Note: Descriptions are shown in the official language in which they were submitted.
9 7 ~
P'IELD OF TH~3 1NV~;N~LIOI~
This invention relates to voice and data
transmission in wireless communications, and
particularly between fixed and portable transmitters
and receivers.
CL~IM TO COP~RIG~T
A portion of the disclosure of this patent
document contains material which is subject to
copyright protection. The copyright owner has no
objection to the facsimile reproduction by anyone of
the patent document, as it appears in the Patent and
Trademark Office patent file or records, but otherwise
reserves all copyright rights whatsoever.
2 0 BACKGROU~D AND SUMMARY OF THE~ v~!ih lL IoN
This patent document presents a new multiple
access technique for Personal Communication Networks
(PCN). Personal communication networks are networks
that allow individuals and equipment to exchange
information with each other anywhere at anytime
through voice, data or video. PCN typically include a
number of transceivers, each capable of transmitting
and receiving information (voice, data or video) in
the form of electromagnetic signals. The transceivers
may be fixed or portable, and may be identical or one
or more of them may be more complex.
The system must allow the transceivers to
access each other to enable the exchange of
information. When there are a number of transceivers,
multiple access, that is, access by more than one
transceiver to another transceiver, must be allowed.
One of the constraints of designing a PCN is
that a transceiver, or portable radio unit, must be
2 ~ 7 ~
small in size. The smaller the unit, the better for
portability. The small size of the units means only
small and light-weight power sources can be used. If
the portable is to be used for any length of time, it
must therefore consume minimal power.
Also, to allow use of the radio frequency
spectrum without obtaining a license in North America,
the system must use a spread spectrum and satisfy
federal regulations. In part, these regulations impose
limits on the power and the frequency spread of the
signals exchanged between the transceivers. An object
of an aspect of this invention is to satisfy those
requirements.
Also, transceivers talk to each other over
a fixed bandwidth. Because of the limited availability
of the RF spectrum, the system must be bandwidth
efficient yet at the same time maintain high quality
exchange of information at all times in one o~ the
most hostile channels known in communication. The new
multiple access technique proposed here addresses all
these issues.
The new access technique has a low Bit Error
Probability (BER) as well as a low probability of
dropped and blocked calls. This is due to the fact
that the access technique is robust against multipath,
Doppler shifts, impulse noise and narrowband
interference. It has a low cochannel interference and
little or no intersymbol interference.
The new access technique can offer up to 38
times the capacity of analog FM. It includes in one
aspect wideband orthogonal frequency division
multiplexing of the information to be exchanged, and
may include slow Frequency Hopping (FH). The technique
is implemented using Digital Signal Processors (DSP)
7 ~
replacing conventional analog devices. The system
operates with relatively small cells. In other
aspects, dynamic channel allocation and voice
activation may be used to improve the capacity of ~he
system.
Advantages of the present invention include:
1. It can be used indoors as well as outdoors using
the same transceivers. If data is to be exchanged, as
opposed to voice, the transceiver preferably contain~
an estimator to allow pre-distortion and post-
distortion of the transmitted signal.
2. The system, as compared with prior art systems
omits the clock or carrier recovery, automatic gain
control or passband limiter, power amplifier, an
equalizer or an interleaver-deinterleaver, and
therefore has low complexity.
3. The system offers good speech quality, as well as
low probabilities of dropped and blocked calls. It is
robust against Doppler and multipath shifts. It is
also robust against both impulse noise and narrowband
interference.
4. The system is flexible, such that at the expense
of increased complexity of the DSP receiver it can be
applied over noncontiguous bands. This is accomplished
by dividing a 100 MHz (in one of the exemplary
embodiments described here) band into several subbands
each accommodating an integer number of voice
channels.
5. The system offers low frame delay (less than 26.2
ms in the exemplary cellular embodiment described
here ? . The transceiver requires low average
transmitted power (of the order of 20~W in the
exemplary cellular embodiment described here) which
2 ~ 7 ~
means power saving as ~ell as enhanced biological
safety.
6. The system offers up to a 38 fold increase in
capacity over ~he North American Advanced Mobile Phone
System (AMPS) which uses analog frequency modulation.
Operation of the system in accordance with
the techniques described in this disclosure ma~ permit
compliance with technical requirements ~or spread
spectrum systems.
There is therefore disclosed in one aspect
of the invention a method for allowing a number of
wireless transceivers to exchange information (data,
voice or video) with each other. In the method, a
first frame of information is multiplexed over a
number of frequency bands at a first transceiver, and
the information transmitted to a second transceiver.
In a cellular implementation, the second transceiver
may be a base station with capacity to exchange
information with several other transceivers. The
information is received and processed at the second
transceiver. The frequency bands are selected to
occupy a wideband and are preferably contiguous, with
the information being differentially encoded using
phase shift keying.
A signal may then be sent from the second
transceiver to the first transceiver and de-processed
at the first transceiver. In addition, after a pre-
selected time interval, the first transceiver
transmits again. During the preselected time interval,
the second transceiver may exchange information with
another transceiver in a time duplex fashion.
The processing of the signal at the second
transceiv~r may include estimating the phase
20~ 7~
differential of the transmitted signal and pre-
distorting the transmitted signal.
The time intervals used by the transceivers
may be assigned so that a plurality o~ tlme intervals
are made available to the first transcei~er for each
time interval made available to the second transceiver
while the first transceiver is transmitting, and for
a plurality of time intervals to be made available to
the second transceiver for each time interval made
available to the first transceiver otherwise.
Frequencies may also be borrowed by one base station
from an adjacent base station. Thus if one base
station has available a first set of frequencies, and
another a second set of distinct frequencies, then a
portion of the frequencies in the first set may be
temporarily re-assigned to the second base station.
In an implementation of the invention for a
local area network, each transceiver may be made
identical except for its address.
Apparatus for carrying out the method of the
invention is also described here. The basic apparatus
is a transceiver which will include an encoder ~or
encoding information, a wideband frequency division
multiplexer for multiplexing the information onto
wideband frequency voice channels, and a local
oscillator for upconverting the multiplexed
information. The apparatus ma~ include a processor for
applying a Fourier transform to the multiplexed
information to bring the information into the time
domain for transmission.
9 ~ ~
BRIEF DESC~IPTION OF T~E DRAWIN~S
There will now be described a preferred
embodiment of the invention, with reference to the
drawings, by way of illustration, in which like
numerals denote like elements and in which:
Figures la and lb are schematics of a prior
art receiver and transmitter respectivelyi
Figure 2 is a schemàtic showing the use of
the available frequencies according to one aspect of
the invention for use with cellular applications;
Figure 3a is a schematic showing an
idealized pulse for transmission over a cellular
system;
Figure 3b is a schematic showing a modified
version of the pulse shown in Figure 3a;
Figure 3c is a schematic showing a further
modified version of the pulse shown in Figure 3a;
Figure 4 is a schematic showing an exemplary
protocol for cellular communication;
Figure 5a is a block diagram showing the
structure and function of an embodiment of the
transmitter of a cellular portable in accordance with
the invention;
Figure 5b is a block diagram showing the
structure and function of an embodiment of the
transmitter and receiver of a cellular base station in
accordance with the invention;
Figure 5c is a block diagram showing the
structure and function of an embodiment of the
receiver of a cellular portable in accordance with the
invention;
Figure 6a is a flow diagram showing the
function of the processor in either of Figures 5a or
5b;
'
.
' .
2 ~
Figure 6b is a schematic showing the
function of the deprocessor in either of Figures 5b or
5c;
Figure 6c is a schematic further
illustrating the operation of the processor and
deprocessor shown in Figures 6a and 6b;
Figure 7a is a schematic showing the
structure and function of the channel est.imator in
Figure 5b;
Figure 7b is a flow chart showing the
operation of the channel estimator of Figures 5b and
7a;
Figures 8a, 8b and 8c are respectively
schematics of 126, 63 and 7 cell reuse patterns;
Figures 9a and 9b are schematics showing
transmit protocols according to one aspect of the
invention;
Figure 10 is a schematic showing the use of
the available frequencies according to another aspect
of the invention for use with local area network
applications;
Figure lla is a schematic showing an
idealized pulse for transmission over a local network
system;
Figure llb is a schematic showing a modi~ied
version of the pulse shown in Figure lla;
Figure llc is a schematic showing a further
modified version of the pulse shown in Figure lla;
Figure 12 is a schematic showing a preferred
protocol for local area network communication;
Figure 13a is a block diagram showing the
structure and function of an embodiment of the
transmitter of a local area network transceiver
according to the invention;
2 ~ 7 ,7
Figure 13b is a block diagram showing the
structure and function of an embodiment of a further
local area network transceiver according to the
invention;
Figure 13c is a block diagram showing the
structure and function of an embodiment of the
receiver of a local area network transceiver according
to the invention;
Figure 14a is a flow diagram showing the
function of the processor in either of Figures 13a or
13b;
Figure 14b is a schematic showing the
function of the deprocessor in ei~her of Figures 13b
or 13c;
Figure 14c is a schematic further
illustrating the operation of the processor and
deprocessor shown in Figures 14a and 14b; and
Figure 15 is a schematic showing the
structure and function of the channel estimator in
Figure 13b.
DLTAILED DES~RIPTION OF PR~FERR~ EMB~DIMENTS
Introduction
The benefits of the invention can be readily
appreciated with reference to Figure 1, which shows a
prior art transmitter/receiver configuration for a
portable unit. The transmitter includes a vocoder 110,
an interleaver 112, a modulator 114, a filter 116,
local oscillator 118, power amplifier (PA) 120 and
antenna 122. The receiver includes an LNA 124, a local
oscillator 126, a filter 128, automatic gain control
(AGC) 130 with an associated passband hardlimiter not
separately shown, carrier recovery 132, sampler 134,
clock recovery 136, adaptive (or fixed) equalizer 138,
demodulator 140, deinterleaver 142 and decoder 146.
206~7S
g ~
With implementation of the present invention, several
of the blocks shown in Figure 1 are not required.
These are the interleaver 112, dein~erleaver 142,
power amplifier 120, automatic gain control 130 with
passband hard-limiter, clock recovery 136 and carrier
2 ~ 7 :.~
recovery 132, and the equalizer 138. It will now be
explained how the proposed system obtains the omission
of these blocks without impairing khe quality and
capacity of the sy6tem.
In this disclosure there will be described
~wo systems as examples of the implementation of the
invention. The system described first here will apply
to a cellular system with a number of portable
transceivers and base stations (BS). Then will be
described a local area network implementation. A local
area network will typically be a system of equal
transceivers. The invention may also be implemented
with combinations of cellular and local area n~tworks,
or to a system with a number of equal transceivers and
a master or controlling transceiver. "Equal" as used
here means that the transceivers have more or less the
same processing equipment and processing capabilities.
The system described here is primarily for the
exchange of voice information.
Link set-up and termination protocols
between transceivers, and the equipment required to
implement them, are well understood in the art as well
as the basic structure of radio transceivers that may
be used to implement the invention. Hence these
elements are not described here. What is described
here are the novel operational, functional and
structural elements that constitute the invention.
Cellular Implem~ntation of ~i~eb~n~ Modulation
The present invention proposes in one
embodiment a wideband modulation scheme for exchange
of information between transceivers such as portables
and base stations.
2 ~ 7 ~
Wideband in this patent document i6
described in the context of Wideband-Orthogonal
Frequency Domain ~odulation (W-OFDM or wideband OFDM).
In OFDM, the entire available bandwidth B is divided
into a number of points K, where adjacent points are
separated by a frequency band Qf, that is B = K~f. The
K points are grouped into a fram~ of K1 points and two
tail slots of K~ points each, so that K = K1 + 2K2. The
frame carries thQ information intended for
transmission under the form of multilevel differential
phase shift keying (MDPSK) symbols or differential
quadrature amplitude modulated (DQAM) symbols. Thus
each point in the frame corresponds to one information
symbol. The two tail slots act as guard bands to
ensure that the out-of-band signal is below a certain
power level. For example, when a pulse P(f) is
selected for pulse shaping and the out-of-band signal
has to be ydB or less relative to the in-band signal,
K2 is selected such that
20-1og1olP(f)/P(0)¦ < y for f > K2~f.
When the pulse is a raised-cosine pulse with a roll-
off ~ and when the number of levels each 6ymbol can
take is M, the bit rate is equal to K1log2M/(~t + (1
+ ~?/~f) where (1 + ~)/hf is the duration of the ~rame
and ~t is the guard time required to take into account
the delay of arrival and the delay spread due to
multipath. In this case, the bandwidth efficiency,
which is defined as the ratio between the bit rate and
the bandwidth, is equal to:
log2M/((1 + ~ + ~t~f)(l + 2K2/K1))
In wideband-OFDM, both K and ~f are selected
sufficiently large to achieve a high throughput as
well as to reduce the effects on the BER of the clock
error, the Doppler shift and the frequency offset
%~97~
1~
between the LO in the transmitter and the one in the
receiver. To show what is meant by ~K and ~f are
selected sufficiently large", consider the effect of
increasing K and ~f on ~1) the clock error, (~) the
Doppler shift and (3) the frequency offset between the
LO in the transmitter and the L~ in the receiver.
(1) When a clock error at a transceiver of
value ~ occurs in the time domain, it causes a shift
in the phase difference between adjacent symbols in
the frequency domain of value 2n~f~. When ~ is equal
to xT where T is duration of one time domain sample
and X is any real value, the shift is equal to 2n~fxT.
Hence, ~ causes a shift in the phase difference
between adjacent symbols of value 2nx/Kl since T is
equal to 1/(KlQf). By doubling the number of symbols
from K1 to 2Kl the shift in the phase difference is
reduced by half from ~%/K1 to ~X/Kl- Thus, the effect
of the clock error on the BER is reduced by increasing
~C .
~2) When there is relative motion between
the transmitting transceiver and the receiving
transceiver, a Doppler shift occurs with a maximum
absolute value ¦V/A¦ where V is the relative velocity
between the two transceivers and A is the wavelength
of the travelling wave corresponding to the carrier
frequency fc (i.e. fc is the frequency corresponding
to the middle point in the frame). Such a Doppler
shift causes a sampling error in the frequency domain
of the same amount, or equivalently, it causes a
sampling error of V/(A~f) relative to one symbol
sample. Thus, the effect of the Doppler shift on the
BER is reduced by increasing df.
(3~ When a frequency offset between the LO
in the transmitter and the one in the receiver occurs
20~ 7~
with a value fO, it causes a sampling error in the
frequency domain of the same amount, or equivalently,
it causes a sampling error of fO/~f relative to on~
symbol sample. Thus, the effect on the BER of the
frequency offset between the L0 in transmitter and khe
one in the receiver is reduced by increasing ~f.
