U.S. patent application number 11/165297 was filed with the patent office on 2006-12-28 for estimating bit error probability (bep) in an edge wireless system.
Invention is credited to Kaushik Ghosh.
Application Number | 20060291591 11/165297 |
Document ID | / |
Family ID | 36954907 |
Filed Date | 2006-12-28 |
United States Patent
Application |
20060291591 |
Kind Code |
A1 |
Ghosh; Kaushik |
December 28, 2006 |
Estimating bit error probability (BEP) in an edge wireless
system
Abstract
Distribution parameter mapping determines the bit error
probability (BEP) of a burst transmitted from a base station to the
mobile station using a modulation and coding scheme (MCS) specified
in the EDGE standard. Depending on whether the multi-bit soft
decisions of the burst most resemble a Gaussian or a Rician
distribution, the statistical parameters .mu. and .quadrature. or A
and .quadrature. are determined. The ratio .mu./.quadrature. or
A/.quadrature. is mapped to an empirically determined BEP in a
Gaussian or Rician lookup table, respectively. The BEPs are not
influenced by the degree of code redundancy in the MCS. The BEPs
for the four bursts in a radio block are then averaged, filtered
and quantized according to the EDGE standard. The quantization
level of the average BEP is reported to the base station so that
subsequent radio blocks can be transmitted using an MCS that is
appropriate for the estimated BEP.
Inventors: |
Ghosh; Kaushik; (San Diego,
CA) |
Correspondence
Address: |
QUALCOMM INCORPORATED
5775 MOREHOUSE DR.
SAN DIEGO
CA
92121
US
|
Family ID: |
36954907 |
Appl. No.: |
11/165297 |
Filed: |
June 22, 2005 |
Current U.S.
Class: |
375/340 ;
375/332 |
Current CPC
Class: |
H04L 1/203 20130101;
H04L 1/0003 20130101; H04L 1/206 20130101; H04L 1/0009 20130101;
H04L 1/0026 20130101 |
Class at
Publication: |
375/340 ;
375/332 |
International
Class: |
H04L 27/06 20060101
H04L027/06; H04L 27/22 20060101 H04L027/22 |
Claims
1. A method comprising: (a) receiving I and Q samples, wherein the
I and Q samples exhibit a bit error rate (BER); and (b) using
distribution parameter mapping to estimate the BER.
2. The method of claim 1, wherein the I and Q samples have been
modulated and coded with a modulation and coding scheme (MCS) that
conforms to a standard for Enhanced Data rates for GSM Evolution
(EDGE).
3. The method of claim 1, further comprising, between (a) and (b):
(c) demodulating the I and Q samples to obtain demodulated I and Q
samples; and (d) equalizing the demodulated I and Q samples to
obtain soft decision bits, wherein the soft decision bits have a
statistical distribution, and wherein the using the distribution
parameter mapping in (b) involves determining a type of the
statistical distribution.
4. The method of claim 3, wherein the type of the statistical
distribution is taken from the group consisting of: a Gaussian
distribution, a Rice distribution, a Rayleigh distribution, a
Poisson distribution and a Laplace distribution.
5. A method comprising: (a) equalizing demodulated I and Q samples
to obtain a plurality of multi-bit soft decisions, wherein the
demodulated I and Q samples exhibit a bit error probability (BEP),
wherein the plurality of multi-bit soft decisions has a
distribution, and wherein the distribution has a mean and a
variance; (b) determining a type of the distribution; (c)
calculating the mean and the variance of the distribution; and (d)
estimating the BEP based on the mean and the variance of the
distribution.
6. The method of claim 5, wherein calculating the mean and the
variance in (c) is performed based on the type of the
distribution.
7. The method of claim 5, wherein the type of the distribution is
taken from the group consisting of: a Gaussian distribution and a
Rician distribution.
8. The method of claim 5, further comprising: (e) deinterleaving
the plurality of multi-bit soft decisions; and (f) convolutionally
decoding the deinterleaved plurality of multi-bit soft decisions to
obtain single-bit hard decisions.
9. The method of claim 5, wherein the multi-bit soft decisions
comprise symbols, and wherein each symbol has three bits.
10. The method of claim 5, wherein a frame payload comprises the
plurality of multi-bit soft decisions.
11. The method of claim 5, wherein a radio block is comprised of
four pluralities of multi-bit soft decisions.
12. The method of claim 5, wherein the estimating the BEP in (d)
involves finding the BEP in a lookup table using a ratio equaling
the mean divided by the variance.
13. The method of claim 5, further comprising before (a): (e)
demodulating I and Q samples to obtain the demodulated I and Q
samples, wherein the demodulating involves a modulation scheme
taken from the group consisting of: Gaussian minimum shift keying
(GMSK) and octal phase shift keying (8-PSK).
14. The method of claim 5, further comprising: (e) equalizing
second demodulated I and Q samples to obtain a second plurality of
multi-bit soft decisions, wherein the second demodulated I and Q
samples exhibit a second BEP; (f) determining a mean BEP, wherein
the mean BEP is an average of a plurality of bit error
probabilities, and wherein the plurality of bit error probabilities
includes at least the BEP and the second BEP; and (g) filtering the
mean BEP to obtain a filtered mean BEP.
15. The method of claim 14, further comprising: (h) quantizing the
filtered mean BEP.
16. A processor-readable medium for storing instructions operable
in a wireless device to: (a) equalize demodulated I and Q samples
to obtain a plurality of multi-bit soft decisions, wherein the
demodulated I and Q samples exhibit a bit error probability (BEP),
wherein the plurality of multi-bit soft decisions has a
distribution, and wherein the distribution has a mean and a
variance; (b) determine a type of the distribution; (c) calculate
the mean and the variance of the distribution; and (d) estimate the
BEP based on the mean and the variance of the distribution.
17. The processor-readable medium of claim 16, wherein the mean and
the variance are calculated in (c) based on the type of the
distribution.
18. The processor-readable medium of claim 16, and further for
storing instructions operable in the wireless device to: (e)
deinterleave the plurality of multi-bit soft decisions; and (f)
convolutionally decode the deinterleaved plurality of multi-bit
soft decisions to obtain single-bit hard decisions.
19. The processor-readable medium of claim 16, wherein the BEP is
estimated in (d) by finding the BEP in a lookup table using a ratio
equaling the mean divided by the variance.
20. The processor-readable medium of claim 16, and further for
storing instructions operable in the wireless device to: (e)
demodulate I and Q samples to obtain the demodulated I and Q
samples, wherein the I and Q samples are demodulated using a
modulation scheme taken from the group consisting of: Gaussian
minimum shift keying (GMSK) and octal phase shift keying
(8-PSK).
21. The processor-readable medium of claim 16, and further for
storing instructions operable in the wireless device to: (e)
equalize second demodulated I and Q samples to obtain a second
plurality of multi-bit soft decisions, wherein the second
demodulated I and Q samples exhibit a second BEP; (f) determine a
mean BEP, wherein the mean BEP is an average of a plurality of bit
error probabilities, and wherein the plurality of bit error
probabilities includes at least the BEP and the second BEP; and (g)
filter the mean BEP to obtain a filtered mean BEP.
22. The processor-readable medium of claim 21, and further for
storing instruction operable in the wireless device to: (h)
quantize the filtered mean BEP.
23. A device comprising: (a) means for equalizing demodulated I and
Q samples to obtain a plurality of multi-bit soft decisions,
wherein the demodulated I and Q samples exhibit a bit error
probability (BEP), wherein the plurality of multi-bit soft
decisions has a distribution, and wherein the distribution has a
mean and a variance; (b) means for determining a type of the
distribution; (c) means for calculating the mean and the variance
of the distribution; and (d) means for estimating the BEP based on
the mean and the variance of the distribution.
24. The device of claim 23, wherein the means in (c) calculates the
mean and the variance based on the type of the distribution.
25. The device of claim 23, further comprising: (e) means for
deinterleaving the plurality of multi-bit soft decisions; and (f)
means for convolutionally decoding the deinterleaved plurality of
multi-bit soft decisions to obtain single-bit hard decisions.
26. The device of claim 23, further comprising: (e) means for
demodulating I and Q samples to obtain the demodulated I and Q
samples, wherein the means in (e) demodulates the I and Q samples
using a modulation scheme taken from the group consisting of:
Gaussian minimum shift keying (GMSK) and octal phase shift keying
(8-PSK).
27. The device of claim 23, further comprising: (e) means for
equalizing second demodulated I and Q samples to obtain a second
plurality of multi-bit soft decisions, wherein the second
demodulated I and Q samples exhibit a second BEP; (f) means for
determining a mean BEP, wherein the mean BEP is an average of a
plurality of bit error probabilities, and wherein the plurality of
bit error probabilities includes at least the BEP and the second
BEP; and (g) means for filtering the mean BEP to obtain a filtered
mean BEP.
28. A circuit comprising: a distribution analyzer that receives a
distribution of multi-bit soft decisions, wherein the distribution
of multi-bit soft decisions exhibits a distribution type, and
wherein the distribution analyzer determines the distribution type;
a bit error probability estimator that receives the distribution of
multi-bit soft decisions, wherein the bit error probability
estimator calculates statistical parameters of the distribution of
multi-bit soft decisions; and a lookup table, wherein the bit error
probability estimator determines a bit error probability (BEP) by
mapping the statistical parameters to the BEP in the lookup
table.
29. The circuit of claim 28, wherein the statistical parameters of
the distribution of multi-bit soft decisions approximate a
signal-to-noise ratio of the multi-bit soft decisions, and wherein
the lookup table correlates the statistical parameters of the
distribution of multi-bit soft decisions to bit error probabilities
of the multi-bit soft decisions.
30. The circuit of claim 28, wherein the distribution type is
Rician, wherein the statistical parameters include a mean (A) and a
variance (sigma), and wherein the bit error probability estimator
determines the BEP by mapping a quotient A/sigma to the BEP in the
lookup table.
31. The circuit of claim 28, wherein the distribution type is
Gaussian, wherein the statistical parameters include a mean (mu)
and a variance (sigma), and wherein the bit error probability
estimator determines the BEP by mapping a quotient mu/sigma to the
BEP in the lookup table.
32. The circuit of claim 28, wherein the distribution analyzer and
the bit error probability estimator are dedicated hardware in a
digital baseband processor.
33. The circuit of claim 28, wherein the bit error probability
estimator is a processor executing instructions stored on a
processor-readable medium.
34. The circuit of claim 28, further comprising: an equalizer that
outputs the distribution of multi-bit soft decisions.
35. The circuit of claim 28, further comprising: a convolutional
decoder that outputs hard decision bits based on the multi-bit soft
decisions.
36. A circuit comprising: an equalizer that receives demodulated I
and Q samples and outputs multi-bit soft decisions; and means for
estimating a bit error probability (BEP) based on the multi-bit
soft decisions.
37. The circuit of claim 36, wherein the means estimates the BEP
based on a statistical distribution of the multi-bit soft
decisions.
38. The circuit of claim 36, wherein the demodulated I and Q
samples are demodulated using a modulation and coding scheme (MCS)
that conforms to a standard for Enhanced Data rates for GSM
Evolution (EDGE).
Description
BACKGROUND
[0001] 1. Field
[0002] The present disclosure relates generally to wireless
communication devices and, more specifically, to a method for
estimating the bit error probability (BEP) in a wireless channel
between a base station and a mobile station.
[0003] 2. Background
[0004] As mobile telecommunications evolves, increasing speeds of
data transmission to mobile stations enables new types of services
to be offered to mobile subscribers. Usage of these services, in
turn, generates a demand for ever increasing data rates. The
European Telecommunications Standards Institute (ETSI) introduced
the General Packet Radio Service (GPRS) as an initial standard to
increase data rates by providing packet-switched data to mobile
stations based on the Global System for Mobile communications
(GSM). Then as an enhancement to GSM data services, ETSI
promulgated the Enhanced Data rates for GSM Evolution (EDGE)
standard, with a packet-switched portion called Enhanced GPRS
(EGPRS). Together, EDGE and EGPRS are described in the
TIA/EIA-136-370 standard published by the Telecommunications
Industry Association (TIA). Further enhancements to high-speed data
transmission based on GSM include the GSM/EDGE radio access network
(GERAN) standard specified by the 3.sup.rd Generation Partnership
Project (3GPP). The TIA has described the GERAN enhancements in the
TIA/EIA-136-370-A revision to its EGPRS-136 standard. For
simplicity, the EDGE, EGPRS, TIA/EIA-136-370 and TIA/EIA-136-370-A
standards are collectively referred to herein as the "EDGE
standard."
[0005] The physical layer dedicated to packet data traffic in the
EDGE standard is called the Packet Data Channel (PDCH). The
physical layer of the EDGE standard is specified in ETSI standard
TS 145.008 (3GPP TS 45.008). Both signaling and traffic channels
are transmitted over the PDCH. One of the signaling channels is the
Packet Associated Control Channel (PACCH). The traffic channel
transmitted over the PDCH is called the Packet Data Traffic Channel
(PDTCH).
