RU2216871C2 - Method for receiving information about signal quality in receiver - Google Patents

Abstract

FIELD: digital data exchange over digital phase-modulation communication channels; digital data transmission systems. SUBSTANCE: information about signal quality in receiver is obtained by executing following operations with signal: first estimate of phase φ₁ is generated in receiver for each respective phase-modulated digital character; second estimate of phase φ₂ is calculated in receiver for each phase-modulated digital character from formula φ₂ = φ₁+φ₃, where φ₃ = const; decisions are taken using first and second calculated estimates of phases φ₁ and φ₂ on respective first and second m-bit codes, where m = log₂M; distance is found according to Hamming as difference in number of bits between first and second m-bit codes and distances found according to Hamming for preset number of phase-modulated digital characters are summed up; information about quality of phase-modulated digital characters is determined by sum obtained. EFFECT: enhanced precision and equal time for measuring all values of input signal-to- noise ratio. 1 cl, 3 dwg

Description

 The invention relates to the field of discrete information exchange over communication channels using digital phase modulation (FM).

 There is a method for determining the signal-to-noise ratio of M-position phase-modulated signals, in which the first and second signals are formed from the input signal, the mean of the first signal and the variance of the second signal are measured, and the signal-to-noise ratio is calculated from the measured values, which is a measure of signal quality ( see ed. St. 1798738, MKI 5 G 01 R 29/26, published in BI 8, 02/28/93) [1].

 The first disadvantage of this method is its excessive complexity associated with the need to perform a large number of complex calculations and, consequently, large computational costs during implementation.

 The second disadvantage of this method is that the device for its implementation must be performed parametrically, since the optimal structure of the device itself must change when changing over a wide range of signal-to-noise ratios.

 Of the known technical solutions, the closest to the patented method is a method of obtaining signal quality information in a receiver, in which a plurality of M-position phase-modulated digital symbols are received by the receiver, a corresponding phase estimate for each phase-modulated digital symbol is generated in the receiver, and the corresponding phase error signal (see RF patent 2113061, IPC 6 H 04 L 27/22, publ. in BI 16, 06/10/98) [4].

 The disadvantages of the known prototype method are as follows.

Firstly, the known method has low instrumental accuracy at high signal to noise ratios, i.e. with high signal quality and, as a result, with high signal to noise ratios, the method requires averaging over an excessively large number of characters for reliable measurement of signal quality. This is due to the fact that the dependence
R _osh (R _I )
where R _Osh - the probability of an error in the bit;
R _I - the input signal-to-noise ratio is essentially nonlinear, sharply decreasing with increasing R _I.

поскольку при этом одна ошибка в бите (или, что эквивалентно, одна ошибка по фазе) формируется на 10³ принятых символов и для статистически достоверного подсчета такого количества ошибок потребуется усреднение по (3÷5)•10³ битам.So, for example, in order to reliably evaluate (in any units) the quality of the FM-2 signal at R _in ≈6.8 dB, averaging over the number of symbols of the order

since in this case one error in a bit (or, equivalently, one error in phase) is generated by 10 ³ received symbols, and for a statistically reliable calculation of such a number of errors, averaging over (3 ÷ 5) • 10 ³ bits will be required.

At high signal-to-noise ratios required in connected applications, the probability of error is usually normalized at the level of 10 ^-6 ÷ 10 ^-9 and lower (even with the use of coding), which will require an increase in memory to implement moving averaging by more than 10 ⁵ times.

 Secondly, the known method is intended to determine information about the reliability of a bit of a demodulated FM signal for cases of decoding by a Viterbi convolutional decoder with a “soft” solution. Accordingly, the efficiency of using the known method is reduced when using other types of decoders (other than convolutional ones, or when using discrete message signals in the transmission system without error-correcting coding), since the method involves the use of the angular distance as a measure of the reliability of the decision made.

