ADSL Analytical Test

1

In a DMT-based ADSL system using FDM duplexing, the available bandwidth is divided into 256 sub-carriers with 4.3125 kHz spacing. Channels 1-5 are unused guard bands, channels 6-30 (25 channels) are allocated for upstream, and channels 31-255 (225 channels) for downstream. Given that each sub-carrier can modulate 0-15 bits using QAM, calculate:

(a) The theoretical maximum downstream data rate assuming all downstream sub-carriers carry 15 bits and the frame rate is 4000 symbols/second.
(b) Explain why the practical maximum downstream rate is limited to approximately 8.16 Mbps despite the theoretical calculation.

10 Marks

✓ Model Answer & Explanation

(a) Theoretical Maximum Downstream Data Rate:

$$ \text{Total downstream sub-carriers} = 255 - 31 + 1 = 225 \text{ channels} $$ $$ \text{Less: Sub-carrier 256 (Nyquist) and sub-carrier 64 (Pilot) are unavailable} $$ $$ \text{Available sub-carriers} = 254 $$ $$ \text{Data Rate} = 254 \times 15 \times 4000 = 15,240,000 \text{ bps} = \mathbf{15.24 \text{ Mbps}} $$

(b) Practical Limitation to 8.16 Mbps:

The practical limitation arises from Reed-Solomon (RS) coding constraints:

RS codeword size is limited to 255 bytes maximum
With S=1 (one data frame per RS codeword), the maximum is ~8.16 Mbps
Even with S=1/2 (two RS codewords per FEC output frame), allowing ~16 Mbps, system architecture constraints limit practical rates
At least one byte per codeword is used for framing overhead, reducing effective rate to 8.128 Mbps

Key Technical Concepts:

DMT frame rate is fixed at 4000 baud (250 μs symbol period)
Sub-carrier spacing $\Delta f = 4.3125$ kHz (total bandwidth ~1.104 MHz)
SNR-dependent bit loading means not all sub-carriers can carry 15 bits
Reed-Solomon coding adds redundancy for error correction

2

An ADSL system employs the bit-loading algorithm where the number of bits $b_n$ allocated to sub-carrier $n$ is given by: $$ b_n = \log_2\left(1 + \frac{\text{SNR}_n \cdot CG}{\Gamma \cdot M}\right) $$ where $CG$ is coding gain (3 dB), $\Gamma$ is the SNR gap to capacity (9.75 dB for $P_e = 10^{-7}$), and $M$ is the noise margin.

(a) If a sub-carrier has an SNR of 30 dB and the system requires a 6 dB noise margin, calculate the number of bits loaded onto this sub-carrier.
(b) Explain the difference between margin-adaptive and rate-adaptive bit loading algorithms, and identify which approach ADSL typically uses during initialization.

10 Marks

✓ Model Answer & Explanation

(a) Bit Loading Calculation:

Step 1: Convert all dB values to linear scale:

$$ \text{SNR} = 30 \text{ dB} \rightarrow 10^{30/10} = 1000 $$ $$ CG = 3 \text{ dB} \rightarrow 10^{3/10} \approx 1.995 \approx 2 $$ $$ \Gamma = 9.75 \text{ dB} \rightarrow 10^{9.75/10} \approx 9.44 $$ $$ M = 6 \text{ dB} \rightarrow 10^{6/10} \approx 3.98 \approx 4 $$

Step 2: Calculate effective SNR:

$$ \text{Effective SNR} = \frac{1000 \times 2}{9.44 \times 4} = \frac{2000}{37.76} \approx 52.97 $$

Step 3: Calculate bits per sub-carrier:

$$ b_n = \log_2(1 + 52.97) = \log_2(53.97) \approx \mathbf{5.75 \text{ bits}} $$ $$ \text{Practical allocation: } \mathbf{6 \text{ bits}} \text{ (rounded down to integer, or 5 with conservative margin)} $$

(b) Bit Loading Algorithms:

Margin-Adaptive Algorithm:

Minimizes Bit Error Rate (BER) for a given fixed bit rate
Optimizes power allocation to achieve target rate with maximum margin
Used when data rate is fixed and reliability is priority

Rate-Adaptive Algorithm:

Maximizes bit rate for a fixed BER and power constraints
Reformulation of Shannon capacity formula
Allocates bits based on "water-filling" principle across sub-carriers

ADSL Implementation: ADSL uses rate-adaptive bit loading during initialization (channel analysis phase). The system measures SNR on each sub-carrier and allocates bits accordingly to maximize data rate while maintaining the target BER of $10^{-7}$ and specified noise margin (typically 6-10 dB).

