`
`4:3;
`
`m ,
`7
`
`",
`
`HILIPS
`
`----~-~
`
`VHTC EX1023, Page 1
`
`|PR2018—1556
`
`IPR2018-1556
`HTC EX1023, Page 1
`
`
`
`......
`".. '...
`
`'.
`
`Philips Journalof Research
`
`Philips Journalof Research, published by the Philips Research Laboratories,
`Eindhoven, The Netherlands,
`is a bimonthly publication containing papers on
`research carried out in the various Philips laboratories. Volumes 1 to 32 have
`appeared under the name of Philips Research Reports.
`
`subscription price for Volume 39 is Dil. 75, payable in advance.
`Annual
`Payments
`should be made only after
`receipt of an invoice. Correspondence
`should be addressed
`to: Philips
`Journalof
`Research, Philips Research
`Laboratories, Building WBp, Room No. 42, Eindhoven, The Netherlands.
`
`Editorial Board
`
`A. van Oostrom (General Editor) Philips Reseàrch Laboratories,
`5600 JA Eindhoven, The Netherlands
`
`Y. Genin, Brussels, Belgium
`R. Kersten, Aachen, Germany
`R. Memming, Hamburg, Germany
`R. F. Milsom, Redhill, Surrey, U.K.
`W. A. Smith, Briarcliff Manor, N.Y., U.S.A.
`W. T. Stacy, Sunnyvale, CA., U.S.A.
`J. B. Theeten, Limeil-Brévannes, France
`
`Cover design based on a visual representation
`word "Philips" .
`
`of the sound-wave associated with the spoken
`
`© Philips
`illustrations
`1984. Articles or
`International B.V., Eindhoven, The Netherlands,
`reproduced in whole or in part must be accompanied by a full acknowledgement of the source:
`Philips Journalof Research.
`
`IPR2018-1556
`HTC EX1023, Page 2
`
`
`
`t,~PH'UPS' GLOElLAMPfNfAB,
`Research
`
`Philips
`
`Journalof
`
`Volume 39
`1 984
`
`© Philips International B.V., Eindhoven, The Netherlands, 1984.
`Articles or illustrations reproduced,
`in whole or in part, must be accompanied
`by full acknowledgement of the source: Philips Jo~rnal of Research.
`
`IPR2018-1556
`HTC EX1023, Page 3
`
`
`
`CONTENTS
`
`'PHILIPS
`
`JOURNAL OF RESEARCH, VOL. 39
`
`R 1077 R. J. Murray and
`.
`R. W. Gibson
`
`R 1078 P. C. Zalm and
`L. J. Beekers
`
`for
`A coherent digital demodulator
`minimum shift
`key and related
`modulation schemes
`.
`.
`.
`.
`..
`
`In quest of the spike: energy depen-
`dence of the sputtering yield of zinc
`bombarded with neon and xenon
`ions
`.
`.
`.
`
`R 1079 J. Pergrale, C. Berche,
`D. Iachetti and
`G.Normand
`
`between attentuation
`Comparison
`correction methods
`in transaxial
`single photon emission tomography
`
`R 1080 G. G. P. van Gorkom
`andA.M.E.
`Hoeberechts
`
`R 1081 P. C. Zalm and
`L. J. Beekers
`
`R 1082 J. L. C. Daams and
`K. H. J. Buschow
`
`R 1083 K. H. J. Buschow and
`P. G. van Engen
`
`An efficient silicon cold cathode for
`high current densities.
`.
`.
`..
`
`Secondary electron yields from dean
`polycrystalline metal surfaces bom-
`barded by 5-20 keV hydrogen or
`noble gas ions
`.
`.
`.
`.
`.
`..
`
`The crystal structure of LaNiSn .
`
`Magneto-optical properties of rare
`earth cobalt compounds and amor-
`phous alloys.
`.
`.
`.
`.
`.
`.
`.
`.
`
`R 1084 W. J. W. Kitzen and
`P.M. Boers
`
`Applications of a digital audio-signal
`processor in TV. sets.
`.
`.
`.
`.
`.
`
`Pages
`
`1- 10
`
`11- 23
`
`24- 50
`
`51- 60
`
`61- 76
`
`77~ 81
`
`82- 93
`
`94-102
`
`R 1085 H. G. R. Maas and
`J.A.Appels
`
`PABLO a versatile VLSI technology 103-108
`
`H. B. G. Casimir
`
`Preface.
`
`.
`
`.
`
`.
`
`.
`
`109-110
`
`R 1086 A. Blokhuis and
`J. J. Seidel
`
`R 1087 J. Boersma and
`R. Nijborg
`
`An introduetion to multilinear al-
`gebra and some applications..
`
`111-120
`
`Some electrostatic problems
`spindle.
`.
`.
`.
`.
`.
`.
`..