In summary, OFDM with a K and a ~f large
enough to be able to achieve a specific throughput and
large enough to be able to avoid using either a clock
or a carrier recovery device without substantially
affecting the BER is referred to here as Wideband-
OFDM. As an example, let us assume that MDPSK is used
in an OFDM system with the number M of levels, with a
carrier frequency fc~ with a raised cosine pulse of
roll-off ~, with the LO at the receiver having a
frequency offset fO relative to the LO at the
transmitter (so that the frequency offset between the
carrier frequencies in the first and second
transceivers of the multiplexed information is fO)~
with a given maximum expect~d clock error ~ = xT at
the receiving transceiver, where T is the duration of
one time domain sample, and with a maximum expected
relative velocity V between the transceivers. Thus,
in order to ensure that the out-of-band signal is ydB
or less relative to the in-band signal and to be able
to avoid using either a clock or a carrier recovery
device without substantially affecting the BER we have
to:
1. Find the acceptable sampling error ~f', relative
to one symbol sample, which does not
substantially affect the BER. This can be done
using the following rules:
When 0.2 < ~ < Q.3, ~f' = 7.50%
When 0.3 s ~ s 0.4, ~f' = 10.0%
.
2 ~ 7 ~
14
When 0.4 < ~ ~ 0.5, ~f' = 12.5%
When 0.5 ~ ~ ~ 0.6, ~fl - 15.0%
2. Find ~f such that:
V/(~Q~) + ~O~f s af~
3~ Find K2 such that
20 l0g10lP(f)/P(0)¦ < y for f > K2wf
4. Find K1 such that
2~X/Kl < rt/M
In this case, we refer to OFDM as Wideband-OFDM.
Element 4 is a necessary condition for wideband OFDM,
and given a sampling error, the sampling error may be
corrected with the methods described in this patent
document.
To implement wideband modulation, Orthogonal
Frequency Division Multiplexing (OFDM) is preferred in
which the information, for example encoded speech, is
multiplexed over a number of contiguous frequency
bands. Wideband OFDM forces tha channel to be
frequency selective and causes two types of linear
distortion: amplitude distortion and phase distortion.
To reduce the effect of amplitude distortion the
modulation is preferably phase modulation, while the
effect of phase distortion is reduced by employing
differential phase modulation. Hence the modulation
~5 may be referr0d to as Differential OFDM (DOFDM).
Unlike in other proposed schemes, neither pilot tones
nor diversity are required in DOFDM. Possibly/
quadrature amplitude modulation might be used, but
amplitude modulation makes it difficult to equalize
the distorting effects of the channel on the signal.
To implement wideband modulation in a
cellular system with a plurality of portables and one
or more base stations, a 100MHz band is divided in~o
409~ points, as shown in Figure 2, plus two tail slots
2 ~
of 195.3 KHz each. The 4096 points represent N voice
channels (vc). Adjacent points are separated by 24.414
KHz and each point represents a Differential eight
Phase Shift Keying (D8PSK~ Symbol ej~(n), where ~(n) =
~(n-l) + ~(n) + x(n). ~(n) takes one of the eight
values {0, 2~/8, 4n/8, ..., 14n/8} with equal
probability for n = 1, 2, ..., 4096 and ~(0) takes an
arbitrary value. x(n) also takes an arbitrary value.
x(n) may be used as a security key and will be known
only to the transmitter and receiver. Information in
the form of output bits from a vocoder are mapped onto
~(n). Vocoders are well known in the art and do not
need to be described in detail here. The focus here is
to transmit the bits with an acceptable Bit Error
Rate, i.e. with a BER <10-2 for voice and ~10-8 for
data.
Ideally, 40.96 ~s (=1/24.414 KHz) is the
minimum duration required for one frame to be
transmitted without frequency domain intersymbol
interference. This can be achieved using a Rais0d
Cosine (RC) pulse with zero roll-off, as shown in
Figure 3a. Figure 3a illustrates a rectangular (time
domain) window corresponding to the RC (frequency
domain) pulse. Such a pulse, however, requires an
infinite frequency band. To alleviate such a
requirement, an RC pulse with a 20% roll-off (i~e. ~
= 0.2) may be used as shown in Figure 3b. The frame
duration has increased by 20% to 49.152 ~s. The two
tail slots of 195.3 KHz each (i.e. 3 points each)
ensure that the signal outside the entire band of
100.39 MHz is below -50 dB. To allow the frame to
spread over the time as a consequence of the multipath
nature of the channel, an excess frame duration of
2 0 6 ~ 9 7 ~
16
2.848 ~s is provided as shown in Figure 3c, making the
frame duration 52 ys in total.
Since the frame duration is 52 ~s, the frame
rate is 252 frames per 13.104 ms or equivalently, 126
full duplex frames may be transmitted/received every
13.104 ms. The reason for pre-selecting an interval of
13.104 ms is to ensure a transmission delay to allow
one transceiver to communicate with other transceivers
at the same time, but must not be so long that the
delay becomes unacceptable to the user. Delays longer
than about 40 ms are too great for voice, and lt is
preferable to be lower. For data, the delay may be
longer and still be acceptable.
In the exemplary embodiment described here,
three bit rates are considered for the vocoder: 18.77
K~ps, 9.16 Kbps and 6.18 Kbps. Table I displays the
structure of a vc slot and the number N of vc for each
vocoder rate. The control symbols in each vc slot are
required for handoff and power control. Figure 2 shows
that N vc can be transmitted simultaneously. This is
known as Frequency Division Multiple Access. Figure 3c
shows that 126 full duplex frames can be transmitted
every 13.104 ms in a Time Division Multiple Access
fashion (TDMA). The total number of Full Duplex voice
channels (FDvc) is therefore 126xN and is shown in
Table I.
To ensure that the channel is slowly fading,
a Time Division Duplex protocol for exchange of
information between the portable and the base station
is proposed as illustrated in Figure 4. The protocol
is as follows:
1. The portable transmits a frame 410 over one vc
slot. See the discussion in relation to Figure 5a
below.
2 ~
2. The Base Station (BS) receives the frame 410 from
the portable and processes (analyzes) it as shown and
discussed in relation to Figure 5b belo~.
3. Based on the received signal, ~he BS predistorts
a frame 420 and transmits it to the portable over the
same vc slot, 520 ~s or some other suitable time
inter~al later in which the channel does not chan~e
substantially. The time interval will depend on
factors such as the frequency, speed of the
transceiver and other environmental factors.
4. The portable receives the frame from the BS. See
the discussion in relation to Figure 5c below.
5. Steps 1 through 4 are repeated, as for example by
the transmission of the next frame 430, eve~y 13.104
ms until the call is terminated.
During 520 ~s, a portable travelling outdoor
at 100 km/hr moves 1~44 cm, which leaves the outdoor
channel largely unchanged. Indoors, a portable moving
at 2 m/s moves 0.1 cm again leaving the channel
unchanged. Assuming that the channel is reciprocal and
stationary over 520 ~s, a predistorted signal,
transmitted by the BS, should reach the portable
undistorted.
From Figure 4, one can see that the portable
transmits/receives one FDvc every 13.104 ms, while the
BS can transmit/receive up to 21 frames or
equivalently up to 21xN FDvc every 13.104 ms. The
frames 440 labelled frame 2...frame 21 are frame~ that
may be transmitted to other portables. This implies
that while one BS is processing its data over 520 ~s,
six other BS can communicate to their corresponding
portables in a Time Division Multiple Access (TDMA)
fashion using the same frequency bands. Also, during
the 13.104 ms, or such other preselected time interval
206~9 1~
1~
that is suitable, the BS may communicate with one or
more other portables.
When a portable i8 statlonary during a call,
it is possible with high probability to have the
transmitted signal centered with several deep
(frequency domain) nulls, hence, causing speech
degradation. Also, narrowband interference over the vc
slot can deteriorate the speech. In order to avoid
both situations, the signal is preferably frequency
hopped into a new vc slot within the same (frequency
domain) frame. This frequency hopping is ordered by
the BS which is constantly monitoring the channel
frequency response. Monitoring techniques, as well as
frequency hopping, are known in the art, and not
described here further. When an unacceptable speech
degradation is first noticed by the BS a probation
period is initiated and maintained for at least 10
cycles (i.e. lOx13.104 ms) unless speech degradation
has ceased. In other words, the probation period is
terminated if speech degradation has ceased. Frequency
hopping is then ordered at the end of the probation
period. The period of 10 cycles is long enough to
indicate the portable stationarity and is short enough
to allow speech interpolation between unacceptable
speech frames, hence maintaining good speech quality.
As known in the art, the BS ensures that no collisions
take place between hopping portables.
Digital Signal Proce~sing
The transmitter/receiver block diagrams
corresponding to the protocol in Figure 4 are shown in
Figures 5a, Sb and 5c. Figure 5a coxresponds to step
1 in the protocol described above. Speech is provided
to a vocoder 510 where the speech i8 digitized and
2 ~ 7 ~
19
coded to create bits of information. The bits are
provided to the modulator 512 which turns them into
D8PSK symbols, with three bits per symbol. The D8PSK
symbols are then processed in the processor 514 which
is described in more detail in Figure 6aO The output
from the processor is then filtered in low pass filter
516, upconverted to RF ~re~uencies using local
oscillator 518 and transmitted by antenna 520. Figure
5b corresponds to steps 2 and 3.
In Figure 5b, the received signal at the
base station is filtered in a bandpass filter 522, and
down converted by mixing with the output of a local
oscillator 524. The average power of the downcoverted
signal is monitored by a power controller 525 that
ad~usts the average power to the specifications
required by the sampler 526. The adjusted
downconverted bits are then sampled in sampler 526 to
produce bits of information. The bits are then
processed in the deprocessor 528, descrihed in more
detail in Figure 6b. An estimate of the phase
differential of the received signal is taken in the
channel estimator 530, as described in more detail in
relation to Figure 7a and 7b below, and the estimated
phase differential is supplied to a decoder-
~emodulator 532 to correct the received bits. Theestimated phase differential is also supplied to a
pre~distorter 534 in the transmitter. ~t the
transmitter in the Base Station, the same blocks are
incorporated as in the portable transmitter except
that a pre-distorter is used to alter the phase of the
D8PSK symbols to make the channel appear Gaussian
(ideal) as opposed to a fading channel. The pre~
distorter 534 receives a signal corresponding to the
estimated phase differential of the channel. On the
7 ~
(believed reasonable) assumption that ~he channel is
reciprocal, the signal being transmitted is pre-
distorted with the estimated phase differential 60
that the received signal at th0 portable with which
the BS is communicating will be corrected for any
phase distortion over the channel. The advantage of
rendering the channel Gaussian is a large saving in
the power required to achieve an acceptable BER. The
initial power control 525 also sends a signal to th0
pre-dlstorter 534 to adjust the transmitted power to
an appropriate signal level for the sampler 526 in the
portablels receiver depending on the average power of
the received signal. Thus if the average power is too
low, the transmitted power is increased and i~ the
average power is too high, the transmitted power is
decreased. The power controller 25 may also be used in
frequency hopping to monitor the average power o~ the
received frequency and determine when frequency
hopping need take place.
Figure 5c corresponds to step 4, and shows
the receiver of the portable, which is the same as the
receiver in the BS except it does not include an
estimator or a power controller. These are not
required in the portable on the assumption that the BS
will carry out the phase estimation and the power
control. However, if desired, the portable may include
these functions.
Figures 6a, 6b and 6c illustrate the
function and structure of the processor and the
deprocessor respectively in the transmitter and
receiver. Software for modelling the function of the
processor in a general purpose computer is attached
hereto as Appendix A, and for modelling the function
of the deprocessor as Appendix B.
206497~
21
Figure 6a shows that the processor is a DSP
~mplementation of an RC pulse shaping filter with a
20% roll-off, followed by an inverse Fourier
~ransform. The proce6sor first inverse Fourier
tran6forms the 4096 D8PSK modulated 6~mbols output
from the modulator. The transformed symbols are then
triplicated as a group eo that the total number of
samples is tripled, with three consecutive groups each
consisting of the 4096 transformed symbols. The
triplication of the signal is illustrated in Figure
6c, where the symbols are shown as first delayed and
addbd together. Next, as shown in Figures 6a and 6c,
the three groups are windowed by a Raised Cosine
window with a roll-off of 0.2 centered in the middle
of the three groups. In other words, the processor
takes D8PSK symbols in, pulse shapes them and inverse
Fourier transforms them. On the other hand, the
deprocessor undoes what th~ processor did, i.e. it
remove6 the pul~e shaping, then Fourier transforms the
received signal to obtain the original D8PSK symbols.
The first two block6 in Figure 6b are similar to the
second two blocke in Figure 6a except for two
differences. The two differences are as follows. In
the first block of the deprocessor, the repeated
groups of symbols are partially overlapped as shown in
Figure 6c on the right hand side. In the second block,
a rectangular window is used instead of the Raised
Cosine. In the preferred implementation, the blocks
are repeated three times but other numbers of
repetition may be used.
Figures 6a, 6b and 6c show that the DSP
blocks used in the processor are identical to the ones
used in the deproce6sor, except for a 6mall change in
the two transforms and a small change in the ehapes of
2~6~7~
22
the two windows. Thus the same hardware can be used by
both the processor and the deprocPssor.
Figure 7a shows a block diagram of an
example of a preferred channel estimator, and Figure
7b is a flow chart showing the operation of the phase
estimator. Each of the steps is carried out in a
computing means that may be a special purpose computer
or a general purpose computer programmed to carry out
the digital signal processing described here, as for
example with the software that is attached hereto as
appendix C. Other methods of estimating the channel
may be used that obtain an estimate of the channel
group delay or phase differential of the transmitted
symbols. However, a preferred implementation is
described here.
The first block in Figure 7a estimates the
envelope A(n) for n = 1, ..., 4096 of the (frequency
domain) samples transmitted over the fading channel as
output from the deprocessor. The estimate A~(n) is
the square-root of the sum of the squares of the
quadrature (Q) and inphase (I~ samples output from the
deprocessor. The second block performs the operation:
~ln(A'(t)) = (A'(n) - A~(n-1))/A'(n), for n = 2, ....
4096, where A~(n) is the estimate of A(n). The third
block performs a Hilbert transform operation
H[~ln(A~(t))] on the result of the second block.
H~ln(A'(t))] is an estimate of l~(n)l for n = 2,
..., 4096, where ~(n) is the phase differential of
the transmitted signal (~ is the phase of the signal).
The Hilbert transform is preferably carried out by
taking the discrete fast Fourier transform of the data
record, multiplying the positive frequency spectrum of
the transform by -i (square root -1), and the negative
frequency spectrum of the transform by i, and taking
2 ~ 7 ~
the inverse discrete fast Fourier transform. The
result is a set of symbols representing an estimate of
the phase differential of the received signal, as
determined from its sampled amplitude envelope.
Instead of a Hilbert transform, a different
estimation may be made to estimate the phase
differential. In this case, firstly, after the
electromagnetic signal has been sampled, a ser.ies of
data frames of a number of consecutive amplitude
samples (A(t)) of the electromagne~ic signal are
constructed. These data frames are then segmented into
segments [t1lt2]l where the amplitude of the
electromagnetic signal is at least a predetermined
number of dB less th~n its running mean, for example,
lOdB. The following calculation is then applied to
these segments of the amplitude samples:
~(t) ~1/to -1
1 ~ (t'/~o)
where t t - tmin, tmin is the time in [t1l t2] when
A(t) reaches its minimum, t is the time from the
beginning of the segment, and to is the time from the
instant the amplitude of the electromagnetic signal
reaches its minimum during the segment until the
amplitude reaches double its minimum during the
segment. In other words, the phase differential may be
calculated from
~(t) ~ -to/(to~ + t'2).
The polarity of ~(n) is extracted using the
last block shown in Figure 7a. The estimate so
calculated does not provide the sign of the
differential. This may be determined by known
techniques, for example by adding the phase
differential to and subtracting the phase differential
2 ~ 7 ~
24
from the received phase (tan~1 tQ/I~) and taking the
sign to be positive lf the addition results in the
smaller Euclidean distance to the expected value and
negative if the subtraction results in the smaller
Euclidean distance to the expected value.
Equivalently, for each sample n, the ideal
phase closest to ~(n)+~(n) is determined and labelled
~+(n), and the ideal phase closest to ~(n)-~(n) is
determined and labelled ~_(n). The two sums P =
~¦~+(n) - {~(n)+~(n)}¦ and N = ~ - {~(n)-~(n)}¦
are calculated. If P < N, then ~(n)+~(n) is used to
correct the signal, and if not then ~(n)-~(n) is used
to correct the signal.
For simplicity of the estimator, the
determination of the sign need only be carried out for
phase differentials greater than a predetermined
threshold. This will be in the vicinity of a fade and
may be accomplished by segmenting the data record into
a segment in which the phase differential is larger
than a selected threshold and setting the remainder of
the data record to zero. This computation may be
carried out with a simple discrimination circuit or
equivalent computing means in the estimator.
The bias ~ of the channel group delay is
estimated by averaging ~(n) over n for n = 1~ ....
4096 where ~'(n) is the measured value of ~(n). The
estimates ~(n) and ~(n) are used directly in the
predistortion filter in Figure 5b, while the estimates
~(~(n) and ~o of the unbiased channel group delay and
of the bias of the channel group delay respectively
are used in the demodulator.
The complexity of the processor-deprocessor-
channel estimator is displayed in Table II. Complexity
is measured in Mega Instructions Per Second (MIPS)
2 ~
where one instruction is defined as one complex
addition, one complex multiplication and a storage of
one complex number. It does not include overhead.
The complexity of the processor-deproces~or-
channel estimator in the BS i~ computed from thecomplexity of ths Inver6e Fast Fourier Trans~orm
(IFFT)/Fast Fourier Transform (FFT)/Hilbert Transform.
The complexity is 4096 x 12 x 4 x 21/13.194ms for the
BS. For the portable, it is computed from the
complexity of the FFT/IFFT per vc: (32 x 5 + 64 ~ 128
~ 256 + 512 ~ 1024 + 2048 + 4096)2/13.104ms for the
portable with a 6.18 Kbps vocoder. Such a complexity
assumes that the A/D converter operates at 100 MHz
wlth 12 bit precision. As seen in Table II, the
portable has smaller complexity due to the fact that
the portable transmits/receives one vc in 13.104 ms
and the BS transmits/receives up to 21xN vc in 13.104
ms.
Raducing An~log Complexity
Comparing Figure 1 (prior art) and Figure 5,
it will be seen that ~everal conventional blocks are
not used in the present invention, namely the
interleaver-deinterleaver, the Power Amplifier (PA),
both the clock and the carrier recovery, both the AGC
with its associated Passband hard limiter, as well as
the equalizer.
From the BS point of view, the interleaver-
deinterleaver is not required since the signal is
predistorted before transmission forcing the received
samples to be independent. From the portable point of
view, the interleaver-deinterleaver i6 not required as
a separate entity from the vocoder due to the fact
that the channel is highly frequency selective, hence
20~97~
26
the interleaving/deinterleaving can be applied
implicitly in the vocoder over one vc, without a need
for a separate time domain interleaver/deinterleaver.
This eliminates excess speech delays as60ciated with
interleaving/deinterleaving between frames.
The PA is not required since the cells can
have, as shown later, a radius of up to at least 250
m outdoors and 30 m indoors, if the transmitted power
i8 Up to 6 dBm. Such a power can be generated by the
Local Oscillator (LO) without a need for a PA. It i~
important to avoid using a PA since DOFDM generates a
time domain signal with non constant envelope. A power
efficient class C PA cannot be used without distorting
the signal. A class A PA can be used at the expense of
power efficiency.
A clock recovery device is not required
since a sampling error in the time ~ -; n is
equivalent to a phase shift in the frequency domain.
The phase shift is a linear function of frequency. It
contributes to the bias in the channel group delay.
Such a bias can be easily estimated and removed as
mentioned previously by averaging ~(n) over n in the
frequency domain. Such an estimate is accurate as long
as the sampling error is less than 0.2 ys or
equivalently less than 20 samples (since in this case,
the corresponding phase shift is less than n), and as
long as the number of point~ in one vc is large enough
as it is here.
A carrier recovery device is not required
since a carrier offset in the time domain is
equivalent to a sampling error in the frequency
domain. For the chosen RC pulse, a sampling error of
up to 10% of the duration of one pulse i6 acceptable.
This implies that a frequency offset of up to 2.414
2~ 7~
KHz is acceptable regardless whether it is due to
carrier offset as low as 1 part in a million, i.e. ~s
low as 1 KHz per 1 GHz. When a carrier ~requency
higher than 2.414 GHz is required, one can decrease in
Figure 2 the number of points per 100 MHz or one can
use an RC pulse with a rolloff larger than 20~.
Neither an AGC nor a Passband hard-limiter
are required since khe level of the received power may
be controlled constantly. This i5 achieved as follows:
The portable transmits a frame. The BS receives the
frame and predistorts a frame intended for
transmission accordingly, assuming that the channel is
reciprocal and stationary over 520 ~s. This includes
controlling the transmitted power according to the
received power. The BS transmits the predistorted
frame and simultaneously orders the portable to
control its power. The order is conveyed using the
control symbol in the vc slot ~See table I). The
degree of power control may be determined using the
power controller 525, and the instruction for the
inclusion of a power control symbol in the vc may be
sent from the power controller 525 to the pre-
distorter 534.
One advantage of wideband modulation over
narrowband modulation is that the wideband signal does
not experience short term fading the same way the
narrowband one does. The wideband signal is mainly
affected by shadowing and other long term effects
which vary slowly and are easily monitored from one
frame to the other as long as the same vc slot is used
by the portable to transmit and receive (i.e. as long
as TDD is employed).
Finally, conventional equalization, whether
it is linear or nonlinear, is not required simply
2 l1 ~ 7 e~
28
because there is little or no ISI. Also, from the
port~ble point of view, each received vc is
predistorted by the BS. Hence, the channel can be
modeled approximately as an ideal memoryless Additive
White Noise Gaussian (AWGN) channel, assuming channel
reciprocity and stationarity over 520 ~s. From *he BS
point of view, since the received si~nal is nst
predistorted by the portable prior transmission, the
channel estimator is used to reduce the e~ect of the
channel group delay.
Smaller cells
As mentioned previously, the LO generates a
6 dBm average power, hence ~he signal power
transmitted by the BS over one vc slot is (6 dBm -
lOlog10N dB) while the signal power transmitted by the
portable over one vc slot is 0 dBm. Also, since the
noise power over a 100 MHz band is -94 dBm, it is (-94
dBm - lOlog10N dB) over one vc. A typical noise figure
at the receiver is 7 dB. The penalty for not using a
matched filter in the receiver is 1 dB. ~ombining
together the above figures provides the portable with
an (86 dB - path loss in dB) received signal to noise
ratio (SNR), while it provides the BS with an (86 dB
+ lOlog10N dB - path loss in dB) received SNR.
For a path loss of 75 dB, the radius of the
urban cell can be 250 m while it can be 30 m for the
indoor cell. Such a path loss provides the portable
with a 17 dB received SNR, while it provides the BS
with an (11 dB + lOlog10N dB) received SNR. From the
portable point of view, the channel can be modeled
approximately as an ideal AWGN channel, hence the 17
dB received SNR re~ults in a 2x10 3 BER. On the other
hand, the channel can be pessimistically modeled as a
2~6~7~
2~
Rayleigh fading channel from the BS point of view. The
corresponding BER are displayed in Table III which
shows that the achieved BER i6 < 4x10-3. A BER < 10 2
is acceptable for voice.
Cell Patt~rn Rause
From Table I, the number of Full Duplex
voice channels (FDvc) that can be tran6mitted/received
per frame is 136 over 100 MHz, for a 6.18 Kbps
vocoder. If the bandwidth is halved to 50 MHz, the
number of FDvc per frame is reduced to 68, the noise
floor is reduced by 3 dB and the number of full duplex
frames that a BS can transmit/receive is doubled to
42, leaving the frame duration, the number of frames
per 13.104 ms and the processsr/deprocessor complexity
unchanged.
Reducing the available bandwidth directly
affects the cQll pattern reuse. This can be explained
as follows, assuming that we are required to offer a
minimum of 136 FDvc per cell, that the vocoder rate is
6.18 Kbps and that the cell .adius is fixed at 1 km
outdoors and 30 m indoors. For a 100 MHz band, we
assign one frame per cell and offer 136 FDvc per cell~
In this case, the cell pattern reuse consists of 126
cells as shown in Figure 8a which displays a seven
layer structure. For a 50 MHz band, we assign two
frames per cell and offer 136 FDvc per cell, hence
reducing our cell pattern reuse to a 63 cell pattern
as shown in Figure 8b which displays a five layer
structure. If the available bandwidth is as low as
5.86 MHz, we have 8 vc per frame. Hence we have to
assign 18 frames per cell in order to offer the
minimum required number of FDvc per cell. This reduces
the cell pattern reuse to as low as a 7 cell pattern
~6~7~
as shown in Figure 8c which displays a two layer
structure.
In Figures 8a, b and c, a shaded area is
shown around the center of the pattern, indicating l9,
38 and 126 full duplex frames that the central BS can
transmit/receive respectively. Tables IVa, b and c
show the number of cell layers in each cell pattern
reuse, the coverage area in Km2 of the pattern reuse
for both the indoor and the urban environments, as
well as the carrier to interference ratio (CIR) in dB,
for the 100 MHz, 50 MHz and 5.96 MHz bands,
respectively. In all cases, the CIR is large enough to
sustain a toll quality speech.
Transmission/Xe~eption Protocol
Since the number of FDvc a portable can
transmit/receive is one, while the number of FDvc a BS
can transmit/receive is much larger as shown in Table
V for each of the three vocoder rates, we have chosen
the following transmission/reception protocol:
1. The portable transmits a frame over a vc.
2. Seven adjacent BS receive the frame from the
portable.
3. One BS transmits to the portable, depending for
~5 example on the strength of the received signal by each
of the BS.
The control of this protocol may use any of
several known techniques. For example, the commonly
used technique is to have the portable monitor the
channel and determine which of several base stations
it is closest to. It can then order the nearest BS to
communicate with it. Another technique is to use a
master control which receives information about the
strength of the signal on the channel used by the
2 0 6 4: 9 r~ ~j
portable and controls the BS accordingly. Such
techni~ues in themselves are known and do not form
part of the invention.
Such a protocol has several advantages. For
instance, the location of the portable can be
determined with high accuracy based on the received vc
at the seven adjacent BS. Locating the portable can
assist in the BS hand-off. A BS hand-off and a
portable hand-off do not necessarily occur
simultaneously, contrary to other prior art systems.
In the present invention, when a portable roams from
one cell X to an adjacent cell Y, a new vc is not
raquirad immediately. What is required is a BS hand-
offl meaning that BS Y (associated with cell Y) must
initiate transmission to the portable over the same
vc, while the BS X (associated with cell X) must
terminate its transmission to the portable.
A BS hand-off occurs without the knowledge
of the portable and can occur several times before a
portable hand-off is required. A portable hand-off is
required only when the CIR is below a certain level.
In this case, the Mobile Telephone Switching Office
(not shown) calls for a portable hand-off in
accordance with known procedures. Reducing the
portable hand-off rate reduces the probability of
dropped calls. This is because a dropped call occurs
either because the portable hand-off is not successful
or because there are no available channels in cell Y.
The present invention allows the use of post-
detection diversity at the BS, and the use of dynamic
channel allocation (DCA).
2 ~
Dynamic Channel Allo~atio~
DCA is made pos6ible by each BS having
capability to transmit/rec0ive more than the num~er of
FDvc allocated to it6 cell, namely seven times the
number of FDvc for a 5.86 MHz band and up to twenty-
one times the number of FDvc for a 100 MHz as well as
a 50 MHz band. The DCA protocol simply consists of
borrowing as many FDvc as needed from the adjacent
cells, up to a certain limit. The limit for the case
when we employ a 6.18 Kbps vocoder, a 5.86 MHz band
and 18 frames per cell is obtained as follows. The
cell reuse pattern consists of 7 cells. Each cell is
preassigned 144 FDvc. Assuming that at peak hours, 75
FDvc are used on the average and 5 FDvc are reserved
at all times, then we are left with 64 idle channels
which represent the limit on the number of FDvc one
can borrow from the cell.
One should distinguish between the limit on
the channels borrowed and the limit on the
nonpreassigned channels a BS can use. For instance, if
a call originates in cell X and the portable roams
into an adjacent cell Y where no preassigned cells are
available, BS Y does not need to borrow immediately a
new channel from an adjacent cell. It can use the
original channel as long as the level of CI~ is
acceptable. If on the other hand, a portable wants to
initiate a call in cell Y where all preassigned
channels are used, BS Y can borrow a channel fr~m an
adjacent cell up to a limit of 64 channels per cell.
The main advantage of DCA over Fi~ed Channel
Allocation (FCA) is the increase in traffic handling
capability. For FCA, a 7 cell pattern each with a
preassigned 144 Fdvc can carry a total traffic of
880.81 Erlang at 0.01 Blocking Probability (BP). For
2 ~
DCA, a 7 cell pattern consists of 6 cells each with 80
FDvc that can carry a total traffic of 392.17 Erlang,
combined with one cell with 528 FDvc that can carry
501.74 Erlang. The total traffic is therefore 893.91
Erlang. This increase appears to be marginal (1.5~).
However, if 501.74 Erlang are actually offered to one
cell in the FCA system ~with 14 FDvc/cell), while the
six other cells carry 392.17/6 = 65.36 Erlang per
cell, the BP at that busy cell 0.714 while it is
negligible at the six other cells. The total blocked
traffic (i.e. lost traffic) in the FCA system is then
equal to (6x65.36xO.0 + lx0.714x501.24) = 358.24
Erlang. This represents a 0.4 average BP. If the DCA
is allowed such a loss, its traffic handling capacity
would increase to 1768.04 Erlang which represents a
100% increase in traffic handling capacity over the
FCA system, or equivalently a 160% increase in the
number of available FDvc. The DCA system thus
represents a marked improvement over the FCA sy~tem.
Voi~e Acti~ation
Voice activation is controlled by the BS
according to techniques known in the art. At any
instant during a conversation between a BS and a
~5 portable, there are four possibilities:
1. BS talks while the portable listens.
2. BS listens while the portable talks.
3. BS and portable talk simultaneously.
4. BS and portable listen simultaneously.
The BS controls the voice activation
procedure by allocating in cases 1, 3 and 4 three
slots (frames 1.1, 1.2 and 1.3) to the BS and one slot
the portable (frame 1~ every four slots as shown in
7 ~
34
Figure 9a. Likewise up to 21 portables may co~municake
with the base station in like fa6hion.
In case 2, on receiving a signal from the
portable, the BS allocates three slots (frames 1.1/
1.2 and 1.3) to the portable and one slot (frame 1) to
the ss every four slots as shown in Figure 9b.
Likewise, up to 21 other portables may communicate
with the base station in like fashion. Consequently,
instead of transmitting two full duplex voice frames
over four slots as in Figure 4, voice activation
allows us to transmit three full duplex voice frames
over four slots. Hence, voice ac-tivation provide~ a
50~ increase in the number of available FDvc at the
expense of increasing DSP complexity.
Capacity
The capacity of Code Division Multiple
Access (CDMA) may be defined as the number of half
duplex voice channels (HDvc) effectively available
over a 1.25 MHz band per cell. Based on such a
definition, Table IV displays the capacity of analog
FM and of the present system with a 6.18 Kbps vocoder,
5.86 MHz band, 1 frame per cell and DCA. As shown in
Table IV, the capacity of analog FM is 6
HDvc/1.25MHz/cell while for the present system it is
150 HDvc/1.25MHz/cell. The 6.25 MHz band consists of
5.86 MHz plus two tail slots. When voice activation is
used, the capacity of the present system is increased
by 1.5 times to 225 HDvc/1.25MHz/cell, a 38 fold
increase over analog FM.
2 a ~ 3
Local Area Networks
The invention may also be applied to produce
a 48 Mbps wireless LAN, which also satisfies the
technical requirements for spread spectrum.
For wireless LAN, wideband differential
orthogonal frequency division multiplexing is again
employed. The LAN will incorporate a plurality of
transceivers, all more or less equal in terms of
processinq complexity, and possibly with identical
components except for addresses.
To implement wideband modulation for a LAN,
a 26 MHz band is divided into 128 points, as shown in
Figure 10, plus two tail slots of 1.48 MHz each within
the 26 MHz band. Adjacent points are s parated by 180
KHz and each point, as with the application described
above for a portable-base station, represents a D8PSK
symbol. The transmitter components will be the same as
shown in Figure 5b, with suitable modifications as
described in the following, and will include an
encoder. The output bits from the encoder are mapped
onto the D8PSK symbols.
The frame duration for the symbols is
illustrated in Figure 11. A rectangular time domain
window corresponding to a RC frequency domain pulse
has a 5.55 ys duration, and includes a 25% roll-off
and excess frame duration of 0.26 ~s, making a total
7.2 ~s duration for the frame.
For such a wireless local area network
~LAN), in which the transceivers are equal, the Time
Division Duplex protocol is as illustrated in Figure
12 (assuming there are at least a pair of
transceivers):
1. A first transceiver transmits a signal (frame 0)
over the entire frame.
2 ~ 7 ~
36
2. A second transceiver receives the signal from the
first transceiver and processes (analyzes) it.
3. Based on the received signal, the second
transceiver predistor~s and transmits nine frames
(frames 1 - 9) to the first transceiver immediately.
Each transceiv~r has transmitter components
similar to those illustrated in Figure 5b, with
suitable modifications to the internal structure to
allow the use of the particular frequency band and
frame duration employed.
The transmitter/receiver functional and
structural block diagrams are shown in Figures 13a,
13b and 13c for the exchange of data. Data i~ provided
to an encoder 810 where the speech is digitized and
coded to create bits of information. The bits are
provided to the modulator 812 which turns them into
D8PSK symbols, with three bits per symbol. The D8PSK
6ymbols are then processed in the processor 814 which
is described in more detail in Figure 14a. The output
from the processor is then filtered in low pass filter
816, upconverted to RF frequencies using local
oscillator 818 and transmitted by antenna 820.
In Figure 13b, the received signal at the
base station is filtered in a bandpass filter 822, and
down converted by mixing with the output of a local
oscillator 824. The average power of the downcoverted
signal is monitored by an initial power control 825
that adjusts the average power to the specifications
required by the sampler 826. The adjusted
downconverted signal is then sampled in sampler 826 to
produce bits of information. The bits are then
processed in the deprocessor 828, described in more
detail in Figure 14b. An estimate of the phase
differential i6 taken in the channel estimator 830, as
37
described in more detail in rela~ion to Figure 7
above, and the estimated phase differential is
supplied to a decoder/demodulator 832 to correct the
received bits. The estimated phase dif~erential is
also supplied to a pre-distorter 834 in the
transmitter. .~t the transmitter in the Base Station,
the same blocks are incorporated as in the portable
transmitter except that a pre-distorter is used to
alter the phase of the D8PSK 6ymbol8 to make the
channel appear Gaussian (ideal) as opposed to a fading
channel. The initial power control 825 also sends a
signal to the pre-distorter 834 to ad~ust the
transmitted power to an appropriate signal level for
the sampler 826 in the first transceiver. It will be
appreciated that a pre-distorter will be included in
the first transceiver~s transmitter but that it will
not be operable, except when the first transceiver i6
operating as a base station.
Figure 13c shows the functional blocks of
the receiver of the first transceiver, which is the
same as the receiver in the second transceiver except
it does not include an estimator. The processor is
illustrated in Figure 14a and 14c and the deprocessor
in Figure 14b and 14c. The processor first inverse
Fourier transforms the 128 D8PSK symbols output from
the modulator. The transformed symbols are then
triplicated as a group so that the total number of
samples is tripled (see the left side of Figure 4c~,
with three consecutive groups each consisting of the
128 transformed symbols. Next, the three ~roups are
windowed by a Raised Cosine window with a roll-off of
0.25 centered in the middle of the three groups. In
other words, the processor takes D8PSK symbols in,
pulse shapes them and inverse Fourler transforms them.
~0~7~
38
On the other hand, the deprocessor undoes what the
processor did, i.e. it removes the pulse shaping, then
Fourier transforms the received ~ignal to obtain the
original D8PSK symbols. The first two blocks in Figure
14b are similar to the ~econd two blocks in Figure 14a
except for two differences as follows. In the first
block shown in Figure 14b, the repeated groups of
symbols are partlally overlapped, as shown in Figure
14c. In the second block, a rectangular window is used
instead of the Raised Cosine to produce 128 output
samples corresponding to the 416 input samples.
The phase estimator is the same as that
shown in Figure 7, except that there are only 128
input samples, and the same description applies.
For both the LAN and cellular networks, the
present system is designed to operate as a spread
spectrum system preferably over such bands as are
permitted, which at present are the 902 ~ 928 MHz
band, 2.4 - 2.4835 GHz and 5.725 - 5.85 MHz. The
carrier frequency in the local oscillator shown in
Figures 5a, b and c will then be 915 MHz in the case
of the 902 - 928 MHz band, and the frequencies used
~or modulation will be centered on this carrier
frequency.
Alternative Embodiments
A person skilled in the art could make
immaterial modifications to the invention described
and claimed in this patent without departing from the
essence of the invention.
For example, a system may consist of one or
more central controllers (comparable to the Base
Stations in the exemplary cellular system described)
and some slave units (comparable to the portables~.
.,
2 ~
39
The slave unit executes tha commands it receives from
the central controller. The commands may be requesting
the slave unit to transmit a receive acknowledge, a
status code or information that the slave has acc~ss
to. The command may also be to relay the command or
the information to another slave unit.
2 ~ 7 ~
N ~ ~'
~' E ~ u~
.
C~ ~ ~
2 ~ ~ ~
~ ~ ~ _
C) C~ V -
~, ~ ~ r~r3 Q
O
~I ~ r~
oo ~, a
_ R.
Q ;L Q O_
a~ Y ~ Y o li
0 ~ CD ~~
,.
~ . .
'
2 ~ 7
BS cornplexi~y of processor/ vocoder complexity of processor/
de-processor/estirnator rate de-proc~ssQr for portable
18.77 Kbps 1.35 Mips
630 Mips 9.1 S Kbps 7 .29 Mips
6.18 Kbps 1.26 Mips
~ Mips = Mega Instruction Per S~cond.
t 1 Instruction = 1 complex add.+1 complex multi.~1 stara~e.
Table 11
Vococler received BER
rate SNR
18.77 Kbps 27.8 dB 4 X 10-
9.15 Kbps 30.8 dB 2 X 10
6.18 Kbps 31.33 dB 2 X 10
Table 111
2 ~ 7
100 MHz, 1 frame/cell
urban indoor CIR
~, , ar~a~ ar~a 1~
C~311S l~y~rs in Km~ in Km~ In UD
11 ~6 7 24~6 ().355 37.~3
(~)
50 Mtlz, 2 frames/c~13
~ of ~ urban indoor CIR
cells lay~s in Km2 in Km2 in dB
63 5 11.94 0.172 25.8
(b)
5.86 VlHz, 18 frames/cell
# Of # of urban indoor C:IR
cells layers in Km2 in Km2 in ~3
7 2 1.~3 0002 13~8
(c)
Tabl~ IV
- 100 IV Hz 50 IV Hz ~.86 h~l k
. . .
Vocsder ~=# of # of FDvc t~ of ~ of FDvc N=# of # o7 FDvc
FDv~l BS Tx/Rx FDv~/ BS Tx/Rx FDv~/ Bt:: Tx/P~x
Rate fr~me (21X1XN) fr~m~ 2XN ) frame (7~18XN)
18.77 Kbp~ 48 1008 24 1008 2 252
9.15 Kbps 95 1995 47 1974 5 630
- ~ - 6.1g Kbps 136 2856 68 2856 8 1008
Table V
~6~
~3
~7
C~
~3
,; .
~ 0X
.
~ ~r~ ~
al~ -
t o ~
a~
~) I N ~ ~
tlS ~ 11 I
~ Q V
.
- '
.. . , :
2 ~
APPEWDI~ A
Copyright ~ 1991 University Technoloyie~ Inc.
CCC~CCCCCCCCCCC(~CCCCCCCCCCCCCCCCCCCCCCCCCCC~CCCCC
CCC
C
C C
C A LISTING FOR THE "P5:~0CESSOR" C
C WRITTEN IN FORTRAN C
C C
CCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCC~CCCCCC~CCCCC
CCC
~1 1 1 1 1 1 1 1 1 1 1 1 1 1 ~ I I I I I I I I I I I I I I I I I I I I I I i I I I I I I I I I I I I I I I I I
C NAME : BOXNRM(ISTRM)- FUNCTION +
C FUNCTION : SIMULATES NORMALY DISTRIBUTED DEVIATES +
C ( POLAR METHOD OF BOX & MULLER IS USED) +
C REF: +
C CALLING PROG/SUBPROG: +
C CALLED SUBPROG : URAND +
C +
C INPUTS : ISTRM - PASSED FROM THE CALLING ROUTINE +
C OUTPUTS : STD. NORMALY DISTRIBUTED RANDOM DEVIATES
+
C PRINCIPAL VARIABLES : +
C DATE : 26 APRIL, 1985. +
ClIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII
FUNCTION BOXNRM(ISTRM)
C DECLARATION:
REAL URAND, V1, V2, R, FAC, GSET, BOXNRM
INTEGER ISTRM, ISET
SAVE GSET
DATA ISET/0/
C SHOULD WE GENERATE A PAIR OF DEVIATES ?
IF(ISET .EQ. 0) THEN
C YES, WE DON'T HAVE AN EXTRA DEVIATE HANDY, SO PICK
C TWO UNIFORM NUMBERS IN THE SQUARE EXTENDING FROM
2 ~ 7 ;3
APPENnIa~ A
Copy~ight @ 1991 University Technologie~ Inc.
C -1 TO +1 IN EACH DIFIECTI()N
Vl = 2. ~ URAND(ISTRM) - 1.
V2 = 2. ~ URAND(ISTRM) - 1.
C SEE IF THEY ARE IN THE UNIT CIRCLE, IF THEY ARE NOT GO TO 1
R = V1~2 + V2~2
IF( R .GE. 1.) GO TO 1
C NOW MAKE THE BOX-MULLER TRANSFORMATION TO GET TWO
NORMAL DEVIATES
FAC = SQRT(-2. ~ ALOG(R)/R)
C RETURN ONE AND SAVE ONE FOR NEXT TIME
GSET = Vl ~ FAC
BOXNRM = V2 ~ FAC
C SET FLAG
iSET = 1
ELSE
C WE HAVE AN EXTRA DEVIATE HANDY, SO RETURN IT
BOXNRM ~ GSET
C UNSET THE FLAG
ISET = 0
ENDIF
RETURN
END
2 ~ 7 ~
AP~DI~ ~
Copyright ~ 1991 University Technologies Inc~.
subroutine demod16(delta,cc)
complex cc
common /pie1/piover4,piover8,piover16
if(delta.ge.31.*piover16.or .delta.lt. 1.~piover16)cc=(+1,+1)
if(delta.ge.1.*piover16.and.delta.1t. 3.~piover16)cc=(+1,-1)
if(delta.ge. 3.*piover16.and.delta.1t. 5.*piover16)cc=(-1,-1)
if(delta.ge. 5.~piover16.and.delta.1t. 7.~piover16)cc=(-1,+1)
if(delta.ge. 7.~piover16.and.delta.1t. 9.*piover16)cc=(-1,+3)
if(delta.ge. 9.~piover16.and.delta.1t.11.~piover16)cc=(-1,-3)
if(delta.ge.11.*piover16.and.delta.1t.13.~piover16)cc=(+1,-3)
if(delta.ge.13.~piover16.and.delta.1t.15.*piover16)cc=(+1,+3)
if(delta.ge.15.~piover16.and.delta.1t.17.~piover16)cc=(+3,+3)
if(delta.ge.17.~pioverl 6.and.delta.1t.19.~piover16)cc=(+3,-3)
if(delta.ge.19.*piover16.and.delta.1t.21.~piover16)cc=(-3,-3)
if(delta.ge.21.~piover16.and.delta.1t.23.*piover16)cc=(-3,+3)
if(delta.ge.23.~piover16.and.delta.1t.25.~piover16)cc=(~3,+1)
if(delta.ge.25.*pinver16.and.delta.1t.27.*piover16)cc=(+3,~1)
if(delta.ge.27.*piover16.and.delta.1t.29.*piover16)cc=(+3,-1)
if(delta.ge.29.~piover16.and.delta.1t.31.~piover16)cc=(-3,-1)
return
end
g 7 ~.~
APPE~DI~C A
Copyright Q 1991 IJniversity Technologi~s Inc.
subroutine demod4(delta,cc)
complex cc
comrnon /pie1/piover4,piover8,piover16
if(delta.ge.O.*piover4.and.delta.1t.2.~piover4)cc=cmplx(~1 ,~1 )
if(delta.ge.2.*piover4.and.dslta.1t.4.~piover4)cc=cmplx(+1 ,-1 )
if(delta.ge.4.~piovera~.and.delta.1t.6.~piover4)cc=cmplx(-1 ,-1 )
if(delta.ge.6.~piover4.and.delta.1t.8.~piover4)cc=cmplx(-1 ,+1 )
return
end
7 ~
APPENDIX A
Copyright ~ 1991 University Technologies Inc.
subroutine demod8(delta,cc)
complex cc
common /pie1/pisver4,piover8,piover16
i~delta.ge.15.~piover8.or.delta.1t. 1.~piover8)cc=(+1,+1
if(delta.ge. 1 .~piover8.and.delta.1t. 3.~piover8)cc=(+1,-1 )
if(delta.ge. 3.~piover8.and.delta.1t. 5.~piovsr8)cc=(-1,-1)
if(delta.ge. 5.~piover8.and.delta.1t. 7.~piover8)cc=(-1,+1)
if(delta.ge. 7.tpiover8.and.delta.1t. 9.~'piover8)cc=(-3,+1)
if(delta.ge. 9.~piover8.and.delta.1t~ piover8)cc=(+3,+1 )
if(delta.ge.11 .~piover8.and.delta.1t. 1 3.~piover8)cc=(+3,-1 )
if (delta.ge. 1 3.~piover8.anci.delta.1t. 1 5.~piover8)cc=(-3,-1 )
return
end
206~:97 j
APP~5~7DIX A
Copyright ~ 1991 univer~ity Technologies Inc.
c SUBROUTINE FFT() ~ REFER TO THE BOOK BY 'GONZAI EZ 8i WINTZ'
c f is the input array to the fft
c f is also the output array from the fft with -tve freq. a~ the begining
c and -ve freq. at the end
c fl is the output array from the fft with -ve freq. at the begining
c and +ve freq. at the end.
c ind is the index array for f1.
c if ix=0 the F is the input array.
c if ix=1 the F1 is the input array.
subroutine fft (ix,f,ind,f1, In)
integer In, n, nv2, nm1, j, i, k, le, le1, ip,ind(2048)
real pi
complex f(2048), fl (2048),u, w, t
c
pi = 4.0 ~ atan(1.0)
n = 2~1n
nv2 = n/2
nm1 = n-1
if(ix.eq. 1 )call deswap(f,f 1, n)
j = 1
do 3 i= 1 , nm1
if (i.ge~j) go to 1
t = f~)
f(j) = f(i)
f(i) = t
k= nv2
2 if (k.ge.j) go to 3
j =j-k
k = ic/2
~oto2
3 j = j+k
do 5 I= 1 , In
ie = 2~1
le1 = le/2
u = (1.0, 0.0)
w = cmplxt cos(pi/le1), -sin(pi/le1 ) )
do 5 j= 1 , ie1
do 4 i- j, n, le
APPEND I a~ A
Copyright @ 1991 University Technologies Inc.
ip = i + le1
t = f(ip) ~ u
4 f(i) = f(i) + t
5 u=u~w
do 6 i= l, n
f(i) = f(i) / float (n)
6 ind(i) = i-n/2-1
call swap(f,f1,n)
return
end
2 ~
APPE~JDIX A
Copyright Q 1991 Univer6ity T~chnologies Inc.
C ~ * ~ * ~ * ~ ~ ~ * ~ * ~ ~
C Inverse fft transform
CA ~*~ A A A A~**~**~**~*~*~*~*~ *~ r~*~
c if ix=0, f is the input array to the Ifft with +ve freq. at the begining
c and -ve freq. at the end.
c if ix=1, f1 is the input array to the Ifft with -ve freq. at the begining
c and ~ve freq. at the end.
subroutine tdft (ix,xft,ind,xft1,nft)
integer nft, i, mft,ind(2048)
complex xft(nft),xft1(2048), buff(2048), buff1(2048)
mft= nint(alog10(float(nft))/alog10(2.0))
print ~, ' nft and mft are:', nft, mft
if ~ix.eq.1)call deswap(xft,xft1,nft)
do 7 i= 1, nft
7 buff(i) = conjg(xft(i))
c fft() does 1 -D Fourier transform
call fft(O,buff,ind,buff1, mft)
do8i=1,nft
xft(i) = nft*conjg(buff(i))
8 xft1 (i) = nft*conjg(buff1 (i))
return
end
- 206~9rl~
APPENDIX A
Copyright ~ 1991 University Technologies Inc.
C ~ * * * ~ * ~ b * * l- * ~ * ~ * ~ ~ * )~
subroutine swap(f,f1,n)
complex f(2048), f1(2048)
do i=-n/2,-1
f1 (i+n/2+1 ) = f(n+i~1 )
enddo
do i=O,n/2-l
f1 (i+n/2+1 ) = f(i+1 )
enddo
return
end
C ~ ~ * * ~ ~ ~ * * ~ * * .t * ~ R ~ . * ~ * l~ * * * .~ ~ * t~ * * ~ ~ * ~
subroutine deswap(f,fl,n)
complex f(2048), f1(2048)
do i=-n/2,-1
f(n+i+1 ) = f1 (i+n/2+1 )
enddo
do i=O,n/2-1
f(i+1 ) = f1 (i+n/2+1 )
enddo
return
end
2 ~
APP13ND I X A
Copyright @ 1991 University Technologie6 Inc.
real pi,TWOPl,arRe(0:512),arlm(0:512),norm,p,
& curphase(0:512),oldph,ph,psi(0:512),RC(-2048:2048)
integer ind(2048),counter
complex z(0:512),r(0:512),c(0:512),d(0:512),cc
complex F(2048),F1(2048)
character string~0
common /pie/pi
common /pie1/piover4,piover8,piover16
common /dec/no
open(100,file-'transmitted')
open(200,file='received')
open(400,file='BER')
open(800,file='envelope')
open(310,file='RC')
open(350,file='data')
open(351,file='random')
pi = 4. ~ atan(1.0)
piover4= pi/4.
piover8 = pi/8.
pioverl 6 ~ pi/16.
TWOPI = 2. ~ pi
read(~,102)string
write(~,102)string
read '*,~)number
read ~,~)M
read ~A,*)MM
read(~,~)no
1 02 format(a50)
C open(101,~ile=string)
C numb=255
LL= MM/2 -1
counter- 0
call raisedcos(MM,RC)
C do i=1,numb
c READING ~AA ~ A~AAAAAA AAA~AAA~ *~
C read(101,~)p,arRe(i),arlm(i)
C norm = ( 900. / p )~2
C arRe(i) = arRe(i)/norm
C arlm(i) = arlm(i)/norm
C z(i)=cmplx(arRe(i),arlm(i))
C write(800,~)i,sqrt(real(Z(i)~conjg(Z(i))))
C enddo
C z(0) =cmplx~arRe(1),arlm(1))
C z(MM)=cmplx(arRe(numb),arlm(numb))
,
2 ~
APPE~DIX A
Copyright @ 1991 University Technologies Inc.
do ii=1,number
C F I R ST LO O P A A A A ~ A A A ~ ~ A A A A ~ ~ A A A A ~ A
curphase(0) = urand(3)~TW0PI
d(0)= cmplx(cos(curphase(0)),sin~curphase(0)))
C~ STARTING THE DATA TRANSMISSION A~A~Ab~ AAAAA
C~ DATA GEN~RATIoN
do 10 I=1,MM,1
if(M.eq.4)call modu4(cc,delta)
if(M.eq.8)call rnodu8(cc,delta)
if(M.eq.1 6)call modu1 6(cc,delta)
write(500,~)1 ,cc,delta
C(l) = CC
psi(l) = urand(4)~TWOPI
write(351 ,~)psi(l)
curphase(l) = curphase(l~ delta ~ psi(l)
d(l) = cmplx(cos(curphase(l)),sin(curphase(l)))
enddo
C~ MODULATOR~
do k=0,MM,1
write(1 00,~)k,sqrt(real(d(k)~conjg(d(k))))
C r(k) = d(k)~(k)
r(k) = d(k)
write(451 ,~)k,r(k)
enddo
C ~ FIND THE TIME RESPONSE l~ AAAA~AAA
j=O
do k=1,MM,1
j=j+1
F1 (j) = r(k-1 )
enddo
CALL tdft(1 ,F,ind,F1 ,MM)
open(300,file='time_RCSF')
open(301 ,flle='timeO')
open(302,file='time_NOSF')
call window(MM,LL,RC,F1,counter)
C goto1 01
C~ DEMODULATOR~ t~
oldph = atan2(aimag(r(0)),real(r(0)))
~~ 206~7a
APPENDI~ A
Copyright Q 1991 University Technologies Inc.
do k=1,MM-1
ph = atan2(aimag(r(k)),real(r(k)))
delta= ph - oldph - psi(k)
if 'delta.gt.~WOPl)delta=delta-TWOPI
if delta.gt.~WOPl)delta=delta-TWOPI
if delta.gt.TWOPl)delta=delta-TWOPI
if ~delta.ltØ)delta=delta+TWOPI
if(delta.ltØ)delta=delta~TWOPI
if(delta.lt.û.)delta=delta~TWOPI
if'M.eq.4)call demod4(delta,cc)
if M.eq.8)call demod8(delta,cc)
if ~M.eq.1 6)call demud1 6(delta,cc)
write(501 ,~)k,cc,delta
if(real(cc).ne.real(c(k)))errorRe=errorRe~1
if(aimag(cc).ne.aimag(c(k)))errorlm=errorlm+1
write(400,~)k,' ',real(cc),real(c(k)),
& ' ',aimag(cc),aimag(c(k))
oldph = ph
enddo
print~ ,'errorRe=',errorRe,' ','errorlm=',errorlm
1 0 continue
1 01 continue
enddo
rewind(301 )
call quantize(LL,counter,prod)
write(306,~)prod
stop
end
2~6~
APPENDI~ A
Copyright ~ 1991 University Techn~logieg Ins.
subroutine modu16(cc,delta)
complex cc
common /pie1/piover4,piover8,piover16
s1 = urand(1)
s2 = urand(2)
s3 = urand(1)
s4= urand(2)
if(s1 .geØ5)s1 -+1
if(s1 .ItØ5)s1 =-1
if(s2.geØ5)s2=+1
if(s2.1tØ5)s2=-1
if(s3.geØ5)s3=+1
if(s3.1tØ5)s3=-1
if(s4.geØ5)s4=+1
if(s4.1t.U.S)s4=-1
cc = cmpix(s1 +2~s3,s2+2~s4)
write(350,~)cc
if(cc.eq.(+1 ,+1 ))delta=O.~piover8
if(cc.eq.(+1,-1 ))delta=1 .~piover8
if(cc.eq.(-1 ,-1 ))delta=2.~piover8
if(cc.eq.(-1 ,+1 ))delta=3.~piover8
if(cc.eq.(-1 ,+3))delta=4.fpiover~
if(cc.eq.(-1 ,-3))delta=5.~piover8
if(cc.eq. (+1 ,-3))delta=6.~piover8
if(cc.eq.(+1 ,+3))delta-7.~piover8
if(cc.eq.(+3,+3))delta=8.~piover8
if(cc.eq.(+3,-3))delta=9.~piover8
if(cc.eq.(-3,-3))delta=1 O.~piover8
if(cc.eq.(-3,+3))delta=1 1 .~piover8
if(cc.eq.(-3,+1 ))delta=1 2.~piover8
if(cc.eq.(+3,+1 ))delta=1 3.~piover8
if(cc.eq.(+3,-1 ))delta=1 4.~piover8
if(cc.eq.(-3,-1 ))delta=1 5.~piover8
return
end
,
APPENDIX A
Copyright @ 1991 University Technologies Inc.
subroutine modu4(cc,delta)
complex cc
cornmon /piellpiover4,piover8,piover16
s1 = urand(1 )
s2 = urand(2)
if(s1 .geØ5)s1 =+1
if(s1 .ItØ5)s1 =-1
if(s2.geØ5)s~=+1
if(s2.1tØ5)s2=-1
cc= cmplx(s1,s2)
writel350,~)cc
if(cc.eq.(+1,~1 ))delta=1 .~piover4
if(cc.eq.(~ 1 ))delta=3.bpiover4
if(cc.eq.(-1 ,-1 ))delta=5.~piover4
if (cc.eq. (-1,+1 ))delta=7.~piover4
return
end
APPENDIX A
Copyright @ 1991 Univer6ity Technologies Inc.
subroutine rnodu8(cc,delta)
complex cc
common Ipie1/piover4,piover8,piover16
s1 = urand(1 )
s2 = urand(2)
s3 = urand(1 )
if(s1 .geØ5)s1 -+1
if(s1 .ItØ5)s1 =-1
if(s2.geØ5)s2=+1
if(s2.1tØ5)s2=-1
if(s3.geØ5)s3=+1
if(s3.1tØ5)s3=-1
cc= cmplx(s1+2~s3,s2)
pp = urand(4)~TWOPI
write(350,~)cc
if(cc.eq.(+1 ,+1 ))delta=0.~piover4
if(cc.eq.(+1,-1 ))delta=1 .~piover4
if(cc.eq.(-1 ,-1 ))delta=2.~piover4
if(cc.eq. (-1,+1 ))delta=3.~piover4
if(cc.eq.(-3,+1 ))delta=4.~piover4
if(cc.eq.(+3,+1 ))delta=5.~piover4
if(cc.eq.(+3,-1 ))delta=6.~piover4
if(cc.eq.(-3,-1 ))delta=7.~piover4
return
end
2 ~
APPENDI~ A
Copyri~ht @ 1991 University l'echnologie~ Inc.
subroutine quantize(LL,counter,prod)
integer Fre,Fim,counter
xmax=-100.
xmin=1 G0.
do j=1,counter
read(301,~)i,arRe,arlm
if(arRe.gt.xmax)xmax=arRe
if(arlm.gt.xmax)xmax=arlm
if(arRe.lt.xmin)xmin=arRe
if ~arlm.lt.xmin)xmin=arlm
enddo
if(abs(xmax).gt.abs(xmin))xxx=abs(xmax)
if(abs(xmax).lt.abs(xmin))xxx=abs(xmin)
prod= 128./xxx
sumRe = 0.
sumlm = 0.
surRe = 0.
surlm = 0.
rewind(301)
do j=1,counter
read(301,~)i,arRe,arlm
Fre = 128 + prod~arRe
Fim = 128 + prod~arlm
C if(Fre.gt.255)Fre=255
C if(Fre.lt.O)Fre=0
C if(Fim.gt.255)Fim=255
C if(Fim.lt.O)Fim=0
write(401,~)i,Fre,Fim
C write 402,~)j,Fre
C write 403,~)j,Fim
sumF e = sumRe + (arRe-(Fre-128.~/prod)~*2
sumlm = sumlm + (arlm-(Fim-128.)/prod)~2
surRe = surRe + (arRe)~2
surlm = surlm + (arlm)~2
enddo
print~,'quantization noise=',(sumRe + sumlm)/(surRe + surlm)
print~,prod
return
end
2 ~ 3
APP~NDIX A
Copyright @ 1991 University Technologies Inc.
subroutine raisedcos(MM,RC)
real RC(-2048:2048j,pi
common /pie/pi
do i=-MM/2-MM/8,-MM/2+MM/8
t = -float(i)/float(MM)
RC(i)=(û.5~(1 .-sin(4.~pi~(t-0.5))))
write(31 0,~)i,RC(i)
enddo
do i=-MM/2+MM/8+1,MM/2-MM/8-1
t = float(i)/float(MM)
RC(i)=1 .
write(31 0,~)i,RC(i)
enddo
do i=MM/2-MM/8,MM/2+MM/8
t = float(i)/float(MM)
RC(i)=(0.5~ sin(4~pi~(t-0.~))))
write(31 0,*)i,RC(i)
enddo
return
end
2 ~6~ J
APPENDI~ A
Copyright @ 1991 Univer~ity Technologie~ Inc.
subroutine demod4(delta,wRe,wlm)
y = cmplx(cos(delta),sin(delta))
if(real(y).ge.O)then
wRe=+1 .
else
wRe=-1 .
endif
if(aimag(y).ge.O)then
wlm=~1 .
else
wlm=-1 .
endif
return
end
.
'
2 ~ 7 ~
APPE~DIX A
Copyright ~ 1991 University Technologies Inc.
C NAME : URAND(ISTRM) - FUNCTION +
C +
C FUNCTION : GENERATES STANDARD UNIFORMLY DiSTRlBUTED
+
C RANDOM NUMBERS. +
C ( USES THE RECURSION: SEED(ISTRM) = +
C 16807 ~ SEED(ISTRM) MOD (2~f(31) - 1) ) +
C SOME COMPILERS REQUIRE THE DECLARATION:
+
C INTEGER~4 ISTRM, K1 +
C REF:
C +
C CALLING PROG / SUBPROG: +
C +
C CALLED SUBPROG : NIL +
C +
C INPUTS : IS T RM - THE STREAM NUMBER +
C 0 < SEED(ISTRM) < 2147483647 +
C +
C OUTPUTS : STD. UNIFORMLY DISTRIBUTED RANDOM NUMBERS
+
C NEW VALUE OF SEED(ISTRM) +
C +
C PRINCIPAL VARIABLES : +
C~ +
C DATE :26APRIL,1985. +
C +
C
FUNCTION URAND(ISTRM)
C
C DECLARATION:
C
INTEGER SIZE3
PARAMETER(SIZE3 = 20)
REAL URAND
INTEGER SEED(SIZE3),1STRM,ISTRG,K1
C
save seed
data seed(1)/456789/ .
data seed(2)/1037625857/
data seed(3)/102899405/
data seed(4)/535417/
data seed(~)/211229747/
data seed(6)/2g52853/
data seed(7)/2987570/
~0~7~
APP~DIX A
Copyright ~ 1991 University Technologies Inc.
data seed~8)/45376256/
data seed(9)/17830857/
C
C GENERATE A U(0,1) VARIATE
C
ISTRG = ISTRM
K1 = SEED(ISTRG)/127773
SEED(ISTRG) = 16807 ~ ( SEED(ISTRG) - K1 ~ 127773) - K1 b 2836
IF( SEED(ISTRG) .LT. 0) SEED(ISTRG) = SEED(ISTRG) + 2147483647
URAND = FLOAT(SEED(ISTRG)) ~ 4.656612875E-10
C
RETURN
END
~6~
APPENDI~ A
Copyright Q 1991 U~iversity Technologie~ Inc.
subroutine window(MM,LL,RC,F,counter)
complex F(2048),F0(-20~:2048),Fm(-2048:2048),Fp(-2048:2048),
& Fsum(-2048:2048),FRC(-2048:2048)
real RC(-2048:2048)
integer counter,count1
common /dec/no
save count1
data countl/0/
countl = count1 + 1
do i=-MM-LL,MM+LL+1
FO(i)= (0.,0.)
Fm(i)= (0.,0.)
Fp(i)= (0.,0.)
0nddo
do i=-LL,LL+1
FO(i) = F(i+LL+1)
Frn~i-MM) = FO(i)
Fp(i+MM) = FO(i)
write(304,~)i,FO(i)
enddo
do i=-MM-LL,MM+LL+1
Fsum(i) = FO(i) ~ Fm(i) + Fp(i)
write(302,~)i,sqrt((real(Fsum(i)))*~2+(aimag(Fsum(i)))~2)
enddo
do i=-LL - (LL+1)/4-8,-LL - (LL+1)/4-1
counter = counter + 1
FRC(i) = cmplx(0.,0.)
write(301,*)counter,0. ,0.
enddo
do i=-LL- (LL+1)/4,LL+1 + (LL+1)/4
counter = counter + 1
FRC(i) = Fsum(i)~RC(i)
write(300,~)counter,sqrt((real(FRC(i)))~*2+(aimag(FRC(i)))~*2)
write(301,~)counter,real(FRC(i)),airnag(FRC(i))
write(202,~)counter,real(FRC(i))
write(203,~)counter,aimag(FRC(i))
enddo
do i=LL+1 + (LL+1)/4+1,LL+l + (LL+1)/418
counter = counter + l
FRC(i) = cmplx(0.,0.)
write(301,~)counter,0. ,0.
enddo
2a~s~
APPE~DIX A
Copyright @ 1991 Universi~,y Technologies Inc.
if(countl .eq.no)then
do i=-LL- (LL+1)/4-8,LL~1 ~ (LL~l)/4+8
counter = counter + 1
FRC(i) = cmplx(0.,0.)
write(301 ,~)counter,0. ,0.
write(202,~)counter,real(FRC(i))
write(203,~)counter,aimag(FRC(i))
enddo
endif
return
end
APPE~IDIX A
Copyright @ l9gl University Technologies Inc.
FFLAGS =
FFILES = main.f urand.f ffl.f window.f raisedcos.f quantize.f \
modu4.f moud8.f modul6.f demod4.f demod8.f demod16.f
OFILES = main.o urand.o fft.o window.o raisedcos.o quantize.o \
modu4.o modu8.s modu~6.o demod4.o demod8.o demod16.o
a.out: ${0FILES}
f77 ${FFLAGS} ${0FILES}
2 ~
APP~DI~ A
Copyright ~ l99l University Technologies Inc.
../../../../..IHF_LFITWOCHANNEU900/test. 1
16
64
2 ~
APPEN~I~ B
Copyright @ 1991 University Technologie~ Inc.
5~ C
C WRITTEN iN FORTRAN C
C
C ALISTING FORTHE"DEPROCESSOR" C
C C
GCCCCCCCCCC~CCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCC
CCCC
subroutine dernod16(delta,cc)
complex cc
common /pie1/piover4,piover8,piover16
if(delta.ge.31 .~piover1 6.or .delta.lt. 1 .~piover1 6)cc=(+~ ,+1 )
if(delta.ge. 1 .*piover1 6.and.delta.1t. 3.~piover1 6)cc=(+1,-1 )
if(delta.ge. 3.*piover1 6.and.delta.1t. 5.*piover1 6)cc=(-1,-1 )
if(delta.ge. 5.~piover1 6.and.delta.1t. 7.~piover1 6)cc=(-1,+1 )
if delta.ge. 7.*piover16.and.delta.1t. s~piover16)cc=(-1~+3)
if delta.ge. 9.~piover16.and.delta.1t.11.~piover16)cc=(-1,-3)
if ;delta.ge. 1 1 .~piover1 6.and.delta.1t. 1 3.*piover1 6)cc=(+1,-3)
if(delta.ge.1 3.~piover1 6.and.delta.1t.1 5.~piover1 6)cc=(+1,+3)
if 'delta.ge. 1 5.~piover1 6.and.delta.1t. 1 7.~piovar1 6)cc=(+3,+3)
if delta.ge.1 7.~piover1 6.and.delta.1t.1 9.*piover1 6)cc=(+3,-3)
if ~delta.ge.1 9.tpiover1 6.and.delta.1t.21 .~piover1 6)cc=(-3,-3)
if(delta.ge.21 .~piover1 6.and.delta.1t.23.~piover1 6)cc=(-3,+3)
if(delta~ge.23.~piover1 6.and.delta.1t.2S.*piover1 6)cc=(-3,+1 )
if(delta~ge~25~pioverl 6~and~delta~1t~27.~piover1 6)cc=(+3,+1 )
if(delta.ge.27.~piover1 6.and.delta.1t.29.*piover1 6)cc=(+3,-1 )
if(delta.ge.29.~piover1 6.and.delta.1t.31 .~piover1 6)cc=(-3,-1 )
return
end
'-' 2 0 ~ ~ ?9 7 ~
APPENDI~ B
Copyright ~ 1991 University Technologie~. Inc.
subroutine demod4(delta,cc)
complex cc
common /pie1/piover4,piover8,piover16
if(delta.ge.O.~piover4.and.delta.1t.2.~piover4)cc=cmplx(+1 ,+1 )
if(delta.ge.2.'~piover4.and.delta.1t.4.~piover4)cc=cmplx(+1,-1 )
if(delta.ge.4.~'piover4.and.delta.1t.6.~piover4)cc=cmplx(-1 ,-1 )
if(delta.ge.6.~piover4.and.delta.1t.8.~piover4)cc=cmplx(-1 ,+1 )
return
end
~, ,
2~9~
APPENDI~ B
Copyright @ 1991 Un.iv~rsity Technologies Inc.
subroutine dernod8(delta,cc)
complex cc
common /pie1/piover4,piover8,piover16
if(delta.ge.15.~piover8.or.delta.1t. 1.~piover8)cc=(+1,+1)
if(delta.ge. 1 .~piover8.and.delta.1t. 3.~piover8)cc=(+1,-1 )
if(delta.ge. 3.~piover8.and.delta.1t. 5.~piover8)cc=(-1,-1)
if(delta.ge. ~ piover8.and.delta.1t. 7.~piover8)cc=(-1,+1)
if(delta.ge. 7.~piover8.and.delta.1t. 9.~piover8)cc=(-3,+1)
if(delta.ge. 9.~piover8.and.delta.1t.11.~piover8)cc=(+3,+1)
if(diolta.ge. 1 1 .~piover8.and.delta. It. 1 3.~piove r8)cc=(+3,-1 )
i~(delta.ge.1 3.~piover8.and.delta.1t.1 5.~piover8)cc=(-3,-1 )
return
end
% ~ 6 4 9 r~ ~
-
APPE~DI~ B
Copyright @ 1991 University Technologie~ Inc.
subroutine dewindow(MM,numb,F)
integer shift
complex F(2048),F0(-2048:2048),Fm(-2048:2048),
Fsum(-2048:2048),Fp(-2048:2048),fprime
common /SHHH/shift
oommon /SDHH/prod
do i=1,numb
read(101,~)j,arRe,arlm
F(i) = cmplx(arRe-shift,arlm-shift)
enddo
if(coun1.eq.no)then
do i=1,numb
read(101,~)j,arRe,arlm
enddo
endif
LL=MM/2-1
do i=-MM-numb/2+1,MM+numb/2
FO(i) = (0.,0.)
Fm(i) = (0.,0.)
Fp(i) = (0.,0 )
enddo
do i=-numb/2+1,numb/2 '
FO(i) = F(i+numb/2)
Fm(i-MM) = FO(i)
Fp(i+MM) = FO(i)
enddo
do i=-MM-nurnb/2+1,MM+numb/2
Fsum(i) = FO(i) + Fm(i) + Fp(i)
enddo
sum =0.
sup= O.
do i=-LL,LL+1
write(3û3,~)i,sqrt((real(Fsum(i)))'t~2
& +(aimag(Fsum(i)))~2)
F(i~LL+1) = Fsum(i)
write(304,~)i,Fsum(i)
C read(604,~jj,Fprime
C sum = sum + conjg(Fsum(i)/prod-Fprime)~
C & (Fsum(i)/prod-Fprime)
C write(602,~)i,abs(real(Fsum(i)/prod))
9 7 ~
APPENDXX ~
Copyright Q 1991 University Technologies Inc.
C & -abs(real(Fprime))
C write(603,~)i,abs(aimag(Fsum(i)/prod))
C & -abs(aimag(Fprime))
C vvrite(605,~)i,Fsum(i)/prod-Fprime
C
C sup - sup + conjg(Fprime)~(Fprime)
enddo
C print~,'quantization error=',sum,sup,sum/sup
return
end
20~7~
APPENDIX B
Copyright @ 19~1 Univer~ity Technologies Inc.
cSUBROUTINE FFT()-REFER TO THE BOOK BY'GONZALEZ& WINTZ'
c f is the input array to the m
c f is also the output array from the fft with +ve freq. at the begining
c and -ve freq. at the end.
c f1 is the output array from the fft with -ve freq. at the begining
c and +ve freq. at the end.
c ind is the index array for f1.
c if ix=0 the F is the input array.
c if ix=1 the F1 is the input array.
subroutine fft (ix,f,ind,f1, In)
integer In, n, nv2, nm1, j, i, k, le, le1, ip,ind(2048)
real pi
complex f(2048), f1 (2048),u, w, t
c
pi = 4.0 ~ atan(1.0)
n = 2~1n
nv2= n/2
nm1 = n-1
if(ix.eq.1)call deswap(f,f1,n)
j= 1
do 3 i= 1 , nm1
If (i ge.j) go to 1
f(j) = f(i)
~(i) = t
k= nv2
2 if (k.ge.j) go to 3
J =J~k
k= k/2
go to 2
3 j= j~k
do 5 I= i , In
le= 2~1
le1 = le/2
u = (1.0, 0.0)
w - cmplx( cos(pi/le1), -sin(pi/le1) )
do 5 j= 1 , le1
do 4 i= j, n, le
2~6~
APPENDIX B
Copyright @ 1991 University Technologias Inc.
ip=i~lel
t = ~(ip) ~ u
f(ip) = f(i) - t
4 f(i) = f(i) ~ t
S u=u~w
do 6 i= 1 , n
f(i) = f(i) / float (n)
6 ind~i) = i-n/2-1
call swap(f,fl,n)
return
~nd
9 7 ~
APPEN~IX B
Copyright Q 1991 University Technologies Inc.
C~ A ~ ~ ~ b ~ ~ h *~ ****l~'lrb*11~t*'1t*~**~***~**
C Inverse ffl transform
C~ ~ A ~ **~*t.*~**~*****~*~**lt~**~*h
c if ix=0, f is the input array to ths Ifft with ~ve freq. at the begining
c and -ve freq. at the end.
c if ix=1, f1 is the input array to the Ifft with -ve freq. at the begining
c and +ve freq. at the end.
subroutine tdft (ix,xft,ind,xft1 ,nft)
int0ger nft, i, mft,ind(2048)
complex xft(nft),xft1(2048), buff(2048), buff1(2048)
mft= nint(alog10(float(nft))/al3g10(2.0))
print *, ' nft and mft are:', nft, mft
if(ix.eq.1)call deswap(xft,xft1,nft)
do7i=1, nft
7 buff(i) = conjg(xft(i))
c ffl() does 1 -D Fourier transform
call fft(O,buH,ind,buff1, mft)
do8i= 1, nft
xft(i) = nft*conjg(buff(i))
8 xft1 (i) = nft*conjg(buff1 (i))
return
end
9 ~ ~
APPENDI~ ~
Copyright @ 1991 University Technologies Inc.
C~r**~0~ *t~ *~b*~*1~ *~ *.t*~*1~*~r~*~ **~ 1r*~ *
subroutine swaptf,f1,n)
complex f(2048), f1(2048)
do i=-n/2,-1
fl (i+n/2+1 ) = f(n+i+1 )
enddo
do i-O,n/2-1
f1 (i+n/2+1 ) = f(i+1 )
enddo
return
end
~' . .'
7 ~
APPB~ilDI~ IES
Copyright @ 1991 University Technologies Inc.
1 0
C~ ~ S * ~ S L ~ A ~ ~ ~ h ~ S ~ Jr ~ S ~ ~ ~ S ~ ~ A ~ L ~ ~ ~ S ~ S S ~
subroutine deswap(f,f1,n)
complex f(2048), f1(2048)
do i=-n/2,-1
f(n+i+1 ) = f1 (i+nt2+1 )
enddo
do i=0,n/2-1
f(i~1 ) = f l (i+n/2+1 )
enddo
return
end
.:
2 ~
APPENDIX B
Copyright @ 1991 Univ~rsity Technologies Inc.
real pi,TWOPl,oldph,ph,psi(0:512)
complex r(0:512),c(0:512),cc,cl
complex F(2048),F1(2048)
integer ind(2048),shift
character string~40,string1~401string2~40
common /pie/pi
common /pie1 /piover4,piover8 piover16
common /SHHH/shift
common /SDHH/prod
open(100,file='transmitted')
open~200,file='received')
open(400,file='BER')
open(800,file='envelope')
open(310,file='RC')
pi = 4. ~ atan(1.0)
piover4= pil4.
piover8= pi/8.
piover16 = pil16.
T~IVOPI = 2. ~ pi
read(~,102)string
write(~,102)string
read(~,102)string1
write(~,102)string1
read(~,102)string2
write(*,102)string2
read(h,~)number
read '~,~)M
read ~,~)MM
read~ )ni
1 02 format(a40)
open(101,file=string)
open '350,file=string1)
open 351,file=string2)
open 604,file='..lfort.304')
open ~704,file='. ./fort.501 ')
shift- 0
prod= 1.
if(string.eq.'../fort.401 ')shift=128
if(string.eq.'. .Ifort.401 ')read(306,~)prod
numb=MM~1.25 + 16
c~A~ READING ~ *~ *~
sum =0.
.
7 ~
APPE27DIX B
Copyright @ 1991 University Technologie~ Inc.
do ii=1, number
call dewindow(MM,numb,F)
call fft(0,F,ind,F1,ni)
j=O
do k=0,MM-1,1
jj+1
r(k) = F1 (j)
write(451 ,~)k,r(k)
read(350,~)c(k+1 )
read(351 ,~)psi(k+1 )
enddo
C t DEMODuLAToR ~ A~AAAAA~AAAAAAA1~AAt~t/~b*
oldph = atan2(aimag(r(0)),real(r(0)))
do k=1,MM-1
ph = atan2(aimag(r(k)),real(r(k)))
delta = ph - oldph - psi(k) + pi
if(delta.gt.TWOPl)delta=delta-TWOPI
if(delta.gt.~WOPl)delta=delta-TWOPI
if(delta.gt.TWOPl)delta=delta-TWOPI
if(delta.ltØ)delta=delta+TWOPI
if(delta.ltØ)delta=delta+TWOPI
if(delta.ltØ)delta=delta+TWOPI
if(M.eq.4)call democl4(delta,cc)
if(M.eq.8)call demod8(delta,cc)
if(M.eq.1 6)call demod1 6(delta,cc)
write(501 ,~)k,cc,delta
read(704,~)1 ,cl,deltap
diff = delta-deltap
if(diff.gt.+pi)diff=diff-TWOPI
if(diff.lt.-pi)diff=diff+TWOPI
if(k.ne.MM/2.and.k.ne.MM/2+1 )sum = sum + difftt2
write(606,t)k,diff
if(real(cc).ne.real(c(k)))errorRe=errorRe+1
if(aimag(cc).ne.aimag(c(k))3errorlm=errorlm+1
write(400,~)k,' ',real(cc),real(c~k)),
' ',aimag(cc),aimag(c(k))
oldph = ph
enddo
print~,'errorRe=',errorRe,' ','errorlm=',errorlm
print~,'sum=',sum/float(MM-3)
1 0 continue
enddo
AP2ENDI~ B
Copyright @ ï991 University Technologies Inc.
1 3
stop
~nd
2 ~ 7 ~5
.
~PPENDlX B
Copyright @ 1991 Univer6ity Technologies Inc.
1 4
FFLAGS =
FFILES = main.f fft.f dewindow.f demod4.f demod8.f demod16.f
OFiLES = main.o fft.o dewindow.o demod4.o demod8.o demod16.o
a.out: $~0FILES}
f77 ${FFLAGS) ${0FILES}
7.~
APP~NDI~ B
Copyright @ 1991 University Technologies Inc.
../fort.401
../data
/random
16
64
.
. . . ~ . . . . .
2~6~75
APPENDI~ C
Copyright @ 1991 AGT Limited
SOF~W~RE: FOI~ SIMUL~'rIl:)N ON DSP32C
Thc followin~ sof~w;urc proL~r;lm w;ls dcvclopccl ;uld run usin~ D3SIM of ATScT. It
simul;~lcs ~IIC opcr;l~ion of lhc cqu;tlizcr on t DSP32C.
~includc <nn;t~lt.h>
~includc ~stdio.h>
t~includc ~lib;tp.h>
~*~ t~ *~ *~*~**~*~*********~ *~t~t~**
*t**~t**~ t~*t~***~ *~******************~*t~ ***
doublc twopi-6.283185372;
/* List thc ori~in31 tr~n~ n~ cd symbols */
illt jj2[514] =( -1, -1,
1, -1,
1, -1,
1,
1,
1, 1,
1, 1, .
-1, 1,
1, -1,
-1, -1,
1, 1,
-1, -1,
-1, -1,
1, 1,
1, 1,
-1, 1,
-1, -1,
1, -1,
-1, -1,
1, -1,
1, -1,
'1, 1,
-1, 1,
1, -1.
l, -I,
~1, -1,
1, -1,
1,
-1, 1,
2 ~ 7 ~
~PPl~NDI~ ~
Copyright @ 19 91 AGT Limited
-1, -1,
-1, 1,
1, -1,
1, -1,
-1, 1,
1 1
1, -1,
1, 1,
-1, -1,
-1, 1,
1, -1,
1, 1,
1, 1,
-1, 1,
1, -1,
-1, -1,
1, -1,
-1, 1,
-1, -1,
,
1 1
-1, -1,
-1, 1,
-1, 1,
1, -1,
1, -1,
-1, 1,
1, -1,
-1, -1,
-1, 1,
-1, -1,
1, -1.
1 1 ; , ,
-1, -1,
1, -1,
1, -1,
-1, -1,
-1, -1,
-1, -1,
-1, 1,
1, -1,
-1, -1,
1, 1,
-1, -1,
-1, -1 ,
-1, 1,
1, 1,
1, -1.
-1, -1,
APPI:NDI~ C
Copyright Q 19 91 AGT Limited
1 1
, l.
1 -
-1 1
l 1
- 1,
, 1,
, 1,
1 1
, 1,
1, 1,
1, -1,
-
1 1
-
1 1
1 , 1,
i, -1,
1 1
-1, -1,
2 0 ~ 4 9 7 r5
APPENDI~ C
Copyright @ 1991 AGT Limited
1, 1, .
1, -1,
- 1,
-1
-1, 1,
-1
1, -1,
-1'
1 11
-1, -1,
:' 1' 1 . . ..
1, 1
-1, 1,
1' 1'
-1, -1, ,
1 -1
1, 1, ,
-1, 1,,
-1 , -1,
-1,
,
-1 1
2 ~
APP~D:C~ C
Copyright @ 1991 AGT Limited
1 ~ -lt
-1, -1,
1, 1,
-1, 1,
-1, 1,
1 , -1,
1, 1,
1, 1,
1, 1,
-1, -1,
1, -1,
-1, -1,
1, 1,
1, -1,
,
-I, -1,
-1, 1,
-1, 1,
1, 1,
,, 1, ~,
-1, 1,
-1, -1,
-1, 1,
1, -1,
1, -1,
1, -1,
1, 1,
-1, 1,
-1, 1,
-1, -1.
-1, 1.
1, 1,
1, -1,
1, -1,
1, -1,
I, -1,
1, 1,
- I , - 1 ,
-1, 1,
-1, 1,
-1, -1.
1, -1,
1, -1,
1, 1, .
1, -1,
2 ~
APP13ND I ~C C
Copyr ight @ 19 91 AGT Limited
-I -I
1, 1,
1, -1,
1, 1,
1, -1,
-1, -1,
1, -1,
-1, -1,
~, 1,
1, 1
I* List ~IG rcccivcd I ;uld Q */
doublc z[514]=(
1.8$26647996902 ,0~7904273271561,
-0.98167~02880096 , -1.7151~88065720,
- 1.9221851825714, -0.4~4524636268~)2,
- 1 ~6370518207550 , 1.0X9839100~377,
-0.3161578476~290 ,1.92~X703718185,
-15235337018967 , 1.1806204319000,
0 16620002685977 ,-1.8899723291397,
-1.3761~2474670~, 1.2521643638611l
' 7.168105803~301~-03 ,-1.8177816867828,
- 1.30318343G3937 , - 1.1972546577454,
-0.15820230543613 , -1.7096272706985,
-0.95022~41453552 , 1.332978725~333,
-1.5678428411484, 0.327Q0437307358,
-1.3414373397827, -0.75913363695145,
1.3957004547119, -0.4962155X189392,
1.3290159702301, 0.5087554~548035,
-1.197313~047623, 0.6627529~634964,
- 1.2967063188553, -0.2~4~5547163~ 86,
0.82353335618973, 0.9775187~692~78,
2.796626091003~-02 , 1.2459809780121,
0.97553628683090, 0.74172759056091,
0.3024236~598679 , 1.1787307262421,
1.1158645153046, 0.49576562643051,
1.0~71860885620, -0.57270908355713,
-0.2456~725990295 , 1.241Y062686920,
-0.83266~7$437424 , - ~ .0038361549377,
-1.3508954048157 , 2.358290X324X97E-03,
-0.90133684873581 , 1.0763~28211212,
-1.4~09677982330, 0.2~2~0809679031,
-0.79~42151975632 , 1.2981982231140,
-1.5102118253708, 0.46876969933~10,
- 1.4932200908661, -0.67981 ~52398682,
1.5572143793106, -0.676236629~X60X,
1.GS70891141891, 0.56G12873077393,
-1.SX09999704361, 0.8602319955~258,
-1.7~3020896912, -0.~5382~132680X~,
-1.5810590982~37 , 1.0169~8~52539,
0.3~510~06851768 , -1.87X~730102539,
~ 4 9 I ~
APPENDIX C
Copyright ~ 19 91 AGT Limi ted
1.1434341fi69083 , 1.55736~9937592,
-1.93141400~14Q6,-0.241~7317~90578,
-1.51040184497~3 , 1.2377147674561,
1.~451805353165, 0.14569547772408,
1~298800~684~48 , 1.4411~46776581,
-1.9200929403305 ,-5.7761263102293E-02,
-1.35098~8117828 , 1.32672~5552658,
1.8577233552933 ,-2.11265SS278897~-02,
-1.3225439786911 ,-1.2418869733810,
1.7606525421143 ,-9.0550~1462002E~02,
1.116155505180~ ,-1.28824G6316223,
1.6323~74132538,-0.15046791732311,
0.97648859024048 ,-1.2~67235517502,
0.201178565621~ , 1.477~09~61~0,
-0.~258~977979660 , 1.1416391134262,
1.3004480fiO0357,-0.~32921677827~,
0.66765332221985 ,-1.0373107194~
0.27767503261566 , 1.107~35345fi~97,
0.50518971681S95,-0.91855829954147,
0.90428864955902,-0.30539751052856,
0.79054325819016, 0.34204551577S68,
-0.69724625349045, 0.32767096161~42,
-~.18177013099194, 0.6585951~47486(),
0.3457849621772~, 0.4925813376~035,
-0.52803617715836 ,-2.7841679751873E-02,
-0.29G41786217690, 0.36104345321GSS,
-0.404005944728~5, 0.116421423XS244,
0.114550895988~4,-0.37470200657845,
0,29129487276077,-0.24794004857540,
-4.7727126628160E-02,-0.3879120051X608,
0.19418933987fil7,-0.36395436S253~5,
-0.40166890621185, 0.18578046560287
-0.46209335327148,-0.11633513122797
-~.4i677090525627, 0.29576727747917,
-6.0622435063124E-02, 0.54043716192245,
-0.43378528952599, 0.37475159764290,
-0.59756731987000 ,-2.9096489772201E-02,
-0.42078813910484,-0.45302fi80134773,
0.63260388374329 , 2.2896632552147~-02,
0.47454643249512,~0.43297693133354
O.G~523047208786 , ~.2224980890751E-02,
-0~41148558259010,-0.~98132646~83~3,
8.6346767~42770E-02,-0.63570129~71368,
0.523321?4777985,-0.3575389981269X,
~0.15361997485161, 0.60483527183533,
0.549~2035675049,-0.2733777761~59~,
0~4155624210835, 0.55399334430695,
0.57553601264954,-0.1621826~91~843,
0.485040~282310~, 0.34690761566162,
O.GOOG1538219452 ,-2.7974145486951E-02,
0.4657788574695G,-0.40029266~76631,
-0.62349128723145,-0.12451778352261,
-0.302453~7751846,-0.59375917911530,
7 ~
APPE~DI% C
Copyright @ 19 91 AGT Limited
-0.642~2871~52057 -0.290002733~G901,
0.72607111930847 -0.19453555345535,
0.65767973661~23, 0.46279X23780060,
0.85773074626923 , -7.9774774610996E-02,
0.6fi653263568878, 0.63700795173645,
0.98371291160583 , 3.8463655859232E-02,
-0.668365478~ 1~62, -0.~0670833587646,
- 1.0991182327271, -0.15678122639656,
0.96613508462906, -0.66218781471252,
-0.27184593677521 , 1.19933533G6852,
1,1098629236221, -0.64718896150589,
1.28019189834S9, 0.38049080967903,