Key Technical Concepts:

SNR gap $\Gamma$ accounts for practical QAM implementation vs. theoretical Shannon limit
Coding gain $CG$ comes from Reed-Solomon and Trellis coding
Noise margin $M$ protects against crosstalk and interference variations
Bit loading is performed during modem training/initialization
ADSL2 improves efficiency by loading one extra bit per tone pair

3

The ADSL POTS splitter is a critical component that enables simultaneous voice and data transmission over a single copper pair.

(a) Design a passive LC low-pass filter specification for the POTS splitter that passes voice frequencies (300-3400 Hz) with less than 1 dB insertion loss while attenuating ADSL frequencies (above 25.875 kHz) by at least 55 dB. Specify the cutoff frequency and minimum filter order required.
(b) Explain why the splitter must present high impedance to ADSL frequencies at the telephone port, and describe the consequences if this impedance requirement is not met.

10 Marks

✓ Model Answer & Explanation

(a) LC Low-Pass Filter Design:

Requirements Analysis:

$$ \text{Passband: } 300-3400 \text{ Hz (POTS voice band)} $$ $$ \text{Passband ripple: } < 1 \text{ dB} $$ $$ \text{Stopband: } > 25.875 \text{ kHz (ADSL band start)} $$ $$ \text{Stopband attenuation: } > 55 \text{ dB} $$

Cutoff frequency ($f_c$) selection:

$$ f_c \text{ is typically placed between } 4-10 \text{ kHz} $$ $$ \text{Selected: } f_c = 10 \text{ kHz (}-3\text{dB point)} $$

Filter order calculation (Butterworth approximation):

$$ A = 10 \log_{10}\left[1 + \left(\frac{f_{stop}}{f_c}\right)^{2n}\right] $$ $$ 55 = 10 \log_{10}\left[1 + \left(\frac{25.875}{10}\right)^{2n}\right] $$ $$ 10^{5.5} = 1 + (2.5875)^{2n} $$ $$ 316227 \approx (2.5875)^{2n} $$ $$ \log_{10}(316227) = 2n \times \log_{10}(2.5875) $$ $$ 5.5 \approx 2n \times 0.413 $$ $$ n \approx 6.65 \rightarrow \mathbf{\text{Minimum 7th order filter required}} $$

Practical implementation: 7th order Butterworth or Chebyshev LC ladder network

(b) High Impedance Requirement:

The splitter must present high impedance ($> 600$ Ω equivalent) to ADSL frequencies (25 kHz - 1.1 MHz) at the telephone port for two critical reasons:

Loading Effect Prevention: Telephone equipment and wiring have capacitive and inductive properties. If the splitter doesn't isolate these, they create additional load on the line, attenuating the high-frequency ADSL signal.
Reflection Minimization: Impedance mismatches cause signal reflections, creating inter-symbol interference (ISI) and degrading the SNR on DMT sub-carriers.

Consequences of Poor Impedance Isolation:

ADSL signal attenuation (additional 10-20 dB loss)
Increased bit errors on affected sub-carriers
Reduced data rate due to lower SNR
Potential modem synchronization failures
Audible noise in telephone handset (high-frequency bleed-through)

Key Technical Concepts:

POTS splitter is a passive device (no power required) - critical for lifeline voice service during power outages
Series inductors present high impedance to high frequencies
Shunt capacitors bypass high frequencies to ground
Return loss specifications ensure impedance matching (typically $> 6$ dB)
DC resistance must be $< 100$ ohms to support loop current (20-100 mA)

4

Crosstalk is a major limitation in ADSL systems, particularly Near-End Crosstalk (NEXT) and Far-End Crosstalk (FEXT).

(a) Derive the relationship showing why FEXT dominates in ADSL downstream reception while NEXT is the primary concern for upstream. Include the frequency-dependent characteristics of crosstalk coupling.
(b) In a binder group with 25 ADSL lines, if the coupling coefficient for NEXT is $K_{NEXT} = 10^{-14}$ and for FEXT is $K_{FEXT} = 10^{-17}$, calculate the approximate crosstalk power ratio relative to the received signal at 100 kHz for a line length of 3 km. Assume the crosstalk transfer function follows $f^{3/2}$ dependence.

10 Marks

✓ Model Answer & Explanation

(a) NEXT vs FEXT in ADSL:

System Architecture:

ATU-C (Central Office): Transmits high-power downstream, receives low-power upstream
ATU-R (Remote/Customer): Receives downstream, transmits upstream

NEXT (Near-End Crosstalk):

Occurs when a transmitter at one end couples into a receiver at the same end
In upstream: ATU-C receivers are co-located with ATU-C transmitters → strong NEXT from adjacent downstream transmitters
Coupling increases with frequency as $f^{3/2}$ (capacitive coupling dominates)
Power transfer: $P_{NEXT} \propto K_{NEXT} \cdot f^{3/2} \cdot d \cdot N^{0.6}$

FEXT (Far-End Crosstalk):

Occurs when a transmitter couples into a receiver at the opposite end
Coupled signal travels through the line, experiencing attenuation
Power transfer: $P_{FEXT} \propto K_{FEXT} \cdot f^2 \cdot d \cdot N^{0.6} \cdot |H(f,d)|^2$
$H(f,d)$ is the line transfer function (attenuation)

Why FEXT dominates downstream: Downstream signal travels from CO to CPE. FEXT from other CO transmitters couples into the downstream receiver at the CPE end. However, in ADSL, downstream uses higher frequencies (above 138 kHz) where line attenuation is significant, but FEXT is still problematic because all downstream signals are synchronized from the same DSLAM.

(b) Crosstalk Power Calculation:

Given parameters:

$$ f = 100 \text{ kHz} = 10^5 \text{ Hz} $$ $$ d = 3 \text{ km} = 3000 \text{ m} $$ $$ N = 25 \text{ lines (disturbers)} $$ $$ K_{NEXT} = 10^{-14}, \quad K_{FEXT} = 10^{-17} $$ $$ \text{Frequency dependence: } f^{3/2} $$

NEXT Power Ratio:

$$ \frac{P_{NEXT}}{P_{sig}} = K_{NEXT} \times f^{3/2} \times d \times N^{0.6} $$ $$ = 10^{-14} \times (10^5)^{1.5} \times 3000 \times (25)^{0.6} $$ $$ = 10^{-14} \times 10^{7.5} \times 3000 \times 9.09 $$ $$ = 10^{-14} \times 3.16 \times 10^7 \times 3000 \times 9.09 $$ $$ \approx 10^{-14} \times 8.62 \times 10^{11} $$ $$ \approx \mathbf{0.0862 \text{ or } -10.6 \text{ dB}} $$

FEXT Power Ratio (simplified):

$$ \frac{P_{FEXT}}{P_{sig}} = K_{FEXT} \times f^2 \times d \times N^{0.6} \times |H(f,d)|^2 $$ $$ \text{Assuming line attenuation } H(f,d) \approx e^{-\alpha\sqrt{f} \cdot d} \text{ where } \alpha \approx 0.0012 $$ $$ \text{At } 100 \text{ kHz: attenuation } \approx 40 \text{ dB over } 3 \text{ km} $$ $$ |H|^2 \approx 10^{-4} $$ $$ \frac{P_{FEXT}}{P_{sig}} = 10^{-17} \times (10^5)^2 \times 3000 \times 9.09 \times 10^{-4} $$ $$ = 10^{-17} \times 10^{10} \times 2.727 $$ $$ \approx \mathbf{2.73 \times 10^{-7} \text{ or } -65.6 \text{ dB}} $$

Conclusion: NEXT (-10.6 dB) is significantly stronger than FEXT (-65.6 dB), explaining why NEXT is the dominant impairment in upstream reception at the central office.

Key Technical Concepts:

NEXT is strongest at the transmitter end and increases with more disturbers ($N^{0.6}$)
FEXT is attenuated by line loss, making it less severe in long loops
ADSL uses FDM (Frequency Division Multiplexing) to avoid NEXT between upstream and downstream
Bit loading algorithms account for crosstalk by reducing bits on affected sub-carriers
Spectral shaping (PSD masks) limits crosstalk injection

5

ADSL framing structure uses a superframe consisting of 68 data frames followed by one synchronization symbol.

(a) Calculate the overhead percentage due to the synchronization symbol and explain its purpose in maintaining system stability.
(b) Describe the function of the Fast Byte and Interleaved Byte in the ADSL framing structure, and explain how the system provides different levels of error protection for different types of traffic (e.g., voice vs. video vs. data).
(c) If the system operates with a 4 kHz DMT symbol rate and uses Reed-Solomon coding with 16 parity bytes per 240 data bytes on the interleaved path with depth $D=4$, calculate the effective data throughput and latency introduced by the interleaver.

