telecommunications switching systems 13

7/29/2019 Telecommunications Switching Systems 13

1/16

DPCM, DM

L-13

DPCM

Designed to take advantage of sample-to-sample redundancies.

Since the range of differences is less than therange of individual samples, fewer bits arerequired to encode difference samples.

Sampling rate, band limiting filter and smoothingfilter are identical to those used in conventionalPCM systems.


2/16

Concept

Store the previous sample using S/Hcircuit and use a subtractor to measure thechange.

The change in the signal is then quantizedand encoded for transmission.

DPCM

The previous input value is reconstructed by afeedback loop that integrates the encodedsample differences.

In essence the feedback signal is an estimate ofthe input signal as obtained by integrating theencoded sample differences.

Thus, the feedback signal is obtained in thesame manner used to reconstruct the waveformin the decoder.


3/16

Functional Block Diagram - DPCM

Advantage of feedback loop

Feedback loop at transmitter side does notallow quantization errors to getaccumulated.

If the feedback signal drifts from the inputsignal, as a result of an accumulation of

quantization errors, the next encoding ofthe difference signal automaticallycompensates for the drift.


4/16

DPCM (Analog Integration)

31001 (1)1.123.1

52010 (2)1.734.7

20000 (0)0.322.3

22010 (2)1.701.7

X+PXBDPN

DPCM (Digital Integration)

001 (1)1.123.1

010 (2)1.734.7

000 (0)0.322.3010 (2)1.701.7

BDPN


5/16

DPCM (Digital Differencing)

2 (010)

1 (001)0 (000)

2 (010)

D

3233.1

5354.7

2222.3

2021.7

REG = D+PPBN

DPCM Decoders The decoders are exactly like the feedback

implementations in the encoders. This reinforces the fact that the feedback loop

generates an approximation of the input signal(delayed by one sample).

If no channel errors occur, the decoder output(before filtering) is identical to the feedbacksignal.

Thus the closer the feedback signal matches theinput, the closer the decoder output matches theencoder input.


6/16

DPCM

A/D process can be uniform or companded

Some DPCM systems also use Higher OrderPrediction.

The feedback signal of DPCM systemrepresents first order prediction of the next

sample value, an the sample difference is aprediction error.

Higher order prediction

DPCM concept can be extended to incorporatemore than one past sample value into theprediction circuitry.

The additional redundancy available from allprevious samples can be weighted and summedto produce a better estimate of the next inputsample.

With a better estimate, the range of theprediction error decreases to allow encodingwith fewer bits.


7/16

Third Order Prediction

For systems with constant predictorcoefficients, results have shown that mostrealizable improvement occurs when usinglast three sample values.

DPCM (3rd Order Prediction)


8/16

DPCM

DPCM systems with first-order predicationtypically provides a 1-bit-per-sample reduction incode length relative to PCM systems withequivalent performance.

Extended DPCM systems utilizing third orderprediction can provide reductions of 2 bits persample.

Thus a standard DPCM systems can provide 64-kbps PCM quality at 56-kbps and third-orderprediction can provide comparable quality at 48-kbps.

ADPCM

ITU-T established a 32 kbps ADPCMstandard (Recommendation G.721)

The 32-kbps rate imples a 2:1 savings inchannel BW w.r.t standard PCM.

The G.721 standard is conceptually similar

to DPCM (3rd

order prediction) but uses 8th

order predictor, adaptive quantization andadaptive prediction.


9/16

MOS

The mean opinion score(MOS) is a method ofevaluating speech quality.

The MOS method usestrained listeners toevaluate the speechquality on a scale of 1:5.

Subjective evaluation ofG.721 algorithm usingMOS method of

evaluating speech quality(with carbon microphone)is shown:

Delta Modulation

DM can be considered as a special case ofDPCM using only 1-bit per sample of thedifference signal.

The single bit specifies merely the polarity of thedifference sample and thereby indicates whetherthe signal has increased or decreased since thelast sample.

In this way the input is encoded as a sequenceof ups and downs in a manner resembling astaircase.


