CN103873168A - Multiple bandwidth sampling-based data preprocessing method in EVM (Error Vector Magnitude) measurement - Google Patents

Abstract

The invention discloses a multiple bandwidth sampling-based data preprocessing method in EVM (Error Vector Magnitude) measurement, and belongs to the field of EVM measurement of the communication system. The multiple bandwidth sampling-based data preprocessing method in the EVM measurement is implemented in the steps: firstly calculating a model of a cross correlation coefficient of a receiving sequence and a training sequence; then setting sequence variables in a model value greater than the threshold value as a set A; in the set A, setting elements of which the interval of the training sequence period number is the training sequence period as sets; setting non-empty header elements of the sets as a set B; searching the set B for N continuous elements of which the interval is 1 and setting the N elements as a set C, wherein N is a ratio of the IQ rate to the bandwidth; generating a preprocessing data sequence by the receiving sequence and the set C; finally calculating and compounding a corresponding new preprocessing data sequence. According to the multiple bandwidth sampling-based data preprocessing method in the EVM measurement, all the information of the original data is used, and the accuracy of the EVM measurement is effectively improved.

Description

Data preprocessing method during EVM based on many times of bandwidth samplings measures

Technical field

The present invention relates to the data preprocessing method in a kind of EVM measurement based on many times of bandwidth samplings, belong to the EVM fields of measurement of communication system.

Background technology

Communication system as TD, LTE in, conventionally use Error Vector Magnitude (EVM) to represent signal modulation quality, the immediate cause of EVM is phase error and the range error of receiving terminal modulation signal.The generation of EVM is because radio-frequency (RF) device local oscillator (LO) exists local-oscillator leakage, phase noise and power amplifier nonlinear distortion on the one hand, because algorithm performance difference (comprising channel estimating, input etc.) occurs that error causes on the other hand, so EVM value also can embody the quality of some algorithm performances.

In many times of bandwidth sample rate systems, sampled data IQ speed is N times of stream rate, at general Data processing, normally N haplotype data is carried out to N down-sampling doubly, then carry out data subsequent treatment, the method is little on the impact of decoding in OFDM receiver, because the N-1 number being rejected is according in fact or can correct in scope, can correct by the channel estimating in receiver processing module and Viterbi decoding, but the impact that this data processing method is measured for EVM is larger, reason is a part of information that data that sampling obtains only contain initial data, although can not impact decoding, but can increase the distance that sample point departs from planisphere standard point, thereby can affect the EVM test performance of tester.

Summary of the invention

The present invention is directed to the deficiency in background technology, propose the data preprocessing method in a kind of improved EVM measurement, cause EVM to test affected problem to solve the full detail that does not comprise initial data because of data from the sample survey.

The method comprises the steps:

Step 1: the mould r (n) that calculates the cross-correlation coefficient of receiving sequence and training sequence:

$r\left(n\right)=\left|\frac{{\mathrm{\Σ}}_{m=0}^{L-1}s(n+m)\·p^{*}\left(m\right)}{\sqrt{{\mathrm{\Σ}}_{m=0}^{L-1}{[s(n+m)-\stackrel{\‾}{s}\left(n\right)]}^2}\·\sqrt{{\mathrm{\Σ}}_{m=0}^{L-1}{[p\left(m\right)-\stackrel{\‾}{p}]}^2}}\right|$

Wherein: s (n) is receiving sequence; P (m) is training sequence; p ^*(m) be the conjugation of training sequence; L is the training sequence cycle, lower same; for the average of L continuous input sample point;

for the average of training sequence;

Step 2: obtain set by step 1

wherein δ _tfor default threshold value;

Step 3: in set A, M element composition that is spaced apart L gathered, by the header element composition set of qualified all set

above-mentioned M is training sequence periodicity, lower same;

Step 4: find out continuous N the element that is spaced apart 1 in set B, and the collating sequence of this N element composition is formed to set $C=\{b\left(m_{q_0}\right),b\left(m_{q_1}\right),...,b\left(m_{q_{N-1}}\right)\},$ Wherein $b\left(m_{q_1}\right)-b\left(m_{q_0}\right)=1,b\left(m_{q_2}\right)-b\left(m_{q_1}\right)=1,...,$ Above-mentioned N is the ratio of IQ speed and bandwidth, lower same;

Step 5: form preprocessed data sequence S by receiving sequence s (n) and set C ₀, S ₁..., S _n-1, wherein:

$S_x=\{s\left(b\left(m_{q_x}\right)\right),s\left(b(m_{q_x}+N)\right),s\left(b(m_{q_x}+2N)\right),...\}(x=\mathrm{0,1},...,N-1)$

Step 6: the preprocessed data sequence of step 5 gained is calculated to synthetic corresponding new preprocessed data sequence S:

S=α ₀·S ₀+α ₁·S ₁+...+α _N-1·S _N-1

Wherein: ${\mathrm{\α}}_k=\frac{\frac{1}{{\mathrm{\β}}_k}}{\frac{1}{{\mathrm{\β}}_0}+\frac{1}{{\mathrm{\β}}_1}+...+\frac{1}{{\mathrm{\β}}_{N-1}}}(k=\mathrm{0,1},...,N-1)$

${\mathrm{\β}}_k=\frac{1}{M}\·{\mathrm{\Σ}}_{m=0}^{M-1}r(C\left(k\right)+m\·L)(k=\mathrm{0,1},...,N-1).$

Further, in described step 1

expression formula be:

in described step 1

expression formula be: $\stackrel{\‾}{p}=\frac{1}{L}\·{\mathrm{\Σ}}_{m=0}^{L-1}p\left(m\right).$

Technique effect:

Preprocessing algorithms during the perfect EVM based on many times of bandwidth samplings measures, combine the N dot information of Data processing, make data from the sample survey include the full detail of initial data, thereby reduce the distance that sample point departs from planisphere standard point, EVM is measured more accurate, improved the EVM index that tester is tested.

Embodiment

The invention will be further described below.

The inventive method is the improvement of data preprocessing part during existing EVM is measured, and mainly comprises the steps:

Step 1: the mould r (n) that calculates the cross-correlation coefficient of receiving sequence and the training sequence that adopts:

$r\left(n\right)=\left|\frac{{\mathrm{\Σ}}_{m=0}^{L-1}s(n+m)\·p^{*}\left(m\right)}{\sqrt{{\mathrm{\Σ}}_{m=0}^{L-1}{[s(n+m)-\stackrel{\‾}{s}\left(n\right)]}^2}\·\sqrt{{\mathrm{\Σ}}_{m=0}^{L-1}{[p\left(m\right)-\stackrel{\‾}{p}]}^2}}\right|$

Wherein: s (n) is receiving sequence, n=0,1,2 ...; P (m) is training sequence, m=0, and 1 ..., L-1; p ^*(m) be the conjugation of training sequence; L is the training sequence cycle, lower same;

for the average of L continuous input sample point;

for the average of training sequence;

expression formula be $\stackrel{\‾}{s}\left(n\right)=\frac{1}{L}\·{\mathrm{\Σ}}_{m=0}^{L-1}s(n+m);$

expression formula be

because adopted training sequence is known, therefore

can calculate in advance as constant and use;

Step 2: the mould r (n) that recording step 1 calculates gained is greater than threshold value δ _ttime n value corresponding to r (n), and these n values are formed to set A, that is:

Obtain set by step 1

δ _tfor default threshold value, δ _t∈ (0,1);

Step 3: in set A, M element composition that is spaced apart L gathered, can be formed several qualified set in set A, by the header element composition set B of qualified all set of composition;

header element be designated as b (m), k _{m '}=[m ', m '+L ..., m '+(M-1) L] (m '=0,1 ...);

above-mentioned M is training sequence periodicity, lower same;

Step 4: find out continuous N the element that is spaced apart 1 in set B, if there is such one group of N element, represent to have found corresponding collating sequence, the collating sequence of this N element composition is formed to set C, that is:

$C=\{b\left(m_{q_0}\right),b\left(m_{q_1}\right),...,b\left(m_{q_{N-1}}\right)\},$ Wherein $b\left(m_{q_1}\right)-b\left(m_{q_0}\right)=1,b\left(m_{q_2}\right)-b\left(m_{q_1}\right)=1,...,$ Above-mentioned N is the ratio of IQ speed and bandwidth, lower same;

Step 5: form preprocessed data sequence S by receiving sequence s (n) and according to set C ₀, S ₁..., S _n-1, wherein:

$S_x=\{s\left(b\left(m_{q_x}\right)\right),s\left(b(m_{q_x}+N)\right),s\left(b(m_{q_x}+2N)\right),...\}(x=\mathrm{0,1},...,N-1)$

Step 6: the preprocessed data sequence of step 5 gained is calculated to synthetic corresponding new preprocessed data sequence S:

S=α ₀·S ₀+α ₁·S ₁+...+α _N-1·S _N-1

Wherein:

${\mathrm{\α}}_k=\frac{\frac{1}{{\mathrm{\β}}_k}}{\frac{1}{{\mathrm{\β}}_0}+\frac{1}{{\mathrm{\β}}_1}+...+\frac{1}{{\mathrm{\β}}_{N-1}}}(k=\mathrm{0,1},...,N-1);{\mathrm{\β}}_k=\frac{1}{M}\·{\mathrm{\Σ}}_{m=0}^{M-1}r(C\left(k\right)+m\·L)(k=\mathrm{0,1},...,N-1).$

Method test of the present invention:

Based on the relevant agreement of OFDM, utilize 802.11a comprehensive test instrument to implement the performance test of this method, other algorithm links in EVM test remain unchanged, only change preprocessing algorithms part, utilize short training sequence in test, relevant parameter is set as: N=2, L=16, M=10.

The test of scene 1:WiFi module data

By Analysis On Constellation Map, utilize EVM value corresponding to existing method to be-29.08296dB, utilize EVM value corresponding to the inventive method to be-30.791124dB.

Scene 2: comprehensive test instrument loopback test

By Analysis On Constellation Map, utilize EVM value corresponding to existing method to be-31.1202dB, utilize EVM value corresponding to the inventive method to be-32.8273dB.

Visible, the inventive method is effectively in EVM measures, and can improve the EVM index that tester is tested.

Claims (2)

1. the data preprocessing method in the EVM measurement based on many times of bandwidth samplings, is characterized in that comprising the steps:

Step 1: the mould r (n) that calculates the cross-correlation coefficient of receiving sequence and training sequence:

$r\left(n\right)=\left|\frac{{\mathrm{\Σ}}_{m=0}^{L-1}s(n+m)\·p^{*}\left(m\right)}{\sqrt{{\mathrm{\Σ}}_{m=0}^{L-1}{[s(n+m)-\stackrel{\‾}{s}\left(n\right)]}^2}\·\sqrt{{\mathrm{\Σ}}_{m=0}^{L-1}{[p\left(m\right)-\stackrel{\‾}{p}]}^2}}\right|$

Wherein: s (n) is receiving sequence; P (m) is training sequence; p ^*(m) be the conjugation of training sequence; L is the training sequence cycle, lower same;

for the average of L continuous input sample point;

for the average of training sequence;

Step 2: obtain set by step 1

wherein δ _tfor default threshold value;

Step 3: in set A, M element composition that is spaced apart L gathered, by the header element composition set of qualified all set

above-mentioned M is training sequence periodicity, lower same;

Step 4: find out continuous N the element that is spaced apart 1 in set B, and the collating sequence of this N element composition is formed to set $C=\{b\left(m_{q_0}\right),b\left(m_{q_1}\right),...,b\left(m_{q_{N-1}}\right)\},$ Wherein $b\left(m_{q_1}\right)-b\left(m_{q_0}\right)=1,b\left(m_{q_2}\right)-b\left(m_{q_1}\right)=1,...,$ Above-mentioned N is the ratio of IQ speed and bandwidth, lower same;

Step 5: form preprocessed data sequence S by receiving sequence s (n) and set C ₀, S ₁..., S _n-1, wherein:

$S_x=\{s\left(b\left(m_{q_x}\right)\right),s\left(b(m_{q_x}+N)\right),s\left(b(m_{q_x}+2N)\right),...\}(x=\mathrm{0,1},...,N-1)$

Step 6: the preprocessed data sequence of step 5 gained is calculated to synthetic corresponding new preprocessed data sequence S:

S=α ₀·S ₀+α ₁·S ₁+...+α _N-1·S _N-1

Wherein: ${\mathrm{\α}}_k=\frac{\frac{1}{{\mathrm{\β}}_k}}{\frac{1}{{\mathrm{\β}}_0}+\frac{1}{{\mathrm{\β}}_1}+...+\frac{1}{{\mathrm{\β}}_{N-1}}}(k=\mathrm{0,1},...,N-1)$

${\mathrm{\β}}_k=\frac{1}{M}\·{\mathrm{\Σ}}_{m=0}^{M-1}r(C\left(k\right)+m\·L)(k=\mathrm{0,1},...,N-1).$

2. the data preprocessing method in the EVM measurement based on many times of bandwidth samplings according to claim 1, is characterized in that:

In described step 1 expression formula be:

In described step 1 expression formula be: