Note: Descriptions are shown in the official language in which they were submitted.
CA 02572236 2006-12-22
WO 2006/012245 PCT/US2005/022335
1
RECEIVER FOR USE IN WIRELESS COMMUNICATIONS AND METHOD
AND TERMINAL USING IT
Field of the Invention
This invention relates to a receiver for use in wireless
communications and a method and terminal using it. In
particular, the invention relates to a direct conversion
receiver capable of demodulating a frequency modulated
(FM) RF (radio frequency) signal by resolution and use of
in-phase (I) and quadrature (Q) components of the
modulated signal.
Background of the Invention
Conventional FM wireless receivers built using direct
conversion architectures to detect I and Q components of
a received signal have an underlying problem. As
illustrated later, such receivers can develop an error
in relative phase and amplitude between the I and Q
components. This error, sometimes referred to as
`quadrature imbalance', can cause a distortion in the
resulting output audio signal. The distortion may be
unacceptable to users particularly under conditions when
the received signal is subject to Rayleigh fading (herein
`fading') and/or has a low signal to noise ratio. The
prior art does not provide a satisfactory solution to the
problem of quadrature imbalance.
The present invention is concerned in particular with the
amplitude imbalance component of quadrature imbalance.
CA 02572236 2009-08-04
2
US5705949 proposes a procedure for removing amplitude or
gain error between I and Q components. The procedure
requires complex processing capacity and is unlikely to be
satisfactory in a fading environment.
Summary of the Invention
In accordance with the present invention there is provided
a wireless receiver for receiving and demodulating a
frequency modulated RF signal by a direct conversion
procedure. The receiver includes an input signal path for
delivering an RF input received signal, a circuit for
producing in-phase and quadrature components of the
received signal, and an estimator for periodically
estimating an imbalance in amplitude between the in-phase
and quadrature components and for applying an relative
adjustment in amplitude to compensate for the detected
imbalance. The estimator is operable to:
(i) divide samples I, of the in-phase component I(t) and
corresponding samples Qi of the quadrature component Q(t)
into blocks;
(ii) calculate for each block a block power value I,
corresponding to a summation of values of squares of the
samples I, and a block power value Qn corresponding to a
summation of values of squares of the samples Q;;
(iii) calculate from the block power values I, and Q1za
block amplitude imbalance value A,,= and
CA 02572236 2009-08-04
2A
(iv) calculate for a set of the block amplitude
imbalance values an average value.
In another aspect, the set of block amplitude imbalance
values for which an average value is calculated by the
estimator is a selected subset of a larger set of block
amplitude imbalance values.
In another aspect, the estimator is operable to reject at
least one other subset of block amplitude imbalance values.
In another aspect, the estimator is operable to sort the
block amplitude imbalance values in terms of their sizes
and to reject (i) a first subset of block amplitude
imbalance values greater in size than a first threshold and
(ii) a second subset of block amplitude imbalance values
less in size than a second threshold, wherein each of the
first and second thresholds corresponds to a subset having
a predetermined number of block amplitude imbalance values.
In another aspect, the estimator is further operable to
select from the sorted set a third sub-set of block
amplitude imbalance values having block amplitude imbalance
values which are less than those of the first sub-set and
greater than those of the second sub-set.
In another aspect, the estimator is operable to calculate a
geometric average of the block amplitude imbalance values.
CA 02572236 2009-08-04
2B
In another aspect, the receiver is operable to select a
received signal of predetermined form to estimate the
current amplitude imbalance between the in phase and
quadrature components.
In another aspect, the estimator is operable to select a
size for the blocks of samples according to at least one of
a detected property of the received signal.
In another aspect, the estimator is operable to select a
size for the blocks of samples according to whether the
signal to noise ratio or the received signal strength of
the received signal is detected to be at least one of a)
above or below a given threshold and b) equal to or above a
given threshold and to select a second greater size for the
blocks of samples according to whether the signal to noise
ratio or signal strength is below a given threshold.
In another aspect, the receiver includes means for
periodically detecting an imbalance in phase between the
in-phase and quadrature components and for applying an
adjustment in relative phase to compensate for the detected
imbalance.
Embodiments of the present invention will now be described
by way of example with reference to the accompanying
drawings, in which:
CA 02572236 2009-08-04
2C
Brief Description of the Drawings
FIG. 1 is a schematic block circuit diagram of a known
direct conversion RF receiver.
FIG. 2 is a schematic block circuit diagram of a direct
conversion RF receiver embodying the invention
I
CA 02572236 2006-12-22
WO 2006/012245 PCT/US2005/022335
3
Description of embodiments of the invention
FIG. 1 shows a known RF direct conversion FM receiver 100
illustrating the problem to be addressed by the present
invention. An incoming FM signal x(t)is delivered via an
input path 101 having branched connections 103, 105
respectively to two mixers 107, 109. A local oscillator
111 generates a reference signal having the same
frequency as the carrier frequency of the incoming signal
x(t). A first component of the reference signal is applied
directly to the mixer 107 where it is multiplied with the
input signal x(t). A second component of the reference
signal is applied to a phase shifter 113 and a phase
shifted output of the phase shifter 113 is applied to the
mixer 109 where it is multiplied with the input signal
x(t). Although the phase shifter 113 in combination with
the mixers 107 and 109 is intended to introduce a phase
shift of 90 degrees with unity gain between the
components of the reference signal applied to the mixers
107 and 109, in practice a phase shift slightly different
from 90 degrees and a gain slightly different from unity
are introduced. An output signal from the mixer 107 is
passed through a low pass filter (LPF) 115 to produce an
output in-phase component signal 1(t) and an output signal
from the mixer 109 is passed through a low pass filter
(LPF) 117 to produce an output quadrature component
signal Q(t). The imbalance in amplitude introduced into
the output of the mixer 109 is shown in block 119 as an
imbalance gain A.
CA 02572236 2006-12-22
WO 2006/012245 PCT/US2005/022335
4
A mathematical analysis of the arrangement shown in FIG.
1 is as follows:
The input signal may be represented as
x(t) = cos(wt + q5(t) + y)
where (.o is RF carrier frequency of the input RF signalx(t),
yis oscillator arbitrary phase and q5(t)is the frequency
modulation of x(t) to be detected.
In addition, x(t) = I(t)+j*Q(t), where 1(t) and Q(t) are in-phase and
quadrature components of x(t).
1(t) = 2 cos(wt + q(t) + )/) cos(wt) =
= cos(2wt + O(t) + y) + cos(q5(t) +y) = cos(q(t) + y)
after
LPF
Q(t) = 2A cos(wt + O(t) + y) sin(wt + a)
= A sin(2wt + q5(t) + y + a) + A sin(q$(t) + )/ + a) A sin(O(t) + y + a).
after
LPF
where A represents the amplitude imbalance and
arepresents the phase imbalance angle between the phase
angles of I(t) and Q(t).
In accordance with an embodiment of the present invention
to be described the components I(t) and Q(t) are processed in
a manner to be described to estimate and apply an
adjustment to eliminate the amplitude imbalance A. The
phase imbalance is also estimated and eliminated, e.g. as
described in Applicant's copending UK patent application
number 0411888.1 The resulting adjusted components are
CA 02572236 2006-12-22
WO 2006/012245 PCT/US2005/022335
combined to construct the modulation signal q5(t)to provide
an audio signal output.
FIG. 2 is a block schematic diagram of a circuit 200
5 embodying the invention for use in a direct conversion FM
receiver. Components having the same reference numerals
as components in FIG. 1 have the same function as such
components and will not be further described.
The output signal I(t) passed by the low pass filter (LPF)
115 is sampled by a connection 201 and the output signal
Q(t) passed by the low pass filter (LPF) 117 is sampled by
a connection 203. The respective sampled signals obtained
by the connections 201 and 203 are provided as respective
inputs to a processor 204 which operates an amplitude
imbalance algorithm to be described in detail later. An
output signal from the processor 204 is an amplitude
imbalance correction signal indicating a value of 1/A.
This correction signal is applied via a connection 202 to
an amplitude modifier 205 which modifies the amplitude of
Q(t) by a factor of 1/A to eliminate the detected
amplitude imbalance A.
A phase adjustment processing circuit (not shown) using
samples of I(t) and Q(t) estimates a phase imbalance between
I(t) and Q(t), e . g . in the manner described in Applicant's
copending UK patent application number 0411888.1, and
generates a phase shift control signal corresponding to
an equal and opposite value of this estimated phase
imbalance. The phase adjustment signal estimated in this
way is applied by a phase shifter 207. A signal
CA 02572236 2006-12-22
WO 2006/012245 PCT/US2005/022335
6
corresponding to the quadrature component Q(t) is applied
from the low pass filter 117 via a connection 226 to the
phase shifter 207. The phase shifter 207 thereby applies
a phase angle adjustment which compensates for the
detected phase imbalance angle a.. An output from the
phase shifter 207 corresponding to a phase adjusted value
of Q(t) is applied to a processor 209. A signal
corresponding to the in-phase component I(t) is also
applied as an input to the processor 209 via a connection
224. The processor 209 calculates a value of the
quotient Q(t)lI(t) from its respective inputs and
supplies a signal representing the result to a processor
211. The processor 211 calculates the value of the
arctangent (arctg) of the quotient parameter represented
by the input signal from the processor 209. An output
signal from the processor 211 is applied to a further
processor 213 which calculates the differential with
respect to time t of the input signal to the processor
213. Finally, an output signal from the processor 213 is
applied to an audio output 215. The audio output 215
includes a transducer (not shown specifically) such as an
audio speaker which converts an electronic signal output
from the processor 213 into an audio signal, e.g. speech
information.
The amplitude imbalance algorithm operated by the
processor 204 is as follows. Samples of the components
I(t) and Q(t) are taken at a frequency of 20
ksamples/sec. So the length of time of each sample is
1/20k = 50psec. The samples are taken over a sampling
period of 500 msec. So the total number of samples is
CA 02572236 2006-12-22
WO 2006/012245 PCT/US2005/022335
7
500msec/50psec = 10000 samples. The samples are divided
into blocks. The block size is selected according to the
operating conditions, e.g. the received signal strength
or S/N (signal to noise ratio). For example, for received
signal strength or received S/N equal to or greater than
a threshold value, the block size may be set to a first
value and for received signal strength or S/N less than
the threshold value, the block size may be set
for a second, higher value. For example, for S/N
equal to or greater than the threshold value of
15dB, there may be 15 samples per block. So there
are 10000/15 = 666 blocks in the sampling period.
For a S/N lower than the threshold value of 15dB,
there may be 100 samples /block. So there are
10000/100 = 100 blocks. The algorithm performs
better in a fading environment with a small block
size.
For each block of samples, a value of the power
block size
of I (Iõ _ Iiz ) and a value of the power of Q
i=1
block size
( Q,~ _ ~i2) is calculated.
For each block, an amplitude imbalance value is
calculated from the power of I and the power of Q
using the following calculation: A,,=
Q" .
In
CA 02572236 2006-12-22
WO 2006/012245 PCT/US2005/022335
8
Thus, the block value of An is the square root
value of the power of Q divided by the power of
I.
The values of amplitude imbalance found for each of the
blocks in a given set of blocks, say 1000 blocks, are
sorted in order from lowest to highest.
A sub-set of 45% of the highest block amplitude imbalance
values in the sorted set and a sub-set of 45% of the
lowest block amplitude imbalance values in the sorted set
are rejected leaving only a sub-set of the 10% block
amplitude imbalance values between the rejected sub-sets.
So for example where there are 1000 blocks in the set,
the 450 highest and the 450 lowest block amplitude
imbalance values results are rejected leaving 100 block
amplitude imbalance values which are further processed.
From the remaining K imbalance results, where K is the
number of blocks in the remaining subset, e.g. 10 per
cent of the set of blocks evaluated in the above example,
a geometric average value is obtained using the following
calculation:
x
Acorr K A,1
n=1
where Acorr is the amplitude imbalance to be corrected for.
Thus, Acorr is equal to the kth root of the product
of the k sample results for A multiplied by each
other.
CA 02572236 2006-12-22
WO 2006/012245 PCT/US2005/022335
9
A signal corresponding to l/A,o,. is issued by the
processor 204 to be applied by the amplitude modifier
205. The algorithm is performed continuously and
adaptively on the received FM modulated signal during
periods when a signal is received.
The processor 204 may be operated when any received
speech signal plus associated sub-audio signalling is
received by the receiver 200. However, if desired, the
algorithm may be operated selectively only when a
specific input signal is received by the receiver 200.
For example, the receiver may operate on a known analogue
FM signal received from a RF transmitter. This may for
example be a standard FM modulated signal in accordance
with the industry standard TIA 603.
Division into blocks of the samples processed by the
processor 204 in the manner described earlier is
beneficial for processing a signal received in a fading
environment. If division into blocks is not applied there
is no possibility to reject results that are not correct
due to fading. In a fading environment there are fast
variations in signal envelope. When a signal is in a deep
fade the result for the quotient Q/I (for the block
processed when this applies) can be very large (I is
close to zero) or very small (Q is close to zero).
Division into blocks, sorting and rejecting high and low
results allows incorrect results caused by fading not to
be included in the amplitude imbalance estimation. In
practice,.a wireless terminal is always likely to work in
fading environment.