[0006] Unlike basic GSM, several of the higher-speed versions of
GSM transmit data at multiple data rates. For example, data is
transmitted at nine different data rates over the PDTCH. In a
process called "link adaptation," the data rate over the wireless
channel is adjusted based on the channel condition. When the
channel condition is good and the signal-to-noise ratio of the
wireless channel is high, data can be transmitted at higher data
rates. Conversely, when the channel condition is poor and the
signal-to-noise ratio is low, data must be transmitted at slower
data rates. Transmitting data using a particular modulation and
coding scheme (MCS) at a data rate that is too high for the
channel's signal-to-noise ratio can result in a loss of data. Link
adaptation increases overall data throughput by using the highest
data rate that can dependably be supported using a particular MCS
at the signal-to-noise ratio that momentarily exists on the
wireless channel. The EDGE standard requires the mobile station
periodically to report the channel condition in the PACCH to the
base station. The condition of the channel between the base station
and the mobile station is expressed in terms of the bit error
probability (BEP). The BEP is the expected value of the actual Bit
Error Rate (BER) of a signal received by the mobile station over
the wireless channel. The base station then transmits data in the
PDTCH to the mobile station at the appropriate data rate depending
on the channel condition as indicated in the PACCH.
[0007] Link adaptation can most effectively be performed when the
mobile station reports a BEP that most accurately estimates the
actual BER. One way to estimate the BEP is to attempt to calculate
the BER itself. A "re-encoding" method is based on determining the
number of bit errors that are corrected in the decoding process.
Error control decoding, such as that performed by a convolutional
decoder, attempts to correct bit errors that are introduced in the
wireless channel. Frames that are output from the block
deinterleaver and the convolutional decoder of the mobile station
are re-encoded and re-interleaved. The resulting re-encoded bits
are then compared to the bits received by the block deinterleaver
to determine the number of corrected bit errors. The re-encoding
method, however, yields inaccurate results because it relies on the
assumption that the error control decoding corrects all of the
errors that have been introduced by the wireless channel.
Therefore, the BEP obtained using the re-encoding method varies
depending on the degree of redundancy employed by the various MCS
schemes used to transmit the bits over the wireless channel. Even
with a poor channel condition, a high redundancy level of the data
allows the error control decoding to decode all of the bits
correctly and thus yields a more accurate estimated BER. On the
other hand, if the channel condition is poor and redundancy level
of the data is low, the error control decoding is unable to correct
all of the erroneous bits, and an inaccurate estimate of the BER
results. Thus, a better channel quality is required to estimate the
BER accurately using a lower redundancy MCS scheme, such as MCS9,
than using a higher redundancy MCS scheme, such as MCS5.
[0008] FIG. 1 (prior art) compares the estimated BEP obtained using
the re-encoding method on data from two channels modulated with
different MCSs at different redundancy levels of the data. Less
error is introduced by the channel modulated with a higher
redundancy code. A curve 10 shows the relationship between the
signal-to-noise ratio and the BEP of a channel modulated with
Gaussian minimum shift keying (GMSK) at a redundancy level of 1.89.
Another curve 11 shows the relationship between the signal-to-noise
ratio and the BEP of a channel modulated with GMSK at a redundancy
level of 1.0. The re-encoding method indicates that at higher noise
levels the BEP of the channel modulated at a redundancy level of
1.0 is lower and thus less accurate than the BEP of the channel
modulated at a redundancy level of 1.89. Thus, the estimated BEP at
a given signal-to-noise ration is not independent of the redundancy
level of the data, as required by the EDGE specification.
[0009] FIG. 2 (prior art) compares the BEP obtained using the
re-encoding method on data transmitted at three different
redundancy levels and modulated with octal phase shift keying
(8-PSK). A curve 12 shows the relationship between the
signal-to-noise ratio and the BEP for a channel with a redundancy
level of 2.70. A curve 13 shows the relationship between the
signal-to-noise ratio and the BEP for a channel with a redundancy
level of 1.32. A curve 14 shows the relationship between the
signal-to-noise ratio and the BEP for a channel with a redundancy
level of 1.0. Curves 12-14 show that the re-encoding method
inaccurately indicates that the BEP decreases, and the channel
condition improves, as the redundancy level decreases.
[0010] A second way of estimating the BEP involves first measuring
the signal-to-noise ratio of the radio frequency (RF) signal that
carries the PDCH. The relationship between the measured
signal-to-noise ratio and the BER of the PDCH received by the
mobile station is empirically determined in a laboratory. The
values of BER that vary as a function of the measured
signal-to-noise ratio are then stored in a lookup table on the
mobile station. This method requires the mobile station to have an
estimator of the signal-to-noise ratio in the RF signal. The BEP is
determined by using the estimated signal-to-noise ratio to look up
the corresponding BER in the lookup table. The accuracy of the BEP
in this method depends on the accuracy of the estimated
signal-to-noise ratio of the RF signal. Where the channel condition
is affected by signal interference and fading, an accurate
determination of the signal-to-noise ratio of the RF signal can be
difficult, and the BEP estimation is prone to inaccuracy.
[0011] A method is sought for accurately determining the bit error
probability (BEP) without requiring a direct estimation of the
signal-to-noise ratio of the RF signal and without re-encoding the
output of the convolutional decoder of the mobile station.
Moreover, a method is sought for determining the BEP that is not
influenced by the degree of redundancy in the modulation and coding
scheme (MCS) used to transmit the data over the wireless
channel.
SUMMARY
[0012] A distribution parameter mapping method estimates the bit
error probability (BEP) of bits in a burst transmitted in a radio
frequency (RF) signal from a base station to a mobile station using
one of the nine modulation and coding schemes (MCSs) specified in
the EDGE standard. The BEP estimated using the distribution
parameter mapping method is not influenced by the degree of code
redundancy in the particular MCS used to modulate data over the RF
signal. The circuitry determines whether the multi-bit soft
decisions that were equalized from demodulated I and Q samples of
the burst most resemble a Gaussian distribution or a Rician
distribution. The statistical parameters for the mean (.mu.) and
the variance (.sigma.) are determined for soft decisions having a
Gaussian distribution. The statistical parameters A and .sigma. are
determined for soft decisions having a Rician distribution. The
signal-to-noise ratio of the RF signal is represented by the ratio
.mu./.sigma. for a Gaussian distribution of soft decisions and by
the ratio A/.sigma. for a Rician distribution of soft decisions.
