Learn more about Signal Processing Blockset

Uniform Encoder - Quantize and encode floating-point input into integer output

Library

Quantizers

dspquant2

Description

The Uniform Encoder block performs the following two operations on each floating-point sample in the input vector or matrix:

Quantizes the value using the same precision
Encodes the quantized floating-point value to an integer value

In the first step, the block quantizes an input value to one of 2^B uniformly spaced levels in the range [-V, (1-2^1-B)V], where you specify B in the Bits parameter and you specify V in the Peak parameter. The quantization process rounds both positive and negative inputs downward to the nearest quantization level, with the exception of those that fall exactly on a quantization boundary. The real and imaginary components of complex inputs are quantized independently.

The number of bits, B, can be any integer value between 2 and 32, inclusive. Inputs greater than (1-2^1-B)V or less than -V saturate at those respective values. The real and imaginary components of complex inputs saturate independently.

In the second step, the quantized floating-point value is uniquely mapped (encoded) to one of 2^B integer values. When the Output type is set to Unsigned integer, the smallest quantized floating-point value, -V, is mapped to the integer 0, and the largest quantized floating-point value, (1-2^1-B)V, is mapped to the integer 2^B-1. Intermediate quantized floating-point values are linearly (uniformly) mapped to the intermediate integers in the range [0, 2^B-1]. For efficiency, the block automatically selects an unsigned output data type (uint8, uint16, or uint32) with the minimum number of bits equal to or greater than B.

When the Output type is set to Signed integer, the smallest quantized floating-point value, -V, is mapped to the integer -2^B-1, and the largest quantized floating-point value, (1-2^1-B)V, is mapped to the integer 2^B-1-1. Intermediate quantized floating-point values are linearly mapped to the intermediate integers in the range [-2^B-1, 2^B-1-1]. The block automatically selects a signed output data type (int8, int16, or int32) with the minimum number of bits equal to or greater than B.

Inputs can be real or complex, double or single precision. The output data types that the block uses are shown in the table below. Note that most of the Signal Processing Blockset blocks accept only double-precision inputs. Use the Simulink Data Type Conversion block to convert integer data types to double precision. See Working with Data Types in the Simulink documentation for a complete discussion of data types, as well as a list of Simulink blocks capable of reduced-precision operations.

Bits	Unsigned Integer	Signed Integer
`2` to `8`	`uint8`	`int8`
`9` to `16`	`uint16`	`int16`
`17` to `32`	`uint32`	`int32`

The Uniform Encoder block operations adhere to the definition for uniform encoding specified in ITU-T Recommendation G.701.

Examples

The following figure illustrates uniform encoding with the following parameter settings:

Peak = 2
Bits = 3
Output type = Unsigned

The real and complex components of each input (horizontal axis) are independently quantized to one of 2³ distinct levels in the range [-2,1.5] and then mapped to one of 2³ integer values in the range [0,7].

-2.0  is mapped to  0
-1.5  is mapped to  1
-1.0  is mapped to  2
-0.5  is mapped to  3
 0.0  is mapped to  4
 0.5  is mapped to  5
 1.0  is mapped to  6
 1.5  is mapped to  7

The table below shows the results for a few particular inputs.

Input	Quantized Input	Output	Notes
`1.6`	`1.5+0.0i`	`7+4i`
`-0.4`	`-0.5+0.0i`	`3+4i`
`-3.2`	`-2.0+0.0i`	`4i`	Saturation (real)
`0.4-1.2i`	`0.0-1.5i`	`4+i`
`0.4-6.0i`	`0.0-2.0i`	`4`	Saturation (imaginary)
`-4.2+3.5i`	`-2.0+2.0i`	`7i`	Saturation (real and imaginary)

The output data type is automatically set to uint8, the most efficient format for this input range.

Dialog Box

Peak: The largest input amplitude to be encoded, V. Real or imaginary input values greater than (1-2^1-B)V or less than -V saturate (independently for complex inputs) at those limits.
Bits: Specify the number of bits, B, needed to represent the integer output. The number of levels at which the block quantizes the floating-point input is 2^B.
Output type: The data type of the block's output, Unsigned integer or Signed integer. Unsigned outputs are uint8, uint16, or uint32, while signed outputs are int8, int16, or int32.

References

General Aspects of Digital Transmission Systems: Vocabulary of Digital Transmission and Multiplexing, and Pulse Code Modulation (PCM) Terms, International Telecommunication Union, ITU-T Recommendation G.701, March, 1993

Supported Data Types

Double-precision floating point
Single-precision floating point