JP5278944B2 - Signal detection device - Google Patents

Description

The present invention relates to a signal detection apparatus that can detect a weak signal based on statistical properties of internal noise.
This signal detection device can be used mainly in the field of electronic circuits for receiving circuits for wireless communication and wired communication, and for their integration.

In recent years, wireless communication technology has been developed rapidly, and electronic devices are required to be smaller and lighter, and to have multiple functions and higher performance. In accordance with this requirement, further miniaturization, higher speed, and lower power consumption are desired for the built-in integrated circuit. Currently, CMOS RF circuit technology that can be expected to improve performance due to miniaturization is actively developed as a technology that meets such demands.
However, due to process miniaturization and low power consumption, the influence of noise inside the circuit becomes relatively large, and it is difficult to secure a dynamic range with the conventional reception method using A / D conversion. .

Therefore, as described in Non-Patent Document 1, a high-sensitivity detection technique that can detect even a weak signal having a noise level or less using noise statistics has been recently studied. This technique presupposes that the response of the comparator has a distribution based on thermal noise and statistical characteristics of characteristic variations.
In the technique of Non-Patent Document 1, one comparator that compares the input signal s (t) with a threshold value is prepared, the comparison output is obtained a plurality of times for each time, and their statistical properties are used. The waveform of s (t) is estimated.
G. Tapang, C. Saloma, “Dynamic-Range Enhancement of an Optimized 1-Bit A / D Converter” IEEE TRANSCATIONS ON CIRCUITS AND SYSTEMS-II : ANALOG AND DIGITAL SIGNAL PROCESSING, VOL. 49, NO. 1, JANUARY 2002

However, the technique of Non-Patent Document 1 has the disadvantage that it can handle signals that change in time periods but cannot handle non-periodic signals.
In addition, internal noise is not white noise, but there is a case where there is a correlation between noises with different sampling times in the same circuit, such as 1 / f noise. It becomes difficult to distinguish gender.

  An object of the present invention is to provide a signal detection device capable of sufficient signal detection even for a weak signal that is below a noise level and below a threshold, based on the statistical properties of internal noise.

Signal detection apparatus of the present invention is a signal detector for detecting the input signal, N (N is a natural number of 2 or more) each input signal in parallel into, the N comparing each with the threshold value A comparator, an arithmetic unit for calculating the number N + of comparators in which each input signal exceeds the threshold value, and a comparison in which each input signal exceeds the threshold value based on the output of the arithmetic unit Analysis of calculating the ratio of the number N + of the comparators to the number N of the comparators: N + / N and estimating the waveform of the input signal by using the ratio N + / N and the internal noise statistics Means (claim 1).

  In this signal detection apparatus, the same signal is input to all the comparators, and the input signal is estimated based on noise statistics from the number N + of comparators in which each input signal exceeds the threshold value. is there. That is, when the input signal and noise are mixed, the larger the ratio N + / N of the comparator that detected the input signal, the larger the input signal waveform that cannot be buried in the noise can be estimated. The smaller the signal ratio N + / N, the smaller the input signal waveform buried in the noise.

The “noise” is not noise included in the input signal but noise generated from a circuit inside the signal detection apparatus. In the present invention, the type of “noise” is regarded as thermal noise, and its standard deviation (sometimes referred to as “noise intensity” in this specification) is written as σn. “Statistics of noise” refers to the statistical nature of this thermal noise, that is, the statistical nature according to a Gaussian distribution.
The “threshold value” may be represented by an anonymous number normalized by the noise intensity σn. The minimum amplitude of the input signal that can be detected by the signal detection device is referred to as “sensitivity”. In this specification, when the minimum amplitude of the input signal that can be detected is small, the sensitivity is good, and when the minimum amplitude of the input signal that can be detected is large, the sensitivity is bad.

  In one aspect of the signal detection device of the present invention, the input signal is a signal that takes positive and negative values by changing with time, the threshold value is a set of positive and negative values, and the computing unit is Calculating the number N + of comparators in which each of the input signals exceeds the positive threshold and the number N− of comparators in which the input signal falls below the negative threshold, Based on the output of the computing unit, the ratio N + / N: N + / N and the ratio N− / N: N− / N are calculated. The waveform of the input signal is estimated using the ratios N + / N, N− / N and the statistical properties of noise. Thus, the detection reliability can be further improved by using two threshold values.

The signal detection apparatus of the present invention may include N amplifiers that amplify the parallel input signals (claim 3). The amplification factor of the amplifier can be arbitrarily selected (may be 1). The thermal noise generated inside the amplifier is also included in the “internal noise” of the present invention.
The analysis means can estimate the waveform of the input signal using, for example, an inverse function of the error function erf (Claim 4). The reason why it is reasonable to use the error function erf is that when a signal buried in noise is detected by a comparator, the probability that the signal exceeds the threshold is

This is because the error function erf is expressed as follows. Here, σn is noise intensity, B is a threshold value, s is an input signal, and x is a signal with noise.
The analysis means can also estimate the waveform of the input signal by using the difference between N + / N and N− / N, taking advantage of the use of two thresholds. This is an approximate representation of the inverse function of the error function of [Equation 1]. As a result, the statistical processing method, which is generally difficult to implement in hardware, can be substituted, and the signal can be restored using a simple arithmetic circuit.

The ratio of the threshold value to the noise intensity (B / σn) is preferably in the range of approximately 1 to 2 (claim 6). If the ratio (B / σn) is within this range, the sensitivity of the signal detection device can be optimized.
Preferably, the input signal is a periodically changing signal, and the comparator samples m times (m> = 2) in one cycle. What is needed more than once is due to the sampling theorem.

It is preferable to change the number of times of sampling m in accordance with the expected input signal strength, that is, sensitivity. For example, the sampling frequency m is increased for weak input signals. For large input signals, the sampling frequency m is reduced. Thereby, useless power consumption can be eliminated.
It is preferable to change the number N of elements according to the expected input signal strength, that is, sensitivity. Many comparators and amplifiers are operated for weak input signals, and few comparators and amplifiers are operated for large input signals. That is, the number N of comparators to be used is changed according to the expected input signal strength. Thereby, the sensitivity can be changed according to the input signal intensity. Further, by reducing or blocking the bias current to the unused circuit, the power consumption of the circuit is reduced in the case of a large input signal, and wasteful power consumption can be eliminated.

As described above, according to the present invention, it is possible to detect a weak signal buried in noise power by arranging the comparators in parallel in an array. By applying the present invention to a communication receiving circuit, the dynamic range of the A / D converter can be greatly improved. As a result, it is possible to expect effects such as reducing the transmission output of the mobile phone terminal and the base station and reducing the number of optical fiber communication repeaters.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
<Signal detection device>
FIG. 1 is a block diagram showing the configuration of the signal detection apparatus of the present invention. The input signal s (t) to be detected by the signal detection device is an analog signal that takes positive and negative values by changing over time, but a signal that takes only positive values or only a negative value. It is also applicable to signals that take

  The signal detection apparatus includes N amplifiers 41 for amplifying each of the input signals s (t) paralleled in N (N is a natural number of 2 or more), and a set of positive and negative threshold values for the output signal of each amplifier. N comparators 42 that obtain three values of 1, 0, -1 by comparing with B and -B, respectively, and the number N + of comparators whose output signals exceed the positive threshold B are calculated. At the same time, an arithmetic unit 43 that calculates the number N− of the comparators 42 that is below the negative threshold −B, and a ratio between the number N of elements and the number N of elements based on the output of the arithmetic unit 43: N + / N, the ratio of the number of elements N− and the number of elements N: N− / N is calculated, and the ratio of the input signal is calculated by using these ratios N + / N, N− / N and the internal noise statistics. And a statistical processing block 44 for estimating the waveform.

When the input signal s (t) is simultaneously input to the N amplifiers 41, thermal noise generated in each amplifier 41 is added to each signal output. The same applies to the comparator 42. The noise generated here is uncorrelated between the N amplifiers 41 and also uncorrelated between the N comparators 42.
The noise converted at the input position of the comparator 42 is expressed as noise n (t). Noise n (t) is added to the input signal s (t). The sum of the input signal s (t) and the noise n (t), s (t) + n (t) is expressed as x (t).

In the present embodiment, the amplitude of the input signal s (t) is smaller than the standard deviation of noise n (t) (sometimes referred to as “noise intensity” in this specification) σn and equal to or less than a threshold value B. And Here, the noise intensity σn is a constant determined in the circuit.
The comparator 42 compares the noisy signal x (t) with both the threshold values B and -B, and converts it into a digital signal of 1, 0, -1. The computing unit 43 obtains the number N + of the comparators 42 determined that the signal is larger than B and the number N− smaller than −B among the N parallel elements by looking at the statistics of the digital output. As a result, two outputs, N + and N-, are obtained for an input signal for a predetermined period (for example, one cycle). The statistical processing block 44 reproduces the original signal by using statistical processing on these two outputs.

All or part of the functions of the statistical processing block 44 is realized by the computer of the signal detection apparatus executing a program recorded in a predetermined storage medium such as a memory, a CD-ROM, or a hard disk. However, as will be described later, the function of the statistical processing block 44 can be realized by a hardware circuit.
FIG. 2 is a diagram illustrating the influence of noise superimposed on an input signal. The horizontal axis represents voltage, and the vertical axis represents probability density. The input signal s (t) and the threshold value B are indicated by vertical lines. The noise x (t) is shown as a Gaussian distribution with variance σn ² and average value s. In this example, a signal exceeding the threshold value B is recognized as “1”, and a signal not exceeding it is recognized as “0”.

  In this graph, the input signal s (t) is a value smaller than the threshold value B, but because of the noise n (t), a part of x (t) exceeds the threshold value B. It is sticking out. The probability density of the protruding portion is represented by P (x> B). The probability density P (x> B) is

It is represented by Here, erf is the error function, σn is the noise intensity, and s is the amplitude of the signal. In this way, the signal may exceed the threshold due to the addition of noise. Since the probability P (x> B) exceeding this threshold B can be calculated from the temporal statistics and ensemble statistics of the output of the comparator 42, the signal s (t) is estimated based on the probability P (x> B). It is possible.
Statistical processing block 44 by the inverse process of formula [Formula 2], a time with a signal s (t) (t = iT S, T S: sampling period, i is the number of sampling times) the sample value S (i in ) Can be restored. The restoration formula is

become that way. From this equation, if the threshold value B, the noise intensity σn, and the ratio N + / N of the number of elements N + and the total number N of elements where x (t) exceeds the threshold value B: N + / N, the signal S is obtained. You can see that it can be restored. The relationship between threshold and noise can be normalized. When the threshold is B and the standard deviation of noise is σn, the normalized threshold is represented by B / σn.
The threshold value B is preferably set such that the normalized threshold value (B / σn) is in the range of 1 to 2 when normalized with the noise intensity σn, as described in the embodiment.

The above analysis was for the case where the threshold value B is one, but when the input signal s (t) is a signal that takes positive and negative values due to temporal changes (for example, phase-modulated). It can be easily extended to the waveform of an AC signal.
FIG. 3 depicts an input signal s (t) that takes positive and negative values by changing with time, and a signal x (t) on which the input signal s (t) is loaded with noise n (t). It is a graph. A set of threshold values B, -B is shown. In the graph, the noise intensity σn is 1, the amplitude of the input signal s (t) is 0.3, and the threshold level is 2 (both are normalized by the value 1 of the noise intensity σn). The number of elements is N = 128. The gain of the amplifier 41 is 1.

  This graph is for one cycle of the input signal, and this is compared by the comparator 42 m times. This m is called the number of oversampling. In this embodiment, m = 8. Based on the comparison result of each sampled comparator 42, the number N + of the comparators 42 in which each input signal exceeds the positive threshold value B is calculated for each sampling, and below the negative threshold value -B. The number N− of the comparators 42 is calculated. Then, a ratio between N + and the number N of the comparators 42: N + / N and a ratio between N− and the number N of the comparators 42: N− / N are calculated.

  (A) When N +> N− (when the input signal is estimated to be positive),

To restore the original input signal.
(2) When N + <N− (when the input signal is estimated to be negative),

To restore the original input signal.
(3) When N + = N− (when the input signal is estimated to be 0),

Estimated.
Based on the above [Equation 4] to [Equation 6], the original input signal is restored.
If the restored signal S is graphed in this way, the result is as shown in FIG. As a result, it can be confirmed that the restored signal S has a 95.1% correlation with the original signal s (t).
As described above, in this signal detection device, the number of elements N + of the comparator 42 determined that the signal is larger than B and the number of elements N− of the comparator 42 determined that the signal is smaller than −B are input. The recovered signal can be restored.

<Hardware implementation of statistical processing block>
Real-time processing is indispensable when this signal detection apparatus is used for detecting communication signals and the like. However, since the statistical processing block 44 handles complicated expressions including error functions and complementary error functions such as [Equation 4] and [Equation 5], real-time processing is difficult.
Note that [Equation 2] can be approximated as [Equation 7].

