Note: Descriptions are shown in the official language in which they were submitted.
WO 91/20137 PCI/US91/03481
206~758
COMMUNICATION SIGNAL HAVING A TIME DOMAIN
PILOT COMPONENT
Technical Field
This invention relates generally to communication
methodology, and more particularly to communication
signals having information components that require the
presence of a pilot component in order to facilitate
recovery of the information components.
BackQround of the Invention
Various communication systems are known in the
art. Pursuant to many such systems, an information
signal is modulated on to a carrier signal and
transmitted from a first location to a second location.
At the second location, the information signal is
demodulated and recovered.
Typically, the communication path used by such a
system has various limitations, such as bandwidth. As a
result, there are upper practical limitations that
restrict the quantity of information that can be
supported by the communication path over a given period
of time. Various modulation schemes have been proposed
that effectively increase the information handling
capacity of the communication path as measured against
other modulation techniques. For example, a 16 point
WO 91/20137 2 0 6 ~ 7 S 8 PCI`/US91/03481
quadrature amplitude modulation (QAM) approach
provides a constellation of modulation values
(distinguished from one another by phase and amplitude)
wherein each constellation point represents a plurality
5 of information bits.
Such QAM signals are typically transmitted in
conjunction with a pilot component. For example, the
information components of the QAM signal can be
broadcast in conjunction with one or more pilot tones
10 that are offset in frequency from the information
content itself. These pilot components can be utilized to
support synchronization, and to otherwise support
recovery of the information component in a variety of
ways.
Unfortunately, such frequency offset pilot
components themselves consume bandwidth, thereby
reducing the amount of bandwidth available in a
communication path to support the information
components. If the information components are
themselves parsed into frequency offset data packages,
the problem increases as further spectrum must be
utilized to support the multiplicity of pilot references
that are typically required to allow recovery of the
various information packets.
In partial response to this situation, the prior art
has proposed the use of time domain pilot components.
For example, the information components of a particular
QAM transmission are combined with an inband
predetermined pilot reference component that appears in
a periodic manner. (Since the pilot component appears
only from time to time, the component is referred to as
WO 91/20137 PCI/US91/03481
20647~8
existing in the time domain, as distinguished from the
frequency domain pilot components discussed above.)
Though suitable for many applications, QAM
transmissions that include time domain pilot
5 components are not satisfactory in all applications. For
example, in an RF communication environment, where
communication units may be in motion with respect to
one another, such prior art time domain pilot reference
QAM methodologies may provide unacce~table
10 performance. In particular, the land-mobile radio channel
is characterized by multipath fading that causes the
channel phase and amplitude to vary over time as the
receiving or transmitting unit moves about. Such
variations must be compensated or otherwise allowed
15 for in order to provide proper reception. Typically, phase
and frequency modulation schemes avoid the need for
compensation since channel amplitude variations can be
ignored and differential or discriminator reception
techniques can automatically account for the channel
20 phase variations. I lowevor, phase and frequency
modulation are not very bandwidth efficient. While QAM
techniques can introduce bandwidth efficiency by
comparison, QAM requires more complicated channel
compensation methods, such as those prior art
25 techniques that use one or more pilot tones in
association with the information content.
Another problem associated with the multipath
nature of the radio channel is that of frequency-
selective fading. This occurs whenever the delay
30 difference between the various multipath components
that arrive at the reciever become large enough relative
to the signalling rate in the channel. When this happpens,
WO 91/20137 2 0 6 4 7 S 8 PCI /US91/03481
the channel's frequency response will no longer appear to
be flat in the band of interest, but will exhibit phase and
amplitude variations with frequency, which in turn will
vary with time as the transmitter or receiver moves
5 about. This frequency-selective effect causes signal
distortion that is present independent of the strength of
the received signal. In data communication systems, this
distortion manifests itself as an irreducible bit error
rate, or error floor, that persists regardless of how
10 strong the received signal becomes. In addition, the
distortion effect worsens as the information capacity of
the signal increases.
Accordingly, a need exists for a communication
methodology that will provide efficient use of QAM (and
15 the like) modulation techniques while simultaneously
substantially avoiding spectral inefficiencies that may
occur through use of certain prior art pilot component
techniques and other multipath compensation techniques.
This technique will preferably remain substantially
20 robust in a varying multipath operating environment.
Summary of the Invention
These needs and others are substantially met
25 through provision of the communication techniques
disclosed herein. Pursuant to this invention, an original
information signal is converted into a parallel plurality
of processed information signal sample sequences. At
least one of these sequences is then combined with a
30 reference sequence containing at least one
predetermined sample, which sample serves as a time
domain pilot reference that a receiver utilizes to
WO 91/20137 PCI/US91/03481
5 206~758
effectively recover a signal corresponding to the
original information signal.
In one embodiment of the invention, the original
information signal can be in the form of a serial data
stream, and the conversion step operates upon
preselected serial portions thereof.
In one embodiment of the invention, the conversion
step further includes converting groups of bits that
comprise the original information signal into
corresponding multibit symbols. In a further
embodiment, a predetermined plurality of these symbols
constitutes a processed information signal sample
sequence.
In one embodiment of the invention, the combining
step includes combining the predetermined sample
(which represents the time domain pilot reference) with
at least two of the sample sequences. In another
embodiment, all of the sequences are combined with a
pilot tone reference in thls manner.
In yet another embodiment, the time domain pilots
can be provid~d in some, but not all, of a group of
subchannels. To provide for channel compensation in the
subchannels that do not have a pilot, the time domain
pilots that are provided can be utilized to provide an
estimation of a pilot for that subchannel. In effect, then,
the occassiQnaly sent pilots can be utilized to
interpolate both over time and over frequency to allow
for channel compensation of the information signals.
Brief DescriDtion of the Drawin~s
WO 91/20137 PCI/US91/03481
2064758
Fig. 1 comprises a block diagram depiction of a
signal processor suitable for use in a transmitter in
accordance with the invention;
Fig. 2 comprises a depiction of a 16 QAM symbol
co nstel lation;
Fig. 3 comprises a depiction of a symbol
constellation wherein one of the symbols constitutes a
predetermined pilot reference symbol;
Figs. 4a-c comprise timing diagrams
representative of a series of symbol sequences as
provided in various embodiments in accordance with the
mventlon;
Fig. 5 comprises a spectral diagramatic
representation of a plurality of sample sequences, each
1~ having been combined with a predetermined symbol, in
accordance with the invention;
Figs. 6a-b comprise block diagrams depicting a
receiver suitable for use in receiving a signal in
accordance with the invention; and
Fig. 7 comprises a graph illustrating interpolated
channel gains as determined in accordance with the
inventlon .
Best Mode For Carrying Out The Invention
A signal processor for preparing a signal for
transmission in accordance with the invention is
generally depicted in Fig. 1 by the reference numeral
100. Though depicted in block diagram format for the
convenience of explanation and understanding, it should
be understood that the invention can be practiced in a
variety of embodiments; in particular, a digital signal
WO 91/20137 PCI`/US91/03481
206g758
processor, such as from the Motorola DSP 56000 or DSP
96000 families, is readily programmable to accomplish
the functions set forth below. Also, although described
below in the context of a 16#QAM application, it should
5 also be understood that the teachings herein are also
applicable for use with other modulation schemes as
well .
A processing unit (102) receives an original
information signal (101). In this particular embodiment,
10 this information signal constitutes a serial bit stream
having an ef~ec~i~/e baud rate of 53.2 kilobits per second.
This bit stream can represent, for example, true data,
digitized voice, or other appropriate signals.
The processing unit (102) functions to convert
15 groups of16 serial bits of the original information signal
into four 16 QAM complex signal points (symbols). For
example, Fig. 2 depicts a 16 QAM complex signal symbol
constellation (200). Each symbol in the constellation
represents a different combination of four serial bits.
20 For example, a first one of these symbols (201)
represents the bits "0001." A second symbol (202), on
the other hand, represents the bits ~0100," all in
accordance with well understood prior art methodology.
