Digital Telephony1
2 Analog/digital systems Analog signal -voltage -speech -pressure SP Analog Sampler Discrete signal F s F max Quantiz- er Error is introduced Digital signal A/D converter or data from -tape -simulations -digital devices DSP Digital Signal Processor -digital computer -dedicated dig. hw -programmable hw Digital signal D/A Analog signal
Digital Telephony 3 Issues F Reconstruction accuracy Conditions for perfect reconstruction F Digital signal is not just an approx. representation of an analog signal Could be generated digitally The processing being performed may not be realizable in analog F The theory of discrete time signal processing is independent of continuous
Digital Telephony 4 Digital vs. analog processing F DSP implementations are flexible, programmable and modular F More precise and repeatable F Performance and cost effectiveness (riding the microelectronics wave) F Direct mapping of mathematical expressions with less approximation possible (enables sophisticated algorithms)
Digital Telephony 5 Digital vs. analog... F Digital hardware can be multiplexed better than analog. Allows integration of multiple operations and services on a h/w platform F Digital storage is more reliable, cheaper and more compact
Digital Telephony 6 On the other hand F Analog SP still offers higher bandwidth F Higher dynamic range F Can be very low power
Digital Telephony 7 Analog to Digital Conversion F To convert “real-world” analog signals to digital signals for processing F Sampling F Quantizing and coding Xa(t)Xa(t) X [n] Xq[n]Xq[n] Sampler Quantizer and Coder Analog signal Discrete signal Digital signal
Digital Telephony 8 Sampling F Uniform One sample every T seconds (ideally) x[n] = x a (nT), n Sampling period: T Sampling frequency: F s =1/T Assume: x a (t) = Acos( 2 Ft+ ) = Acos( t+ ) Then: x[n] = Acos[2 FnT+ ] = Acos[ Tn+ ] = Acos[ n+ ], where T is called the normalized or discrete domain frequency
Digital Telephony 9 F f = F/ F s must be rational in order for x[n] to be periodic F If f = k/N, then x[n] is periodic with period N Now, x a (nT) = Acos( Tn+ ) = Acos(( +2 k/T Tn+ ) is periodic in with period 2 /T Also, x[n] = Acos[ n+ ] = Acos[( +2 k) n + ] is periodic in with period 2
Digital Telephony 10 xa(t)xa(t) n= S(t) = (t nT) x s (t) = x a (nT) (t nT) n= convert to discrete sequence x[n] = x a (nT)
Digital Telephony 11 F Let us look at the continuous time Fourier transform of x s (t) X s (j ) = X a (j ) * S(j ) S(j ) = k s X s (j ) = X a (j kj s ) 2T2T k= 1 2 1T1T k=
Digital Telephony 12 Thus, X a (j ) must be bandwidth limited If the max frequency in X a (j ) is N, then the sampling rate s 2 N ensures no information is lost due to aliasing F This sampling rate is known as Nyquist rate F A lower sampling rate causes a distortion of the signal due to Aliasing F If no Aliasing occurs, the signal can be perfectly reconstructed by passing through an ideal low pass filter with
Digital Telephony 13 Reconstruction X r (j ) = H r (j ) X s (j ) if N c ( s N ) then X r (j ) = X c (j ) Hr(j)Hr(j) c cc s >2 Xs(j)Xs(j)
Digital Telephony 14 Reconstruction Frequency response of ideal reconstruction filter c cc T Impulse response of ideal reconstruction filter H r (j ) = { T, c c 0, otherwise h r (t)= sin t/T t/T
Digital Telephony 15 Reconstruction X r (j ) = H r (j ) X s (j ) F x r (t) = x s (t) * h r (t) = [ k x a (kT) t-kT h r (t) = k x a (kT) h r (t-kT) k x a (kT) sin t-nT)/T t-nT)/T
Digital Telephony 16 xa(t)xa(t) xs(t)xs(t) hr(t)hr(t) xr(t)xr(t)
Digital Telephony 17 Sampling theorem If the highest frequency contained in a signal x a (t) is 0 and the signal is uniformly sampled at a rate s 0, then x a (t) can be exactly recovered from its sample values using the interpolation function and then x a (t) = k x a (kT) h r (t-kT), where {x a (kT) } are the samples of x a (t), and T=2 s h r (t)= sin t/T t/T
Digital Telephony 18 Quantization and coding F Quantization: Converting discrete time signal to digital x q (n) =Q [x(n)] Quantization step
Digital Telephony 19 x Q(x)
Digital Telephony 20 Quantization F Rounding: Assign x[n] to the closest quantization level F Quantization error e q [n] = x q [n] - x[n] e q [n] Uniformly distributed u mean = 0 variance =
Digital Telephony 21 Quantization F Range of quantizer: x max -x min F Quantization levels: m F Assuming uniform quantization = X m / (m-1) where X m = (x max -x min )/2 is called the full-scale level of the A/D converter m-1 x max -x min
Digital Telephony 22 Coding F Coding is the process of assigning a unique binary number to each quantization level Number of bits required log 2 m F Alternatively, given b+1 bits x max -x min )/2 b+1 =X m /2 b F For A/D devices, the higher F s and m, the less the error (and the more the cost of the device)
Digital Telephony 23 F Assuming dynamic range of A/D converter is larger than signal amplitude SNR = 10 log 10 ( x e ) = 10 log 10 ( x ) = 10 log 10 (12.2 2b x / X m ) =6.02b log 10 (X m / x ) Quantizer + x(n) x q (n) x(n) x q (n) e q (n)
Digital Telephony 24 Uniformly Encoded PCM X/Xm Number of bits per sample dB Signal to Quantiiation Noise Ratio (dB)
Digital Telephony 25 Example F What is the minimum bit rate that a uniform PCM encoder must provide to encode a high fidelity audio signal with a dynamic range of 40 dB? Assume the fidelity requirements dictate passage of a 20-kHz bandwidth with a minimum signal-to-noise ratio of 50 dB. For simplicity, assume sinusoidal input signals.
Digital Telephony 26 Companding F Companded PCM with analog compression and expansion A/D Compression Linear PCM Encoder Input Signal D/A Linear PCM Decoder Expansion Output Signal Compressed Digital Codewords
Digital Telephony 27 Segment Approximation Input Sample Values Uniform quantization
Digital Telephony 28 T1 Channel Bank A/D D/A T1 transmission Line Analog Inputs Eigth bits per PCM code word companding functions with mu=255
Digital Telephony 29 Performance of a Encoder dB Signal Power of sinewave (dBm0) Signal-to-quantization noise ratio (dB) 8 bit bit 100 Piecewise linear 8 bit
Digital Telephony 30 Total Noise Power Signal Power relative to full-load signal (dBm0) dB at which persons find communication difficult Signal-to-total noise noise ratio dB dB required for good communication 40 dB range of possible signals
Digital Telephony 31 Error Performance F Fewer than 10% of 1 min intervals to have BER worse than 10E-6 F Fewer than 0.2% of 1 sec intervals to have BER worse than 10E-3 F 92% error free sec
Digital Telephony 32 DS1 Signal Format (8x24)+1=193 bits in 125 s F 193 x 8000 = Mbs F Bit “robbing” technique used on each sixth frame to provide signaling information
Digital Telephony 33 Plesiochronous Transmission Rates 64 kbits/s Japanese StandardNorth America Standard European Standard 1544 kbits/s 2048 kbits/s 8448 kbits/s kbits/s kbits/s kbits/s 6312 kbits/s kbits/s kbits/s kbits/s kbits/s x24 x30 x4 x3 x4 x3 x4 x5x7 x6 x4 x3
Digital Telephony 34 Plesiochronous Digital Hierarchy
Digital Telephony 35
Digital Telephony 36 Plesiochronous Digital Hierarchy F The output of the M12 multiplexer is operating 136 kbs faster than the agragate rate of four DS vs 4x1.544=6.176 F M12 frame has 1176 bits, i.e. 294-bit subframes ; each subframe is made of up of 49-bits blocks; each block starts with a control bit followed by a 4x12 info bits from four DS1 channels
Digital Telephony 37 Makeup of a DS2 Frame M C F C C F M C F C C F Bit stuffing F 4 M bits (O11X X=0 alarm) F C=000,111 bit stuffing present/absent F nominal stuffing rate 1796 bps, max 5367
Digital Telephony 38 Regenerative Repeaters Amplifier Equalizer Input Timing recovery Regenerator Output F Spacing between adjacent repeaters
Digital Telephony 39 Digital Transmission Systems
Digital Telephony 40 PCM System Enhancements F North America Superframe of 12 DS0’s has a sync sequence for odd ( for even frames) F Extended superframe 24 frames - (4 S bits for frame allignment signal); 6 S bits for CRC-6 check; the rest 12 constitute 4 kbs data link