`
`for a
`
`121-142
`
`ii
`
`Phillps Journalof
`
`Research Vol.39 No.6
`
`1984
`
`IPR2018-1556
`HTC EX1023, Page 4
`
`
`
`R 1088 A. R. Calderbank and
`.
`J.-M. Goethals
`
`Three-weight
`schemes.
`.
`
`codes and association
`.
`.
`.
`.
`.
`.
`.
`.
`
`Pages
`
`. 143-152
`
`R 1089 P. Branquart
`
`R 1090 P. J. Courtois and
`P. Semal
`
`R 1091 Ch. Couvreur and
`Ph. Piret
`
`R 1092 M. Davio
`
`R 1093 P. Delsarte, Y. Genin
`and Y.Kamp
`
`R 1094 C. Dierieck and
`F. Crowet
`
`R 1095 K. H. J. Buschow
`
`A method
`algorithmic
`
`of code generation
`languages
`.
`.
`.
`
`for
`.
`
`. 153-177
`
`decomposition
`Block
`in stochastic matrices
`
`and iteration
`...
`.
`
`. 178-194
`
`Codes between BCH and RS codes
`
`195-205
`
`Algorithmic
`design
`.
`
`.
`
`aspects of digital system
`.
`.
`.
`.
`.
`.
`.
`.
`
`. 206-225
`
`of the index theory of
`Application
`pseudo-lossless
`functions
`to
`the
`Bistritz stability test
`.
`.
`.
`.
`. 226-241
`
`.
`
`decomposition
`Helmholtz
`tiply connected domains
`
`on mul-
`
`242-253
`
`Formation,
`physical
`3d-based
`
`thermal
`properties
`alloys.
`
`.
`
`and
`stability
`of
`amorphous
`.
`.
`.
`.
`.
`
`. 255-274
`
`R 1096 J. L. C. Daams and
`J.}{.lV.
`van Vucht
`
`Contribution
`Hg
`.
`.
`
`.
`
`to the system Mg-Au-
`.
`.
`
`R 1097 W. J. van Gils
`
`of optimal binary
`Some constructions
`linear unequal error proteetion codes
`
`R 1098 C. Ronse
`
`Networks
`
`for sorting with fusion
`
`Author
`
`to index volume 39
`
`275-292
`
`293-304
`
`305-316
`
`317-319
`
`Supplement No. 1
`J. J. G. Willems
`
`Metal hydride electrodes
`components.
`LaNi5-related
`
`stability of
`..
`.
`
`1- 94
`
`Phillps Journal of Research Vol.39 No.6
`
`1984
`
`iii
`
`IPR2018-1556
`HTC EX1023, Page 5
`
`
`
`Philips J. Res. 39, 1-10, 1984
`
`R 1077
`
`A COHERENT DIGITAL DEMODULATOR FOR
`MINIMUM SHIFT KEY AND RELATED MODULATION
`SCHEMES
`
`by R. J. MURRAY and R. W. GIBSON
`Philips Research Laboratories, RedhilI, Surrey, RH] 5HA, U.K.
`
`Abstract
`We show that in a coherent digital demodulator both the clock and carrier
`signals can be recovered from the hard-limited outputs of the two qua-
`drature channels. Clock and carrier are recovered simultaneously thus per-
`mitting fast-acquisition direct-conversion radio receivers.
`MSK, TFM, GMSK and similar schemes can be demodulated. Measured
`Bit-Error-Rates
`for TFM were within 0.5 dB of that obtained with a
`reference clock and carrier. Acquisition within 30 bits was achieved with a
`degradation in performance of less than 1dB: shorter acquisition times are
`possible with some further loss of performance.
`EECS numbers: 61,64.
`
`1. Introduetion
`in
`The possibility of fully integrated radio receivers has renewed interest
`direct demodulation (zero i.f.) techniques. At the same time applications are
`beginning to arise for receivers which are required to handle digital signals
`only. One class of digital signal particularly suited to radio is Minimum Shift
`Key (MSK) and its derivatives Tamed Frequency Modulation (TFM) and
`the
`It may be observed that
`Gaussian Minimum Shift Key (GMSK) 1,2).
`coherent demodulator
`for this class of signals is almost identical in layout to a
`direct demodulation radio receiver. Hence the possibility arises of combining
`the two functions.
`Figure 1 shows the form of a combined receiver and demodulator. The
`modulated r.f. signal is shifted down to baseband by a pair of quadrature
`mixers and then lowpass filtered. Since the lowpass filters provide the selec-
`tivity of the receiver it follows that both carrier and clock recovery must be
`performed downstream of these filters. Thus we have to recover both clock
`and carrier from the baseband signal; this is different from many versions of
`the coherent demodulator which recover the carrier by a phase locked loop at
`r.f.
