Scribd
Upload
508 views

Finite Word Length Effects

1. When analog signals are converted to digital signals using an analog-to-digital converter (ADC), the signals are sampled, quantized, and assigned a finite number of bits. 2. This quantization process introduces quantization error/noise, where the quantized value differs from the actual value. 3. The quantization error depends on the number of bits used and the type of quantization (e.g. truncation vs. rounding). More bits and rounding reduce the error. 4. Quantization error is modeled as additive noise in digital signal processing systems, where the quantized signal is the sum of the actual signal and error/noise.

Uploaded by

ohmshankar
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
508 views

Finite Word Length Effects

1. When analog signals are converted to digital signals using an analog-to-digital converter (ADC), the signals are sampled, quantized, and assigned a finite number of bits. 2. This quantization process introduces quantization error/noise, where the quantized value differs from the actual value. 3. The quantization error depends on the number of bits used and the type of quantization (e.g. truncation vs. rounding). More bits and rounding reduce the error. 4. Quantization error is modeled as additive noise in digital signal processing systems, where the quantized signal is the sum of the actual signal and error/noise.

Uploaded by

ohmshankar
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 9

FINITE WORD LENGTH

EFFECTS
QUANTIZATION NOISE
• MOST OF THE ENGINEEERING APPLICATIONS USE THE INPUT
SIGNAL AS CONTINOUS TIMR OR ANALOG SIGNAL
• FOR CONVERTING ANALOG SIGNAL TO DIGITAL SIGNAL WE
USE ADC

x(t) x(n) xq(N)


SAMPLER
QUANTIZER

x(t) IS SAMPLED AT REGULAR INVERALS t = n T where n=0,1,2,3


• THEN NUMERIC EQUIVALENT OF EACH SAMPLE X(n) IS
EXPRESSED BY A FINITE NO OF BITS GIVING THE SEQUENCE
Xq(n)
• Difference between the e (n) =xq(n)- x (n)
is called quantizer noise
• Let us assume a sinusoidal signal
between +1 and -1 having a dynamic
range 2
• No of levels for quantizing x(n) is 2b+1
• if employs (b+1) bits
• The intervals bit successive levels
q=2/2b+1=2-b
• q= quantization step size
b= 3 bits then q = 2-3= 0.125

There are two methods for quantization


1.Truncation 2. rounding

Truncation: the process of discarding all the bits less significant than the least
significant bit that retained is called truncation

Eg: 0.00110011 (8bit) to 0.0011 (4 bit)


1.01001001 (8bit) to 1.0100 (4bit)

ROUNDING : rounding of a number of b bits is


accomplished by choosing the round result as the b bit
number to the original number un rounded
Eg : 0.110100 0.110 or 0.111
0.110111111 is round off 0.11011111 or 0.1110000
QUANTIZATION ERROR

• Quantization error occurs when a continuous signal is


converted into digital signal
Analog signal Digital signal

Different types of quantization error


(i) input quantization error (ii) steady state input noise power
(iii) Steady state out put noise power (iv) product quantization error
(v) co efficient quantization error
Input quantization error
The quantization error is given by
e (n) = x q (n) – x (n)
Where x q (n) = sampled quantized value
x (n) = sampled un quantized value
Depending up on the x (n) is quantized different distribution
of quantization may obtained. If rounding off a number is
used to get x q (n) the error signal satisfies the relation
-q/2 < e (n) < q/2
Because quantized signal may be greater or less than the
actual signal
For example let x (n) = (0.70)10=(0.10110011)2
• After rounding x (n) to 3 bit
x q (n) =0.101+
1
0.110 =(0.75)10
Now the error e (n) =x q (n) – x (n) = 0.05
The probability density function p (e) for round off error
and quantization characteristics is show in fig
x q (n)

-q/2
x (n) -q/2 q/2
STEADY STATE INPUT NOISE
• In digital signal processing of analog signals ,the
quantization error is commonly viewed as an additive
noise signal ,that is
x q (n) =x (n) + e (n)

X (t) SAMPLER x (n T) x q (N)


QUANTIZER

x (t) x (n T)
SAMPLER
x q (n) =x (n) + e (n)
+
e (n)

You might also like