CA 02572236 2006-12-22
WO 2006/012245 PCT/US2005/022335
Block size is also significant: for fast fading a small
block size is optimal, for low S/N a larger block size
would be optimal. The receiver 200 may constantly measure
5 the received signal power using a known RSSI (Received
Signal Strength Indicator). The result may be provided to
the estimator 204 which may be operable to adjust the
block size automatically using the result provided. It
may be assumed that there is a relationship between
10 received signal power and received S/N so that for high
received power the S/N is also high. A threshold received
power value is used to determine whether a small block
size is to be used when the received power value is equal
to or above the threshold or.a greater block size is to
be used when the received power value is below the
threshold.
Using only a selected subset of block amplitude imbalance
results, e.g. only 10% of the results as in the example
described above, also has the benefit of significantly
reducing the complexity of the algorithm operated by the
processor 204 and therefore the amount of signal
processing required. This results in reduced consumption
of power from the battery of the terminal in which the
receiver 200 is incorporated.
Finding the geometric average has been found to be better
than finding the arithmetic average, because the latter
was found to introduce a bias to the results and gave an
incorrect amplitude imbalance estimation. An example to
illustrate this is as follows:
CA 02572236 2006-12-22
WO 2006/012245 PCT/US2005/022335
11
Say for block 1 Q1=3 and I1=2, and for block 2 Q1=2 and
I1=3 and A is 1. A1=3/2 and A2=2/3. Using an arithmetic
average for A will give 0.5(3/2+2/3) = 1.0833 - an
incorrect result. However, using a geometric average for
A: will give sqrt( (3/2)*(2/3))=1 - a correct result
(where `sgrt' is the square root).
Various processors are shown in FIG. 2. These processors
may be separate processors as shown or the functions of
two or more of the processors may be combined into a
single processor, e.g. digital signal processor
programmed with computational software, as will be
apparent to those skilled in the art.
Results
The algorithm operated by processor 204 was tested on
simulated and actual analogue FM recorded signals. The
actual signals were recorded using a Direct Conversion
receiver operating in the manner described with reference
to FIG. 2. We measured amplitude imbalance (Amp IM) in %.
Amp IM [%] = 100e where e is given by
Q(t) = A sin(O(t) + y + a) = (1 + e) sin(O(t) +,v + a)
Target performance for the error in applying the
algorithm is a maximum error in Amp IM = 0.5%.
For a variety of recorded real 60dbm signals in a fading
environment at 450MHz, the error in Amp IM was measured
and the results ranged from 0.08% to a maximum of 0.45%
with an average error of 0.2%. Similarly, the error in
CA 02572236 2006-12-22
WO 2006/012245 PCT/US2005/022335
12
Amp IM was estimated for various simulated signals with a
signal to noise ratio (SNR) ranging from 15dB to 35dB and
the error ranged from 0.08% (35dB SNR) to 0.2% (15dB
SNR).
In contrast, we also estimated the amplitude imbalance
error using the known calculation A =.V Z Q
T . In a fading
I
signal environment we obtained an average error value in
Amp IM of 4% using the known procedure.
Where the invention is used in a radio receiver, a memory
of the radio may be programmed following manufacture to
store a table of initial imbalance values versus RF
frequency. During operation of the radio the imbalance
values (amplitude and phase) will change with time. Thus,
updated imbalance information may be gathered in use as
described in the above embodiments and used to provide
suitable compensation to maintain a suitable quality of
audio output signal. The updated imbalance information
may also be stored in the memory of the radio to replace
the originally stored information.
Summary
In summary, an improved method for adaptive amplitude
imbalance compensation in a direct conversion receiver
has been provided together with a receiver operating
using the method.
CA 02572236 2006-12-22
WO 2006/012245 PCT/US2005/022335
13
The method is gives a substantial improvement to
estimating the required imbalance compensation in
conditions where the received signal is subject to noise
and/or fading.
A look up table of initial amplitude imbalance values vs.
RF frequencies may be programmed in a memory associated
with the receiver, e.g. in a memory of a mobile station
in which the receiver is used. This may be for example
the so called codeplug which stores the operating
programs and data of the mobile station.
During use of the receiver, the amplitude imbalance as a
function of frequency will change gradually with time.
Information gathered by the processor 204 may be used to
update the stored information in the memory.
The invention gives improved audio performance in a
wireless terminal having a receiver operating on an FM
analogue signal in a direct conversion mode.