The technical result is the provision of high instrumental accuracy of the method and the same measurement time for all values of the input signal-to-noise ratio; the possibility of using the method by any consumers of information about the signal quality in discrete message transmission systems is achieved by performing the following operations on the signal:
- the receiver receives a plurality of M-position phase-modulated digital symbols;
- generate in the receiver the corresponding first phase estimate φ ₁ for each phase-modulated digital symbol;
- calculate in the receiver for each phase-modulated digital symbol the second phase estimate φ ₂ according to the formula φ ₂ = φ ₁ + φ ₃ , where φ ₃ = const;
- according to the first and second estimates, the phases φ ₁ and φ ₂ make decisions about the first and second m-bit codes corresponding to them, where m = log ₂ M;
- find the Hamming distance between the first and second m-bit codes;
- summarize the resulting Hamming distance for a predetermined number of phase-modulated digital symbols;
- according to the received amount, information about the signal quality of phase-modulated digital symbols is determined;
- depending on the determined information about the signal quality of phase-modulated digital symbols, a decision is made to change the number φ ₃ .
This is achieved by the fact that according to the method of obtaining information about the signal quality at the receiver, the receiver receives a plurality of M-position phase-modulated digital symbols, and a corresponding first phase estimate φ ₁ for each phase-modulated digital symbol is generated at the receiver.

According to the invention, for the second phase-modulated digital symbol, the second phase estimate φ _{2 is} additionally calculated in the receiver according to the formula φ ₂ = φ ₁ + φ ₃ , where φ ₃ = Const, according to the calculated first and second phase estimates, φ ₁ and φ ₂ decide on the corresponding the first and second m-bit codes, where m = log ₂ M, find the Hamming distance as the difference in the number of bits between the first and second m-bit codes, summarize the resulting Hamming distance for a predetermined number of phase-modulated digital symbols, according to the sum about the information on the signal quality of the phase-modulated digital symbols is distributed, depending on the determined information on the signal quality of the phase-modulated digital symbols, a decision is made to change the number φ ₃ .
In FIG. Figure 1 shows the signal points with the codes assigned to them for the case of a 4-position FM signal. Figure 2 explains the algorithm for calculating the value of R _I , figure 3 explains the process of iterative adaptation according to the parameter φ ₃ .
According to the patented method:
1) The receiver receives a plurality of M-positional FM digital signals (hereinafter, simply FM signals). Moreover, each of the M reference vectors (symbols) of such an ensemble transmits m = log ₂ M bits of information.

при расположении опорных векторов

так, как того требуют оценивание по максимуму правдоподобия.2) A corresponding estimate of φ ₁ for each FM symbol is generated in the receiver. This procedure can be implemented by a sufficiently large number of methods, including calculating the quadrature components of the received symbol X ₁ and Y _{1 by a} coherent (correlation) demodulator and calculating the function

when positioning support vectors

as required by maximum likelihood assessment.

3) Calculate the second estimate of the phase φ ₂ according to the formula φ ₂ = φ ₁ + φ ₃ , where φ ₃ is a constant in a certain range of input signal-to-noise ratios R _d . Practice shows that for most applications it is enough to choose a constant φ ₃ from a set of 3 ÷ 4 values.

4) Based on the calculated values of φ ₁ and φ ₂ make decisions about the corresponding first and second m-bit codes. This procedure corresponds to the decision on the criterion of the optimal observer Kotelnikov-Siegert in accordance with the selected manipulation code. For the considered example of the FM-4 signal (Fig. 1), this operation can be set in the form of table 1 (see the end of the description).

 5) Find the Hamming distance between the first and second m-bit codes. This procedure consists in calculating the number of bits that distinguish two m-bit codes for each of the FM symbols.

 6) Summarize the resulting Hamming distance for a predetermined number of FM symbols, i.e. implement a rolling summation procedure over a predetermined number of characters.

The implementation of procedures 5) and 6) allows us to judge the quality of the input signal. In fact, the implementation of procedure 3) is equivalent to a decrease in the noise immunity of coherent reception by an amount in the signal-to-noise ratio associated with the value of φ ₃ . This can be explained as follows.

где Р_c - вероятность ошибки в символе;
R - входное отношение сигнал/шум;
erf - функция интеграла ошибки.For M-positional FM signals it is known that the probability P _{C of an} error in a symbol is limited by the inequalities [2, p. 99]:

where P _c is the probability of error in the symbol;
R is the input signal-to-noise ratio;
erf is the error integral function.