0.62276816368103 , 1.2329781055450,
-0.47980448603630 , 1.33809399604~0,
0.58856463432312 , 1.3312312364578,
0.567219734191B9 ,-1.3701ql2G77765,
-1.401169896125g, 0.54447442293167,
1.3742272853851, 0.64058619737625,
-1.~402492046356, 0.~9066120386124,
1.3491046428680, 0.69823592901230,
-0.42755731940269 , - 1.4469082355499,
-0.73903095722198 , 1.2944310903549,
-1.42062079~0646, 0.355XS281252861,
-1.2107812166214, -0.76239341497421,
-1.3619115352631, 0.27647992968559,
1.0996322631836, 0.76~32592487335,
1.2723374366760, -0.19058661162853,
0.96331912279129, 0.75740247964859,
1.15443992G 1475 , -9.950599~ 194218~-02,
-0.80496501922607, -0.7307~96070861 ~,
-1.0116646289825 , ~,7180349938571E-03,
-0.69000792503357, 0.62838578224182,
-0.84825158119202 , -9.2191144X2~028E-02,
0.63727533817291, -().~3797433376312,
-0.66910350322723, ~0.189~7243859768,
0.57503563165665, -0.23856720328331,
0.28575962781906, -0.47963011264801,
3.S2~8123955727~-02, 0.50607603788376,
0.37911093235016, -0.28557923436165,
0.43339198827744, 0.1 G655600070953,
-9.2856124043465E-02, -0.~6805140376091,
-0.36173328757286, 0.36008G02380753,
-0.55111116170883 ~ -9.2662073671818~-02,
0.54514288902283, -0.28926~51990891,
0.26530939340591, -0.62695705890656,
0.71201461553574, -0.22391571104527,
-0.69442093372345, -0.~1981008648872,
-0.16683530807495, -0.8580374121666~,
-0.55144131183624, 0.75251996517181,
0.12049946933985, 0.97948336601257,
0.80047225952148, 0.65618216991425,
1.0733104944229 ,-8.698831q986229E-02,
0.83770149946213, 0.73083782196045,
A.PP~S~IDI~ e
Copyright @ 19 91 AGT Limited
-6.7907817661762E-02, -1.1372~5S93~78~l,
-0.86384069919586, -0.7731391191q825,
1.169844~50082 , -6.43284916~7775I~-02,
-0.87872487306595, -0.78181076049~305,
-1.1705212593079, 7.67~10'15X607fi7~1E-02,
-0.75661289691925, 0.882379412651OG,
-0.10~02639412880 , -1.13955~977~170,
0.87500M1074371, 0.6983'1393262863,
1.0780709981918, -0.14845751225948,
-0.60881644487381, 0.8569556a,746857,
0.20570~44411850, 0.98799550533295,
-0.49079838395119, 0.8287196159362X,
-0.$71~8609113693, ().27489036321640,
0.3~792~7391~239, -0.7908919'1536209,
0.35360851883888, 0.73334228992462,
-0.18458327651024, 0.74415004253387,
-0.57590019702911, 0.4390312433242~,
5.7776970788836E-03 ,-0.68923103809357,
0.~2798008918762, 0.4039097428321 ~,
-0.62690776586533, 0.183030828~3358,
0.22190481424332, -0.6170297265052X,
0.55797028541565, -0.37609174847G03,
0.70261883735657, 3.456574G7X'1210E-02,
-0.56754517555237, -0.4832079410SS30,
-0.78116804361343, 0.15341848134995,
-0.40339776873589, 0.7516025900g'10~,
0.~4919923543930 -0.3374855220317X,
-0.31929352879524 0.92272567749023,
-0.51332873106003, -0.90345317125320,
0.23162299394608 , - 1.0757941007614,
-0.94100481271744, 0.67701315879822,
-1.2062630653381, -0.14108607172966,
-0.95936530828476, 0.82507961988449,
1.3102972507477 4.8358015716076E-02,
-0.95462840795517 -0.95657634735107,
1.3848868608475 , -4.5906782150269E-02,
- 1.0633902549744, -0.93128252029419,
1.4279329776764, -0.1410651355981X,
1.1497G60875320, 0.88279163837433,
0.2364769~309452, 1.4383053779602,
-0.81110906600952, 1.2128539085388,
0.331~9531483650, 1.4158697128~96,
-1.2524482011795, -0.7169552'1'154117,
0.42545720934868, 1.3614822626114,
-0.60175400972366, 1.2690199613571,
1.2769542932510, -0.51767551898956,
0.46760299B01826 , -1.2636756896973,
0.60743457078934, l.lfi49851799011,
1.2380973100662, 0.31721931695938,
0.69398546218872, 1.02907121181~9,
0.15386630594730 , -1.1944670677185,
-0.77655005455017, -0.87338465452194,
-1.8741229549050E-02 , -1.1353790760040,
~ 3
A2PEN~ ~ ~ C
Copyri~ht @ 1991 AGT Limited
-0.7026374936103~, O.S5~1322671S9026,
-0.1~654582440853 , - 1.0637385845184,
0.~264~ 190259933, 0.52192509174347,
0.98~657253742~2, -0.37526~36991310,
-0 99220246076584, -0.33656212G87492,
0 55053806304932, 0.89534473419189,
1.0506724119186, 0.151~1125869751,
Q 71802806854248, 0.80500024557114,
i.1011143922806 , -2.6790169999003E-02,
0.873591~6343002, 0.71470791101456,
-0.19459256529808 , -1.14280831~1381,
-1.0133891105652, -0.6~,73406147956X
-0.34699076414108 , -1.1751155~53271
0.54547435045242 , - 1.13401591777~0,
0 48006322979927 , 1 1975069046021
-i.2326081991196, -0.47131305932999,
0.59059274196625 , 1.2095X56666565,
0.40663030743599 , - 1.30693X7674332,
0.67615872621536 , 1.2111126184~()4,
1.3554904460907, 0.3527274~297981,
0.73520946502686 , 1.2020268440247,
- 1.3775057S~4986 -0.31040790677071,
-l.lS24613809586 0.76710045337677
-0.27996954321~61 , 1.3730301X56995
0.77210420370102 , 1.1527576446533,
1.3428838253021, 0.26121476~92610
-0.75139236450195 , -1.1134694814682,
1.2886705398560, 0.25347581505775,
-1.0653619766235, 0.70698553323746,
-1.2127152681351, -0.25565749406815,
1.0094038248062, -0.64167791604996,
1.1179939508438, 0.26628923416138
0.94675332307816, -0.55X93~45014954,
0.28359553217888 , - 1.00804257392~8,
-0.46277087926865, -0.87873333692551,
-0.88684558868408, -0.305564731359qX,
-0.80680304765701, 0.35761266946793,
0.75870722532272, 0.33003318309784,
0.24814304709435, 0.73252141~75677,
-0.6~811541557312, -0.35476726293564,
0.65750586986542, -0.1391437G497269,
-0.49959596991539, -0.37754675745964,
0.58338809013367 ,-3.5333048552275E-02,
0.37756574153900, 0.39624589681625,
0.S1176452636719 , 5.879380181431~-02,
0.~0890917181969, -0.26618802547~55,
-0.44414889812469, -0.13911458849907,
-0.41381889581680, 0.16923630237579,
-0.38192304968834, -0.20201~73779297,
8.9968837797642E-02, 0.40955q 15368080,
-0.24452674388885, 0.326292qS51963X
0.39503705S01556 , -3.1020458787680~-02,
0.2643900215625X, -0.27824369072914,
1 0
7 ~
~PPENDIX C
Copyright @ 1 9 9 1 AGT Limi ted
-0.36956810951233 , -5.6~73057037592E-03,
-0.2601644992~284, 0.238510G8317890,
-0.33284601569176, -1.90100260078911~-02,
-0.231259~6521759, 0.2075'16010613'1'1,
-0.28497639298439 , -8.6228372529149E-03,
O.17795602977276, -0.185502409935û0,
-0.22646434605122 , 2.4967700242996E-02);
~loublc frl[514];
~ *~ **********~*:h***~**~*~ ***~**~**~ **/
m~inO
int crrorrc=O,crrorim=O,numbcr,lllf~ id[257]~sl257]~ k;
int jre,jim,l;
doublc mctricp,mctricm,milli l ,lllilli2~ jcy;
doublc dcltphas";rc, Xilll, CUrl)l1;15~129J;
3sm(".~10b31 dccl3rcd");
;Ism ("dccl;lrcd:");
numbcr= 12~;
/*~ rc3d ~ c Q compollcll~s S; filld ~llcir ;ullpli~udc. "rrl" **~***~****/
for (j=0; j <= numbcr, j~
curph3s~;] = 3t~ll2(~2*j~ [2*il);
frl [2*j] = sqr~(~[2*j]*;G[2*jj ~z[2*j~1]*:~2*j~ 1]); 3
;Ism(".~lob31 rc3d d;.t;-");
3sm ("rc;ld_d3t;0");
/t*~*t~ Esliln;l~c ~I~G Group dc13y "rrl" frolll ~lc ;lmpli~udc "frl" *~*******~*~/
for (k=l; k <= numbcr ;1; ~
frl[2*k-1] = (frl[2*k] - frl[2*k-2])/frl[2~k];
frl [2*k-y =0.0; )
frl[2*k] = 0.0;
cstim3tc lllc L~roup dcl;ly "frl" usinL~ c hilbcr~ tr;lllSfOrm ~ *
1ll = 7;
3sm(".~1Ob31 b_fft");
;Ism ("b_fft:");
fh(numbcr,m,frl);
;Ism(''.~lob31 3_fft");
;Ism (".~_fft:");
for(k=l; k~= numbcr/2; k~
jcy = frl[2*k-11/nulllbcr,
frl~2~k 1]=0.0-frl[2*1;]/numbcr;
frl [2*k]=jey;
for (k=nulllbcr/2-~1; k <- nulllbcr; k
2 ~ 7 ~
APPEND I ~ C
Copyright @ 1991 AGT Limitecl
jey = ~rl[2*k~ umbcr,
~1~2~k-1~= frl[2~k~/nuMber;
rrl~2*k] =Ojcy;)
;lsm(".~lob.ll b_iffi");
~sm("b ifft:");
~ ff~(numbcr,m,~rl);
3sm(".~10b;11 a_ifft");
;Ism(".l_ifft:"); ,.
/~*~rcsllol~ 6 ~ sc~ cl~ **~****~
The purpose of this seetion is ~o de~ermille ~hc sCECmCllts of ihc reccivc~l ~rcquency fr~me
WllCrC lhC Csti~ rtl ~roup dcl;~y is l;lr~er ~h;u~ ;~ ccru~ rcshold. Tllcsc sc~mcn~s .~rc
l;lbelled
~*~*1
id[l~=l;
m -0;
illit = O;
for (k=1; ke=numbcr; k~
: if(( frl~2*X-1~ > 0.07 )Il( frl[2*1;-l] e -0,07))~
frl[k~ = ~rl[2*k-1];
if(init==O)~
M+~;
id[m] = k;
init = 1;) )
else ~
frl~k~ = O.;
if(init==1)
mt~;
idlm~ = k;
illit = O; ) )
if(illi- -1)(
m~;
id~ml = number ~ l; )
;~sm(".~lob;ll lhresllold");
~ ~sm~"thre,shokl:");
t~ Group (Icl;ly Si~ ori~llm*~ tt~*~ t~
The purpose of this p~rt of lhe softw.Lre is lo delermine ll~e proper si~n of eslimn~d ~roup
del~y~ This is done throu~ll definin~ two 1-13mmin~ disl~nees one if we ~ssuem the
es~ cd ~roup del;ly w;ls posilive 3nd Ihe olher if it is 3ssullle~1 ne~3live ~nd ehoosin~ the
si~n th3~ Ie~ds lo the sm.lller disl3llee
~t*q~ tt~ *~ *:~******~****~ ***:~*~ t~ **~/ ~
for (k=l; k<=m; k~
metriep = O.;
2 ~ 7 ~
APPEND~ C
Copyright @ 1991 AGT Limit0d
mc~ricm = O.
if( (id[kl==id[k~ l) 1l (i(l[k]==id[1;~1]-2) )(
s[k]=O;
~o~o ~wcnty;)
for (j =id[k~l; j<=id[l;+l~
dcltph~s = curph;lsU] - curph;~j-l] - rrl[j~;
xrc = 1.4142136377*cos(dcltpll.ls);
xim = 1.4142136377*~in(dcl~ph;ls);
mini l - l .-xrc
minil- millil*minil;
if(minil>((l.~xrc)*(l.~-xrc))) millil=(l.~xrc)*(l. ~xrc);
mini2 = l.-xim;
mini2 = mini2*mini2;
if(mini2>((1.~xim)*(1.~xim))) milli2~ xilll)*(l.~-xim);
mcLricp ~= millil ~ milli2;
* ~ t ~ * ~ * ~ 'k l' * ~ * * * 'k ~ * ~ t * ~ t t ~ ~/
dcllpll~s = curpll;lsCj] - curph.ls[j-l] ~ frl~
xrc = 1.4142136377*cos(dcl~ph~s);
xim - 1.4142136377*sin(~1clîpll;ls),
minil = l.-xrc;
minil - minil * millil;
if(minil~((l.~xrc)*(l.~xrc))) millil=(l.~xrc)*(l.~xrc);
mini2 = l.-xim;
mini2 = rnini2 * mini2;
if(mini2>t(1.~xim)*(1.-~xim))) 1llilli2=(l.~xilll)*~ xilll),
mctricm ~= millil ~ milli2;
S[~C]=Q- l;
if(mctricp ~ mc~riun) s[k~=l;
/* printf("s[%d] = %~l",k,s[k]);*l
lwcn~y: k~; )
for (k=l, k<=m; k~
for a =id[k]; j<=id[k~l]-l; j~) frl~j] *= s[k];
/*k~;*l
~sm(".~lob~ fs");
3sm("~_fs:");
t~ tt'~ DcGodc ~llc Si~ J ~*~***~:I**~ **'P*~t~ *~
APPENDIX C
Copyright @ 1991 AGT Limited
1=1;
whilc (1 < num~cr~
dclll7hils = curph;ls[l] - curpll;~s[l- 1] - rrl [1~;
xrc = cos(dcltph~s);
xim = sin(dcltph;ls);
jrc = xrc/f~bs(xrc~;
jim = xim/f~bs(xim);
if (jrc == jim) (jrc = O jrc;
jim = O jim;
if(l>l) 1
if(jrc != iJ2~2*1~) crrorrc ~;
if(jim != jj~[2*1~1]) crrorim ~-;
printf("%d %d ~Oc~l",i,crrorrc,crrorim);
~sm(".~lob~l cnd");
~sm("cnd~
/*cnd of m;lill*/
1 4
