Single Sideband Technologies for Optimization of VHF Aerospace Communications

1

Single Sideband Technologies for Optimization of VHF Aerospace Communications

Behnam Kamali, Mercer University, Macon GA Akshay Patel, ARRIS Group, Suwanee, GA. (Milieu International)

Abstract The highest priority in air travel operation is

justifiably placed on flight safety which is critically linked to the availability of reliable communications and navigation systems. The current aeronautical VHF spectrum dedicated to civil aviation communications in the US and the rest of the world is19 MHz wide. With the 25-kHz commercial AM technology, this band supports 760 radio channels. In the United States, owing to the rise in the number of airplanes in both commercial and general aviation sectors, spectral congestion in the aeronautical VHF band is rapidly becoming a predicament. This article proposes single side band as an alternative technology that enables efficient use of the available spectrum while requiring minimal change in the infrastructure. It is shown that SSB in conjunction the Weaver modulation method presents an economically feasible solution to spectral congestion in VHF aeronautics, with which the system capacity can be increased by a factor of 3, 4, 5, 6, and even 7 (X-factor). Hardware costs increments for implementing either the analog or digital Weaver method are modest among schemes with different X-factors. This implies that it is possible to increase the present capacity of 760 to up to 5320 voice channels; which meets the challenge of required number of civil aviation radio channels for at least two decades.

Introduction As aviation traffic increases around the world,

congestion within the limited bandwidth allocated for aerospace voice communication rises. The current aeronautical VHF spectrum dedicated to civil aviation communication in the US and the rest of the world is 19 MHz wide. With the current commercial AM technology, this band supports 760 radio channels of 25 kHz bandwidth. The U.S. and European Union are in dire need of a simple and inexpensive method of alleviating the bandwidth congestion.

In the late 1990’s the Europeans proposed a scheme in which 25 kHz spacing band is reduced to 8.33 kHz (8.33-AM) and thereby the number of available radio channels is tripled to 2280. The 8.33-AM format ran to some standardization problems and was not implemented in the United States although it was deployed in Europe. Many groups have proposed various schemes of a digital solution known as VHF Digital Links (VDL)1. This article proposes Single Sideband (SSB) technology as an alternative solution to the VHF aeronautical spectral congestion problem. Since current voice communications in aerospace uses analog AM modulation, SSB modulation may be implemented concurrently with the existing systems by using simple switching techniques. The factor by which the VHF aeronautical communication capacity is increased, directly affects the Weaver filter quality factor and complexity. Results show that by using either the analog or digital versions of the Weaver Method, the current system capacity can be increased by up to seven folds. This effectively increases the present channel capacity from 760 to up to 5320 channels.

Single Side Band Modulation with Weaver method

It is well known that single sideband has three principle advantages over DSB commercial AM.

• It requires half the bandwidth of DSB signal.

• SSB is highly efficient in the usage of the transmitter’s primary power source.

• The power efficiency is 100 percent for SSB.

1-4244-1216-1/07/$25.00 ©2007 IEEE.

2

Since only one sideband is transmitted in this scheme, the power consumption is almost half (3dB gain) of the amount consumed by commercial AM. The low power consumption also facilitates less bulky and lighter system designs. SSB modulation has no carrier frequency, which implies that the entire peak power is dedicated to the information-bearing content of the signal2. One of the disadvantages in SSB modulation is that the receiver structure is more complex than that of commercial AM. SSB system is also more susceptible to spurious emissions. However, these shortfalls can be overcome by a careful system design procedure.

Of the three standard methods available for SSB radio implementation; Weaver method is shown to be the best choice for this application. The “filter method” requires bandpass filtering in the carrier frequency range i.e. a complex bank of filters have to be designed to match various channels; as such this method is costly and not feasible for this application. The “phase shift method”, on the other hand; requires the approximation of non-causal Hilbert transform filters. The Weaver method is implemented using a low pass filtering scheme in the baseband. It requires only a pair of filters for all channels, thus dramatically reduces the system cost. The Weaver method can be implemented using both analog and digital design schemes. Despite the need for higher level of computational complexity; the Weaver method proves to be the optimum choice in practicality, complexity, and feasibility senses. The Weaver method implementation of SSB signal is illustrated in Figure 1. The Weaver technique splits the input signal )(tm into upper and lower branch signals. The upper branch shifts the central frequency of )(tm to 0 Hz by modulating complex signal )(ts ; resulting in signal )(1 tm . Centering

)(1 tm about 0 Hz shifts the upper sideband and the lower sideband spectra such that they spread over equal positive and negative frequencies. The other sideband is then filtered using a regular lowpass filter, thus preserving a single sideband

signal )(1 tx . Signal )(1 tx is then modulated with the carrier signal )(tc to produce the high frequency modulated signal )(1 tr . The lower branch signal goes through the same procedure; however, the complex signals )(ts and )(tc are shifted by 1800, resulting in )(2 tr . The product of )(1 tr and )(2 tr generates the desired SSB modulated signal )(tr .