где P_C(Ф) - вероятность ошибки в символе при наличии статической фазовой ошибки Ф;

Сравнивая системы неравенств (1) и (2) и учитывая, что при кодировании кодом Грея одна ошибка в пользу смежного символа приводит к появлению одной ошибки в соответствующем m-разрядном коде [3, с.162], можно считать, что вычисление второй оценки фазы φ₂ происходит с отношением сигнал/шум, худшим в

раз, чем при вычислении первой оценки φ₁.
Выбирая величину φ₃ такой, чтобы вероятность ошибки в бите увеличивалась примерно не менее чем на два порядка, можно пренебречь ошибками, содержащимися в первом m-разрядном коде и считать, что сумма расстояний между кодами по Хэммингу примерно пропорциональна числу истинных ошибок при когерентной демодуляции заданного числа символов. Другими словами, при выборе коэффициента пропорциональности (и, соответственно, связанного с ним угла φ₃) используют следующие соображения. Разобьем все значения R_ВХ на диапазоны. Для достижения одинаковой инструментальной точности как для высоких, так и для низких отношений сигнал/шум (высокого и низкого качества входного сигнала), потребуем, чтобы в пределах одного диапазона точность измерения Р_ош путем реализации процедур 5) и 6) была не хуже δ(%). Зададимся, например δ=1%. Тогда величина φ₃ должна быть такой, чтобы вероятность истинной ошибки во втором коде была более чем на два порядка выше, чем в первом коде (для того, чтобы при вычислении расстояния по Хэммингу ошибки в первом коде не влияли на результат).At the same time, it is known that in case of inaccuracy in the reference signals at the angle Ф, the probability Р _{C is} limited by the following inequalities [2, p.268]:

where P _C (Ф) is the probability of an error in a symbol in the presence of a static phase error Ф;

Comparing the systems of inequalities (1) and (2) and taking into account that when coding with the Gray code, one error in favor of the adjacent symbol leads to the appearance of one error in the corresponding m-bit code [3, p.162], we can assume that the calculation of the second estimate phase φ ₂ occurs with the signal-to-noise ratio, the worst in

times than when calculating the first estimate of φ ₁ .
Choosing a value of φ ₃ such that the probability of an error in a bit increases by about at least two orders of magnitude, we can neglect the errors contained in the first m-bit code and assume that the sum of the distances between the Hamming codes is approximately proportional to the number of true errors during coherent demodulation of a given number of characters. In other words, when choosing the coefficient of proportionality (and, accordingly, the angle φ ₃ associated with it), the following considerations are used. We divide all the values of R _BX into ranges. To achieve the same instrumental accuracy for both high and low signal-to-noise ratios (high and low quality of the input signal), we require that, within the same range, the measurement accuracy of _Рш by implementing procedures 5) and 6) is not worse than δ ( %). We set, for example, δ = 1%. Then the value of φ ₃ should be such that the probability of a true error in the second code is more than two orders of magnitude higher than in the first code (so that when calculating the Hamming distance, the errors in the first code do not affect the result).

 7) Summarize the resulting Hamming distance for a predetermined number of FM symbols.

Let us define the number of FM symbols for which the Hamming sliding summation procedure will be performed. Let this number be, for example, N = 10 ³ characters. From statistical considerations of the consistency of the assessment, we assume that the optimal value of the sum, which preserves the instrumental accuracy of the method, is SUM opt, and
SUM _lower ≤SUM _opt ≤SUM _upper (3)
For definiteness, let
SUM _bottom = 5;
SUM _top = 100.

Based on the selected number of characters over which averaging is carried out, we can say that when implementing the method, the probability of a bit error in the second m-bit code of the order of 10 ^{-3 is} allowed. Therefore, at the same time, the probability of error in the first m-bit code can be no more than 1% of 10 ^-3 , i.e. 10 ^-5 , which for FM-4 symbols is equivalent to an input signal-to-noise ratio of the order of 9.5 dB and higher [3, p. 223, 248].

8) Information on the signal quality of the FM symbols is determined from the received SUM. Moreover, taking into account the implemented operations 1) - 7), information about the quality of the input signal is obtained on the basis of the following considerations: since the probability of an error in the first m-bit code is significantly less than the probability of an error in the second m-bit code (10 ^-5 << 0 ^-3 ), the errors in the first m-bit code can be neglected, and the Hamming distance with a given instrumental accuracy δ = 1% allows us to measure errors in the second m-bit code. The statistical reliability of these estimates is ensured if the obtained value of the sum SUM lies within
5 = SUM _lower ≤SUM≤SUM _upper = 100
for the selected number N = 1000 averaged FM symbols. In this case, knowing the specific SUM value, using the formulas (1) and (2), the probability of error in the first m-bit code is recounted and the corresponding value of the input signal-to-noise ratio is found.

A graphical implementation of this procedure is shown in figure 2. From figure 2 it is seen that when adding the angle φ ₃ (equivalent to energy losses ΔR (φ ₃ ), dB) leads to an increase in the magnitude of errors in the second m-bit code.

