Global System for Mobile Communications (GSM) Technology
Investigation on the physical layer technologies employed in the GSM System
Absyarie Syafiq Bin Shahrin
Abstract
Basically in this paper, we intend to give a rundown on GSM (Global System for Mobile Communications) specifically on the technologies employed at the physical layer in the GSM system. The GSM system is a very interesting topic as it revolutionized the way we communicate and it is still being used till this day. It is actually the 2nd Generation (2G) wireless system as it uses digital instead of analog and it also deploys Time Division Multiple Access (TDMA) that is implemented on multiple frequency subbands. Frequency Division Multiple Access (FDMA). GMSK modulation and demodulation technique will also be discussed together with how it works and what their advantages/disadvantages are. The problems with ISI (Intersymbol Interference) in GSM systems will also be addressed together with how to mitigate ISI using channel equalization. With that, we will also give a simple explanation on how speech coding is accomplished in GSM transceivers.
Keywords: Gaussian Pulse, GMSK, ISI, channel equalizer, ISI equalizer, speech coding
I. Introduction
GSM is a standard developed by the ETSI (Europe Telecommunication Standards Institute) to describe the protocols of the 2nd Generation (2G) communication technology used by mobile networks and cell phones. It was first launched in Finland with a data speed of up to 64kbps. The GSM is given the term 2G because it is something completely new compared to the first generation (1G) with the usage of digital signal signals instead of analog. It was designed from scratch with no backward compatibility with the previous 1G technology. Using 124 channels per cell, it can accommodate up to 8 users by using a combination of TDMA and FDD scheme [1], though some of its channels are used for control signals. It also introduces the SIM (Subscriber identity module) card which allows for roaming calls. At first, it was only designed for operation in the 900 MHz band but later it was adapted for 1800Mhz. GSM is a very popular standard used today with over 90% market share, with availability in over 219 countries and territories worldwide.
Originally the GSM was developed with the intention that it will replace the first generation analog networks by having digital, circuit-switched networks which are optimized for full-duplex voice telephony. However as time passes, the GSM system was further developed to include data communications by firstly having it on circuit-switched transport, and then changing it later to packet-switched transport via GPRS ( General Packet Radio Service) and EDGE ( Enhanced Data Rates for GSM Evolution) .
In GSM, Gaussian pulse shaping is used and Gaussian Minimum Shift Keying (GMSK) as a modulation/demodulation technique with a modulation index of 0.5 [2]. This modulation method however gives rise to inter symbol interference. Inter Symbol Interference (ISI) in the GSM system are usually caused by two factors; Multipath propagation and Bandlimited channels. An ISI equalizer is used to solve this problem by implementing the Maximum Likelihood Sequence Estimation (MLSE) via vertibri algorithm. To make things easier to understand, Figure 1 is attached to relate how the GSM system can relate to the OSI (Open System Interconnection) model. We will however, focus more on the Physical Layer of the GSM system.
Figure 1: How the GSM is realized in the famous OSI model [7].
Pulse Shaping
In digital telecommunications systems, we strive to achieve broad spread spectrum with significant low-frequency content. This in return, requires a lowpass channel that has a bandwidth sufficient enough to accommodate the essential frequency content in the data stream. Gaussian function fits this requirement perfectly. The speciality of this waveshape is that, the pulses rise and small smoothly until it settles to a value [14]. This is a valuable asset as it gives a solution to problems such as precursors, overshoot and ringing in a pulse signal [14]. This problems cause uncertainty to the actual value so it is very troublesome. Besides that, it also addresses the two required needs of communication systems which are band-limited channels and reduced Inter-symbol interference (ISI) by applying a Gaussian filter symbol-by-symbol. It is nearly impossible to get the perfect sinc spectrum in the time domain as the bandwidth needs to be infinity. We can only have an approximation or near the same sinc spectrum. ISI can also still happen if control is not exercised over the pulse shaping.
Figure 2: An impulse response of a Gaussian Filter [15]
In GSM, we apply Gaussian filtering for Gaussian Filtering Minimum Shift keying (GMSK) a modulation technique. Basically it is similar as the Minimum Shift Keying (MSK) but the data stream must first go through pulse shaping via Gaussian filter before being applied to the modulator. MSK is already a good modulation scheme as it possess constant envelope and maintains phase continuity. GMSK allows for reduced sideband power which results in the reducing of out-of-band interference between the signal carriers in adjacent frequency channels. The GMSK technique has an advantage of being able to carry data while still maintaining an efficient usage of spectrum. The reduce power in the GMSK is very useful especially for mobile phones as lower battery consumption is needed for operation [16]. The drawback of GMSK is that, it requires more modulation memory in the system and causes ISI.
We have two ways to generate GMSK modulation. The most basic way is to apply Gaussian filter on the input signal and then apply a frequency modulator with a modulation index of 0.5 [2] [16]. The problem with this method is that it must have an exact modulation index of 0.5. In the real world, this is impossible as component tolerance drift can vary[16].
Figure: Flow chart of GMSK modulation using a Gaussian filter and Voltage controlled oscillator
The second method is more realistic and widely used. This GMSK method uses the Quadrature (I-Q) modulator. The operation starts by having the Gaussian filtered data separated into two parts, in-phase I and quadrature phase (Q). The I and Q components will then be mixed up to the frequency of the RF carrier to have a modulated RF signal. This kind of modulator can maintain 0.5 modulation index without having any modifications. The performance of this quadruple modulation depends on the accurate creation of I and Q components. For demodulation, this scheme can be used in reverse [16].
X – mixer or multiplier
LO – Local oscillator
Figure 3: Block diagram of I-Q modulator
Inter symbol interference and channel equalization
ISI in the GSM system is mainly caused by multipath propagation. Multipath propagation is a result when signals arrive at different times (delay) because it is does not travel in line of sight (LOS). In reality, connection will never be in LOS all the time so the signals will go through different paths by being reflected or refracted from different objects to reach the destination. When the signals travel through multiple paths, they will arrive at different times depending on the route they used. It is also possible for reflected signals to overlap with the subsequent signals [13]. This in addition, results in distortion to the received signals because all the signals have different delay. This situation happens either from mobile station to base station or vice versa. Since the delay spread is more than the symbol time, frequency selective fading occurs.
Figure 4: An example of multipath propagation
Figure 5: ISI as a result of multipath distortion [13]
To combat the problem with multipath propagation, we use and ISI equalizer. This equalization technique is based on the MLSE which uses the Viterbi Algorithm [3] [10]. Figure below shows the block diagram of the ISI equalizer.
Figure 6: Block diagram on how ISI equalizer is used in GSM environment
When the base station or the mobile station transmits a TDMA burst, not all of is user data. Instead, 26 bits are allocated for the training sequence and they are known by their receivers (either mobile station or base station). Each of the known sequence bits unique for a certain transmitter is unique for a certain transmitter and this sequence bits is also repeated in every transmission burst. The figure below shows the normal burst structure in the GSM burst.
Figure 7: GSM Normal Burst Structure
A channel estimator is needed because to perform MLSE, we require information on the CIR (Channel Impulse Response). The channel estimator will estimate the CIR for each of the bursts by comparing the transmitted bits with the received signal to produce he(t) [10]. Channel estimation in GSM uses Linear MMSE (Minimum mean square error) [11]. Since the match filter is in time domain, r(t) will be convoluted with the signal obtained from the channel estimation, he(t) to create a model signal Y(t). The output model signal obtained can then be used to estimate the transmitted bits based on the bits received by performing MLSE. The last process uses Viterbi Algorithm hence the process, Viterbi equalisation [2] [9].
Speech coding in GSM transceivers
Speech is originally analog in nature and GSM is a digital system. In order to use the speech information, we need to run to a series of process known as speech processing. Figure shows how the speech processing is done in a GSM system.
In speech coding, the GSM system has used a variety of ways to fit in 3.1 kHz audio into between 6.5 and 13 kbit/s. The first two codecs used was called Half Rate (5.8 kbit/s) and Full rate (13 kbit/s) [4]. Both of this codecs use LPC (Linear Prediction Function) where voice signals need to be digitized, and secured using encryption over a narrow voice channel. As time passes, the GSM system was further developed to use the Enhanced Full Rate (EFR) codec which is a 12.2 kbit/s codec and it uses a full-rate channel.
Figure 8: Flow-diagram on GSM speech processing [8]
Full rate speech coder is actually part of the Regular Pulse Excitation – Long Term Prediction (RPE-LTP) coders [4]. Firstly the speech encoder will take an input of 13 bit uniform PCM signal from either the audio part of the mobile station (MS) or the Public Switched Telephone Network (PSTN) side by using 8 bit/A-law to 13 bit uniform PCM conversion. The encoded speech is then delivered to the channel coding function which will then produce an encoded block having 456 bits with a gross bit rate of 22.8 kbps [4] [5]. The remaining 9.8 kbps is used for error protection purposes. The reverse action is performed for decoding. When encoding, 160 frames in 1 sample is encoded to a block of 260 bits with a sampling rate of 8000 samples/s, hence the bitrate of 13kbps [5]. On the decoding part, 260 bits of encoded blocks is mapped back to the 160 frames output reconstructed speech sample.