For each serially received 16 original information
25 bits, the processing unit (102) outputs, in parallel, on
each of 4 signal paths (103-106), an appropriate
representative multibit symbol as described above. A
pilot insertion unit (107-110), located in each signal
path (103-106), inserts a predetermined symbol
30 following receipt of 7 serially received information
symbols from the processing unit (102) pursuant to one
embodiment of a communication methodology in
WO 91/20137 PCI'/US91/03481
8 2064758
accordance with the invention. For exampie, with
reference to the constellation (200) depicted in Fig. 3,
the symbol depicted by reference numeral 301 can, by
way of example, serve as the predetermined symbol
5 inserted by the pilot insertion unit (107-1 10). (Other
symbols within the constellation could of course be
used. Arbitrary signal points not within the
constellation could also be used in an appropriate
application. Furthermore, although a particular symbol is
10 used to represent the pilot reference in this manner, this
does not mean that this same symbol cannot serve as a
multibit symbol for other symbol locations in the symbol
stream. The preferred embodiment would in fact allow
the predetermined symbol to perform this dual function.
15 Lastly, it is not necess~ry that all of the pilot symbols
be identical or spaced in time by a regular, uniform
interval; it is only necess~ry that they be selected in a
predetermined way.)
The resulting output from the pilot insertion units
20 (107-1 10) comprises a symbol stream (in this
embodiment having a symbol rate of 3.8 kilosymbols per
second) that is as generally depicted in Fig. 4a by
reference numeral 400. As depicted, a predetermined
symbol (402) constituting a pilot reference serially
25 appears following each 7 data symbols (401). This
symbol stream forms a composite signal that includes
one pilot reference symbol for every 7 data symbols.
These composite signals are provided to pulse shaping
filters (116-1 19) that appropriately shape the symbols
30 for transmission.
Thereafter, each composite signal is mixed (121-
124) with an appropriate injection signal (126-129) of
WO 91/20137 I PCI/US91/03481
2064758
the form 8 , wherein j is the square root of negative
one, t is time, and fOffk comprises an offset frequency
corresponding to the kth composite signal. All of the
above parameters will be identical for each of the
5 injection signals (126-129) with the exception of the
frequency offset value. In this embodiment, the first
injection signal (126) has an offset frequency value of
minus 6.27 kHz. The second injection signal (127) has an
offset frequency of minus 2.09 kHz. 2.09 kHz comprises
10 the offset frequency for the third injection signal (128),
and 6.27 kHz comprises the offset frequency for the
fourth injection signal (129).
The filtered and offset composite signals are
thereafter combined (131) to form a modulation signal.
15 The real and imaginary parts of this complex modulation
signal are separated (132, 133) and provided to a
quadrature upconverter (134), following which the signal
is amplified (135) and applied to an antenna (136) for
transmission, the latter occurring in accordance with
20 well-understood prior art methodology.
The resultant shaped, frequency offset, and
combined 16 QAM symbol sequences are generally
represented in Fig. 5 by reference numeral 500. As
generally depicted in this spectral diagram, there are
25 four effective sub-channels of symbol information
(501), each being offset from the others in correlation
to the offset frequencies referred to above. In this
embodiment, each subchannel symbol also includes a
time domain pilot reference sequence (figuratively
30 represented by reference numeral 502) embedded
therein. (It is not necessary that each 16 QAM
subchannel symbol of this quad 16 QAM packet include an
WO 91/20137 PCr/US91/03481
2064758
embedded time domain pilot reference. For example,
only one of the QAM signals might include thè pilot
reference, as illustrated in Fig. 4b, with interpolation
techniques being used during reception to provide an
5 estimated pilot reference for use in recovering the
remaining 16 QAM subchannels. In addition, or in the
alternative, pilot sequences for the various subchannels
might be staggered in time relative to each other, as
depicted in Fig. 4c, to allow interpolation over time and
10 frequency of estimated pilot references for use in
recovering symbols for all subchannels. What is
important is that a plurality of QAM signals be
substantially simultaneously provided, in a manner
frequency offset from one another, wherein at least one
15 of the QAM signals includes a time domain pilot
reference.)