The BEP for a burst having a Gaussian distribution of soft
decisions is determined by mapping the ratio .mu./.sigma. to an
empirically determined BEP in a Gaussian lookup table stored in
non-volatile memory on the mobile station. For a Rician
distribution, the ratio A/.sigma. is mapped to an empirically
determined BEP in a Rician lookup table. The estimated BEPs for the
four bursts of each radio block are then averaged, filtered and
quantized into one of thirty-two levels according to the EDGE
standard. The quantization level of the average BEP is then
reported to the base station to permit the base station to transmit
subsequent radio blocks using an MCS that is appropriate for the
estimated BEP of the signal.
[0013] Circuitry in a mobile station that performs distribution
parameter mapping to estimate the BEP includes an equalizer, a
distribution analyzer, a BEP estimator, lookup tables, an averager,
a filter and a non-linear quantizer. The equalizer removes
intersymbol interference from demodulated I and Q samples received
in bursts from a demodulator in the mobile station. For each burst,
the equalizer outputs a distribution of multi-bit soft decisions
that are subsequently processed by the mobile station into
single-bit hard decisions that comprise frames of data. The
distribution analyzer receives the distribution of multi-bit soft
decisions from the equalizer and determines the type of
distribution that the distribution of multi-bit soft decisions
resembles. For example, the distribution of multi-bit soft
decisions can resemble a Gaussian distribution or a Rician
distribution. The distribution analyzer outputs a distribution type
identifier.
[0014] The BEP estimator receives the distribution of multi-bit
soft decisions from the equalizer, as well as the distribution type
identifier from the distribution analyzer. The BEP estimator
calculates various statistical parameters of the distribution of
multi-bit soft decisions, depending on the type of distribution.
When the soft decisions have a Gaussian distribution, the BEP
estimator calculates the statistical parameters for the mean (.mu.)
and the variance (.sigma.). When the soft decisions have a Rician
distribution, the BEP estimator calculates the statistical
parameters A and .sigma.. The BEP estimator also calculates the
ratio .mu./.sigma. for a Guassian distribution and the ratio
A/.sigma. for a Rician distribution. The ratios .mu./.sigma. and
A/.sigma. correlate to the signal-to-noise ratios of the I and Q
samples.
[0015] The BEP estimator estimates the BEP of a burst containing a
Gaussian distribution of multi-bit soft decisions by mapping the
ratio .mu./.sigma. to an empirically determined BEP in a Guassian
lookup table stored on the mobile station. The BEP of a burst
containing a Rician distribution of multi-bit soft decisions is
estimated by mapping the ratio A/.sigma. to an empirically
determined BEP in a Rician lookup table stored on the mobile
station.
[0016] The averager then averages the estimated BEPs from four
bursts and generates a MEAN_BEP. The filter filters the MEAN_BEP
and outputs a filtered MEAN_BEP. The non-linear quantizer quantizes
the filtered MEAN_BEP into one of thirty-two levels and outputs a
value (MEAN_BEP_0 through MEAN_BEP_31) that represents the BEP of
the four bursts on a logarithmic scale.
[0017] Other embodiments and advantages are described in the
detailed description below. This summary does not purport to define
the invention. The invention is defined by the claims.
BRIEF DESCRIPTION OF THE DRAWINGS
[0018] The accompanying drawings, where like numerals indicate like
components, illustrate embodiments of the invention.
[0019] FIG. 1 (prior art) is a diagram plotting bit error
probability (BEP) obtained using a re-encoding method at various
signal-to-noise ratios of data modulated with two different GMSK
modulation and coding schemes (MCSs), each with a different code
redundancy level;
[0020] FIG. 2 (prior art) is a diagram plotting BEP obtained using
the re-encoding method at various signal-to-noise ratios of data
modulated with three different 8-PSK MCSs, each with a different
code redundancy level;
[0021] FIG. 3 is a simplified block diagram of circuitry that
determines BEP using distribution parameter mapping;
[0022] FIG. 4 is a flowchart of steps for performing the
distribution parameter mapping employed by the circuitry of FIG.
3;
[0023] FIG. 5 is a table listing the data transmission rates of
nine MCSs specified in the EDGE standard;
[0024] FIG. 6 is a diagram of equations showing the derivation of
statistical parameters of Gaussian and Rician distributions used in
distribution parameter mapping;
[0025] FIG. 7 is a flowchart illustrating certain steps of the
distribution parameter mapping method of FIG. 4 used to obtain a
quantized, filtered, average BEP for four bursts of a radio
block;
[0026] FIG. 8 is a diagram plotting BEP at various signal-to-noise
ratios obtained using distribution parameter mapping for a static
channel modulated with MCS4 (a GMSK scheme);
[0027] FIG. 9 is a diagram plotting quantized, average BEP at
various signal-to-noise ratios obtained from groups of four
consecutive BEP values of FIG. 8;
[0028] FIG. 10 is a diagram of the probability at various
signal-to-noise ratios that the quantized, average BEP of FIG. 9
falls within a correct quantization level;
[0029] FIG. 11 is a diagram plotting BEP at various signal-to-noise
ratios obtained using distribution parameter mapping for a static
channel modulated with MCS9 (an 8-PSK scheme);
[0030] FIG. 12 is a diagram plotting BEP values at various
signal-to-noise ratios obtained using distribution parameter
mapping for a static channel modulated with MCS9, wherein the BEP
values represent an average of several BEP values that were most
prevalent among the many bursts over which BEP values were
estimated at each signal-to-noise ratio;
[0031] FIG. 13 is a diagram plotting quantized, average BEP at
various signal-to-noise ratios obtained from groups of four
consecutive BEP values of FIG. 11;
[0032] FIG. 14 is a diagram of the probability at various
signal-to-noise ratios that the quantized, average BEP of FIG. 13
falls within a correct quantization level; and
[0033] FIG. 15 is a diagram plotting BEP at various signal-to-noise
ratios obtained using distribution parameter mapping for a fading
channel modulated with MCS9.
DETAILED DESCRIPTION
[0034] Reference will now be made in detail to some embodiments of
the invention, examples of which are illustrated in the
accompanying drawings.
[0035] FIG. 3 is a simplified block diagram of circuitry 20 in a
mobile station that performs distribution parameter mapping to
determine a bit error probability (BEP). The BEP is an estimate of
the bit error rate (BER) of a Packet Data Channel (PDCH)
transmitted over a radio frequency (RF) signal from a base station
to a mobile station using various modulation and coding schemes
(MCSs) that conform to the EDGE standard.
[0036] FIG. 4 is a flowchart showing steps of the by which
circuitry 20 uses distribution parameter mapping to determine the
BEP. The distribution parameter mapping method is not influenced by
the degree of redundancy in the MCS used to modulate data over the
RF signal. The operation of individual elements of circuitry 20, as
shown in FIG. 3, is explained in detail in connection with the
steps listed in FIG. 4. In an initial step 21, an input RF signal
22 is received by an antenna 23 of the mobile station that contains
circuitry 20. In a step 24, an RF receiver 25 converts input RF
signal 22 to digital in-phase (I) and quadrature (Q) samples 26 for
subsequent digital baseband processing. In the embodiment of FIG.
3, the digital baseband layer-1 processing is performed by a
digital baseband processor 27. Digital baseband processor 27 is
part of a digital mobile station modem 28. RF receiver 25 is
incorporated into an RF analog chip 29 that is separate from
digital mobile station modem 28.
[0037] The BEP determined by circuitry 20 is an indication of the
channel condition of the PDCH transmitted over input RF signal 22.
The EDGE physical layer specification (ETSI standard TS 145.008;
3GPP standard TS 45.008) provides that the mobile station
periodically reports the channel condition of the PDCH to the base
station in the Packet Associated Control Channel (PACCH). The base
station polls the mobile station for the channel condition. The
PACCH is transmitted back to the base station over an output RF
signal 30. The mobile station uses the BEP to obtain the channel
condition that is reported to the base station. The channel
condition is expressed as one of thirty-two BEP levels. The base
station then transmits data in the PDTCH over the PDCH back to the
mobile station at the appropriate data rate depending on the BEP
level indicated in the PACCH.
[0038] Depending on the BEP level, data is transmitted at nine
different data rates in the EDGE standard. FIG. 5 lists nine MCSs
that are associated with the nine data rates. The first four MCSs
(MCS1-MCS4) employ the Gaussian minimum shift keying (GMSK)
modulation used by basic GSM. The major enhancement to the GSM
standard to support higher data rates was the introduction in the
EDGE standard of a higher-level modulation technique, known as
octal phase shift keying (8-PSK). The highest five MCSs (MCS5-MCS9)
use 8-PSK modulation. The EDGE standard describes a narrowband
system that uses a combination of frequency division multiple
access (FDMA) and time division multiple access (TDMA). The
frequency band that is allocated to EDGE transmissions is first
divided into various 200-kHz carrier signals. FIG. 5 lists the data
rates achievable with the listed modulation and coding schemes when
using a single 200-kHz carrier and one timeslot. The data rates can
be increased by simultaneously using multiple 200-kHz carriers, for
example, six carriers. The carrier signals are then modulated and
transmitted over an RF signal, such as input RF signal 22 and
output RF signal 30. Each carrier signal is divided into eight
timeslots. The data rates can be further increased by using
multiple timeslots, for example, all eight timeslots. EDGE provides
for the transmission of packet-switched data. Each packet is
composed of frames and includes a data message and control
information. Each frame in turn is transmitted as a burst during an
appropriate timeslot. The frames are transmitted over the carrier
signal in radio blocks. Each radio block is four frames transmitted
as a sequence of four bursts. Each burst is 4.615 ms, and each
radio block is 20 ms.
[0039] The first four MCSs have different coding schemes that
provide for nearly no coding (MSC4) to highly redundant coding
(MSC1). The code rate listed in FIG. 5 is the inverse of the code
redundancy. A higher code redundancy allows data to be recognized
despite channel fading, but results in a lower data rate. For
example, MSC1 has a data rate of 9.05 kbps per channel, and MSC4
has a data rate of 21.4 kbps per channel. By dynamically decreasing
code redundancy during periods of lower fading and noise, a higher
network performance can be achieved. Adapting the code redundancy
and modulation technique to maximize throughput depending on the
channel condition is called "Link Adaptation."
[0040] The highest five MCSs support higher data rates because
8-PSK signals are able to carry three bits per modulated symbol
instead of one bit per symbol with GMSK modulation. Thus, the data
rates of the MCSs employing 8-PSK are approximately three times as
fast. Signal propagation using 8-PSK is diminished, however, in
comparison to GMSK. The coverage area achieved with signals
employing the higher data rates of 8-PSK modulation is therefore
smaller.
[0041] In one mode of link adaptation, the mobile station reports
the BEP level based on the mean BEP for each of the eight timeslots
in a temporary block flow (TBF). The method of FIG. 4 describes
determining a BEP level based on the mean BEP for a particular
timeslot. In a second mode of link adaptation, the mobile station
reports the BEP level based on the mean BEP for the modulation for
which the mobile station has received the largest number of radio
blocks since the previous message. The BEP level is based on the
mean and the coefficient of variation of the BEP measurements for
the primary modulation averaged over all of the timeslots in the
TBF. The EDGE standard provides that a single MCS is used for all
the timeslots allocated to a carrier of a TBF based on the
collective channel condition measurements for all of the
timeslots.
[0042] Digital baseband processor 27 receives the I and Q samples
26 from the RF receiver 25 and outputs frames containing single-bit
hard decisions 31. The single-bit hard decisions 31 are output by a
convolutional decoder 32, such as a Viterbi decoder. The frames are
processed as data or are analyzed as speech in a voice decoder.
Circuitry 20 estimates the signal-to-noise ratio of the PDCH
transmitted over input RF signal 22 without re-encoding the output
of convolutional decoder 32. Circuitry 20 instead analyzes
multi-bit soft decisions 33 that are generated as part of the
digital baseband layer-1 processing to estimate the signal-to-noise
ratio of the PDCH.
[0043] In a step 34, a modulation detector 35 receives the I and Q
samples 26 from RF receiver 25 and determines the type of
modulation scheme by which data was modulated over the carrier
signal on input RF signal 22. According to the EDGE standard, the
modulation scheme is either GMSK or 8-PSK. A detection algorithm is
used to differentiate I and Q samples modulated with either GMSK or
8-PSK based on the different phase characteristics of the GMSK and
8-PSK modulations. One detection method, for example, first assumes
that the data is modulated with GMSK and then performs a
.quadrature.-by-4 rotation. A signal-to-noise ratio is then
estimated for this GMSK hypothesis. A rotation is then performed
assuming that the data is modulated with 8-PSK, and the
signal-to-noise ratio is again estimated. The method determines
that the modulation scheme corresponds to the modulation hypothesis
for which the signal-to-noise ratio was the greatest.
[0044] In a step 36, the I and Q samples 26 are then demodulated.
Depending on the modulation scheme identified in step 34, the I and
Q samples 26 are demodulated by either a GMSK demodulator 37 or an
8-PSK demodulator 38. A GMSK demodulator 37 demodulates I and Q
samples 26 that were modulated with MCS1 through MCS4, which employ
GMSK. An 8-PSK demodulator 38 demodulates I and Q samples 26 that
were modulated with MCS5 through MCS9, which employ 8-PSK. In the
embodiment of FIG. 3, GMSK demodulator 37 and 8-PSK demodulator 38
are dedicated hardware within digital baseband processor 27. In
other embodiments, the GMSK and 8-PSK demodulation performed by
GMSK demodulator 37 and 8-PSK demodulator 38 is performed by a
digital signal processor or a microcontroller that are part of
digital baseband processor 27.
[0045] The demodulated I and Q samples 41 output by GMSK
demodulator 37 and the demodulated I and Q samples 42 output by
8-PSK demodulator 38 constitute symbols in baseband. Depending on
the modulation scheme, a demodulated sample can have various number
of bits, for example, 1, 2 or 10. The demodulated samples represent
positive and negative numbers in GMSK and real and imaginary
numbers in 8-PSK. There are one in-phase sample and one quadrature
sample per symbol bit. In GMSK, there are 116 symbols in each of
the four bursts of a radio block. In 8-PSK, there are 348 symbols
(3.times.116) per burst.
[0046] In a step 39, an equalizer 40 equalizes demodulated I and Q
samples 41 and 42 and outputs the multi-bit soft decisions 33.
Thus, each I and Q sample bit is assigned a multi-bit soft decision
value. The multi-bit soft decisions 33 constitute symbols for which
inter-symbol interference has been removed. Inter-symbol
interference results when one symbol is temporally modulated on top
of another symbol. In one example, each of the multi-bit soft
decisions 33 is a 16-bit 2's complement signed digital value.
[0047] Circuitry 20 estimates the BEP based on the multi-bit soft
decisions 33. The multi-bit soft decisions 33 are also further
processed by digital baseband processor 27 to obtain the single-bit
hard decisions 31 that are included in the frames that contain
voice and data information. A quantizer 41 quantizes the multi-bit
soft decisions 33 into a lesser number of levels than the number of
digital states available from the number of bits of the multi-bit
soft decisions 33. A block deinterleaver 42 receives quantized
symbols 43 from quantizer 41 and output deinterleaved symbols 44.
The convolutional decoder 32 than decodes the deinterleaved symbols
44 and outputs the single-bit hard decisions 31.
[0048] Returning to the distribution parameter mapping method of
estimating the BEP, circuitry 20 next determines the type of
statistical distribution of the multi-bit soft decisions 33. In a
step 45, a distribution analyzer 46 determines the type of
statistical distribution to which the soft decisions 33 of each
burst correspond. Distribution analyzer 46 then outputs a
corresponding distribution type identifier 47. For example, the
distribution of the values of the multi-bit soft decisions 33 may
resemble one of the following distribution types: a Gaussian
distribution, a Rice (Rician) distribution, a Rayleigh
distribution, a Poisson distribution or a Laplace distribution. The
distribution of the multi-bit soft decisions 33 typically resembles
either a Gaussian or a Rician distribution. In a static channel
where the signal-to-noise ratio is not significantly improving or
deteriorating, the distribution of the multi-bit soft decisions 33
typically resembles a Gaussian distribution. On the other hand, if
there is a line of sight path between the base station and the
mobile station, the wireless channel is usually described by the
Rician fading model, and the distribution of the multi-bit soft
decisions 33 typically resembles a Rician distribution.
Distribution analyzer 46 uses well-known algorithms to determine
the statistical distribution type that the distribution of the
multi-bit soft decisions 33 most closely resembles. For example,
the type of distribution can be recognized by the maximum value of
the distribution, the location of the maximum value within the
distribution, and the spread of the distribution.
[0049] A BEP estimator 48 receives the soft decisions 33 for each
burst that are output by equalizer 40. In addition, BEP estimator
47 receives distribution type identifier 47. In a decision step 49,
BEP estimator 48 determines which statistical parameters to
calculate. If the distribution type identifier 47 indicates that
the soft decisions 33 resemble a Gaussian distribution, BEP
estimator 48 proceeds to a step 50 and calculates the statistical
parameters .mu. (mu) and .sigma. (sigma). If the distribution type
identifier 47 indicates that the soft decisions 33 resemble a
Rician distribution, BEP estimator 48 proceeds to a step 51 and
calculates the statistical parameters A and .sigma..
[0050] In the following example of step 50, the statistical
parameters .mu. and .sigma. are calculated from soft decisions
whose distribution is found to resemble a Gaussian distribution in
decision step 49. Thus, the distribution of the soft decisions
resembles the Gaussian probability density function (PDF) 52 shown
in FIG. 6. In Gaussian PDF 52, .mu. represents the mean, and
.sigma. represents the variance of the distribution p(x). In this
example, each of the multi-bit soft decisions 33 output by
equalizer 40 is a 4-bit 2's complement signed digital value. There
are 116 soft decisions in one burst because the soft decisions 33
were equalized from I and Q samples modulated with GMSK. The 116
values are as follows: 15.times.[1100]; 30.times.[1101];
15.times.[1110]; 15.times.[0000]; 30.times.[0001]; 11.times.[0010],
where [1100]=-4; [1101]=-3; [1110]=-2; [0000]=0; [0001]=1; and
[0010]=2. The statistical parameters .mu. and .sigma. are
calculated by first determining the second and fourth moments of
the Gaussian DPF for the sample distribution. The second moment is
defined as the sum of the each element squared, divided by the
number of elements in the distribution. The fourth moment is
defined as the sum of the each element to the fourth power, divided
by the number of elements in the distribution. For the sample
distribution of 116 soft decisions listed above, the second moment
is 5.552, and the fourth moment is 57.897. The second and fourth
moments can also be expressed in terms of the mean (.mu.) and the
variance (.sigma.).
[0051] FIG. 6 shows an equation 53 for the second moment and an
equation 54 for the fourth moment, each expressed in terms of .mu.
and .sigma.. The mean (.mu.) and the variance (.sigma.) are
determined by solving these two equations in two variables. An
equation 55 expresses .mu. in terms of the second and fourth
moments. An equation 56 expresses .sigma. in terms of the second
and fourth moments. For the sample distribution of 116 soft
decisions listed above, .mu. is determined to be 2.039, and a is
determined to be 1.181.
[0052] Returning to the next step in FIG. 4, the BEP is determined
in a step 57 by mapping the quotient .mu./.sigma. to a BEP value in
a lookup table. The quotient of the mean (.mu.) divided by the
variance (.sigma.) is indicative of the signal-to-noise ratio of
the data that comprise a distribution. For the sample Gaussian
distributions the quotient .mu./.sigma. is 1.727. The relationship
between the quotient .mu./.sigma. and the BER for channels whose
data resembles a Gaussian distribution is empirically determined in
a laboratory. The results are then stored in a Gaussian lookup
table 58 in a processor-readable medium 59, as shown in FIG. 3. The
lookup table is then used to estimate the BEP based on the
signal-to-noise ratio estimated by the quotient .mu./.sigma.. BEP
estimator 48 determines a BEP value 60 for each distribution of
multi-bit soft decisions 33 of a burst. For the signal-to-noise
ratio of 1.727 of the sample Gaussian distribution, BEP value 60 is
determined to be 0.050.
[0053] In a decision step 61, circuitry 20 determines whether the
BEP value 60 of each of the four bursts in the radio block has been
determined. If four BEP values have not yet been determined, BEP
estimator 48 determines the BEP for the next distribution of 116
soft decisions on the next GMSK burst. Where the burst has been
modulated with 8-PSK, BEP estimator 48 determines the BEP for a
distribution comprising 348 soft decisions per burst.
[0054] Returning to step 51, the statistical parameters A and
.sigma. are calculated from the sample distribution of soft
decisions listed above assuming that the distribution is found to
resemble a Rician distribution in decision step 49. Thus, in this
example, the sample distribution is found to resemble the Rician
probability density function (PDF) 62 shown in FIG. 6. The
statistical parameters A and a are calculated by first determining
the second and fourth moments of the Rician DPF for the sample
distribution. The values of the second and fourth moments of a
distribution do not change when the distribution is characterized
as resembling a different type of distribution. Therefore, the
values of the second and fourth moments of the Rician distribution
are the same as calculated above for the Gaussian distribution.
[0055] FIG. 6 shows an equation 63 for the second moment and an
equation 64 for the fourth moment, each expressed in terms of A and
.sigma.. These two equations in two variables are then solved to
obtain an equation 65 expressing A in terms of the second and
fourth moments. In addition, an equation 66 expresses .sigma. in
terms of the second and fourth moments. Assuming that the sample
distribution of 116 soft decisions listed above resembles a Rician
distribution, A is determined to be 1.391, and .sigma. is
determined to be 1.345.
[0056] In a step 67, the BEP is then determined by mapping the
quotient A/.sigma. to a BEP value in a lookup table. For the sample
Rician distribution, the quotient A/.sigma. is 1.035. The
relationship between the quotient A/.sigma. and the BER for
channels whose data resembles a Rician distribution is also
empirically determined in a laboratory. The results of the
empirical determination are then stored in a Rician lookup table 68
in processor-readable medium 59. Rician lookup table 68 is then
used to estimate the BEP based on the quotient A/.sigma.. Where the
quotient A/.sigma. of the sample Rician distribution equals 1.035
in this example, BEP value 60 is determined to be 0.079.
[0057] In a step 69, an averager 70 calculates the average of four
BEP values 60 when circuitry 20 determines in decision step 61 that
the BEP of each of the four bursts in a radio block has been
determined. Averager 70 outputs a signal MEAN_BEP 71 that
represents the average of the four BEP values 60.
[0058] In a step 72, a filter 73 receives and filters the MEAN_BEP
71. Filter 73 is a digital low pass filter, such as an infinite
impulse response (IIR) filter. Filter 73 outputs a filtered
MEAN_BEP 74.
[0059] In a step 75, a non-linear quantizer 76 quantizes the
filtered MEAN_BEP 74 into one of thirty-two non-linear levels or
intervals. Non-linear quanitizer 76 outputs one of thirty-two
values MEAN_BEP_0 through MEAN_BEP_31 (77) that represents the
average, filtered BEP on a logarithmic scale. The quantized
MEAN_BEP 77 is then received by an RF transmitter 78 on RF analog
chip 29. In one embodiment, most of the circuitry of digital
baseband processor 27 is part of a digital signal processor (DSP)
79, including distribution analyzer 46, BEP estimator 48, averager
70, filter 73 and non-linear quantizer 76.
[0060] In a step 80, the quantized MEAN_BEP 77
(MEAN_BEP_0--MEAN_BEP_31) of the level of the average BEP is
transmitted back to the base station in PACCH over output RF signal
30. The base station then transmits subsequent radio blocks using
an MCS that is chosen based on the quantized MEAN_BEP 77. For
example, the base station chooses the MCS with the fastest data
rate that can be supported under the channel condition described by
the quantized MEAN_BEP 77.
[0061] FIG. 7 is a flowchart illustrating the various steps
performed by circuitry 20 to obtain the quantized MEAN_BEP 77 for a
radio block. FIG. 7 shows that the steps 50 and 57 (for GMSK) and
the steps 51 and 67 (for 8-PSK) are performed for each of four
bursts of a radio block, whereas the steps 69 (averaging), 72
(filtering) and 75 (quantizing) are performed only once per radio
block.
[0062] FIG. 8 shows the results of using distribution parameter
mapping to determine BEP values for a channel modulated with MCS4
at signal-to-noise ratios ranging from -6 dB to 10 dB. The BEP
values are estimated from bursts transmitted over a static channel
with a constant signal strength exhibiting no fading. Thus, the
distribution of the multi-bit soft decisions 33 used to derive the
BEP values resembles a Gaussian distribution. The BEP values are
obtained using the method of FIG. 4 through step 50, and the BEP
values 60 are determined by mapping the ratio .mu./.quadrature. to
BEP values in the Gaussian lookup table 58. A curve 81 shows the
actual bit error rate (BER) of the channel over the range of
signal-to-noise ratios from -6 dB to 10 dB. The actual BER is
determined by transmitting a known bit sequence over thousands of
radio blocks and comparing the bits from the demodulated I and Q
samples to the known bit sequence. A curve 82 shows the estimated
BEP value 60 obtained at each signal-to-noise ratio using
distribution parameter mapping. The estimated BEP value 60 plotted
in FIG. 8 for each signal-to-noise ratio is the BEP value that
resulted the most number of times from among the thousands of
bursts over which the known bit sequence was transmitted.
[0063] FIG. 9 shows the values of the quantized MEAN_BEP 77
obtained from groups of four consecutive BEP values 60 of FIG. 8.
At lower signal-to-noise ratios, the quantized MEAN_BEP 77 is
assigned a value closer to zero. At higher signal-to-noise ratios,
the quantized MEAN_BEP 77 is assigned a value closer to thirty-two.
A curve 83 shows the estimated, quantized average BEP values
obtained using distribution parameter mapping. A curve 84 shows the
values of the quantization levels that would be output using
demodulated I and Q samples that exhibit the actual BER.
[0064] FIG. 10 shows the probability, at signal-to-noise ratios
from -6 dB to 10 dB, that the MEAN_BEP value 71 will be correctly
determined and reported to the base station as the correct
quantization level. The EDGE standard specifies how to test the
circuitry that generates the values of the quantization levels. The
test requires that a certain percentage of the quantized MEAN_BEP
values 77 reported by the mobile station fall within a narrow range
of correct quantization levels, for example, three quantization
levels. For example, at a signal-to-noise ratio of 5 dB, at least
65% of the quantized MEAN_BEP values 77 must fall within one of the
quantization levels MEAN_BEP_11, MEAN_BEP_12 and MEAN_BEP_13 in
order to pass the test. A dotted curve 85 shows the minimum
probability of achieving an acceptable quantization level when
estimating the BEP of a channel modulated with MCS1 through MCS4
(GMSK) in order to comply with the EDGE standard. A curve 86 shows
the probability that a quantized MEAN_BEP 77 obtained using
distribution parameter mapping falls within an acceptable
quantization level.
[0065] FIG. 11 shows the results of using distribution parameter
mapping to determine BEP values obtained from a channel modulated
with MCS9 employing 8-PSK. The BEP values are estimated from bursts
transmitted over a static channel exhibiting no fading. As in FIG.
8, the BEP values 60 are determined by mapping the ratio
.mu./.quadrature. to BEP values in the Gaussian lookup table 58. A
curve 87 shows the actual BER of the channel over a range of
signal-to-noise ratios from -1 dB to 20 dB. A curve 88 shows the
BEP value 60 obtained at each signal-to-noise ratio using
distribution parameter mapping. The estimated BEP values 60 plotted
in FIG. 11 for each signal-to-noise ratio is the BEP value obtained
the most number of times from the thousands of bursts over which
the known bit sequence was transmitted.
[0066] FIG. 12 shows BEP values obtained from a channel modulated
with MCS9 over the same signal-to-noise ratio as in FIG. 11. The
BEP values and the values of the actual BER, however, fluctuate
over different bursts at each signal-to-noise ratio. The BEP values
and the values of the actual BER plotted in FIG. 12 represent an
average of the three BEP or BER values that were most prevalent in
the multiple bursts over which the known bit sequence was
transmitted at a particular signal-to-noise ratio. A curve 89 shows
the actual BER at each signal-to-noise ratio from -1 dB to 20 dB. A
curve 90 shows the BEP value obtained at each signal-to-noise ratio
using distribution parameter mapping.
[0067] FIG. 13 shows the values of the quantized MEAN_BEP 77
obtained from the BEP values 60 of FIG. 11. The values of the
quantized MEAN_BEP 77 range from zero to thirty-two. A curve 91
shows the estimated, quantized average BEP values obtained using
distribution parameter mapping. A curve 92 shows the quantization
levels obtained from the values of the actual BER.
[0068] FIG. 14 shows the probability at signal-to-noise ratios from
-1 dB to 20 dB that the MEAN_BEP value 71 will be correctly
determined from a channel and reported to the base station as a
correct quantization level. A dotted curve 93 shows the minimum
probability that must be achieved at each signal-to-noise ratio to
comply with the EDGE standard. Dotted curve 93 applies to
quantization levels obtained from average BEP values from channels
modulated with MCS5 through MCS9 (8-PSK). A curve 94 shows the
probability that a quantized MEAN_BEP 77 obtained using
distribution parameter mapping is at the correct quantization level
using the test specified in the EDGE standard.
[0069] FIG. 15 shows the results of using distribution parameter
mapping to determine BEP values obtained from a channel modulated
with MCS9. Unlike the results shown in FIG. 11, the BEP values in
FIG. 15 are estimated from bursts transmitted over a fading
channel. The channel analyzed in FIG. 15 is a TU50 channel, which
is a channel in a typical urban environment where the mobile
station is moving at 50 km/hr. Thus, the distribution of the
multi-bit soft decisions 33 used to derive the BEP values resembles
a Rician distribution. The BEP values are obtained using the method
of FIG. 4 through step 51, and the BEP values 60 are determined by
mapping the ratio A/.quadrature. to BEP values in the Rician lookup
table 68. A curve 95 shows the actual BER of the channel over a
range of signal-to-noise ratios from -1 dB to 27 dB. A curve 96
shows the BEP value 60 obtained at each signal-to-noise ratio using
distribution parameter mapping. At signal-to-noise ratios above
about 7 dB, the BER of the fading channel in FIG. 15 is greater
than the BER of the static channel in FIG. 11, where both channels
are modulated using MCS9. The estimated BEP values 60 plotted in
FIG. 15 for each signal-to-noise ratio is the BEP value obtained
the most number of times from the thousands of bursts over which
the known bit sequence was transmitted.
[0070] Although the present invention has been described in
connection with certain specific embodiments for instructional
purposes, the present invention is not limited thereto. Most of the
circuitry of digital baseband processor 27 is described above as
being part of DSP 79. In other embodiments, some components of
circuitry 20 are implemented as sets of instructions operating on a
processor separate from DSP 79. For example, the separate processor
can be an ARM processor. The instructions are stored on
processor-readable medium 59, and the separate processor reads the
instructions from processor-readable medium 59 before performing
the instructions. Thus, processor-readable medium 59 stores not
only Gaussian lookup table 58 and Rician lookup table 68, but also
program instructions. In this case, processor-readable medium 59 is
a type of non-volatile memory, such as read only memory (ROM). In
one embodiment, for example, each of equalizer 40, distribution
analyzer 46, BEP estimator 48, averager 70, filter 73 and
non-linear quantizer 76 is implemented as a set of instructions
operating on the separate processor.
[0071] The previous description of the disclosed embodiments is
provided to enable any person skilled in the art to make or use the
present invention. Various modifications to these embodiments will
be readily apparent to those skilled in the art, and the generic
principles defined herein may be applied to other embodiments
without departing from the spirit or scope of the invention.
Accordingly, the present invention is not intended to be limited to
the embodiments shown herein but is to be accorded the widest scope
consistent with the principle and novel features disclosed
herein.
* * * * *