Design Specifications For complete design specifications of the

proposed weaver SSB system for aviation communication see reference 8. We venture some comments on the analog and digital filter design; for it enables the determination of X-factors by which the AM-25 kHz aeronautical channels can be divided into SSB channels. The digital technique of implementing the Weaver method has essentially the same components as that of the analog version except that it requires two additional components of analog to digital (A/D) and digital to analog (D/A) signal converters. There are several circuit topologies that are used to implement the required active low pass filters. They include the Butterworth, Chebychev (Type I and II), and the Elliptic filters. Each has various properties that can be advantageous or disadvantageous depending on the application and specifications used to implement them. The analog filter specifications that are required to design the filter are shown in Table 1. Column A shows X-Factor (X = 3 to 8) by which the current aeronautical spectral capacity can be increased e.g. when X-Factor = 4, the number of radio channels in the system is increased to 4 x 760 = 3040. Column B provides information on bandwidth per channel, which clearly depends on the values given in Column A. The channel bandwidth is calculated as follows:

3

Figure 1: Block diagram of implementation of Weaver method

( )125000ColumnA

HzdwidthChannelBan =

The “Guardband”; given in column C, is a 200 Hz frequency band that is provided to compensate for various factors including Doppler shift and governmental requirements. Columns E and F provide critical filter specifications; where Fpass is the 3dB bandwidth throughput, and Fstop is the maximum allowable filter bandwidth. Column G shows the ratio of Fpass to Fstop. In order for a filter to be realizable, the ratio must be

greater than 1. It is normally preferred that the ratio is greater than 2 in order to achieve low filter orders. Fpass is selected to be 1500 Hz, which provides an effective bandwidth for voice signal that is accepted by standard telecommunication networks that carry voice signals. The last row in the table is highlighted to place an emphasis on the fact that a filter with this specification is not realizable.

. Table 1: Analog Filter Specifications

A B C D E F G

LOWPASS FILTER ratio = Fpass/Fstop

X Factor Channel

Bandwidth Gaurdband Actual

Channel bandwidth Fstop Fpass

i 3 8333.33 200 8133.33 4066.67 1500 2.71 ii 4 6250.00 200 6050.00 3025.00 1500 2.02 iii 5 5000.00 200 4800.00 2400.00 1500 1.60 iv 6 4166.67 200 3966.67 1983.33 1500 1.32 v 7 3571.43 200 3371.43 1685.71 1500 1.12 vi 8 3125.00 200 2925.00 1462.50 1500 0.98

4

For digital implementation of the low pass filters, a rather larger catalogue of choices is available. The availability of these options is due to various windowing techniques that might be used, and other algorithms that may be employed for filter response approximation. The “Equiripple” FIR lowpass filter was implemented, for its desirable linear phase response and ease of implementation on digital signal processing (DSP) silicon chips 4,5. Digital filters are generally preferred for availability of cost effective DSP processors,

and the flexibility they provide to the designer to change the system performance and specifications by simply rewriting the software code. The frequency specifications that were determined to design the lowpass Equiripple FIR filters are shown on Table 2. The specifications shown in the table are essentially equivalent to that of the analog filter; except that normalized value columns are added. The digital filters were implemented using MATLAB scripts and the help of MATLAB’s filter design signal processing FDATOOL.

Table 2: Digital Filter Specifications.

LOWPASS FILTER X

Factor

Channel Bandwidt

h Gaurdban

d

Actual Channel bandwid

th Fpass

Fpass normalize

d Fstop

Fstop normaliz

ed

ratio = Fstop/Fpas

s 3 8333.33 200 8133.33 4066.67 0.5000 1500 0.1844 2.71 4 6250 200 6050 3025 0.5000 1500 0.2479 2.02 5 5000 200 4800 2400 0.5000 1500 0.3125 1.6 6 4166.67 200 3966.67 1983.33 0.5000 1500 0.3782 1.32 7 3571.43 200 3371.43 1685.71 0.5000 1500 0.4449 1.12 8 3125 200 2925 1462.5 1500 -

Doppler Shift

The Doppler shift has to be taken into consideration in order to avoid inter-channel interference. In the case of aviation communication the aircraft is in motion with a varying velocity, and the control tower is stationary. The Doppler shift causes a ‘plus’ or ‘minus’ swing from the carrier frequency that could result in inter-channel interference. Depending on the estimated shift, the effect can be compensated by designing a system with a sufficient frequency gap between channels. An alternative method to compensate for the Doppler shift is the application of frequency tracking at the receiver end; however, the price paid is an increase in system complexity. For

design purposes; in order to compensate for Doppler shift, a guard band of 200 kHz is left between channels.