`(or at a conventional non-zero i.f.). Also, since MSK is a constant en-
`velope signal it is advantageous to amplify the filter outputs in hard limiting
`amplifiers,
`thus avoiding problems of a.g.c.
`
`Phillps Journal of Research Vol.39 Nos 1/2 1984
`
`1
`
`IPR2018-1556
`HTC EX1023, Page 6
`
`
`
`R. J. Murray and R. W. Gibson
`
`modulated
`r.f.
`
`demodulator
`
`data out
`
`Q_
`
`Fig. 1. Combined receiver and demodulator.
`
`that
`results,
`We will now proceed to show, both by theory and experimental
`it is possible to recover both the clock and carrier signals
`solely from the hard
`limited baseband signals. Moreover
`the technique
`employed
`allows simultane-
`ous acquisition
`of
`the
`clock and carrier
`and is thus
`suitable
`for
`systems
`requiring
`fast acquisition.
`
`2. Principles of the clock and carrier recovery technique
`solely from the hard
`Since we propose
`to recover both the clock and carrier
`limited outputs of the two quadrature
`channels
`the only information
`available
`to us is contained
`in the timing of the zero crossings. The technique will be
`explained with reference
`to fig. 2 in which the phase
`trajectories
`shown
`represent
`either MSK with no bandwidth
`restrictions
`(the sharp
`angles),
`or TFM, GMSK,
`etc.
`(the smoothed
`curve). The curve
`shows the phase
`of
`the modulated
`r.f.
`signal
`relative to its carrier. But
`the phase of the signal
`is
`transferred
`directly through
`the two quadrature mixers,
`subject only to two
`constant
`shifts separated
`by 90 degrees. We can therefore
`label the phase
`axis
`of the diagram with the equivalent phases
`at the outputs
`of the I and Q chan-
`nels.
`to the trajectory
`then correspond
`in the I and Q channels
`The zero crossings
`which are spaced at 90 degrees of phase. The solid
`crossing the horizontallines
`in the Q channel and the dashed
`lines represent
`the zero and 180 degree phases
`in the I channel. The solid vertical
`lines the zero and 180 degree phases
`lines
`represent
`the instants
`at which the Q channel
`is expected
`to have a transition
`and the I channel
`to be at the centre of its eye. The dashed verticallines
`cor-
`respond
`to the I channel
`transitions
`and the eye centres of the Q channel.
`
`2
`
`Philips Journalof
`
`Research Vol.39 Nos 1/2
`
`1984
`
`IPR2018-1556
`HTC EX1023, Page 7
`
`
`
`Coherent digital demodulator for MSK and RM schemes
`
`4
`
`3
`
`5
`
`Q I
`
`Q phase
`I
`1800
`_L
`900
`00 I
`2700
`1800
`900
`
`+
`
`+
`
`+
`
`+
`
`phase I
`900 _
`00
`2700
`1800
`900 _
`00
`2700
`1800
`900 _
`00
`
`+
`
`+
`
`2
`
`_
`Fig. 2. Phase trajectories.
`
`6
`
`8
`
`9
`
`7
`
`m
`~
`~
`ro
`15
`13
`11
`time {bit periods}
`
`17
`
`On this diagram an error in the carrier phase shows as a vertical shift of the
`trajectory. Similarly, an error in the clock phase shows as a horizontal shift.
`Figure 3 shows errors in both carrier and clock phase;
`the clock is running
`early by x and the carrier oscillator phase is early by y.
`The error observed in the times ofthe zero crossings depends on the slope of
`the trajectory,
`i.e. on whether the instantaneous
`frequency is higher or lower
`than the carrier.
`
`x
`
`Fig. 3. Combined effect of clock and carrier phase errors.
`
`Phillps Journalof Research Vol. 39 Nos 112
`
`1984
`
`3
`
`IPR2018-1556
`HTC EX1023, Page 8
`
`
`
`R. J. Murray and R. W. Gibson
`
`From the diagram
`
`and
`
`whence
`
`and
`
`EL =X+ y,
`
`in principle, on a noise-free, unlimited-bandwidth, MSK signal
`Therefore,
`we would only need one high frequency transition and one low frequency
`transition to be able to set the phase of both the carrier and clock oscillators.
`What
`is more, this can be done in two simple steps as follows:
`Assume errors x and y as above.
`If the first transition is due to a low frequency, then we measure EL = x +y.
`Apply a phase correction of (x +y)/2
`to both clock and carrier oscillators,
`any subsequent low frequency transition sees zero error but the next high fre-
`quency transition sees:
`
`(X-Y)
`EH= -2-
`
`(Y-X)
`- -2- =x-y.
`
`applied to
`
`A correction of (x - y)/2 is applied to the clock and - (x - y)/2
`the carrier, giving zero residual error to both.