As a result of the approximation, this N + / N can be regarded as representing the probability of exceeding the threshold value B. Here, it should be noted that the theoretical probability P (x> B) exceeding the threshold value B and N + / N do not completely coincide with each other (by increasing N, the probability approaches the probability). ), It is necessary to set a standard that allows an error of N + / N. The criteria for allowing this error will be described later.

  Therefore, paying attention to the fact that the signal restoration information is included in the result of subtracting the approximate probability N− / N below the threshold −B from the approximate probability N + / N exceeding the threshold B, the difference in probability

Suggest to take. In the inverse error function erf ⁻¹ of [Equation 4] and [Equation 5], only one of [Equation 4] and [Equation 5] is used, but this method is simple, but both threshold values B and −B are used. I am considering.
A method of restoring the original input signal based on [Expression 4] to [Expression 6] is referred to as a “statistical processing method”. A method of restoring the original input signal based on [Formula 8] (N + / N−N− / N) is referred to as “addition method”.

A hardware circuit for calculating [Equation 8] can be easily made as shown in FIG. 5, for example. In FIG. 5, N +. The value of N− is input to a subtractor 58 to obtain a difference, and is divided by the number N of elements in a divider 59. However, the number of elements N is a fixed value. In this way, (N + / N−N− / N) can be output.
In this way, the statistical processing block 44 can be implemented in hardware (integrated circuit), and circuit mounting is facilitated. Further, the processing time can be shortened, and real-time statistical processing can be performed.

<Increase / decrease in the number of elements>
In the normal reception method, the signal strength necessary for ensuring the resolution of the A / D converter, that is, “sensitivity” is obtained by the pre-stage amplifier, but the intensity of the signal that propagates in communication varies, If the sensitivity is fixed, when a signal having a strength much larger than the threshold value B is received, correct signal reproduction cannot be performed.

In the configuration of FIG. 1, the sensitivity can be varied by varying the gain of the amplifier 41. However, since the internal noise also changes, there is a problem that noise statistical processing is complicated and signal restoration is difficult.
Therefore, by changing the number of elements N, the sensitivity is changed to cope with the increase or decrease of the input signal. The theoretical relationship between the number N of elements and the sensitivity will be described with reference to FIG. 9 in the following [Example].

Specifically, as shown in FIG. 6, a control circuit 45 is provided in the signal detection device, and a signal line C1 for controlling the power supply or bias current of the amplifier 41 is supplied from the control circuit 45 to the amplifier 41. A signal line C2 for controlling the power supply or bias current is supplied to the comparator 42.
The control circuit 45 increases the number of amplifiers 41 and comparators 42 that are operating, that is, the number of elements N, when it is desired to improve the sensitivity of the signal detection device (to decrease the minimum signal intensity that can be detected). When it is desired to lower the sensitivity of the signal detection device (to increase the minimum signal intensity that can be detected), the number of amplifiers 41 and comparators 42 that are operating, that is, the number N of elements is reduced. As a result, the number of comparators 42 and amplifiers 41 to be used can be changed in accordance with the input signal strength to change the sensitivity. Since the sensitivity can be changed even if the intensity of the received signal changes, the dynamic range of the signal detection device can be expanded.

Further, the bias current of the amplifier 41 and the comparator 42 that are not operating is reduced or cut off. Thereby, useless power consumption can be eliminated.
FIG. 7 shows a configuration in which [Equation 8] is calculated by a hardware circuit when the number N of elements is variable. In FIG. 7, the values of N + and N− output from the arithmetic unit 43 are input to a subtractor 58 to obtain a difference, which is divided by the number N of elements in a divider 59. However, the number N of elements is variable and is acquired from the control circuit 45. In this way, (N + / N−N− / N) can be output using a simple hardware circuit.

  Although the embodiments of the present invention have been described above, the embodiments of the present invention are not limited to the above embodiments. For example, the signal detection apparatus of FIG. 1 assumes a comparator that can obtain three values of B, 0, and -B, but it has been confirmed that there is no problem in operation even if a Schmitt trigger circuit is used. When using a Schmitt trigger, it is only necessary to add the outputs, and no difference is required. In addition, various modifications can be made within the scope of the present invention.

-Example 1 (threshold value, sensitivity, number of elements)-
A numerical simulation was performed on the signal detection apparatus shown in FIG.
As already explained, the following three conditions were set. (1) Noise is independent for each amplifier and follows a Gaussian distribution. (2) All internal noises are converted to input before the comparator 42. (3) The gain of each amplifier is 1.

The input signal is assumed to be a phase shift keying (PSK) signal used in wireless communications. In this configuration, noise is not intentionally included in the input signal, but internal noise of the amplifier 41 and the comparator 42 itself is used, so that the noise is determined by circuit design.
In addition, since the circuit is generally evaluated by converting noise to the input side, the second assumption is made.

In addition, the gain of the amplifier is set to 1. This is to reduce the parameter. After clarifying the sensitivity characteristics, the gain of the amplifier is adjusted in consideration of the required sensitivity and noise inside the amplifier. It is good.
Furthermore, in order to determine whether the output signal is correctly restored, the following two criteria (a) and (b) are set.

(A) The correlation between the waveform pattern serving as the input signal and the output signal is examined. In other words, since the PSK signal is assumed in this embodiment, the input signal is a sine wave that is phase-modulated at 0 ° or 180 °. This input signal pattern is already prepared on the output side, the correlation with the output signal is examined, and the signal pattern having the maximum correlation value is set as the restoration signal.
(B) The same process is performed 10,000 times for one amplitude, and if the bit error rate BER is 10 ⁻³ or less, it is determined that the signal of that amplitude has been accurately detected.

In this embodiment, the sensitivity, that is, the minimum amplitude of the input signal that can be accurately detected with a bit error rate of 10 ⁻³ is As, min, and the “normalized sensitivity” is normalized with a threshold value (As, min / B). ).
The normalization sensitivity was investigated while changing the threshold value B with the noise intensity σn fixed.
FIG. 8 is a graph showing the relationship between the normalized sensitivity (As, min / B) and the threshold-to-noise ratio (B / σn). As a result of examining several elements N as a variable, it can be seen that the threshold-to-noise ratio (B / σn) at which the best sensitivity is obtained is in the range of approximately 1 to 2.

  It can also be seen that in a situation where the threshold-to-noise ratio (B / σn) is fixed, the sensitivity changes depending on the number N of elements of the comparator 42 used, as shown in FIG. That is, the greater the number N of elements, the better the sensitivity. When the number of elements N is small, the threshold-to-noise ratio (B / σn) is optimal in the vicinity of 2, but when the number of elements N is large, not a single peak point but a wide range of threshold values with a noise intensity ratio of 1 to 2. Equivalent sensitivity is shown. When the same sensitivity is shown in such a wide range, the vertical axis indicates the signal amplitude normalized by the threshold value, so that it can be seen that a signal having a smaller threshold value can reproduce even a signal smaller than noise. That is, for the same sensitivity, it is desirable to set the threshold value as small as possible.

Next, with the threshold value B fixed, a numerical simulation was performed while changing the number of elements N, the number of oversamplings m within one period, and the modulation method (BPSK, QPSK) of the input signal.
The result is shown in FIG. FIG. 9 shows that the normalized sensitivity is reduced (improved) in proportion to the square root √ (mN) of the product of the number of elements N and m. The change in sensitivity when the input signal is changed from BPSK to QPSK (transition A from the square point to the inverted triangle point in FIG. 9) is the result of halving m for the same number of elements N (square point in FIG. 9). Is consistent with the transition A ′) from the black circle to the black dot.

From this result, it can be seen that the product mN of the number N of elements and the number of oversampling m affects the sensitivity.
From the above results, the relationship of variables when the present invention is put into practical use became clear. That is, (1) the threshold-to-noise intensity ratio (B / σn) may be set in the range of 1-2. (2) In consideration of sensitivity degradation due to advanced modulation schemes, m and the number of elements N corresponding to the desired sensitivity are set. In order to improve the sensitivity, m and the number N of elements may be set large as long as the element can be miniaturized.

-Example 2 (verification of hardware)-
FIG. 10 is a waveform diagram of the amplitude of each signal restored by the statistical processing method. FIG. 11 is a waveform diagram of the amplitude of each signal restored by the addition method. However, (N + / N−N− / N) obtained by the addition method is multiplied by k (k is an integer) for comparison. The value of k is set so that the amplitude of the signal waveform obtained by both methods is closest. In this embodiment, k = 4.

The calculation conditions are: frequency of two input signals: f1 = 125 [MHz], f2 = 117.1875 [MHz], normalized sensitivity: As, min = 0.711, noise intensity: σn = 1, threshold: B = 1.7783, Number of elements: N = 128.
Since two input signals with different frequencies are input, the comparator 42 has a third-order nonlinearity. Therefore, as shown in FIGS. 10 and 11, the “beat” frequency component of the frequency 2f1-f2 or the frequency 2f2-f1 is used. Is output.

10 and 11, it can be seen that the amplitude component of the signal frequency is almost the same in both the statistical processing method and the addition method, and the signal size and shape can be well restored. Considering that a ternary digital value is obtained from the output result of the comparator 42, it can be seen that the quantization error is also suppressed.
Looking at the frequency components of the restored signal in FIG. 11, it can be seen that the S / N ratio of the same degree can be realized although a little nonlinear distortion occurs. Therefore, it can be proved that the addition method is a reliable signal restoration method.

It is a block diagram which shows the structure of the signal detection apparatus of this invention. It is a figure which shows the example from which reading is mistaken for the noise superimposed on an input signal. 6 is a graph depicting a signal waveform of a signal x (t) in which noise n (t) is superimposed on an input signal s (t). It is a graph which shows the waveform of the input signal s (t) and the restored signal. It is a figure which shows the hardware circuit (N: fixed) which calculates the "addition method" which decompress | restores the original input signal based on (N + / N-N- / N). It is a block diagram which shows the structure which concerns on the modification of the signal detection apparatus of this invention. It is a figure which shows the hardware circuit (N: variable) which calculates the "addition method" which decompress | restores the original input signal based on (N + / N-N- / N). It is a graph which shows the relationship between normalized sensitivity (As, min, min / B) and threshold-to-noise ratio (B / σn) while changing the number of elements N. The relationship between the normalized sensitivity (As, min, min / B) and the number of elements N while changing the oversampling count m in one cycle and the modulation method (BPSK, QPSK) of the input signal while fixing the threshold B It is a graph which shows. It is the graph which plotted the amplitude of the signal decompress | restored by the statistical processing method. It is the graph which plotted the amplitude of the signal decompress | restored by the addition method.

Explanation of symbols

41 Amplifier 42 Comparator 43 Adder 44 Statistical Processing Block 45 Control Circuit 58 Subtractor 59 Divider

Claims (9)

A signal detector for detecting the input signal,
N comparators each comparing each input signal paralleled to N (N is a natural number of 2 or more) with a threshold value;
An arithmetic unit for calculating the number N + of comparators in which each of the input signals exceeds the threshold;
Based on the output of the arithmetic unit, a ratio N + / N of the number N + of comparators in which each input signal exceeds the threshold and the number N of elements of the comparator is calculated, and this ratio N + / N and an analysis means for estimating the waveform of the input signal using internal noise statistics. The input signal is a signal that takes positive and negative values by changing with time,
The threshold has a set of positive and negative,
The computing unit calculates the number N + of comparators in which each input signal exceeds the positive threshold value and the number N− of comparators in which the input signal falls below the negative threshold value,
The analysis means calculates a ratio N + / N: N + / N and a ratio N + / N: N− / N based on the output of the computing unit. 2. The signal detection apparatus according to claim 1, wherein the waveform of the input signal is estimated using these ratios N + / N, N− / N and the statistical property of the internal noise.   2. The N amplifiers for amplifying each of the input signals paralleled to N, wherein the N comparators compare the output signals of the amplifiers with the threshold value, respectively. The signal detection apparatus according to claim 2.   The signal detection apparatus according to claim 1, wherein the analysis unit estimates a waveform of the input signal using an inverse function of an error function erf.   4. The signal detection apparatus according to claim 2, wherein the analysis unit estimates a waveform of the input signal using a difference between N + / N and N− / N.   The signal detection device according to any one of claims 1 to 5, wherein a ratio (B / σn) between the threshold value and the noise intensity is in a range of approximately 1 to 2.   7. The signal detection apparatus according to claim 1, wherein the input signal is a signal that changes periodically, and the comparator samples m times (m> = 2) in one period. .   The signal detection apparatus according to claim 7, wherein the sampling number m is set according to sensitivity of the signal detection apparatus.   The signal detection apparatus according to claim 1, wherein the number N of elements is set according to sensitivity of the signal detection apparatus.

Images