Publication number | US20060291591 A1 |

Publication type | Application |

Application number | US 11/165,297 |

Publication date | Dec 28, 2006 |

Filing date | Jun 22, 2005 |

Priority date | Jun 22, 2005 |

Also published as | CN101238675A, CN101238675B, EP1908204A1, EP1908204B1, EP2182668A1, US8532232, US20100290512, WO2007002417A1 |

Publication number | 11165297, 165297, US 2006/0291591 A1, US 2006/291591 A1, US 20060291591 A1, US 20060291591A1, US 2006291591 A1, US 2006291591A1, US-A1-20060291591, US-A1-2006291591, US2006/0291591A1, US2006/291591A1, US20060291591 A1, US20060291591A1, US2006291591 A1, US2006291591A1 |

Inventors | Kaushik Ghosh |

Original Assignee | Kaushik Ghosh |

Export Citation | BiBTeX, EndNote, RefMan |

Patent Citations (8), Referenced by (10), Classifications (10), Legal Events (1) | |

External Links: USPTO, USPTO Assignment, Espacenet | |

US 20060291591 A1

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 μ and □ or A and □ are determined. The ratio μ/□ or A/□ 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.

Claims(38)

(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.

(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.

(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.

(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.

(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).

(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.

(h) quantizing the filtered mean BEP.

(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.

(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.

(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).

(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.

(h) quantize the filtered mean BEP.

(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.

(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.

(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).

(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.

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.

an equalizer that outputs the distribution of multi-bit soft decisions.

a convolutional decoder that outputs hard decision bits based on the multi-bit soft decisions.

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.

Description

1. Field

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.

2. Background

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^{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.”

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).

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.

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 MCS**9**, than using a higher redundancy MCS scheme, such as MCS**5**.

**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.

**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.

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.

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.

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 (μ) and the variance (σ) are determined for soft decisions having a Gaussian distribution. The statistical parameters A and σ are determined for soft decisions having a Rician distribution. The signal-to-noise ratio of the RF signal is represented by the ratio μ/σ for a Gaussian distribution of soft decisions and by the ratio A/σ 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 μ/σ 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/σ 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.

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.

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 (μ) and the variance (σ). When the soft decisions have a Rician distribution, the BEP estimator calculates the statistical parameters A and σ. The BEP estimator also calculates the ratio μ/σ for a Guassian distribution and the ratio A/σ for a Rician distribution. The ratios μ/σ and A/σ correlate to the signal-to-noise ratios of the I and Q samples.

The BEP estimator estimates the BEP of a burst containing a Gaussian distribution of multi-bit soft decisions by mapping the ratio μ/σ 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/σ to an empirically determined BEP in a Rician lookup table stored on the mobile station.

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.

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.

The accompanying drawings, where like numerals indicate like components, illustrate embodiments of the invention.

**4** (a GMSK scheme);

**9** (an 8-PSK scheme);

**9**, 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;

**9**.

Reference will now be made in detail to some embodiments of the invention, examples of which are illustrated in the accompanying drawings.

**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.

**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 **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 **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**.

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.

Depending on the BEP level, data is transmitted at nine different data rates in the EDGE standard. **1**-MCS**4**) 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 (MCS**5**-MCS**9**) 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. **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.

The first four MCSs have different coding schemes that provide for nearly no coding (MSC**4**) to highly redundant coding (MSC**1**). The code rate listed in **1** has a data rate of 9.05 kbps per channel, and MSC**4** 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.”

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.

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

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.

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 □-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.

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 MCS**1** through MCS**4**, which employ GMSK. An 8-PSK demodulator **38** demodulates I and Q samples **26** that were modulated with MCS**5** through MCS**9**, which employ 8-PSK. In the embodiment of **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**.

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 (3×116) per burst.

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.

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**.

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.

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 μ (mu) and σ (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 σ.

In the following example of step **50**, the statistical parameters μ and σ 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 **52**, μ represents the mean, and σ 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×[1100]; 30×[1101]; 15×[1110]; 15×[0000]; 30×[0001]; 11×[0010], where [1100]=−4; [1101]=−3; [1110]=−2; [0000]=0; [0001]=1; and [0010]=2. The statistical parameters μ and σ 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 (μ) and the variance (σ).

**53** for the second moment and an equation **54** for the fourth moment, each expressed in terms of μ and σ. The mean (μ) and the variance (σ) are determined by solving these two equations in two variables. An equation **55** expresses μ in terms of the second and fourth moments. An equation **56** expresses σ in terms of the second and fourth moments. For the sample distribution of 116 soft decisions listed above, μ is determined to be 2.039, and a is determined to be 1.181.

Returning to the next step in **57** by mapping the quotient μ/σ to a BEP value in a lookup table. The quotient of the mean (μ) divided by the variance (σ) is indicative of the signal-to-noise ratio of the data that comprise a distribution. For the sample Gaussian distributions the quotient μ/σ is 1.727. The relationship between the quotient μ/σ 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 **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.

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.

Returning to step **51**, the statistical parameters A and σ 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

**63** for the second moment and an equation **64** for the fourth moment, each expressed in terms of A and σ. 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 σ 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 σ is determined to be 1.345.