`In practice the signal is always bandlimited and the phase trajectory is not a
`set of straight
`lines and sharp angles. The simple two step correction given
`above is therefore not practical. Nevertheless, if we apply partial corrections
`the process will converge on the desired position. The optimum strategy will
`depend on what oompromise between error performance and acquisition time
`we wish to adopt and also on the relative stabilities of the carrier and clock
`oscillators.
`The corrections to the phase of the clock and carrier oscillators can be
`derived from the following rules:
`(a) CLOCK
`If a transition is early, advance the clock.
`If a transition is late, retard the clock.
`(b) CARRIER
`In this case we use the same rule as above or its inverse according to whether
`the transition was due to a high or a low frequency. The rule is thus:
`
`4
`
`Philips Journni of Research Vol.39 Nos 1/2 1984
`
`IPR2018-1556
`HTC EX1023, Page 9
`
`
`
`Coherent digital demodulator for MSK and RM schemes
`
`action
`frequency
`transition
`advance
`low
`late
`retard
`high
`late
`retard
`low
`early
`advance
`high
`early
`to a high or low fre-
`corresponds
`a transition
`We can determine whether
`quency by observing the sign of the other channel at the moment
`the transition
`occurred.
`Since in many cases the carrier phase will need more
`than the clock, we adopted
`the following
`strategy:
`For each transition,
`detected in either
`the I or Q outputs,
`the transition
`is early or late,
`then:
`(i) always adjust
`the carrier phase so as to reduce the error,
`to a frequency
`(ii) only adjust
`the clock phase if the transition
`corresponds
`different from that of the previous
`(different meaning above the
`transition
`carrier
`as compared
`to below the carrier or vice versa).
`This rule ensures that a relatively stable clock oscillator will not be unduly
`disturbed
`by carrier phase fluctuations.
`There is scope for a fuller
`investiga-
`tion as to how necessary
`or advantageous
`this approach
`is.
`
`frequent
`
`correction
`
`determine whether
`
`realisation
`3. Circuit
`rules described above may be imple-
`synchronisation
`The clock and carrier
`mented
`by means of simple digital circuits .'
`The synchronising
`circuits are fed with the hard limited I and Q signals and
`from these produce
`separate
`control
`signals
`for the carrier
`and clock oscilla-
`tors. The control
`signals are positive or negative pulses
`superimposed
`on the
`d.c.
`levels which control
`the frequéncies
`of the clock and carrier oscillators.
`The effect of the pulses is to momentarily
`speed up or slow down the oscillator
`and thus nudge the phase of the oscillator
`in the required direction. Two forms
`a constant nudge demodulator
`of demodulator
`were investigated;
`(the cor-
`of 1 bit period) and a proportional nudge
`rection pulses having fixed duration
`demodulator
`(the pulse duration
`being directly proportional
`to how early or
`late the transitions
`occur).
`
`results
`4. Experimental
`demodulator
`nudge
`A proportional
`series CMOS digital
`integrated
`circuits.
`For
`experimental
`purposes
`a TFM signal was used with a bit
`69.27 kbps
`and carrier
`frequency
`138.5 kHz. The TFM modulator
`
`circuit was constructed
`
`using 4000B
`
`of
`rate
`incor-
`
`IPR2018-1556
`HTC EX1023, Page 10
`
`
`
`R. J. Murray and R. W. Gibson
`
`a ROM and DIA converter
`encoded by
`3). The data was differentially
`porated
`taking the exclusive-OR of the current data bit and the previous data bit from
`the differentially
`encoded output.
`using the
`were each implemented
`The demodulator
`clock and carrier veo's
`to per-
`veo
`of a 4046B. The sensitivities
`of the oscillators
`could be adjusted
`mit phase corrections
`per bit period of more than 60 degrees
`for both clock
`and carrier. We can then define loop gain in terms of:
`
`_
`.
`oop gain -
`
`I
`
`to oscillator
`phase nudge applied
`d
`..
`(h
`)
`timing p ase
`error measure
`
`adjustable.
`The gains of the carrier and clock loops were individually
`The I and Q channel
`arm filters used were low pass 4th order Butterworth
`It should
`filters.
`be emphasised
`that
`these filters
`are non-optimum
`and so
`cause intersymbol
`interference
`and have an excessive noise bandwidth
`and
`cause degradation
`of the demodulator
`performance.
`Two filter bandwidths
`were used:
`(1) 25 kHz, which was found to give best results
`rate measurements,
`to give well defined cross-
`bandwidth
`(2) 30 kHz, which was the narrowest
`overs in the I and Q channel eye diagrams, which is best for fast acquisition.
`
`for the steady state bit error
`
`4.1. Steady state bit error rates
`of SIN ratio were carried out
`Measurements
`of bit error
`rate as a function
`for many different
`loop gains
`(fig. 4). The channel
`filters were set
`to 25 kHz
`(= 0.36 fb). Also shown for comparison
`is the equivalent
`curve using a perfect
`clock and carrier
`(taken directly
`from the modulator,
`suitably
`delayed).
`With perfect
`clock and carrier
`the measured
`bit error
`rate curve is 2.6 dB
`rate of 10-2 and 4.5 dB below MSK optimum
`below MSK optimum at an error
`at an error
`rate of 10-3• The MSK optimum curve includes
`the effect of dif-
`ferential
`encoding.
`Some of the degradation
`is due to the use of non-optimum
`arm filters.
`loop gains of
`for
`show that
`clock and carrier measurements
`The recovered
`bit error rate performance
`is
`0.11 for carrier and 0.028 for clock the measured
`only slightly degraded
`compared
`to the perfect clock and carrier measurement
`(approximately
`0.5 dB). As the loop gains are increased the measured
`bit error
`rate curve moves further
`away from that of the perfect clock and carrier. This
`is as expected since small corrections
`in the presence of noise cause only slight
`jitter
`in the recovered
`clock and carrier and so few errors, whereas
`large cor-
`rections will cause more jitter
`and hence more errors.
`
`6
`
`Phlllps Journalof Research Vol.39
`
`Nos 1/2 1984
`
`IPR2018-1556
`HTC EX1023, Page 11
`
`
`
`Coherent digital demodulator for MSK and RM schemes
`
`-
`
`-0 -
`
`-
`
`perfect clock and carrier
`carrier loop
`clock loop
`gain
`gain
`a = 0.11
`0.028
`b = 0.22
`0.056
`c = 0.33
`0.083
`d = 0.44
`0.11
`c = 0.55
`0.14
`
`filter
`
`bandwidth
`
`25 kHz
`
`MSK optimum with
`differential
`encoding
`
`10-5 l-L---,---'-_L_-'--"__'__':-,---L_'_-l:--"_'---'~:-,-_'___J
`o
`4
`8
`12
`1p
`signal to noise ratio in bit rate bandwidth (dB)
`-
`
`Fig. 4. Steady state BER curves. Filter bandwidth 25 kHz.
`
`With a filter bandwidth of 30 kHz the steady state bit error rate performance
`is slightly degraded (less than 0.5 dB).
`
`4.2. Acquisition measurements
`It is difficult to precisely define the acquisition time of a demodulator, par-
`ticularly in the presence of noise. We therefore measured the bit error rate of
`individual data bits in an acquisition pattern as a function of SIN ratio. The
`measurements were carried out for a number of combinations of clock and
`carrier loop gains. In all cases a bandwidth of 30 kHz was used for the I and Q
`channel arm filters. The acquisition pattern used was .... 000111000111000....
`after differential encoding. The appendix describes why this sequence was
`
`Phillps Journalof Research Vol. 39 Nos 1/2
`
`1984
`
`7
`
`IPR2018-1556
`HTC EX1023, Page 12
`
`
`
`R. J. Murray and R. w: Gibson
`
`chosen and also why the bit error rates for a zero are greater than for a
`one.
`fig. 5 shows the measured bit error
`As an example of the measurements
`rates of the 16th bit (a one) for a variety of clock and carrier loop gains. As the
`loop gains are increased the degradation is initially reduced but eventually
`becomes more severe. The optimum loop gain is approximately 0.33 for car-
`rier and 0.083 for clock, however the optimum is very broad and the case 0.22
`for carrier and 0.11 for clock is also good. For the 30th and 31st bits similar
`behaviour
`is observed,
`in this case all of the curves are closer to the MSK
`optimum than for the 15th and 16th bits. Again a rather broad range of
`optimum loop gains is observed,
`the optimum being approximately
`0.33
`fór the carrier and 0.083 for the clock. Figure 6 shows the measured error
`
`filter bandwidth 30 kHz
`bit no. 16 (11
`carrier
`loop
`gain
`a = 0.11
`b = 0.22
`c = 0.33
`d = 0.44
`e = 0.55
`f = 0.11
`9 = 0.22
`
`clock loop
`gain
`0.028
`0.056
`0.083
`0.11
`0.14
`0.11
`0.11
`
`M5K optimum with
`differential
`encoding
`
`10-5
`
`o
`_
`
`4
`signal
`
`I
`
`16
`~
`8
`to noise ratio in bit rate bandwidth (dB)
`
`Fig. 5. Error rate of the 16th bit in the acquisition sequence.
`
`8
`
`Phillps Journalof Research Vol.39 Nos 1/2
`
`1984
`
`IPR2018-1556
`HTC EX1023, Page 13
`
`
`
`Coherent digital demodulator for MSK and RM schemes
`
`filter bandwidth 30kHz
`carrier loop gain = 0.33
`clock loop gain: 0.083
`a = bit no.15 (Ol
`b = bit nO.16(11
`c = bit no.30(Ol
`d = bit no.31 (11
`e = zero in infinite
`f = one In in finite
`
`sequence
`sequence
`
`MSK optimum with
`differential
`encoding
`
`Fig. 6. Bit error rate for selected bits in acquisition sequence.
`
`rates for the 15th, 16th, 30th and 31st bits for a carrier loop gain ofO.33 and a
`clock loop gain of 0.083.
`
`5. Conclusion
`We have described a coherent digital demodulator which extracts the neces-
`sary synchronisation information from the hard limited baseband signals.
`Experiments
`show that good steady state performance and moderately fast
`acquisition times (approximately 16-30 bits) can be achieved with relatively
`simple versions of the coherent demodulator. The theoretical
`limits of the
`technique have not been investigated.
`Some modifications could be made to the demodulator. These include:
`
`Philips Journalof Research Vol.39
`
`Nos 1/2
`
`1984
`
`9
`
`IPR2018-1556
`HTC EX1023, Page 14
`
`
`
`R. J. Murray and R. W. Gibson
`(,9 the use of optimum filters in the I and Q channels,
`(b) the use of a crystal derived clock, which could be adjusted by direct switch-
`ing of the phase in a digital divider. This would allow virtually instantane-
`ous phase correction.
`(c) integration and filtering of the vea control signals. This would convert
`the controlloops
`from type 1 to type 2 or higher 4).
`
`APPENDIX
`
`Acquisition patterns
`The oscillator synchronisation information is derived from the crossovers
`of the hard limited baseband eye diagrams. All crossovers provide inforrna-
`tion which was used in adjusting the phase of the carrier oscillator but only
`some crossovers (denoted different) provide information to adjust
`the phase
`of the clock oscillator. For fast acquisition it is necessary to use an acquisi-
`tion pattern with many transitions and preferably with mainly different tran-
`sitions.
`(after differen-
`It should be noted that the sequence .... 10101010101010....
`tial encoding) is not always useful as an acquisition pattern with this type of
`demodulator. This is because the phase trajectory of the TFM or GMSK
`signal is substantially a constant 45 degrees and results in I and Q channel
`signals which have no transitions
`and so provide no synchronisation infor-
`mation.
`... 110011001100...
`Three fast-acquisition sequences have been considered,
`(sequence 1),
`... 0001100011000... (sequence 2) and ... 111000111000... (se-
`quence 3). These sequences represent
`the data after differential encoding. For
`all three sequences the data eye opening corresponding to a zero is less than
`that corresponding to a one. So, for all three sequences the demodulator will,
`in the presence of noise, produce more errors for data zeros than for data
`ones. Sequence 1 does not give reliable fast acquisition because it allows an
`unstable false lock with no transitions in either channel. Sequences 2 and 3 do
`not permit a long term false lock and so give better acquisition. Experiment
`shows that sequence 3 results in better acquisition than either of sequences 1
`or 2. For this reason sequence 3 was used for
`the detailed measurements
`described in sec. 4.
`
`REFERENCES
`IEEE Trans. Vol. COM-26 (5),534 (1978).
`and C. B. Dekker,
`1) F. de Jager
`and K. Hirade,
`IEEE Trans. Vol. COM-29 (7), 1044 (1981).
`2) K. Murota
`and L. E. Ze ger s, Philips J. Res. 37,165 (1982).
`3) K. S. Chung
`Phaselock Techniques,
`J. Wiley, New York 1979, Chapter 2.
`4) F. M. Gardiner,
`10
`
`'Philips Journalof Research Vol.39 Nos 1/2 1984
`
`IPR2018-1556
`HTC EX1023, Page 15
`
`
`
`Philips J. Res. 39, 11-23, 1984
`
`R 1078
`
`IN QUEST OF THE SPIKE: ENERGY DEPENDENCE
`OF THE SPUTTERING YIELD OF ZINC BOMBARDED
`WITH NEON AND XENON IONS
`
`by P. C. ZALM and L. J. BECKERS
`Philips Research Laboratories, 5600 JA Eindhoven,
`The Netherlands
`
`Abstract
`Sputtering yields of Zn were determined for bombardment with Ne and Xe
`ions at normal
`incidence in the energy range 0.2-20 keY under ultrahigh
`vacuum conditions. Massselection was employed and the energy spread was
`limited to a few electron volts. High f1uencesof the order of 3x 1017 cm-2
`were used to obtain stationary state yields. The yields were determined from
`accurate weighing of the samples prior to and after
`the irradiation stage.
`The measured yields obtained in low current density bombardments are in
`excellent agreement with predictions from Sigmund's linear cascade theory.
`There is no evidence of an anomalous,
`spike-like, cascade behaviour,
`in
`sharp contrast
`to what could be expected on the basis of experimental sys-
`tematics or theoretical considerations. Yields measured for 20 keY Ne+
`bombardment of Zn at high flux show a sharp deviation from the Sigmund
`estimate. Tests have revealed this to be induced by beam heating of the
`target. The huge yields observed, however, cannot be attributed to a simple
`evaporation process but must be interpreted as tentative evidence of a so-
`called thermal spike.
`PA CS numbers: 79.20.Nc, 61.80.Jh, 34.90. + q.
`
`outline and theoretical preliminaries
`1. Introduction,
`Many observed regularities in the sputtering behaviour of amorphous
`elemental targets bombarded with atomic ions can satisfactorily be accounted
`for by the linear cascade theory as formulated by Sigmund 1). In this model
`the penetrating projectile shares its energy with target atoms initially at rest in
`a series of binary collisions, a process in which fast recoils are created. These,
`in turn, set other target atoms in motion and a continuously increasing num-
`ber of progressively slower atoms results until transferable energies are less
`than the displacement
`threshold. At this stage the elastic collision cascade is
`damped by energy dissipation through e.g. phonon-assisted processes. The
`sputtering yield for an ion with incident energy Ei is given by 1)
`Y(Ei) = A . FD(Ei, X = 0),
`(1)
`where FD stands for the amount of energy deposited at the surface (x = 0) in
`the form of target atom motion and A is a material parameter containing the
`
`Philips Journalof Research Vol. 39 Nos 1/2 1984
`
`11
`
`IPR2018-1556
`HTC EX1023, Page 16
`
`
`
`P. C. Zalm and L. J. Beekers
`
`angle-, depth- and recoil-energy averaged, escape probability for target atoms.
`to the reduced nuclear stopping cross-section andA
`FD is linearly proportional
`to a surface escape barrier U« (usually taken as the
`is inversely proportional
`sublimation energy).
`The assumptions underlying the description leading to eq. (1) are known to
`break down in two cases:
`i)
`for low-energy (Ei $1 keY) light projectiles (e.g. H+, He"), where the
`energy transferred is insufficient to generate energetic recoils;
`ii) for high-energy (Ei;::: 50 keY) heavy ions, where the number of energetic
`recoils is so high that
`locally all atoms are instantaneously moving.
`the so-called spike regime, has attracted considerable
`The latter situation,
`attention in the literature ê'"). In particular
`the possible role of the target
`tem-
`perature (at T -= iTmelt)
`in invoking spikes 3) is still very much a subject of
`debate 4,5).
`Nonlinear effects in heavy-ion sputtering have been studied extensively by
`Andersen and Bay 6). They investigated in particular the projectile atomic num-
`ber dependence (Zp -= 10 - 80) of the sputtering yields of Si, Cu, Ag and Au
`targets bombarded with 45 keV ions. These authors also reported sputtering
`results with diatomic (homonuclear) molecular heavy ions at various energies.
`A very systematic study of "collisionally induced spikes",
`involving both
`projectile atomic number and bombarding energy dependences, has been car-
`ried out by Thompson 7). He collected sputtering yield data for Ag (Uo =
`2.95 eV), Au (3.80) and Pt (5.85) bombarded with a variety of heavy ions
`ranging from P to Bi and clusters thereof at energies from 10 keV to 260 keV
`per atom. Some interesting regularities were observed. First, a sharp break
`away from the predicted eq. (1) was observed above a certain energy deposi-
`tion density or critical yield. Above the critical yield the measured yields were
`found empirically to be more or less proportional
`to the third power of the
`energy deposition distribution function FD(E, 0). Secondly the critical yield, at
`which this anomalous behaviour was seen to occur,
`seemed to be linearly
`related to the sublimation energy of the target (Ycrit = 4.5 Uo leV]-atoms/ion).
`In particular
`the latter conclusion,
`if not sheer coincidence or a measure-
`ment artefact, would allow for à guided search for spike phenomena at lower
`projectile energies. One only needs to select a target material with a low subli-
`mation energy U«, as this both enhances the expected yield according to eq.
`(1), since A oe 1/ U«, and reduces the critical yield, since Ycrit oe U«. In prin-
`ciple frozen noble gases seem ideally suitable for such an experiment. Unfor-
`tunately there is a sharp distinction between the binding forces in metals and
`that
`in Van der Waals crystals, thus rendering any comparison dubious. A
`more feasible alternative for a target could be found in the group I (Na, K, ... )
`
`12
`
`Phlllps Journolof Research Vol.39 Nos 1/2 1984
`
`IPR2018-1556
`HTC EX1023, Page 17
`
`
`
`Energy dependence of zinc bombarded with neon and xenon ions
`
`and group IJ (Mg, Ca, ... ) elements, but then the extreme chemical reactivity
`virtually prohibits their use as' contamination problems will be severe and un-
`avoidable. Yet another problem is that low-sublimation-energy materials and
`ultra high vacuum (URY) conditions,
`required for good quality sputtering
`measurements, are hardly compatible. Moreover, one must be extremely care-
`ful to prevent
`thermal evaporation effects from obscuring the outcome of the
`experiment.
`In this paper we present a dedicated search for spike effects in the bombard-
`ment of Zn (Uo = 1.35 eY) with Ne+ and Xe+ ions at energies ranging from
`0.2-20 keY. We shall see that zinc, when treated with caution,
`is a suitable
`target material
`for providing answers to some open questions in the assess-
`ment of phenomena accompanying or ascribed to spikes. Below we rehearse
`the arguments leading to the conclusion that zinc is a likely spike-prone can-
`didate.
`For numerical evaluation, eq. (1), in the case of perpendicular
`can be recast
`into the form 1)
`
`incidence,
`
`(2)
`where CPI and EpI are characteristic constants depending on projectile and
`target parameters (viz. atomic number, mass and Uo in eV) given by the ap-
`proximate value 10)
`
`and the exact expression
`
`EpI = 3;.5 (1 + ~) z, ZI (Zpi + Zli)! keY.
`
`(3)
`
`(4)
`
`(5)
`
`The nuclear stopping cross-section Sn(e) is well approximated by 11)
`Sn(e) = !In(1 + e) 3 •
`( e )8
`e + 385
`The sputtering yields for Zn predicted a priori from eqs (2-5) range from
`3-10 and 3-29 atoms/ion for bombardment with Ne+ or Xe+ions, respectively,
`in the energy range 0.2-20 keY. From the systematics observed by Thomp-
`son 7) a nonlinear behaviour would be expected above a critical yield Ycrit:::::: 6.
`If this value indeed marked the onset of the above-mentioned third-power-
`of-FD-like behaviour of the yield one could expect a yield of about 500-600
`atoms/ion in the case of Zn bombarded with 20 keY Xe+ ions, a huge effect
`13
`
`Phllips Journalof Research Vol. 39 Nos 1/2
`
`1984
`
`IPR2018-1556
`HTC EX1023, Page 18
`
`
`
`P. C. Zalm and L. J. Beekers
`
`indeed. On the basis of theoretical considerations Sigmund and Claussen 2)
`predicted that pronounced spike contributions
`can be expected only if the
`yield calculated through eq. (1) exceeds 10 atoms/ion.
`Finally, Sigmund 9) stressed the importance of two quantities in determining
`whether or not spike effects will occur and showed that two conditions have to
`be fulfilled simultaneously. The effective maximum energy density ()o in the
`central core of the spike must be larger than the sublimation energy Uo and the
`time constant T governing the quenching (exponential decay) of the spike must
`exceed To, the slowing down time of the projectile. In fig. 1 a graphical repre-
`sentation of these requirements
`is given by depicting the energy dependence
`of ()o and T for 1-100 keY Xe+ bombardment of Zn (calculated with eqs (3),
`(4) and (9) of ref. 9, with m = 1 applying in this case; as approximately 12)
`is
`roughly constant; a more elaborate estimate of To would show an increase
`from 2 X 10-13 sec to 3 X 10-13 sec over this energy range). It is clear from fig. 1
`
`8,,(8) = Ivs. the range R(E;) ex; v'& and consequently To ex; R(E;)/V£";
`
`2
`
`5
`
`5 2
`
`5
`
`2
`
`5 2
`
`10.13
`
`100
`
`2
`
`5
`
`101
`
`2
`
`5
`
`5
`
`2
`
`5 2
`
`J
`
`QJ~
`o
`1-'.
`I-'
`
`Fig. 1. Spike parameters for Xe+ ions incident on Zn, estimated according to ref. 9, versus bom-
`for
`bardment energy. 60 is the effective maximum energy density in the spike, r the time constant
`decay, and '0 the slowing down time of the projectile.
`
`14
`
`Philips Journalor Research Vol.39 Nos 1/2
`
`1984
`
`IPR2018-1556
`HTC EX1023, Page 19
`
`
`
`Energy dependence of zinc bombarded with neon and xenon ions
`
`that spike effects are to be expected in the range of 10 keY Cr ~ ro) to 50 keY
`«(Jo ~ Uo). Similar calculations for Ne+ bombardment of Zn reveal that "col-
`lisionally induced" spikes will certainly not be able to develop.
`
`2. Experimental details
`2.1. Ion beam equipment characteristics
`Beams of single-charged; mass-selected, Ne+ and Xe+ ions at energies of
`0.2-20 keY, with an energy spread limited to a few electron volts, were gen-
`erated in our ion-beam apparatus discussed in detail elsewhere 13). For con-
`venience, some pertinent
`information will be given though.
`The ion beam apparatus, suitable for perpendicular incidence only, operates
`in either of two modes, depending on whether bombarding energies from
`about 10-25 keY or below 2.5 keY are requi