Voice Quality Speech intelligibility is one of the most

critical requirements in aeronautical voice communications. Voice quality lost by limiting the bandwidth for communication channel might lead to catastrophe especially in high distress communications scenarios. In order to decide on filter specifications, voice intelligibility was tested through various studies on samples of male and female speech. Mono recordings were made in a sound-treated booth at the Communication Sciences and Disorders

5

Department at the University of Florida. Raw speech was recorded on a digital audio tape (DAT) and sampled at 16 kHz. The following sentences were recorded by both male and female subjects. These five sentences were chosen carefully to mimic the possible emotions in a communication between an aircraft pilot and an air traffic controller in an airport communications tower. Note that the sentences are listed in order from calm to tense scenarios.

• “This is Delta 1965, requesting

permission to land” • “Ferry in to runway one east-northeast

after nineteen hundred hours” • “Maintain altitude of 10000ft flight

Alpha Gamma 101.” • “Roger, tower. Thank you for guiding

us through the rough storm. Over and out”

• “Mayday! Mayday! Quantas flight from Portsmouth to New York experiencing engine failure”

The signals were then analyzed using Adobe Audition 1.5 to analyze voice characteristics. It was found that the voice spanned approximately 50 Hz to 12 kHz however; the energy in speech at frequencies above 7 kHz is present in fricatives (e.g. /f/, /s/, /sh/) and has little bearing on speech intelligibility or naturalness. Whereas, energy present in frequencies below 300 Hz (e.g. energy in the fundamental frequency) contribute to naturalness. Intelligibility is not lost if the lower frequencies are attenuated. Most relevant frequencies such as the first format and fundamental frequency lie below 700 Hz. Conventional telephone uses a bandwidth that spans 3 kHz, therefore the recorded speech was filtered through a synthesized telephone

channel to study the effects voice quality. As expected a 3 kHz bandwidth of speech proved to be clearly acceptable.

Methodology MATLAB scripts and the available FDATOOL function were used to simulate both the analog and digital filters. The analog filters were designed by writing MATLAB code for each specification that outputs filter orders, bode plots, pole zero plots, and transfer function coefficients. Once the filters were designed successfully, a speech signal was filtered, and its outputs were studied. OrCAD PSPICE was also used to simulate the parameters found from MATLAB simulation. The digital filters were designed with the FDATOOL function available in MATLAB signal processing toolbox. Once again, the speech signal was filtered, and its outputs studied. After all the data was collected, the results were studied. The optimum output that allowed the system to have the highest X factor and best voice quality whilst minimizing cost and complexity was determined.

Results The filer design is the most critical part of the SSB Weaver method, for it determines:

1. The factor by which the number of channels can be increased.

2. The feasibility of implementing the system.

3. The complexity and ease of implementation.

4. Voice quality preservation. The lowpass analog filters were created and simulated by MATLAB. Table 3 illustrates the filter orders that were computed for each individual filter.

6

Table 3: Analog Filter Orders

Filter Order (n) Chebychev X

Factor Butterworth Type I Type II

Elliptic

3 6 4 4 3 4 8 5 5 4 5 12 6 6 4 6 19 8 8 5 7 46 13 13 6 8

As expected, the Elliptic filter with the “X Factor” of 3 had the smallest order of 3, and the Butterworth with the “X Factor” of 7 had the highest order of 46. The magnitude and phase responses were studied to ensure that the filter magnitudes had the widest 3-dB bandwidth,

and the phases did not cause distortion to the voice signals. Figures 2, 3, 4, and 5 show Bode responses for all Butterworth, Chebychev I, Chebychev II, and Elliptic filter specifications, respectively.

102

103

104

105

106

-4320

-3600

-2880

-2160

-1440

-720

0

Phas

e (d

eg)

Bode Diagram

Frequency (rad/sec)

-2000

-1500

-1000

-500

0

500LOWPASS BUTTERWORTH FILTERS

Mag

nitu

de (d

B)

X = 3X = 4X = 5X = 6X = 7

Figure 2: Lowpass Butterworth Filter

7

-600

-500

-400

-300

-200

-100

0LOWPASS CHEBYCHEV I FILTERS

Mag

nitu

de (d

B)

101

102

103

104

105

106

-1260

-1080

-900

-720

-540

-360

-180

0

Phas

e (d

eg)

Bode Diagram

Frequency (rad/sec)

X = 3

X = 4

X = 5

X = 6

X = 7

Figure 3: Lowpass Chebychev I filter

-160

-140

-120

-100

-80

-60

-40

-20

0LOWPASS CHEBYCHEV II FILTER

Mag

nitu

de (d

B)

103

104

105

106

107

-180

0

180

360

540

720

900

1080

1260

Phas

e (d

eg)

Bode Diagram

Frequency (rad/sec)

X = 3X = 4X = 5X = 6X = 7

Figure 4: Lowpass Chebychev II filter

8

-400

-350

-300

-250

-200

-150

-100

-50

0LOWPASS ELLIPTIC FILTER

Mag

nitu

de (d

B)

10-2

10-1

100

101

102

-360

-180

0

180

360

540

720

900

Phas

e (d

eg)

Bode Diagram

Frequency (rad/sec)

X = 3X = 4X = 5X = 6X = 7

Figure 5: Lowpass Elliptic filter

The Bode plot shown in Figure 2 shows the rate of roll offs and the phase responses of each Butterworth filter specification. As expected, X factor = 7 resulted in the best roll off. Although the roll off rate was highest, there is a significant amount of phase distortion. All other Bode diagrams have the same characteristic where X-factor 7 out performs the rest of the X-factors.

The digital Equiripple FIR filters were designed and simulated in MATLAB as well. Table 4 shows the required orders for various X-factors. X-factor 3 results in the smallest filter order of 8, while X-factor 7 requires a filter of order 52. Figures 6 and 7 illustrate the magnitude and phase responses, respectively, for all digital filters.

Table 4: Digital Filter Orders

X FactorFilter

Order (n) 3 8 4 10 5 15 6 23 7 52 8

The phase responses of all the filters are linear. A linear response is desirable for audio processing because it avoids distortion. The overall hardware cost and complexity for each X-factor is equivalent. The only difference in implementation of an X-factor of say 2 or 7 is in software.

9

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9-120

-100

-80

-60

-40

-20

0

20

Normalized Frequency (×π rad/sample)

Mag

nitu

de (d

B)

Magnitude Response (dB)

X = 3X = 4X = 5X = 6X = 7

Figure 6: Magnitude response of digital filters factor X=3 to X=7

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9-2500

-2000

-1500

-1000

-500

0

Normalized Frequency (×π rad/sample)

Phas

e (d

egre

es)

Phase Response

X = 3X = 4X = 5X = 6X = 7

Figure 7: Phase response of digital filters factor X=3 to X=7

Conclusion Single sideband technology with Weaver modulation technique may be implemented;

using either digital or analog filtering techniques, to increase the aeronautical communication system capacity by 3 to 7 folds. Hardware costs for implementing digital

10

Weaver method are almost negligible among all X-factors. The digital Weaver method, if implemented; provides the following features.

• Low Cost: digital implementation of Weaver SSB is cost effective, and system upgrade is achieved merely by altering software code.

• Voice Quality: phase distortion plays a significant role in hampering the quality of speech. Linear phase responses are highly desirable in that regard; FIR digital filters afford linear phase response.

• Integration: the digital Weaver SSB technique can be easily integrated into the current system. A simple switching mechanism can be installed in the existing systems so that they can be used in whatever mode that is preferred.

References 1 Sleighton Meyer & Tom Hausman., 2003, May. VDL Mode 3 White Paper During Integrated Communications, Navigation and Surveillance Conference. Harris Corp., http://www.govcomm.harris.com/news/view_pressrelease.asp?act=lookup&pr_id=1160

2 Sabin, William E., and Edgar O. Schoenike et al., 1987, Single-Sideband Systems & Circuits. New York: McGraw-Hill Book Company, pp. 3-33 3 Weaver, Donald K., 1956, “A Third Method of Generation and Detection of Single-Sideband Signals”, IRE Proceedings, pp. 1703-1705. 4 Carlson, A. Bruce; Crilly, Paul B.; & Rutledge, Janet C., 2002. Communication Systems (4th ed.). 5 Lian, Ken Jin,1993, Jan. Adaptive Antenna Arrays for Satellite Personal Communication Systems. Master of Science THESIS, Virginia Polytechnic Institute and State University. [online]. http://scholar.lib.vt.edu/theses/available/etd-74181839751071/unrestricted/kjlian.pdf 6 O’Shaughnessy, Douglas, 1987, Speech Communication, Human and Machine. Addison-Wesley Publishing Company, pp 140-151 7 Gold, Ben, and Nelson Morgan, 1999, Speech and Audio Signal Processing. John Wiley & Sons, Inc., pp 8 Akshay Patel, 2006, May, Enhancing Aerospace Communications Channel Capacity Using Single Sideband Modulation, Master thesis, Mercer University.

2007 ICNS Conference 1-3 May 2007

Email Addresses [email protected] – Behnam Kamali

[email protected] – Akshay Patel