In a step **67**, the BEP is then determined by mapping the quotient A/σ to a BEP value in a lookup table. For the sample Rician distribution, the quotient A/σ is 1.035. The relationship between the quotient A/σ 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/σ. Where the quotient A/σ of the sample Rician distribution equals 1.035 in this example, BEP value **60** is determined to be 0.079.

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**.

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**.

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**.

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**.

**20** to obtain the quantized MEAN_BEP **77** for a radio block. **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.

**4** 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 **50**, and the BEP values **60** are determined by mapping the ratio μ/□ 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

**77** obtained from groups of four consecutive BEP values **60** of **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.

**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 MCS**1** through MCS**4** (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.

**9** employing 8-PSK. The BEP values are estimated from bursts transmitted over a static channel exhibiting no fading. As in **60** are determined by mapping the ratio μ/□ 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

**9** over the same signal-to-noise ratio as in **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.

**77** obtained from the BEP values **60** of **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.

**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 MCS**5** through MCS**9** (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.

**9**. Unlike the results shown in **33** used to derive the BEP values resembles a Rician distribution. The BEP values are obtained using the method of **51**, and the BEP values **60** are determined by mapping the ratio A/□ 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 **9**. The estimated BEP values **60** plotted in

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.

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.

Patent Citations

Cited Patent | Filing date | Publication date | Applicant | Title |
---|---|---|---|---|

US20020172302 * | Mar 14, 2001 | Nov 21, 2002 | Pantelis Monogioudis | Compensation based bit-error-rate estimation for convolutionally encoded transmissions in wireless systems |

US20030115028 * | Oct 10, 2002 | Jun 19, 2003 | Vpisystems Pty Ltd | Optical error simulation system |

US20040243889 * | Jan 27, 2004 | Dec 2, 2004 | Peng Li | N-dimensional determination of bit-error rates |

US20040253968 * | May 14, 2004 | Dec 16, 2004 | Chang Li Fung | Adaptive channel quality estimation algorithm to support link adaptation |

US20050018794 * | Jul 22, 2003 | Jan 27, 2005 | Xiangguo Tang | High speed, low-cost process for the demodulation and detection in EDGE wireless cellular systems |

US20050182807 * | Feb 12, 2004 | Aug 18, 2005 | Texas Instruments Incorporated | Methodology for designing high speed receivers below a target bit-error-rate |

US20060062363 * | Sep 15, 2005 | Mar 23, 2006 | Sirenada, Inc. | Method and apparatus for interacting with broadcast programming |

US20060203943 * | Mar 7, 2006 | Sep 14, 2006 | Comsys Communication & Signal Processing Ltd. | Single antenna interference suppression in a wireless receiver |

Referenced by

Citing Patent | Filing date | Publication date | Applicant | Title |
---|---|---|---|---|

US7782791 * | Jan 4, 2008 | Aug 24, 2010 | Lg Electronics Inc. | Method of reporting channel quality in EGPRS system |

US7929630 * | Sep 11, 2006 | Apr 19, 2011 | Symbol Technologies, Inc. | Adaptive RFID receiver for QAM signals |

US8385365 | Feb 26, 2013 | Sony Corporation | Communication system, communication apparatus, and packet length control method | |

US8505059 * | Dec 25, 2008 | Aug 6, 2013 | Electronics And Telecommunications Research Institute | Channel capacity estimation and prediction method and apparatus for rate adaptive wireless video |

US8532232 | Jul 26, 2010 | Sep 10, 2013 | Qualcomm Incorporated | Estimating bit error probability (BEP) in an edge wireless system |

US8588356 * | Oct 21, 2010 | Nov 19, 2013 | Stmicroelectronics (Grenoble 2) Sas | Method and device for detecting a phase error of a signal |

US8654834 * | Jan 5, 2010 | Feb 18, 2014 | Electronics And Telecommunications Research Institute | Method for tuning coding rate and applying unequal error protection for adaptive video transmission, and video transmission/reception apparatus using the method |

US20090196343 * | Dec 25, 2008 | Aug 6, 2009 | Electronics And Telecommunications Research Institute | Channel capacity estimation and prediction method and apparatus for rate adaptive wireless video |

US20100171882 * | Jul 8, 2010 | Electronics And Telecommunications Research Institute | Method for tuning coding rate and applying unequal error protection for adaptive video transmission, and video transmission/reception apparatus using the method | |

US20110096876 * | Apr 28, 2011 | Stmicroelectronics (Grenoble 2) Sas | Method and device for detecting a phase error of a signal |

Classifications

U.S. Classification | 375/340, 375/332 |

International Classification | H04L27/06, H04L27/22 |

Cooperative Classification | H04L1/0026, H04L1/203, H04L1/206, H04L1/0009, H04L1/0003 |

European Classification | H04L1/20M |

Legal Events

Date | Code | Event | Description |
---|---|---|---|

Aug 22, 2005 | AS | Assignment | Owner name: QUALCOMM INCORPORATED, CALIFORNIA Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:GHOSH, KAUSHIK;REEL/FRAME:016657/0593 Effective date: 20050802 |

Rotate