer` object to string
`wordlength`	Word length of `quantizer` object

Logging with `quantizer` Objects

`logreport`	Quantization report
`maxlog`	Log maximums
`minlog`	Log minimums
`noperations`	Number of quantization operations by `quantizer` object
`noverflows`	Number of overflows
`nunderflows`	Number of underflows
`resetlog`	Clear log for `fi` or `quantizer` object

Transformations Using `quantizer` Objects

`bin2num`	Convert two's complement binary string to number using `quantizer` object
`hex2num`	Convert hexadecimal string to number using `quantizer` object
`num2bin`	Convert number to binary representation using `quantizer` object
`num2hex`	Convert number to hexadecimal equivalent using `quantizer` object
`num2int`	Convert number to signed integer using `quantizer` object

Relational Operations for `quantizer` Objects

`isequal`	Determine whether real-world values of two `fi` objects are equal, or determine whether properties of two `fimath`, `numerictype`, or `quantizer` objects are equal
`isfixed`	Determine whether input is fixed-point data type
`isfloat`	Determine whether input is floating-point data type
`isquantizer`	Determine whether input is `quantizer` object

Rounding and Error with `quantizer` Objects

`errmean`	Mean of quantization error
`errpdf`	Probability density function of quantization error
`errvar`	Variance of quantization error
`round`	Round `fi` object toward nearest integer or round input data using `quantizer` object

Quantize Fixed Point `fi` Data with `fixed.Quantizer`

`fixed.Quantizer`	Quantize fixed-point numbers
`quantize`	Quantize `fi` values using `fixed.Quantizer` object

Topics

Transformations for Quantized Data
How to transform quantized data between representations.
Manual Fixed-Point Conversion Best Practices
Get from generic MATLAB code to an efficient fixed-point implementation using manual fixed-point conversion.