10/16

DM

Since each encoded sample contains arelatively small amount of information (1-bit), DMsystems require a higher sampling rate thanPCM or multi-bit DPCM systems.

In fact, the sampling rate is necessarily muchhigher than the minimum (Nyquist) samplingrate.

The relatively high sampling rate produces awider separation between replicas of samplespectrum, thus preventing fold-over distortion.


11/16

DM The main attraction of DM is its simplicity. A/D is provided by a simple comparator. A positive difference produces a 1 and a negative

difference produces a 0. D/A function in feedback path and decoder, is a two-

polarity pulse generator. In its simplest for, a capacitor can work as integrator.


12/16

Slope Overload

sqfdt

tdx

)(becauseoccursoverloadSlope

max

)(overloadSlopepreventTo

dt

tdxqfs

Granular Noise

Granular noise is predominant consideration forslowly changing signals, whereas slope overloadis dominant during rapidly changing signals.

Granular noise is small if step sizes are small,but small step sizes increase the likelihood ofslope overload.

The design of DM necessarily involves a trade-off between two types of distortion.


13/16

DM

Slope overload noise has strong componentsidentical in frequency and approximately inphase with a major component of input.

Distortion that is correlated in this manner to thespeech energy is effectively masked by thespeech energy and therefore is less noticeablethen uncorrelated distortion.

In fact, overload noise in much lessobjectionable to a listener than random orgranular noise at an equivalent power level.

Adaptive Delta Modulation


14/16

ADM

In ADM, if DAC o/p amplitude is still belowthe sample amplitude, the next step size isincreased till DAC catches up with analogsignal.

When an alternative 1s and 0s occur, stepsize is reduced.

Dithering Dithering involves the deliberate addition of

noise to our input signal. It helps by smearing out the little differences in

amplitude resolution. The key is to add random noise in a way that

makes the signal bounce back and forthbetween successive levels.

Of course, this in itself just makes the signalnoisier. But, the signal smoothes out byaveraging this noise digitally once the signal isacquired


15/16

Dithering

G.7xx, including G.711, G.721, G.722, G.726, G.727,G.728, G.729, is a suite of ITU-T standards for audiocompression and de-compression.

It is primarily used in telephony.

In telephony, there are 2 main algorithms defined in thestandard, -law algorithm and A-law algorithm.

Both are logarithmic, but the later a-law was specificallydesigned to be simpler for a computer to process.

G.7xx: Audio (Voice) Compression Protocols (G.711, G.721, G.722,

G.726, G.727, G.728, G.729)


16/16

824/40

Extensions ofRecommendation

G.721 adaptivedifferential pulse codemodulation to 24 and40 kbit/s for digitalcircuit multiplicationequipment application

ITU-TG.723

1624/32

Coding at 24 and 32 kbit/sfor hands-freeoperation in systemswith low frame loss

ITU-TG.722.1

16647 kHz audio-coding within

64 kbit/sITU-TG.722

832Adaptive differential pulse

code modulation(ADPCM)

ITU-TG.721

864Pulse code modulation

(PCM)ITU-TG.711

832ADPCMIntel, IMA(ADPCM) DVI

sampling rate

(kHz)

bit rate

(kb/s)description

standardized

byName

Comparison of G.7xx Protocol

IMA: Interactive Multimedia AssociationDVI: Digital Visual/ Video Interface/Interactive

88Coding of speech at 8 kbit/s using conjugate-structure algebraic-code-excited linear-prediction(CS-ACELP)

ITU-TG.729

816Coding of speech at 16 kbit/s using low-delaycode excited linear prediction

ITU-TG.728

8var5-, 4-, 3- and 2-bit/sample embedded adaptivedifferential pulse code modulation (ADPCM)

ITU-TG.727

816/24/32/4040, 32, 24, 16 kbit/s adaptive differential pulsecode modulation (ADPCM)

ITU-TG.726

85.6/6.3Dual rate speech coder for multimediacommunications transmitting at 5.3 and 6.3kbit/s

ITU-TG.723.1

samplingrate (kHz)

bit rate(kb/s)

descriptionstandardized

byName

Comparison of G.7xx Protocol

telecommunications switching systems 13

Documents