A receiver suitable for use in recovering the above
described signal has been set forth in Fig. 6a (600).
Following appropriate reception of the transmitted
20 signal as provided by, for example, an antenna (601),
preselector (602), and quadrature downconverter (603),
a composite signal centered substantially at zero
frequency is provided to a bank of subchannel receivers
(604a-d), for the purpose of recovering the original 16
25 QAM signals.
Operation of the subchannel recievers is further
illustrated in Fig. 6b. The composite signal still
comprising 4 parallel subchannels is mixed (606) with
the appropriate injection signal of the form e d~, in
30 order to center the desired subchannel at approximately zero
frequency (i.e., to remove the frequency offset introduced at the
transmitter).
WO 91/20137 PCI-/US91/03481
_
11 2 064758
A receiver pulse shaping filter (607) receives this
mixed signal and appropriately shapes the received
signal and filters out the other subchannel signals and
noise to produce a single subchannel signal. A symbol
5 sampler (608) then allows individual symbols to be
sampled and provided to both of two processing paths
(609 and 610). The first signal processing path (609)
includes a pilot sampler (611) that selects the pilot
symbols from the composite symbol sequence comprising
10 data and pilot symbols. The pilot samples are then
multiplied (612) by the reciprocal (613) of the original
transmitted pilot symbol (which is known at the
receiver by virtue of having been predetermined), to
provide an estimate of the channel gain corresponding to
15 the pilot sampling instant.
A pilot interpolation filter (614) then processes
this recovered pilot sequence to obtain an estimate of
the channel gain at the intervening data symbol instants.
Compensation of channel phase and amplitude
20 distortion and recovery of the original data symbols are
carried out as follows. Delay (616) provided in the
second processing path (610) serves to time-align the
estimated channel gains with the corresponding data
symbols. The delayed data symbols are multiplied (617)
25 by the complex conjugates (618) of the estimated
channel gains. This operation corrects for channel phase
but results in the symbol being scaled by the square of
the channel amplitude. This is taken into account in the
decision block (619) with appropriate input from a
30 threshold adjustment multiplier (621 ) that itself
utilizes nominal threshold information and a squared
WO 91/20137 PCI`/US91/03481
12 20647S8
representation of the complex channel gain estimate
(622).
The symbols received may have suffered
degradation due to, for example, phase rotation and/or
5 amplitude variations due to transmission and reception
difficulties. By use of information regarding phase
and/or amplitude discrepancies and/or effects that can
be gleaned from the pilot interpolation filter, however,
the symbols as output from the mixer are properly phase
10 compensated. Having been thusly phase compensated,
and given the appropriately adjusted decision thresholds
as are also provided by the pilot filter, a decision can
then be made as to which symbol has been received, and
the detected symbol passed on for further processing as
15 appropriate. Such processing would typically include, for
example, combining detected symbols from different
subchannel receivers, and conversion to a serial format.
Referring to Fig. 7, the function of the pilot
interpolation filter (608) can be described in more
20 detail. Complex channel gain relative to the overall
transmission path can be seen as generally depicted by
reference numeral 701. Pilot samples provide
information regarding channel gain at the various time
instants depicted by reference numeral 702. Based upon
25 this sample information, interpolated channel gain
estimates (703) can be made, which channel gain
estimates are suitable for use in recovering data
samples as described above.
This same methodology could of course be utilized
to support transmission and reception of independent
information signals that are to be sent in parallel with
WO 91/20137 2 ~ 6 4 7 5 8 PCI`/US91/03481
13
one another on a carrier. In effect, pursuant to this
embodiment, the various subchannels described above
would each carry information symbols that are
independent of the other subchannels, but wherein the
5 time domain pilot symbol(s) are interpolated over time
(and frequency, if desired, as described above) to
estimate channel conditions to thereby assist in the
proper recovery of the information symbols from the
various subchannels.
What is claimed is:
