JP2006060386A - Software radio receiver using rf-filter bank - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a software radio receiver using an RF-filter bank capable of eliminating an interference signal on the outside of a band, and capable of forming an arbitrary matched filter having the same high performance as a dielectric filter. <P>SOLUTION: The software radio receiver has the RF-filter bank and a multi-input transversal filter. The RF-filter bank is composed of a plurality of RF filters, AD-converts outputs from each RF filter, and compensates the deviation of each filter characteristic by the multi-input transversal filter. The RF-filter bank further leads out the tap weighting factor of the multi-input transversal filter by the method of a least square by using a training signal and optimally synthesizes the tap weighting factor, and forms the arbitrary matched filter by the RF-filter bank and a digital-signal processing. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a software defined radio receiver, and more particularly, to a software defined radio receiver capable of forming an arbitrary matched filter that can remove out-of-band interference signals and has high performance equivalent to that of a dielectric filter.

  In recent years, research and development of software-defined radio (SDR: software-defined radio) by digitization and softwareization of a mobile radio transmission circuit has been promoted. A software defined radio receiver (SDR receiver) is a receiver that can easily cope with various transmission systems such as an adaptive modulation system, a variable bandwidth, a variable center frequency, and a multiband by software.

  However, since a wireless input signal (RF signal) is generally a high-frequency band signal, the level variation width and the diversity of bandwidth are extremely large. In addition, the interference signal power may be much larger than the desired signal. Therefore, an approach different from general baseband signal processing is required for AD conversion.

  By the way, techniques such as direct conversion, low-IF, filter bank, fractional frequency sampling, and offset frequency sampling are important AD conversion elements in software defined radio (hereinafter also simply referred to as SDR). Technology.

For example, in the “receiver using the low IF method” disclosed in Non-Patent Document 1, an AD converter (ADC) is realized by using a filter bank by a switched capacitor filter (SCF) and fractional frequency sampling. ) Sampling rate and the number of quantization bits can be suppressed, but with the switched capacitor filter (SCF) alone, there is a problem that aliasing occurs in sampling and it is difficult to remove out-of-band interference signals. Further, there is a problem that it is difficult to realize an RF filter.
Sanada Yukitoshi and Ikehara Masaaki, "A Study on Coefficient Error Compensation of Complex Coefficient Filter Bank in Low IF Reception System", IEICE Technical Report, NS2001-79, RCS2001-80, p.7-12, July 2001 Co-authored by WSTYan, RKCMak, HCLuong, “2-V 0.8- / spl mu / m CMOS monolithic RF filter for GSM Receiver “2-V0.8- / spl mu / m CMOS monolithic RF filter for GSMreceivers”, IEEE MTT-S Internal. Microwave Sympo. Digest (Volume 2) , p. 569-572, June 1999 W. Wu, X. Ding, M. Ismail, H. Olsson, “RF Bandpass Filter Design Based on CMOS Active Inductors“ RFbandpass filter design based on CMOS active inductors ””, IEEE Trans. Circuits Syst II, Vol. 50, No. 12, p. 942-949, December 2003 RMBorwick, PAStupar, JF DeNatale, R. Anderson, R. Erlandson, “Variable MEMS” Capacitors Implied Into RF Filter Systems “VariableMEMS capacitors implemented into RF filter systems”, IEEE Trans. On Microwave Theory and Tech., Vol. 51, No. 1, p. .315-319, January 2003 RMYoung, JDAdam, C. R. Vail (CRVale), T. T. Bragggins, T. B. Krishnaswamy (SVKrishnaswamy), C. E. CEMilton, DWBever, LG Chorosinski, Lee-ShuChen, DECrockett, CB, Friedehof ( CBFreidhoff, SH Talisa, E. Capelle, R. Tranchini, J. R. Fender (JRFende), J. M. Rossi Oil (JMLorthioir), A. Co-authored by R. Tries, “Roros Bandpass RF Filter Using MEMS Capacitance Switches to Achieve a One Octave Tuning Grange and Independent Variable Bandwidth “Low-lossbandpass RF filter using MEMS capacitance switches toachieve a one-octave tuning range and independently variable bandwidth” ”, IEEE MTT-S Internal. Microwave Sympos. Digest (IEEEMTT-S Intern. Microwave Symp. Digest), Volume 3, p.1781-1784, June 2003 A.Tombak, J.P.Maria, F.T.Aiguavives, Z.Jin, G.Tauf, A.I.Kingon (AIKingon) and Amir Mortazawi, “Voltage-controlled RF filters and thin-film Barium-Strontium-Titanatetunable capacitors”, IEEE Trans. On Microwave Theory and Tech., Vol.51, No.2, p.462-467, February 2003 N.Sato, H.Suzuki, S.Suyama, K.Fukawa, "Complex Form Bandpass Sampling with Offset Frequency Sampling and Quadra Chua Component Interpolation for Modulated Signals “Complexform bandpass sampling with offset frequency sampling and quadrature component interpolation for modulated signals” ”, IEICE Trans. On Commui., Vol. 86, No. 12, p. .3513-3520, December 2003 S. Haykin, “Adaptive Filter Theory“ AdaptiveFilter Theory ””, Prentice-Hall, 1996 [IEEE Std 802.11a-1999 [ISO / IEC 8802-11: 1999 / Amd 1: 2000 (E)] Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specializations, High Speed Physical Layer in the 5GHz band “IEEEStd 802.11a-1999 [ISO / IEC 8802-11: 1999 / Amd 1: 2000 (E)] Part11: Wireless LANMedium Access Control (MAC) and Physical Layer (PHY) specifications, High-speed Physical Layer in the 5 GHz Band ””, LAN / MAN Standards Committee Off the IEEE Comp. Soc. (LAN / MAN Standards Committee of the IEEE Comp. Soc.), P.9, p.37, 1999

  On the other hand, in recent years, research and development for integrating the entire receiver in the RF band higher than the UHF band has been underway. In order to realize a practical multi-input multi-output (MIMO) communication system, integration of the RF band is indispensable. Among them, an analog filter that can change the center frequency and bandwidth of the RF band is extremely important, and studies using CMOS, MEMS, and the like are underway (see Non-Patent Documents 2 to 6). However, it is very difficult to realize such high performance as a dielectric filter with these semiconductor variable RF filters.

  The present invention has been made under the circumstances described above, and an object of the present invention is to provide an arbitrary matched filter that can remove out-of-band interference signals and has high performance equivalent to that of a dielectric filter. To provide a software defined radio receiver using an RF filter bank.

  The present invention relates to a software defined radio receiver using an RF filter bank. The object of the present invention is a software defined radio receiver including an RF filter bank and a multi-input transversal filter, wherein the RF filter The bank is composed of a plurality of RF filters, and the output of each RF filter is AD-converted, the deviation of individual filter characteristics is compensated by the multi-input transversal filter, and a training signal is used as a minimum. It is effectively achieved by deriving the tap weight coefficients of the multi-input transversal filter by the square method, optimally synthesizing them, and forming an arbitrary matched filter by the RF filter bank and digital signal processing.

  The above-described objects of the present invention include an RF signal / training signal changeover switch, an RF filter bank, a plurality of mixers, a plurality of low-pass filters, a plurality of AD converters, and a multi-input transversal filter. And a weighting factor controller, a demodulator, a reference signal generator, a training signal generator, and a frequency synthesizer, or the RF signal / training signal selector switch includes the RF filter, An RF signal input to the bank and a training signal generated by the training signal generator are switched, and the RF filter bank is composed of a plurality of RF filters, and the RF signal / training signal switching switch Output band signals, and output a plurality of band signals, The Xer multiplies each of the plurality of band signals output from the RF filter bank by a carrier signal having a carrier frequency generated by the frequency synthesizer, and the plurality of low-pass filters include: Only the baseband signal component is extracted from the output of the mixer, and the plurality of AD converters sample the baseband signal that is the output of the low-pass filter and convert it into a digital signal, The multi-input transversal filter linearly synthesizes the output signals of the plurality of AD converters and outputs a detection signal, and the weighting factor controller includes an ideal impulse response between the detection signal and the training signal. The difference is used to control the tap weight coefficient of the multi-input transversal filter. The demodulator demodulates the detection signal, or uses a variable or fixed bandpass filter as the RF filter, or the bandpass filter is a Gaussian bandpass filter. Or by using a chirp signal as the training signal, or the weighting factor controller uses the difference between the detection signal and the ideal impulse response as an estimation error to calculate the absolute value of the estimation error. By controlling by the least square method so that the value square value is minimized, or by using the LMS algorithm or RLS algorithm which is a sequential least square method as the least square method, it is more effective. Achieved.