10 Marks

✓ Model Answer & Explanation

(a) Synchronization Symbol Overhead:

$$ \text{Superframe structure: } 68 \text{ data frames} + 1 \text{ sync symbol} = 69 \text{ symbols} $$ $$ \text{Overhead} = \frac{1}{69} \times 100\% \approx \mathbf{1.45\%} $$ $$ \text{Effective symbol rate} = \frac{68}{69} \times 4000 \approx 3942 \text{ baud} $$ $$ \text{Overhead reduces effective capacity by } \sim 1.45\% $$

Purpose of Synchronization Symbol:

Frame Alignment: Marks superframe boundaries for receiver synchronization
Channel Estimation: Contains known pilot pattern for equalizer adaptation
Timing Recovery: Provides reference for clock recovery at receiver
No User Data: Carries no payload, dedicated to physical layer functions

(b) Fast and Interleaved Paths:

Fast Buffer (Fast Byte):

Low latency path with minimal processing delay ($< 1$ ms)
Contains CRC bits (8 bits per superframe) for error detection
Carries indicator bits (24 bits) for OAM (Operations, Administration, Maintenance)
Used for delay-sensitive traffic: voice, gaming, interactive video
Minimal FEC protection (optional Reed-Solomon)

Interleaved Buffer (Interleaved Byte):

High latency path with deep interleaving (4-16 ms typical)
Strong FEC protection with Reed-Solomon coding
Interleaving spreads burst errors into random errors correctable by FEC
Used for error-sensitive traffic: file transfers, streaming video, data

Dual-Path Architecture Benefits:

Voice over ATM/VoIP: Fast path (low latency critical)
Video Streaming: Interleaved path (error-free delivery critical)
Internet Data: Interleaved path (throughput priority)
Interactive Gaming: Fast path (real-time response)

(c) Throughput and Latency Calculation:

Reed-Solomon parameters:

$$ \text{Data bytes } (K) = 240 $$ $$ \text{Parity bytes } (R) = 16 $$ $$ \text{Codeword size } (N) = 256 \text{ bytes} $$ $$ \text{Interleave depth } (D) = 4 $$ $$ \text{Symbol rate} = 4000 \text{ symbols/sec} $$ $$ \text{Coding overhead} = \frac{16}{256} = 6.25\% $$ $$ \text{Interleave overhead} = \text{Wait for } D \text{ codewords} $$

Effective Throughput:

$$ \text{Raw rate per sub-carrier depends on bit loading} $$ $$ \text{For example, if average 8 bits/sub-carrier on 200 sub-carriers:} $$ $$ \text{Raw rate} = 200 \times 8 \times 4000 = 6.4 \text{ Mbps} $$ $$ \text{After RS coding: } 6.4 \times \frac{240}{256} = \mathbf{6.0 \text{ Mbps}} $$

Interleaver Latency:

$$ \text{The interleaver writes data row-wise into } D \times N \text{ matrix and reads column-wise} $$ $$ \text{Latency} = \frac{(D - 1) \times N \times 8 \text{ bits}}{\text{data rate}} $$ $$ \text{Or simply: Latency} \approx \frac{D \times (N \times 8)}{\text{symbol_rate} \times \text{bits_per_symbol}} $$ $$ \text{For our example (200 sub-carriers } \times 8 \text{ bits = 1600 bits/symbol):} $$ $$ \text{Bits per RS codeword} = 256 \times 8 = 2048 \text{ bits} $$ $$ \text{Symbols per codeword} = \frac{2048}{1600} \approx 1.28 \text{ symbols} $$ $$ \text{Latency} = D \times 1.28 \times 250 \text{ } \mu\text{s} \approx 4 \times 320 \text{ } \mu\text{s} = \mathbf{1.28 \text{ ms}} $$

(Note: Actual latency depends on specific bit allocation across sub-carriers)

Key Technical Concepts:

Superframe structure enables efficient overhead management with minimal capacity loss
Dual latency paths optimize for different application requirements simultaneously
Interleaving depth $D$ trades latency for burst error correction capability
Reed-Solomon coding adds redundancy: $N = K + 2T$ ($T$ = error correction capability)
Fast path uses $S=1$ (one symbol per codeword), Interleaved path uses $S=1,2,4,8,16$
Total latency = Interleaver delay + FEC decoding delay + transmission time

📋 Instructions

Key Technical Concepts:

Key Technical Concepts:

Key Technical Concepts:

Key Technical Concepts:

Key Technical Concepts: