804
`
`IEEE Transactions on Consumer Electronics. Vol. 41, No. 3, AIJGUST 1995
`
`EXPERIMENTAL TEST AND EVALUATION OF A GMSK CHIPSET
`COMPATIBLE WITH THE GSM AND PCS STANDARDS
`Huajing Fu
`Digital Wireless Comm. Lab, Electrical Engineering Dept.
`University of California, Davis, CA 9561 6
`Dr. Kamilo Feher
`Fellow IEEE, Prof. of ECE, University of California, Davis
`and Consulting Group, Digcom Inc., Davis, CA 9561 6
`
`a Quadrature Baseband Processor/QUAD
`by
`modulation architecture, such as illustrated in fig
`in quadrature
`I(b). This property reveals that
`modulated GMSK, the baseband I and Q signals are
`not
`independent signals. Many of the GMSK
`advantages are related to this property. For example,
`cross-correlation lead to GMSK constant magnitude,
`so the subsequent RF power amplifier can operate
`in non-linear mode with a greatly improved power
`efficiency. In today’s wireless cellular PCS systems,
`many devices are driven by battery,
`and large
`is consumed by RF power
`amount of power
`amplifier, so power efficient modulation becoming
`key factor for power saving.
`
`The cross-correlation property could be described as
`a discovery that is “against” the well-established
`wisdom of classical linear communication system
`theories and principles. In classical theory such as in
`Haykin’s[9] textbook, it is demonstrated that for
`theoretical optimum QPSK system, at the end of a
`signaling interval T, the output of the correlater in
`the in-phase channel is
`1, = k - A , T + ~ ( t ) ~0~(2~tf,t)dt
`
`
`
`
`the correlater
`
`in
`
`the
`
`2 *
`‘
`
`i
`I
`
`0
`the output of
`whereas
`quadrature phase channel is
`I, = k - A , T + w(t) sin(2nf, r)dt
`2
`0
`For optimal performance the random variables L1
`and Lz, whose values are denoted by 1, and 12, are
`uncorrelated. They are also Gaussian because they
`are derived from the Gaussian process w(t) by a
`linear filtering operation. Accordingly,
`they are
`
`Abstract
`
`test and
`experimental
`This paper presents
`evaluation results of a baseband processor chipset
`of Gaussian Minimum Shgt Keyed
`(GMSK)
`quadrature modulator. Quadrature GMSK and
`related quadrature Feher ’s QPSK
`(FQPSK)
`modulator structure have been of interest to digital
`communication engineers. In this contribution we
`hightlighted the cross-correlation properties of the
`in-phase(I) and the quadrature-phase(@ signals,
`described some of
`the advantages of cross-
`correlation, gave a simple analytical expression,
`and discussed correlated signal processors initiated
`by Kato/Feher[2] and also described in[l] and
`other recent references.
`
`1. Introduction
`
`The Gaussian Filtered Minimum Shift Keyed
`(GMSK) modulation format has very important
`applications in digital wireless communications.
`European digital cellular systems specified it for the
`Global Mobil System
`(GSM) standard with
`normalized Gaussian lowpass filter bandwidth BTb
`= 0.3. In the Cellular Digital Packet Data (CDPD)
`standard GMSK is specified with BTb=0.5. One of
`the advantages of GMSK is, besides Frequency
`Shift Keyed (FSK) modulation, it can be realized by
`quadrature phase shift keyed modulation. This
`property makes GMSK more attractive
`for
`applications which requires a robust performance.
`
`GMSK has a non-obvious property. There is cross-
`correlation
`between
`the
`In-phase (I) and
`Quadrature-phase (Q) signal when it is implemented
`
`Manuscript received June 12, 1995
`
`0098 3063/95 $04.00 e 1995 IEEE
`
`Qualcomm Incorporated
`Exhibit 1014
`Page 1 of 5
`
`

`

`Fu and Feher: Experimental Test and Evaluation of a GMSK Chipset Compatible With the GSM and PCS Standards
`
`805
`
`fier
`
`(a) GMSK implemented by Frequency
`Shift Keying modulation with FM-VCO
`
`"' amp 1 i fier
`
`t
`
`(
`
`a
`
`Gaussian
`
`(b) GMSK implemented a by quadrature baseband correlated architecture
`
`fc '
`
`Figure 1. Two methods to generate the GMSK signal, the shaded areas in (a) and (b) have the same
`function. The GMSK signal m(t) is same both in (a) and (b).
`
`quadrature phase shift keyed modulation. Fig 1
`shows the two methods. The first method requires
`the frequency deviation factor exactly equals to 0.5,
`that is m= 0.500. It is difficult to meet this demand
`by VCO that must maintain high precision on
`frequency deviation. The second method
`looks
`complicated, but it is relatively easy to implement.
`The second method has another advantage of being
`able to extended to another modulation mode, to
`form the dual compatibility modulation format.
`Most of the practical applications use quadrature
`modulation method to generate GMSK.
`
`statistically independent, thus any form of cross-
`correlation between I and Q will cause performance
`degradation. However, as pointed out in this paper,
`I and Q channel signals
`instead of keeping
`independent, we
`introduced
`cross-correlation
`between
`them, and
`it will
`improve
`some
`characteristics
`in certain applications, especially
`when
`the
`application
`requires
`non-linear
`amplification environment.
`
`This cross-correlation property was initiated and
`included in the Kato/Feher's patent, "The correlated
`signal processor,"'" which gives a good description
`of this property. This patent has many impacts on
`modulation and overall digital communication
`system designs.
`In
`this paper, we use
`the
`measurement results of
`a commercial GMSK
`baseband processor chipset to demonstrate the
`GMSK cross-correlation property
`
`n (5)
`c a,,
`
`In figure 1,
`+cc
`
`a(t) =
`
`1,=--00
`
`T
`b(t) can be expressed as:
`I_
`
`2. Cross-correlation
`
`where
`
`There are two methods to generate GMSK, one is
`the other is
`frequency shift keyed modulation,
`
`Page 2 of 5
`
`

`

`806
`
`IEEE Transactions on Consumer Electronics, Vol. 41. No. 3, AlJGUST 1995
`
`deliberately introduce cross-correlation here. I and
`Q start from the same signal c(t). We noted that the I
`and Q signals are cross-correlated. The reason is
`that we can express the I(t) in term of Q(t).
`I(t) = cos (sin-'(Q(t)))
`
`and also express Q(t) in term of I(t).
`Q(t) = sin ( cos-' (I(t)))
`
`The experimental GMSK baseband processor
`chipset uses the above approach
`to generate its
`baseband signals with Gaussian LPF BTb=0.3. The
`output baseband waveforms are shown in figure 2.
`In these waveforms, there are ripples whenever the
`signals have consecutive '0's or '1's. The signal
`magnitudes are not equal, with ripple bottom is
`about 0.85 of the maximum magnitude. From study
`of this phenomenon, we can observe the effects of
`cross-correlation. The quadrature signal Q reaches
`its highest magnitude if and only if the in-phase I is
`crossing zero, and the quadrature signal Q is at
`lower amplitude when the inphase signal I is at
`bottom of the ripple, or vice versa.
`
`Figure 3 shows the constellation of I and Q
`baseband signals. The encoded output signals have
`amplitudes such that the vector sum of I and Q
`signals are approximately the same at virtually all
`phases of each bit period. The vector magnitude can
`be expressed as,
`m(t) = d
`m
`
`= ,/COS2 (c(t)) + SIN' (c(t)) = 1
`
`Because of the phase ambiguity, we can not tell the
`phase of Q(t) when we know I(t), but we know the
`amplitude of Q(t) if we know I(t). So the normalized
`m(t) will always have the same value as results of
`quadrature modulation.
`
`3. Power efficiency and BER improvement
`
`One of the advantages of cross-correlation is it lead
`to power efficiency. In GMSK, cross-correlation
`makes the modulated RF signal constant envelope,
`suitable for nonlinear power amplification. This
`overcomes the crucial problem of output power
`backoff (OBO). O B 0 normally reduce 6-8 dB
`output power in non-constant envelope modulation
`
`Figure 2. GMSK(BTb =0.3) baseband waveforms,
`the top is PRBS NRZ input data, in the middle is the
`I signal, and bottom is the Q signal.
`
`Figure 3. Constellation of I and Q signals. There are
`4 shiny lines in the constellation graph, it means the
`signal phases stay here longer than other state. The
`shiny lines are well constrained in a circle due to
`cross-correlation property.
`
`and c(t) is
`
`I
`c(t) = b(l)dr
`-a
`I(t), Q(t) signals are,
`Z ( t ) = cos(+))
`Q(t) = sin(c(t))
`
`I and Q are generated by sin( ) and cos( ) of c(t). c(t)
`is used both by I and Q. From our point of view, we
`
`Page 3 of 5
`
`

`

`Fu and Feher: Experimental Test and Evaluation of a GMSK Chipset Compatible With the GSM and PCS Standards
`
`807
`
`of these results is related to the GSM world-wide
`PCS standard, and power efficiency
`in digital
`wireless cellular systems.
`
`References
`
`[I]. K. Feher. "Wireless Digital Communications:
`Modulation and spread spectrum applications"
`Prentice-Hall. 1995
`signal
`"Correlated
`[2]. S. Kato, K. Feher.
`Processor." U S . Patent N0.4~567~602. Jan 1986
`[3]. K. Feher. "GMSK, GFSK and FQPSK
`Implementations of Feher's Patented-Licensed
`Technologies" Proc. 3rd
`Annual Wireless
`Symposium and Exhibition, San Jose, CA Feb.
`1995
`"Digital Communications:
`Feher.
`[4]. K.
`SateIlite/Earth Station Engineering," Prentice
`Hall 1983.
`[5]. C. I. Cook, "Development of Air Interface
`IEEE
`Personal
`Standard
`for
`PCS
`,"
`Communication Magazine, Fourth Quarter, 1994
`[6]. H. Suzuki, K. Momma and Y. Yamao, "Digital
`Portable transceiver using GMSK Modem and
`ADM Codec.". IEEE Journal, Selected Areas in
`Communication, Vol. SAC-2, July 1984.
`[7]. K. Feher, "Advanced Digital Communications,
`System and Signal Processing Techniques",
`Prentice-Hall, 1987
`[SI. Gao Wei, K. Feher I' Performance Evaluation of
`GSM baseband Interface for GMSK system
`Applications",
`International Symposium on
`Satellite Communication and Remote Sensing.
`Xian'an China, SCRS'95.
`[9].
`Simon Haykin,
`"Communication
`Systems", John Wiley & Sons 1978.
`
`Figure 4. GSMK (BTb=0.3) power spectrum density.
`The modulation method is quadrature phase shift
`key. The solid line is linear amplified GMSK PSD,
`and the background line, which is only visible under
`55dB, is non-linearly amplified spectrum. The non-
`linear spectrum is well confined with in 55dB.
`
`format, like QPSK, T Q P S K etc.. We measure the
`4
`spectrum spread in several cases, such as class C
`amplifier, hardlimiter, power amplifier at 1 dB
`compression point and fully saturated state. All of
`the results show that the GMSK is superior to other
`modulation format. Figure 4 is the measured power
`spectrum of GMSK, where the RF power amplifier
`is working at fully saturated state.
`
`The BER measurement is also satisfactory. Under
`additive Gaussian white noise, Eb/N, is about 10.0
`This result is very
`dB, the measured P, is 1 x
`close to the results of computer simulation of
`GMSK performance, which is about 9.8dB when
`error rate is 1 x 1 O-4. It is important to point out that
`this is partly due to the cross-correlation property.
`
`4. Conclusion
`
`In this paper we discussed a neglected fact that there
`is cross-correlation between I and Q signals in
`GMSK modulation format, when implemented by
`quadrature phase shift keyed modulation. The
`experimental measurement of a commercial GMSK
`chipset demonstrate this property. The importance
`
`Page 4 of 5
`
`

`

`808
`
`IEEE Transactions on Consumer Electronics, Vol. 41, No. 3, AUGUST 1995
`
`Huajing Fu received the M.S.EE
`at Beijing University of Posts
`in
`and
`Telecommunications
`1982, he is currently a Ph.D.
`candidate
`at University
`of
`California, Davis. He was a
`senior engineer at Data Comm
`Research Institute of PTT of
`China from 1982 to 1990, led an R&D project on
`Store Program Controlled
`(SPC) TeledData
`Exchange. He received the “China National Science
`and Technology Progressive Award” in 1990, the
`most prestigious award in China for his work in SPC
`switching system. He was with Telecom Australia
`Research Lab from 1990 to 1991, did research on
`ISDN. His current research topics include digital
`wireless
`cellular
`communications,
`digital
`modulation/demodulation, spread spectrum, CDMA,
`fast synchronization.
`
`Dr.Kamilo Feher, Fellow IEEE,
`Professor of ECE, University of
`California, Davis, directs one of
`the most productive experimental
`digital wireless modulation/RF
`design research Laboratories. He
`is author of six books and of many
`R&D
`engineering publication.
`Through Dr. Feher Associated
`Digcom, Inc., and the FQPSK Consortium, he is
`active in consulting, training, technology transfer
`and licensing of his filter, processor, GMSK, GFSK,
`“F-modem,” FBPSK and FQPSK family of patented
`inventions.
`In Feher’s book, Wireless Digital
`Communication: Modulation and spread spectrum
`Applications, published by Prentice hall, May 1995,
`cutting edge information and the last word on
`virtually any wireless digital communications
`TDMA, CDMA, CSMA, modulation, RFIC, design
`and
`implementations
`is provided. This book
`includes the powerful
`“CREATE”
`wireless
`modem design package.Dr. K. Feher, Electrical and
`Computer Engineering Department, University of
`California, Davis, CA 95616. Tel: (916) 752-8127
`Fax:(9 16) 752-8428, or at Digcom, Inc.(916)753-
`0738 Fax: (916) 753-1788.
`
`Page 5 of 5
`
`