EFR (Enhanced Full Rate) is a newer version of the speech codec which uses ACELP (Algebraic Code Excited Linear Prediction) algorithm. The motivation for this development is because of the mediocre / poor quality of the GSM-Full Rate codec. This codec is a step-up from the previous FR because it provides speech quality equivalent or close to wireline telephony which uses 32 kbps ADPCM (Adaptive Pulse Code Modulation) [6]. This codec can provide wireline quality in both error and error-free conditions [6]. EFR which is also a form of traffic channel is bi-directional and can transmit both speech and data [9].
Figure 9: shows how error correction is done at layer 1 of the GSM air interface
Conclusion
All in all, this paper has helped me to better understand the GSM system and how it works in the physical layer. GSM has many sources including but not limited to, books, journals, application notes, lecture notes, documentation as well as survey papers. After reading from various sources, I learned to read efficiently and think critically as the papers written are quite hard and requires a meticulous reading to thoroughly understand what is being presented. I acquired basic research and development (R&D) skills and technical writing skills after almost a month of heavy reading and research. How the physical layer in the GSM system works is also understood. The acquired signal must first be shaped through a Gaussian filter in the GMSK modulator. The Quadruple modulator scheme is used as it does not require modifications to maintain 0.5 modulation index. ISI in the GSM are mostly caused by multipath propagations in which gives frequency selective fading. Frequency selective fading happens when the delay time is spread because symbols arrive at different times. To address the problems with ISI, we need to have an ISI equalizer. ISI equalizer consists of many components such as match filter and MLSE by Viterbi algorithm. I also learned that we have two speech coding options; full rate speech coder and EFR. All this components are essential when building a GSM system.
References
[1] Guifen Gu, Guili Peng “The Survey of GSM Wireless Communication System” International Conference on Computer and Information Application (ICCIA) , 2010
[2] B. Baggini, L. Coppero, G. Gazzoli, L. Sforzini, F. Maloberti, G. Palmisano “Integrated Digital Modulator and Analog Front-End for GSM Digital Cellular Mobile Radio System”, Proc. IEEE 1991 CICC vol. 31, pp.7.6.1{4, Mar. 1991.
[3] M. Drutarovsky, “GSM Channel Equalization Algorithm – Modern DSP Coprocessor Approarch” Radioengineering Vol. 8, No 4, December 1999.
[4] Besacier, L.; Grassi, S.; Dufaux, A; Ansorge, M.; Pellandini, F., “GSM speech coding and speaker recognition,”Acoustics, Speech, and Signal Processing, 2000. ICASSP ’00. Proceedings. 2000 IEEE International Conference on, vol.2, no., pp.II1085,II1088 vol.2, 2000
[5] www.etsi.org, “European digital cellular telecommunications system (Phase 1); Speech Processing Functions; General Description (GSM 06.01)”, GTS 06.01 version 3.0.0, January 1991.
[6] Jarvinen, K.; Vainio, J.; Kapanen, P.; Honkanen, T.; Haavisto, P.; Salami, R.; Laflamme, C.; Adoul, J.-P., “GSM enhanced full rate speech codec,” Acoustics, Speech, and Signal Processing, 1997. ICASSP-97., 1997 IEEE International Conference on , vol.2, no., pp.771,774 vol.2, 21-24 Apr 1997
[7] “Fundamentals: Signalling at the Air-Interface” Rohde and Schwartz Training Center v1.0
[8] http://www.rfwireless-world.com/Tutorials/gsm-speech-processing.html
[9] “GSM Air Interface & Network Planning” Training Document, Nokia Networks Oy, Finland, Jan 2002
[10] Vipin Pathak,“MLSE BASED EQUALIZATION AND FADING CHANNEL MODELING FOR GSM” (Hughes Software systems, Delhi), pp. 100-104, 2003
[11] Manoj Bapat, Dov Levenglick, and Odi Dahan, “GSM Channel Equalization, Decoding, and SOVA on the MSC8126 Viterbi Coprocessor (VCOP)” Freescale Semiconductor Application Note, Rev.0, 2005
[12] Baltersee, J.; Fock, G.; Meyr, H.; Yiin, L., “Linear MMSE channel estimation for GSM,” Global Telecommunications Conference, 1999. GLOBECOM ’99 , vol.5, no., pp.2523,2527 vol.5, 1999
[13] Kang, A. S., and Vishal Sharma. “Pulse Shape Filtering in Wireless Communication-A Critical Analysis.” Pulse 2, no. 3 (2011).
[14] James R. Andrews, “Low-Pass Risetime Filters for Time Domain Applications”, Picosecond Pulse Labs, Application Note AN-7a, March 1999.
[15] http://www.ni.com/white-paper/3876/en/
[16] http://www.radio-electronics.com/info/rf-technology-design/pm-phase-modulation/what-is-gmsk-gaussian-minimum-shift-keying-tutorial.php
[17] Fred Kostedt, James C. Kemerling, “Practical GMSK Data Transmission”, MX.com, INC, Application Note GMSK, 1998.