где N - число символов, по которым ведется суммирование, можно найти истинное значение R_ВХ (c точностью δ%).Since the dependence P (R _I ) has the same character for any φ ₃ , then according to the measured value

where N is the number of characters over which the summation is carried out, you can find the true value of R _BX (with accuracy δ%).

9) If the obtained SUM value does not satisfy the conditions of formula (3), the angle φ _{3 is} changed. Moreover, if the input signal-to-noise ratio is large (signal quality is good), then the angle φ _{3 is} increased so that the probability of error for the second m-bit code lies in the selected neighborhood 10 ^-3 . Conversely, if the signal quality is poor, the angle φ _{3 is} reduced. In this case, the iterative algorithm shown in Table 2 is implemented (see the end of the description).

The adaptive adjustment procedure of the angle φ ₃ shown in figure 3.

(как и слишком большая

приведут к невыполнению при суммировании одного из неравенств системы (3) и по величине

будет произведена адаптация угла φ₃, к входному отношению сигнал/шум для достижения неизменной инструментальной точности измерения качества входного сигнала δ(%)
Итак, патентуемый способ обеспечивает:
а) инвариантную к условиям работы инструментальную точность измерения качества сигнала δ(%), обеспечиваемую за счет адаптации к условиям измерений (изменение величины параметра φ₃);
б) одно и то же время измерений при любом отношении сигнал/шум, поскольку это время определяется только выбранным числом усредняемых сигналов. При выбранном числе символов N=10³ и скорости передачи информации 1 Мбит/с время выдачи результата измерений составляет 10^-3с.Let the measured signal-to-noise ratio lie in the range of R ² d. Then the quantity is too small

(like too big

lead to non-fulfillment when summing one of the inequalities of system (3) and in magnitude

the angle φ ₃ will be adapted to the input signal-to-noise ratio to achieve constant instrumental accuracy of measuring the quality of the input signal δ (%)
So, the patented method provides:
a) instrumental accuracy of measuring signal quality δ (%), which is invariant to the operating conditions, provided by adapting to the measurement conditions (changing the value of parameter φ ₃ );
b) the same measurement time for any signal to noise ratio, since this time is determined only by the selected number of averaged signals. With the selected number of characters N = 10 ³ and the information transfer rate of 1 Mbit / s, the time for issuing the measurement result is 10 ^-3 s.

In the prototype method, in the range of working signal-to-noise ratios (for P from 10 ^-5 to 10 ^-9 ), to ensure the same reliability of measurements, the same time is from 0.5 s to 5 • 10 ³ s, which is absolutely unrealistic most communication applications;
c) obtaining information about the signal quality in the receiver for any consumers of this information (and not just Viterbi soft decoders) by implementing a signal quality measure not in the form of an angular error, but in the form of the probability of a bit error at the output of the demodulator, which is the main a measure of signal quality in a discrete communication channel.

Sources of information
1. A.S. 17989738, MKI 5 G 01 R 29/26, publ. in BI 8, 02.28.93

 2. Stiffler JJ Theory of synchronous communication. - M .: Communication. - 1975 .-- 488 p.

 3. Feer K. Wireless digital communications. Modulation and spreading methods. - M .: Radio and communications - 200. - 520 p.

 4. RF patent 2113061, IPC6 N 04 L 27/22, publ. in BI 16.06.06.98, the prototype.

Claims (1)

A method of obtaining information about the signal quality at the receiver, in which the receiver receives a plurality of M-position phase-modulated digital symbols, produces in the receiver a corresponding first phase estimate φ ₁ for each phase-modulated digital symbol, characterized in that a second estimate is calculated at the receiver for each phase-modulated digital symbol phase ₂ φ from the formula φ ₂ = φ ₁ + φ ₃ wherein φ ₃ = const, calculated by the first and second estimated phase φ ₁ and φ ₂ are solutions of the corresponding first and second m-bit codes, where m = log ₂ M, find the distance Hamming as the difference in the number of bits between the first and second m-bit code, summing the resulting distance Hamming for a predetermined number of phase-modulated digital symbols at received amount determined information about the quality of the signal phase-modulated digital characters as a result of dividing the resulting amount by a predetermined number of phase-modulated digital symbols and, if the resulting amount is more or less than the specified threshold values, reduce, or naturally increase the value of φ ₃ .

Images