  According to the software defined radio receiver using the RF filter bank according to the present invention, an arbitrary matched filter can be formed using an RF filter bank formed of an RF variable filter or the like on a semiconductor. There is an excellent effect that an out-of-band interference signal can be removed and high performance equivalent to that of the dielectric filter can be realized.

  Further, according to the present invention, it is possible to realize a software defined radio receiver that can be integrated on a semiconductor and can easily change the bandwidth, center frequency, and multiband by software.

The best mode for carrying out the present invention will be described below with reference to the drawings.
<1> Configuration and Principle of the Present Invention <1-1> Display of Input Signal Configuration of Preferred Embodiment of Software Defined Radio (hereinafter, also simply referred to as SDR receiver) using RF filter bank according to the present invention Is shown in FIG.

  As shown in FIG. 1, the SDR receiver 1 according to the present invention includes an RF switch 10, an RF filter bank 20, a plurality of mixers 30, a reference signal generator 40, a training signal generator 45, A plurality of low-pass filters (LPF) 50, a plurality of AD converters (ADC) 60, a multi-input transversal filter (TVF) 70, a weight coefficient controller 80, and a demodulator 90 are included. Yes.

  In the SDR receiver 1 of the present invention shown in FIG. 1, the RF switch 10 generates a training signal synchronized with a radio signal (RF signal) received by an antenna and a reference signal which is an output of the reference signal generator 40. The training signal generated by the device 45 is switched. The RF filter bank 20 receives the output band signal of the RF switch 10 and divides the output band signal into K signals (K is a positive integer). K band signals are output through a pass filter (BPF) 25 and a low noise amplifier 28.

  Next, the plurality (K) of mixers 30 are frequency synthesizers 48 that synchronize the K band signals output from the RF filter bank 20 with the reference signal output from the reference signal generator 40. Multiply each of the generated carrier signals at the carrier frequency. A plurality (K) of low-pass filters (LPFs) 50 extract only the baseband signal component from the output of the mixer 30.

  Next, a plurality (K) of AD converters (ADC) 60 sample the baseband signal that is the output of the low-pass filter (LPF) 50 and convert it into a digital signal. The multi-input (K input 1 output) transversal filter 70 linearly synthesizes the output signals of a plurality (K) of AD converters (ADCs) 60 and outputs a detection signal.

  Here, the weighting factor controller 80 uses the difference between the detection signal output from the multi-input (K input 1 output) transversal filter 70 and the ideal impulse response to the training signal to generate a multi-input (K input 1 output). ) Control the tap weight coefficient of the transversal filter 70.

  That is, the weight coefficient controller 80 estimates the difference between the detection signal output from the multi-input (K input 1 output) transversal filter 70 and the ideal impulse response to the training signal as an estimation error (estimation error signal). The tap weight coefficient of the multi-input (K input 1 output) transversal filter 70 is controlled using the least square method so that the absolute value square value of the error is minimized. As a specific method, for example, it is preferably realized by an LMS algorithm or an RLS algorithm which is a sequential least square method.

  The demodulator 90 demodulates the detection signal output from the multi-input (K input 1 output) transversal filter 70.

  More specifically, in the SDR receiver according to the present invention, the RF filter bank is composed of a plurality of RF filters, and in this embodiment, is composed of K band-pass filters (BPF). Here, it is assumed that the bandwidth of each BPF is narrower than the bandwidth of the input RF signal, and the entire RF filter bank covers the bandwidth of the input signal.

  Next, the band signal of the RF filter output is amplified by a low noise amplifier, and further frequency-converted to baseband with a local signal from a frequency synthesizer. From the frequency-converted signal, harmonic components are removed by a low-pass filter (LPF) and converted to a digital signal by an AD converter (ADC). The sampled digital signal is a complex envelope. The complex envelope is optimally synthesized by a transversal filter (TVF), and a received baseband signal is generated and passed to a demodulator.

Here, let _{fc be} the center frequency of the RF input signal s (t) input to the RF filter bank.

但し、ｅ_ｓ(ｔ)は複素包絡線で、Ｅ_ｓ(ｆ)は、複素包絡線ｅ_ｓ(ｔ)のフーリエ変換（Fourier変換）である。⇔はフーリエ変換の関係を表す。

However, e _s (t) is the complex envelope, E _s (f) is the Fourier transform of the complex envelope e _s (t) (Fourier transform). ⇔ represents the relationship of Fourier transform.

The signal x _k (t) from which the k-th RF filter output is sampled is expressed as e _{x, k} (t) ⇔E _{x, k} (f) as follows.

ここで、ｆ_ｒ,φ_ｒは、基準キャリア信号の周波数と位相であり、以下では、ｆ_ｒ＝ｆ_ｃ,０≦φ_ｒ＜２πとする。Ｇ_ｋ(ｆ),φ_ｋ(ｆ)は、それぞれ、ｋ番目のＢＰＦと増幅器の合成された利得と位相を表す。Ｈ_ｋ(ｆ)はＢＰＦの等価低域インパルス・レスポンスｈ_ｋ(ｔ)のフーリエ変換であり、以下では、簡単化のために、中心周波数の利得が１のガウス形ＢＰＦとする。

Here, f _r and φ _r are the frequency and phase of the reference carrier signal. In the following, it is assumed that f _r = f _c and 0 ≦ φ _r <2π. G _k (f) and φ _k (f) represent the combined gain and phase of the kth BPF and the amplifier, respectively. H _k (f) is a Fourier transform of the equivalent low-frequency impulse response h _k (t) of the BPF. Hereinafter, a Gaussian BPF having a center frequency gain of 1 is assumed for the sake of simplicity.

ただし、Ｂ_ｋは各フィルタの３ｄＢ帯域幅であり、

である。

Where B _k is the 3 dB bandwidth of each filter,

It is.

Let {x _k (i)} be a complex envelope obtained by sampling and quantizing x _k (t) with an ADC. However, x _k (i) represents a sampling value of t _i = iΔt, where Δt is the sampling interval. When the maximum frequency of the signal subjected to frequency conversion of s (t) is f _max , Δt = ½f _max . It is also possible to sample at a lower rate by complex sampling or fractional frequency sampling (see Non-Patent Document 7).

<1-2> Optimal synthesis
Since {x _k (i)} is obtained by dividing one signal in frequency, it needs to be synthesized and reproduced. At that time, the BPF overlap part is removed, the gain is adjusted, and the delay distortion is corrected. That is, these are regarded as linear distortions and optimally synthesized by a least square method using a multi-input linear equalizer realized by a transversal filter (TVF).

  In the following, optimal synthesis is performed using a minimum mean square error (MMSE) criterion using K M stages (M is a positive integer) TVF. The composite output y (i) is expressed as follows.

＜１−３＞チャープ信号によるトレーニング信号
ＴＶＦのタップ重み係数を決定するため、チャープ信号ｓ_ｃ(ｔ)により、トレーニング信号を生成する。以下の方法によれば、任意の入力信号に対する整合フィルタを形成することができる。

<1-3> Training signal by chirp signal In order to determine the tap weight coefficient of TVF, a training signal is generated by the chirp signal s _c (t). According to the following method, a matched filter for an arbitrary input signal can be formed.

First, a discrete chirp signal shown in FIG. 2 is generated using a frequency synthesizer. When a certain frequency chirp signal, ADC since samples the L number of discrete signal, that one section is a _{T L =} LΔ _t. The s _c (t) in the n-th (−N _f ≦ n ≦ N _f ) time interval tε [(N _f + n) T _L , (N _f + n + 1) T _L ] is expressed as follows.

ここで、ｇ_ｎ(ｔ)は複素包絡線で、ｆ_ｓは周波数ステップであり、チャープの周波数幅をΔ_Ｆとして、ｆ_ｓ＝Δ_Ｆ／２Ｎ_ｆである。また、整数ｄにより、τ_ｄ＝ｄΔ_ｔで表される時間オフセットは、インパルス応答の有効な時間区間をサンプリングするためのものである。

Here, g _n (t) is a complex envelope, f _s is a frequency step, and f _s = Δ _F / 2N _f where the frequency width of the chirp is Δ _F. Further, the integer d, time offset represented by τ _{d =} dΔ _t is for sampling the valid time interval of the impulse response.

と

とする。 Interval _{n (N f + n) T} L + iΔ a chirp signal at _{t s c} (i), the digital signals x _n, and _k (i) thereto. The origin of this time is shifted as shown in Fig. 2.

When

And

ここで、

は仮想的なインパルス入力信号を表し、

は仮想的なインパルス入力信号に対するＢＰＦや増幅器の応答を表す。したがって、多入力線形等化器の出力が入力信号ｓ(ｔ)のスペクトルと同じになるように、等化器係数を設定すれば、整合フィルタが形成される。ｓ(ｔ)のスペクトルは、シンボルのパルス波形ｐ_ｒ(ｔ−τ_ｄ)で決まるので、そのサンプル値をｐ_ｒ(ｉ−ｄ)で表す。

＜１−４＞逐次形最小２乗法
トレーニング信号に対するベースバンドでの理想インパルス応答を

として、多入力線形等化器のタップ重み係数を、逐次的な最小２乗法であるＲＬＳアルゴリズム又はＬＭＳアルゴリズムにより決定する（非特許文献８参照）。多入力線形等化器の入力信号

に対する相関逆行列をＰ(ｉ)、タップ重み係数ベクトルをｗ、忘却係数をλ(０＜λ≦１)とする。指数重み付評価関数Ｊ(ｉ)を用いる。

here,

Represents a virtual impulse input signal,

Represents the response of the BPF or amplifier to a virtual impulse input signal. Therefore, if the equalizer coefficients are set so that the output of the multi-input linear equalizer is the same as the spectrum of the input signal s (t), a matched filter is formed. Since the spectrum of s (t) is determined by the pulse waveform p _r (t−τ _d ) of the symbol, the sample value is represented by p _r (id).

<1-4> Sequential least squares method Ideal baseband response to training signal

As described above, the tap weight coefficient of the multi-input linear equalizer is determined by the RLS algorithm or LMS algorithm which is a sequential least square method (see Non-Patent Document 8). Input signal of multi-input linear equalizer

Let P (i) be the inverse correlation matrix for w, w the tap weight coefficient vector, and λ (0 <λ ≦ 1) the forgetting factor. An exponential weighted evaluation function J (i) is used.

Ｊ(ｉ)を最小にするｗをｗ(ｉ)とすると、ｗ(ｉ)の更新式は次のようになる。

When w that minimizes J (i) is w (i), the update formula of w (i) is as follows.

上記数式において、時刻ｉをインクリメントして、さらに、

を更新することが出来る。なお、ｉ＝０での相関逆行列Ｐ(ｉ)およびｗ(ｉ)の初期値を以下のように設定する。

In the above formula, the time i is incremented,

Can be updated. Note that the initial values of the correlation inverse matrices P (i) and w (i) at i = 0 are set as follows.

ここで、δは微小な正定数で、ＩはＫＭ×ＫＭの単位行列で、０はＫＭ次元零ベクトルである。

Here, δ is a small positive constant, I is a unit matrix of KM × KM, and 0 is a KM-dimensional zero vector.

Receive the RLS algorithm or LMS algorithm, which is a sequential least-squares method, periodically by using the above-described training, that is, by periodically updating the tap weight coefficient, and receiving by aging and environmental changes. Variations in the filter bank, amplifier characteristics, etc. can be compensated.

<2> Computer Simulation <2-1> Simulation Conditions Computer simulation was performed to evaluate the reception characteristics of the software defined radio receiver using the RF filter bank according to the present invention. A Gaussian bandpass filter (Gaussian BPF) is used for the RF filter bank. The calculations were performed in double precision floating point, except for ADC quantization. Since the training is generated in the receiver, it was assumed that the level was high enough and there was no noise in the input. The phase of the reference carrier is ideal.

The error in the parameters is ± 5% for the 3 dB bandwidth of the Gaussian BPF of the filter bank, ± 1 dB for the combined gain, 0 to 2π for the combined phase, and ± T _s (1 / 8T _s step) for the combined delay. A uniform distribution was assumed. The characteristic indicates the ensemble average for the symbol.

  The simulation conditions are shown in Table 1 below. In the transmission signal, QPSK is used as a modulation method, and the roll-off rate is 0.25. In the filter bank, Gaussian BPFs are arranged at equal frequency intervals.

＜２−２＞シングル・キャリアにおける受信特性
まず、シングル・キャリアの信号受信に関するシミュレーションを行った。送信信号の複素包絡線ｅ_ｓ(ｔ)を下記数２５とする。

<2-2> Reception Characteristics in Single Carrier First, a simulation related to signal reception of a single carrier was performed. The complex envelope e _s (t) of the transmission signal is expressed by the following equation (25).

ここで、１／Ｔ_ｓはシンボル・レートを表し、ｂ(ｌ)は時刻ｊＴ_ｓのＱＰＳＫまたは１６ＱＡＭのシンボルを表す。また、ｐ_ｒ(ｔ)は、送信されるルート・ロールオフ・パルスである。

Here, 1 / T _s represents a symbol rate, and b (l) represents a QPSK or 16QAM symbol at time jT _s . P _r (t) is a route roll-off pulse to be transmitted.

Root roll-off QPSK or 16QAM input signal spectrum is shown in FIG. Note that the signal components are displayed with a shift of −6 dB in order to make the figure easier to see. FIG. 4 shows the eye pattern of the synthesized QPSK signal. As a result of the synthesis, a matched filter is formed in the receiving system, and an ideal waveform without intersymbol interference is obtained. The power spectrum of the combined output is shown in FIG. Regardless of the deviation of BPF center frequency, composite gain, and composite phase, it matched with the spectrum of cosine and roll-off. FIG. 6 shows how the mean square error (MSE) converges in the RLS algorithm. As can be seen from FIG. 6, the matrix elements converge rapidly after 60 steps when the matrix elements converge to some extent. The MSE residual is considered to be the accumulated quantization error of each tap.

<2-3> Reception Characteristics in Multicarrier (Orthogonal Frequency Division Multiplexing (OFDM) Signal) Next, in order to confirm the fine structure for each frequency, the reception characteristics in the multicarrier (OFDM signal) were evaluated. Specifications conforming to the wireless LAN standard IEEE802.11a were set (see Non-Patent Document 9). However, the subcarrier spacing was adjusted to facilitate comparison with the single carrier model.

A root roll-off pulse p _r (t) having a roll-off rate of 0.1 is used for the transmission signal so as to satisfy the spectrum mask. Therefore, the sub-carrier frequency interval of the OFDM as delta _f, also with respect to an odd number the number of subcarriers N, when the _{N h = (N-1)} / 2, OFDM complex envelope _{e s} (t) is It becomes as follows.

ただし、{ｂ_ｎ(ｊ)}は、ｎ番目のサブキャリアの時刻ｊＴ_ｓにおけるＱＰＳＫや１６ＱＡＭなどのシンボルを表す。ガード・インターバル（ＧＩ）が含まれているとする。

However, {b _n (j)} represents a symbol such as QPSK or 16QAM at the time jT _s of the n-th subcarrier. It is assumed that a guard interval (GI) is included.

FIG. 7 shows the power spectrum of the signal that has passed through each Gaussian BPF constituting the filter bank. A power spectrum of a signal obtained by optimally combining these outputs is shown in FIG. Regardless of the deviation of BPF center frequency, composite gain, and composite phase, it matched with the spectrum of cosine and roll-off. It can be seen that the optimum synthesis operates as a matched filter regardless of the band and the reception format.

<2-4> Noise and Interference Characteristics In order to clarify the interference characteristics of a software defined radio receiver using the RF filter bank according to the present invention (hereinafter also simply referred to as a filter bank receiver), an additive white Gaussian Transmission characteristics were examined by superimposing an out-of-band interference wave on a noise (AWGN) channel. FIG. 9 shows the relationship between the desired signal and the interference wave s _i (t). Interference wave is unmodulated signal, and an offset from f _c of the frequency twice the symbol rate. Since it is 4 times sampling, it becomes alias when it is further apart. When the amplitude of the interference wave is A _i , it is expressed by the following equation (28).

比較のために、従来のディジタル・フィルタによる受信機（以下、単に、ディジタル・フィルタ受信機とも称する）の特性も測定した。この場合、整合フィルタは、ディジタル・フィルタだけで形成される。アンチ・エイリアス受信アナログ・フィルタは、±２／Ｔ_ｓの帯域幅の理想フィルタとする。

For comparison, the characteristics of a conventional digital filter receiver (hereinafter also simply referred to as a digital filter receiver) were measured. In this case, the matched filter is formed only by a digital filter. The anti-aliased analog filter is an ideal filter with a bandwidth of ± 2 / T _s .

  The maximum input signal voltage was set to be half the input dynamic range of the ADC. As described above, the gain, phase, and bandwidth deviations can be calibrated by optimum synthesis, and therefore these deviations are not assumed below. The study will proceed based on 8-bit sampling.

  First, FIGS. 10 and 11 show eye patterns of output waveforms of a conventional digital filter receiver using 8-bit ADCs. In this observation, the noise was set to zero. As can be seen from FIGS. 10 and 11, the output waveform of CIR = −24 dB is not crossed at one point due to the quantization error caused by the input of excessive interference waves in both QPSK and 16QAM. In addition, the average value is shifted due to quantization. In order to cross at one point, a 12-bit ADC is required as shown in FIGS. In this case, the average value is almost zero.

  On the other hand, the eye pattern of the output waveform of the filter bank receiver of the present invention is shown in FIGS. As can be seen from FIGS. 14 and 15, even when QPSK and 16QAM are both increased to CIR = −24 dB, the 8-bit ADC eye pattern crosses at one point.

  FIG. 16 shows the bit error rate (BER) characteristics in QPSK. As can be seen from FIG. 16, in the filter bank receiver of the present invention, even when there is an interference wave of CIR = −24 dB, almost the same characteristics as in the case of no interference wave can be obtained by the 8-bit ADC. On the other hand, in the conventional digital filter receiver, in the case of 8-bit ADC, the characteristics are greatly deteriorated at CIR = −18, −24 dB. In order to obtain the same characteristics as when there is no interference wave in the digital filter receiver, a 12-bit ADC is required.

  Next, FIG. 17 shows the influence of interference waves in 16QAM and 64QAM. For reference, the characteristics of QPSK are also shown in FIG. As can be seen from FIG. 17, in the conventional digital filter receiver, CIR = -18 dB, and the amount of interference is suppressed compared to CIR = -24 dB of the filter bank receiver of the present invention. All are 8-bit sampling. The filter bank receiver of the present invention has the same characteristics as when there is no interference. On the other hand, the conventional digital filter receiver is greatly deteriorated.

  As described above, the software defined radio receiver using the RF filter bank according to the present invention is a combination of an RF filter bank composed of a plurality of analog RF filters and an optimum synthesis by a multi-input linear equalizer. A software defined radio receiver that can form a matched filter for any input signal.

  In addition, as described above, through the computer simulation, when the filter bank receiver of the present invention is used, reception of interference signals normally observed in RF band signals is suppressed for signals of various bandwidths and center frequencies. Made it possible. Furthermore, it was shown that the number of quantization bits and the operation speed of the ADC can be reduced. In the present invention, it is considered that the basic configuration for software defined radio including the RF circuit becomes important as the analog variable RF circuit using MEMS or the like is miniaturized.

  In the above-described embodiment, the method of dividing and receiving one signal has been described. However, various reception modes can be considered if the RF filter is made variable. The training method is also effective as a calibration method for the RF filter.

  As described above, the present invention is based on the assumption of a future integrated RF filter, an integrated RF receiver circuit, and a high-performance AD converter (ADC). A method has been disclosed in which a bank is used to divide and receive a single received signal band for optimal synthesis.

  In other words, the present invention discloses a method of forming an arbitrary matched filter by digital signal processing on an RF filter bank formed of an RF variable filter on a semiconductor and the like and an ADC output following each RF filter. . Further, the present invention also discloses a method for performing optimum synthesis by the least square method in order to suppress the influence of the deviation of individual filter characteristics in the RF filter bank.

  Specifically, in the present invention, first, a complex envelope is sampled from the band signal of each RF filter output constituting the RF filter bank, and optimally synthesized by a linear equalizer using a multi-input transversal filter. , Play the signal. A training signal is used to optimize the tap weight coefficient of the multi-input transversal filter. Moreover, it is preferable to use a chirp signal as the training signal.

  Next, by obtaining a complex impulse response by time correction of the output with respect to the training signal (chirp signal) and minimizing the difference from the symbol waveform of the transmission signal, the synthesis circuit can be made a matched filter.

  In short, the software defined radio receiver using the RF filter bank according to the present invention has the following main features.

  First, as a first feature, an arbitrary matched filter is formed by an RF filter bank composed of a plurality of RF filters and digital signal processing. Next, the second feature is that the output of each RF filter is AD-converted, and the deviation of individual filter characteristics is compensated by a multi-input transversal filter. A third feature is that a tap weight coefficient of a transversal filter is derived by a least square method and optimally synthesized. Finally, the fourth feature is that the training signal is used to derive the tap weight coefficient of the transversal filter.

1 is a block diagram showing a configuration of a preferred embodiment of a software defined radio receiver using an RF filter bank according to the present invention. FIG. It is a schematic diagram for demonstrating a discrete chirp signal. It is a graph which shows the power spectrum and Gaussian BPF characteristic of an input signal. It is a graph which shows the eye pattern of a synthetic | combination output. It is a graph which shows the electric power spectrum of a synthetic | combination output. It is a graph which shows the convergence characteristic of MSE in a RLS algorithm. It is a graph which shows the power spectrum in each filter output of an OFDM signal. 7A shows the outputs of the filters # 1, 3, 5, and 7, and FIG. 7B shows the outputs of the filters # 2, 4, 6, and 7. It is a graph which shows the power spectrum of the combined OFDM signal. It is a graph which shows the spectrum relationship between a desired signal and an interference wave. It is a graph which shows the eye pattern of the output waveform of the conventional digital filter receiver when there is an interference wave of CIR = −24 dB with QPSK, 8-bit ADC. It is a graph which shows the eye pattern of the output waveform of the conventional digital filter receiver in case there exists an interference wave of CIR = -24dB by 16QAM, 8-bit ADC. It is a graph which shows the eye pattern of the output waveform of the conventional digital filter receiver in case there exists an interference wave of CIR = -24dB by QPSK and 12-bit ADC. 16 is a graph showing an eye pattern of an output waveform of a conventional digital filter receiver when there is an interference wave of CIR = −24 dB with a 16QAM, 12-bit ADC. 6 is a graph showing an eye pattern of an output waveform of the filter bank receiver of the present invention when there is an interference wave of CIR = −24 dB with QPSK, 8-bit ADC. It is a graph which shows the eye pattern of the output waveform of the filter bank receiver of this invention when there exists an interference wave of CIR = -24dB by 16QAM, 8-bit ADC. It is a graph which shows the influence of the interference wave in QPSK. It is a graph which shows the influence of the interference wave in 16QAM and 64QAM.

Explanation of symbols

1 SDR receiver 10 according to the present invention RF switch 20 RF filter bank 25 Band pass filter (BPF)
28 Low noise amplifier 30 Mixer 40 Reference signal generator 45 Training signal generator 48 Frequency synthesizer 50 Low pass filter (LPF)
60 AD converter (ADC)
70 Multi-input transversal filter (TVF)
80 Weight coefficient controller 90 Demodulator

Claims (8)

A software defined radio receiver comprising an RF filter bank and a multi-input transversal filter,
The RF filter bank is composed of a plurality of RF filters,
A / D conversion is performed on the output of each RF filter, and the deviation of individual filter characteristics is compensated by the multi-input transversal filter,
Using the training signal, the tap weight coefficient of the multi-input transversal filter is derived by the least square method, and is optimally synthesized.
A software defined radio receiver using an RF filter bank, wherein an arbitrary matched filter is formed by digital signal processing with the RF filter bank.   RF signal / training signal selector switch, RF filter bank, multiple mixers, multiple low-pass filters, multiple AD converters, multi-input transversal filter, weight coefficient controller, demodulation A software defined radio receiver using an RF filter bank, comprising: a detector, a reference signal generator, a training signal generator, and a frequency synthesizer. The RF signal / training signal selector switch switches between an RF signal input to the RF filter bank and a training signal generated by the training signal generator,
The RF filter bank is composed of a plurality of RF filters, and an output band signal of the RF signal / training signal selector switch is input, and a plurality of band signals are output,
The plurality of mixers respectively multiply the plurality of band signals output from the RF filter bank by a carrier signal having a carrier frequency generated by the frequency synthesizer,
The plurality of low-pass filters extract only a baseband signal component from the output of the mixer,
The plurality of AD converters sample a baseband signal that is an output of the low-pass filter, and convert it into a digital signal,
The multi-input transversal filter linearly synthesizes the output signals of the plurality of AD converters and outputs a detection signal;
The weight coefficient controller controls a tap weight coefficient of the multi-input transversal filter using a difference between the detection signal and an ideal impulse response to the training signal;
The software defined radio receiver using the RF filter bank according to claim 2, wherein the demodulator demodulates the detection signal.   4. The software defined radio receiver using an RF filter bank according to claim 3, wherein a variable or fixed band pass filter is used as the RF filter.   5. The software defined radio receiver using an RF filter bank according to claim 4, wherein the band pass filter is a Gaussian band pass filter.   The software defined radio receiver using the RF filter bank according to any one of claims 3 to 5, wherein a chirp signal is used as the training signal.   The weighting factor controller is controlled by a least square method so that an absolute value square value of the estimation error is minimized by using a difference between the detection signal and the ideal impulse response as an estimation error. Item 7. A software defined radio receiver using the RF filter bank according to any one of items 6 to 9. 8. The software defined radio receiver using an RF filter bank according to claim 7, wherein an LMS algorithm or an RLS algorithm which is a sequential least square method is used as the least square method.

Images