Subject: Communications 2

Code: EEC 223

Presented By:

Dr. Mohamed Youssef

E-mail:
[email protected]
Agenda
• Challenges.

• DPCM.
• Advantages of DPCM.
• Enhancement of DPCM Performance.
• Estimation (Prediction).

• First Order Linear Predictor.

• Second Order Linear Predictor.
• DPCM Encoder and Decoder.
 ❑ PCM is not a very efficient system because it generates so
many bits and requires so much bandwidth to transmit.

Challenges: ❑ Many different ideas have been proposed to improve the

encoding efficiency of A/D conversion. Differential Pulse
Code Modulation (DPCM) is one such scheme.
✓ The sampling process is a reversable process if we follow the sampling theorem, so
we will reconstruct the samples at receiver if we use reconstruction filter.
✓ The quantization process involves some losses or distortion, and that distortion is
due to the quantization and approximation. And if you recall we found that the
quantization noise power is proportional to step size.
✓ The step size is the difference between two successive levels.
✓ The less the value of step size, the less the quantization error.
✓ However, the less the value of step size, the more the quantization levels, the more
the number of bits per level, the more the required bandwidth which sometimes
cannot be tolerated.
✓ The main motivation behind DPCM is how to address the tradeoff between requiring
higher bandwidth to achieve the better performance in terms of SQNR.
Differential Pulse Code Modulation (DPCM):
In analog messages we can make a good guess about a sample value from knowledge of past
sample values. In other words, the sample values are not independent, and generally there is
a great deal of redundancy in the Nyquist samples. Proper exploitation of this redundancy
leads to encoding a signal with fewer bits.

Instead of transmitting the sample values, we transmit the difference between the successive
sample values. Thus, if m[k] is the kth sample, instead of transmitting m[k], we transmit the
difference.

At the receiver, knowing d[k] and several previous sample value m[k -1] , we can reconstruct
m[k]. Thus, from knowledge of the difference d[k] , we can reconstruct m[k] iteratively at the
receiver.
 Assume that we have the following signal plotted in red. This is a continuous
signal (analog signal) and in this example it is sampled using sampling interval
Example 1: equal one second.
Let us assume that quantization levels = 12 levels, so the step size will be 1 ➔
this for the following signal ranging from 0v to 12v.
Example 2:
Example 3:

The higher the number of samples by reducing the sampling interval, the lower the
maximum amplitude of the differential signal.

Representing the differential signal as showing to us. The signal that represent the
difference is much smaller in amplitude than the original signal without any loss of
information because from this differential signal we can reconstruct the original
signal (just by an accumulation).
Advantages of DPCM:
Now, the difference between successive samples is generally much smaller than the
sample values. Thus, the peak amplitude mp of the transmitted values is reduced
considerably. Because the quantization interval ∆v = mp/L, for a given L ( or n ), this
reduces the quantization interval ∆v, thus reducing the quantization noise. This means
that for a given n (or transmission bandwidth), we can increase the SNR, or for a given
SNR, we can reduce n (or transmission bandwidth).

𝒎𝒂𝒙 −𝒎𝒊𝒏
∆v =
𝑳

Performing the quantization levels for differential signal instead of original signal
which will reduce the quantization error and increasing the SQNR.
Enhancement of DPCM Performance:
We can improve upon this scheme by estimating (predicting) the value of the kth sample m[k]
from a knowledge of several previous sample values. If this estimate is m [k], then we transmit
ෝ
the difference (prediction error) d[k] = m[k] - 𝒎[𝒌].

At the receiver also, we determine the estimate 𝑚[𝑘]

ෝ from the previous sample values, and
then generate m[k] by adding the received d[k] to the estimate 𝑚[𝑘].
ෝ Thus, we reconstruct
the samples at the receiver iteratively. If our prediction is worth its salt, the predicted
(estimated) value 𝑚ෝ 𝑘 will be close to m[k], and their difference (prediction error) d [k] will
be even smaller than the difference between the successive samples.

Consequently, this scheme, known as the differential PCM (DPCM), is superior to the naive
prediction described in the preceding paragraph, which is a special case of DPCM, where the
estimate of a sample value is taken as the previous sample value, that is, m[k] = m[k - 1].

The best reduction of the signal range to improve the performance of DPCM is if we find the
difference between the current signal and a value which is related or close enough to this
current sample.
Estimation (Prediction):
Instead of find the difference between the current sample and the immediate previous
sample, we are going to find the difference between the current sample and all the previous
samples. This estimate might be the closest estimate to the actual value of the signal.

The closer that estimate to the actual value, the less the range of the difference signal.

We create such difference between the current sample and estimate based on the
previous samples.

ෝ
d[k] = m[k] - 𝒎[𝒌]

ෝ
𝒎[𝒌] = σ𝑁
𝑖=1 𝑎𝑖𝑚[𝑘 − 𝑖]
Estimation (Prediction) – Cont.:
Simply the estimation (or prediction) of a certain value would be the weighted average
of the previous n values of the same signal.

ෝ
𝒎[𝒌] = σ𝑁
𝑖=1 𝑎𝑖𝑚[𝑘 − 𝑖]
First Order Linear Predictor

It is based only on the previous sample and only one delay element where, a1=1

m[k] m[k-1] = 𝒎[𝒌]

ෝ
Ts
Second
Order
Linear
Predictor

Second order linear predictor uses two previous samples and has two delay elements

m[k] m[k-1] m[k-2]

Ts Ts
𝒎 𝒌−𝟏 −𝒎[𝒌−𝟐]
ෝ 𝒌 =
𝒎
𝟐
1/2 1/2

ෝ
𝒎[𝒌]
σ
It is not necessary that all weights are equal, in general the weights are decreasing from
a1 to aN.
DPCM Encoder and Decoder:

Type equation here.

We transmit the differential signal d[k], and then when we receive the differential signal
at the receiver, we generate the estimate 𝒎 ෝ 𝒌 and add the received d(k) that will be the
recovered signal.
Tx Rx
ෝ
d[k] = m[k] - 𝒎[𝒌] ෝ 𝒌 + d[k]
m[k] = 𝒎

So, we need to estimate 𝒎[𝒌]

ෝ ➔ to add it to the received differential signal d[k] ➔ to
reconstruct m[k].
 ❑ The main challenge here that we don't transmit d[k], we transmit the
quantized version of dq(k). so, we will have an error due to adding dq(k) instead
of d(k).

Challenge:
Reversable
Version
Example:
Solution:

m[k] 9 9.15 9.37 9.5 9.41 9.23 9.1

d[k] 9 0.15 0.17 0.1 -0.09 -0.17 -0.1

dq[k] 9 0.2 0.2 0.1 -0.1 -0.2 -0.1

mq[k] 9 9.2 9.4 9.5 9.4 9.2 9.1

ෝ q[k]
𝒎 0 9 9.2 9.4 9.5 9.4 9.2
Example (Cont.):
Solution (Cont.):
dq[k] 9 0.2 0.2 0.1 -0.1 -0.2 -0.1

mq[k] 9 9.2 9.4 9.5 9.4 9.2 9.1

ෝ q[k]
𝒎 0 9 9.2 9.4 9.5 9.4 9.2

➢ Compare between m[k] and mq[k] to calculate instant error q[k]

m[k] 9 9.15 9.37 9.5 9.41 9.23 9.1

mq[k] 9 9.2 9.4 9.5 9.4 9.2 9.1

q[k] 0 -0.05 -0.03 0.5 0.01 0.03 0

Thank you
Dr / Mohamed Youssef
01005150491
[email protected]