Exercise 2-4. Companding

Transcription

1 Exercise 2-4 Companding When you have completed this exercise, you will be able to demonstrate companding, the technique used in telephone systems to achieve high-quality voice transmission. Figure 2-15 shows the input/output relationship of the 4-bit A/D converter discussed in the previous exercise of this unit. This converter associates each voltage sample of an analog signal with a quantization interval and outputs the corresponding PCM code. DIGITAL OUTPUT CODES QUANTIZATION INTERVAL INPUT VOLTAGE RANGE

2 At the other end of the system, a 4-bit D/A converter is used to recover the analog signal. This converter produces a voltage whose polarity and value (magnitude) depend on the PCM code received. Table 2-1 indicates the voltage produced by the D/A converter for each of the 16 different PCM codes which can be received Closer inspection of Figure 2-15 and Table 2-1 shows that each voltage which the D/A converter can produce, falls exactly in the middle of each quantization interval of the A/D converter. This indicates that the D/A converter is perfectly adapted to the A/D converter of Figure However, when these two converters are used together, perfect voice digitization and recovery is not achieved due to inherent limitations of the quantization process. For example, when the sampled voltage has any value between and V, it is associated with quantization interval 15, and PCM code 0110 is output by the A/D converter. At the D/A converter, this code is converted into a voltage of V. This clearly shows that the recovered voltage differs slightly from the sampled voltage (unless the sampled voltage value is exactly V), even though both converters are perfectly adapted to each other. This difference is due

3 to the quantization process that separates the analog voltage range into a finite numbers of intervals, thereby introducing a slight inaccuracy in the sampled voltage values. Because of this, the difference between the sampled and recovered voltages is referred to as the quantization error. Figure 2-16 shows the quantization error obtained when a sawtooth-wave signal is digitized and then converted back to analog form using the 4-bit A/D and D/A converters discussed in this exercise. The recovered signal is a staircase signal that differs slightly from the original sawtooth-wave signal. Subtracting the original signal from the recovered signal results in a quantization error signal. This error signal is the distortion introduced in the recovered signal by the quantization process. This distortion is referred to as quantization noise because its effect on the recovered signal is similar to band-limited white noise. Quantization noise is also present in the analog line interfaces of telephone systems, even though 8-bit PCM CODECs providing smaller quantization intervals are used for voice digitization and recovery. Figure 2-16 also shows that the quantization noise amplitude remains constant no matter what the value of the sampled voltage is. This is due to the fact that the quantization is linear, that is, equal quantization intervals are used throughout the input voltage range. When linear quantization is used for voice digitization and recovery, the quantization noise level is constant no matter what the voice signal level is. In other words, the voice signal-to-quantization noise (S/N Q ) ratio decreases as the voice signal level decreases, as shown in Figure This figure illustrates the relationship between the S/N Q ratio and voice signal level for an 8-bit PCM CODEC using linear quantization. If such a PCM CODEC were used in the analog line interfaces of telephone systems, low S/N Q ratios would generally be obtained because voice signal levels tend to be low. This would cause background noise (hissing) to be heard in the receiver of the telephone handset, and thereby, result in poor-quality voice transmission that is unacceptable in modern telephone systems.

4 VOLTAGE [V] ORIGINAL ANALOG SIGNAL RECOVERED ANALOG SIGNAL VOLTAGE [V] TIME SAMPLING INTERVAL (125 µs) QUANTIZATION ERROR SIGNAL (QUANTIZATION NOISE) VOLTAGE [V] TIME

5 50 S/N RATIO [db] Q BIT PCM CODEC VOICE SIGNAL LEVEL [db] To improve the low S/N Q ratios obtained with linear quantization when the voice signal levels are low, the PCM CODEC in the analog line interfaces of telephone systems use non-linear quantization. Figure 2-18 shows an example of a non-linear quantization characteristic (4-bit converter). As can be seen in this figure, the voltage range associated with each quantization interval increases with the input voltage magnitude. This makes the quantization error increase with the sampled voltage value. In other words, non-linear quantization causes the quantization noise to increase proportionally with the voice signal level, and thereby, greatly helps in maintaining a constant S/N Q ratio as a function of the voice signal level. This is illustrated in Figure 2-19 which shows the relationship between the S/N Q ratio and voice signal level for 8-bit PCM CODEC using linear and non-linear quantization. With non-linear quantization, the S/N Q ratio remains virtually constant (35 to 38 db) as the voice signal level varies from 40 to 0 db. This provides a good voice quality (low background noise) over a fairly large range of voice signal levels. Comparison of the curves in Figure 2-19 clearly shows that non-linear quantization increases the S/N Q ratio at low voice signal levels, at the price of a decrease in the S/N Q ratio at high voice signal levels. This is perfectly adapted to telephone systems where low-level voice signals are frequent while high-level voice signals are not.

6 DIGITAL OUTPUT CODES QUANTIZATION INTERVAL QUANTIZATION INTERVAL INPUT VOLTAGE RANGE NON-LINEAR QUANTIZATION 8-BIT PCM CODEC S/N RATIO [db] Q LINEAR QUANTIZATION VOICE SIGNAL LEVEL [db]

7 In telephone systems, the use of non-linear quantization in the PCM CODEC of the analog line interfaces is referred to as companding. This name comes from the early implementations of electronic-type, analog line interfaces. In these interfaces, non-linear quantization was implemented by compressing the analog voice signal before the A/D conversion, and expanding the recovered analog signal after the D/A conversion. The word companding is merely a contraction of the words compressing and expanding. Two types of non-linear quantization (companding) characteristics are used in the PCM CODEC of analog line interfaces: the -law characteristic and the A-law characteristic. The -law companding characteristic is used mainly in North America and Japan while the A-law companding characteristic is used in Europe and most of the rest of the world. Both the -law and A-law companding characteristics improve the S/N Q ratio at low voice signal levels, in a way similar to what is shown in Figure In the first part of the exercise, you will set up a central office with the Telephony Training System (TTS). In the second part of the exercise, you will establish a connection between two telephone sets. On one of the analog line interfaces, you will disconnect the CODEC analog input from the SLIC TXA output, and reconnect this input to a DC source. Without companding on the CODEC, you will measure the quantization interval for analog input voltages of various values (magnitudes). In the third part of the exercise, you will enable companding on the CODEC and measure the quantization interval once again for analog input voltages of various values. In the fourth part of the exercise, you will disconnect the CODEC analog input from the DC source, and reconnect this input to the SLIC TXA output. You will apply a sine-wave sound signal to the telephone handset. You will vary the level of the sinewave sound signal and compare the waveforms of the original and recovered sinewave signals, with and without companding. In the last part of the exercise, you will have a normal phone conversation with and without companding. This will allow you to notice the effect which companding has on the quality of voice transmission. Refer to Appendix A of this manual to obtain the list of equipment required to perform this exercise.

8 * 1. Make sure that the Reconfigurable Training Module, Model 9431, is connected to the TTS Power Supply, Model Make sure that there is a network connection between the Reconfigurable Training Module and the host computer. Install the Dual Analog Line Interface, Model 9475, into one of the analog/digital (A/D) slots of the Reconfigurable Training Module. Connect two analog telephone sets to the Dual Analog Line Interface. Make sure that the tone dialing mode is selected on the analog telephone sets. Connect the AC/DC power converter supplied with each analog telephone set to one of the AC power outlets on the TTS Power Supply. Connect the DC power output jack of each AC/DC power converter to the DC power input connector on either one of the analog telephone sets. The analog telephone set requires an auxiliary DC power source for the digital display to be operative. * 2. Turn on the host computer. Turn on the TTS Power Supply then the Reconfigurable Training Module. * 3. On the host computer, start the Telephony Training System software, then download the CO program to the Reconfigurable Training Module. The CO program configures the Reconfigurable Training Module so that it operates as a central office. If the host computer is unable to download the CO program to the Reconfigurable Training Module, it may not be using the proper IP address. Have your instructor or the LAN administrator check if the host computer uses the proper IP address to communicate with the Reconfigurable Training Module.

9 * 4. On the host computer, make sure that the address of the TSAC in ANALOG LINE INTERFACE A is set to 01. * 5. Lift off the handset of telephone set A and dial the number of telephone set B. Lift off the handset of telephone set B to answer the call and establish a communication. * 6. On the host computer, turn the companding off in the CODEC of ANALOG LINE INTERFACE A. * 7. On the host computer, zoom in on ANALOG LINE INTERFACE A. Set switch S1 so as to connect the CODEC analog input (TP4) to the DC SOURCE output. Connect Oscilloscope Probes 1, 2, and 3 to TP4 (CODEC analog input), TP5 (CODEC digital output), and TP17 (FRAME SYNC. signal), respectively. Probes 1, 2, and 3 are associated with channels 1, 2, and 3 of the Oscilloscope, respectively. * 8. Start the Oscilloscope. Make the following settings on the Oscilloscope: Channel 1 Mode Normal Sensitivity V/div Input Coupling DC Channel 2 Mode Normal Sensitivity V/div Input Coupling DC Channel 3 Mode Normal Sensitivity V /div Input Coupling DC Time Base s/div Trigger Source Ch 3 Level V Slope Positive (+) Display Refresh Continuous

10 * 9. On the host computer, set the output voltage of the DC SOURCE in ANALOG LINE INTERFACE A so that the serial PCM code at the CODEC digital output (TP5) is 0DH ( ). Record the DC SOURCE output voltage (V 1 ) in the following blank space. V 1 = mv Set the DC SOURCE output voltage so that the serial PCM code at the CODEC digital output (TP5) is 1DH ( ). Record the DC SOURCE output voltage in the following blank space. V 2 = mv Calculate the voltage range associated with one quantization interval by dividing the difference between voltages V 1 and V 2 by 16. Quantization Interval = mv (low-level signals without companding) * 10. Measure the quantization interval for medium-level signals. To do so, measure the DC SOURCE output voltages required to set the serial PCM code at the CODEC digital output (TP5) to 3DH ( ) and 4DH ( ), and calculate the quantization interval as in the previous step. Quantization Interval = mv (medium-level signals without companding) * 11. Measure the quantization interval for high-level signals. To do so, measure the DC SOURCE output voltages required to set the serial PCM code at the CODEC digital output (TP5) to 6DH ( ) and 7DH ( ), and calculate the quantization interval as in the previous step. Quantization Interval = mv (high-level signals without companding) When no companding is used, how does the voltage range associated with one quantization interval vary with the magnitude of the signal at the CODEC analog input? * 12. On the host computer, turn the companding on in the CODEC of ANALOG LINE INTERFACE A. * 13. Measure the quantization interval for low-level signals. To do so, measure the DC SOURCE output voltages required to set the serial PCM code at

11 the CODEC digital output (TP5) to 0DH ( ) and 1DH ( ), and calculate the quantization interval as in the previous steps. Quantization Interval = mv (low-level signals with companding) * 14. Measure the quantization interval for medium-level signals. To do so, measure the DC SOURCE output voltages required to set the serial PCM code at the CODEC digital output (TP5) to 3DH ( ) and 4DH ( ), and calculate the quantization interval as in the previous steps. Quantization Interval = mv (medium-level signals with companding) * 15. Measure the quantization interval for high-level signals. To do so, measure the DC SOURCE output voltages required to set the serial PCM code at the CODEC digital output (TP5) to 6DH ( ) and 7DH ( ), and calculate the quantization interval as in the previous steps. Quantization Interval = mv (high-level signals with companding) When companding is used, how does the voltage range associated with one quantization interval vary with the magnitude of the signal at the CODEC analog input? * 16. On the host computer, zoom in on ANALOG LINE INTERFACE A. Set switch S1 so as to reconnect the CODEC analog input (TP4) to the SLIC TXA output. Ask someone to talk very softly into the handset of telephone set B while listening to his (her) voice in the handset of telephone set A. Notice that the quality of the voice transmitted through the telephone system is good even if the voice signal level is low. * 17. Turn the companding off in the CODECs of ANALOG LINE INTERFACEs A and B.

12 Ask someone to talk very softly into the handset of telephone set B while listening to his (her) voice in the handset of telephone set A. Describe the effect that turning companding off has on the quality of voice transmission. * 18. On the host computer, close the Telephony Training System software. Turn off the TTS Power Supply as well as the host computer (if it is no longer required). Disconnect the AC/DC power converters from the TTS Power Supply and the analog telephone sets. Disconnect the analog telephone sets from the Dual Analog Line Interface. Remove the Dual Analog Line Interface from the Reconfigurable Training Module. In this exercise, you learned that perfect voice digitization and recovery cannot be achieved because the quantization process introduces noise in the recovered voice signal. You observed that when a CODEC operates without companding (linear quantization), the quantization interval is fixed. You saw that this causes the quantization noise level to be constant no matter what the analog voice signal level is. You also saw that this results in low S/N Q ratios at low analog voice signal levels, thereby reducing the quality of voice transmission. You observed that when a CODEC operates with companding (non-linear quantization), the quantization interval increases with the analog voice signal amplitude. You saw that this causes the quantization noise level to increase with the analog voice signal level. You also saw that this results in a fairly high and constant S/N Q ratio over a fairly broad range of analog voice signal levels, thereby providing good quality voice transmission. 1. What is the main difference between linear and non-linear quantization?

13 2. Briefly explain why the voice digitization and recovery process introduces quantization noise in the recovered voice signal. 3. Complete the following sentence. Comparing linear quantization to nonlinear quantization shows that non-linear quantization increases the S/N Q ratio at low voice signal levels, but Referring to the curves in Figure 2-19, determine the improvement in the S/N Q ratio due to the use of companding, when the relative voice signal level is30 db. 5. Briefly explain why companding is perfectly adapted to voice digitization and recovery in telephone systems.