SU1755211A1 - Method of fast spectrum analysis of wide-band processes - Google Patents

Abstract

The invention relates to a technique for electrical measurements and rapid spectral analysis of random processes in a wide frequency band. The purpose of the invention is to increase the speed of analysis, which is achieved by a number of specific operations and relationships given in the claims. 1 il.

Description

The invention relates to the field of electrical and radio measurements and rapid spectral analysis of random processes in a wide frequency band.

The aim of the invention is to increase the speed of analysis.

The essence of the method lies in the fact that in the well-known method of spectral analysis, in which discrete Fourier transforms (DFTs) are used to estimate the correlation function of the observed process, the frequency range fmin, fmaxj under study is divided by frequency filters into I subfields fHi. fei ... I, select the sampling time in each sub-band (fBi-fHi) and calculate in each channel m values of the power spectral density, the boundaries of the sub-bands are chosen from the ratios fHi / (fei-fHi) tC fan-i fei, ferfmax, - I, where K is any natural number, estimates of the correlation function Kx (ri Ati),. m are obtained in parallel in all channels, the DFT and the calculation of the estimated power spectral density ex (fm + n Ati), n 1 ... t, at m discrete frequencies are performed separately in each 1-volume

from channels I, and the resolution of the analysis is Afi (fei-fHi) / ni Afi

/ v + iV 1

Y-gt-, and DFT in each i-volume from I kV to /

tax is produced by summing

estimates of the correlation function Kx (hAti), with weight coefficients B "(h) not dependent on subband number i and equal to

WITH/

WITH

Brfhh

0 ,,

pp

cos

m

 ... m

-ABOUT)

if K is an even number and

0,

CJ CJ

N:

Brfh) (-1) h.cos

m

... m

(2)

if K is an odd number and multiplying the sum obtained by 4Ati

If the process under investigation is limited by the frequencies fH. fa such that) is equal to the integer number and sampling time (fe-fH), then the spectral window is an important estimate of the spectral power density

rt ISMf- 3

has only one main maximum in a given range, and the errors of repeated superposition of high-frequency components are absent, therefore, if there are input filters with a sufficiently rectangular frequency characteristic, the estimate of the spectral power density of the filtered implementation at the n-point tjx (fH + n AO, ... m MI., "But consider the estimated spectral power density of the input process at the frequency fH + n Af, Af (fB-fH) / m Formula for estimating the spectral density Gx (f) in this case can be converted to

Gx (fH + nA f) 4, 5Kx (0) +

+ 2 MhA t) cos (fH + nA t) 2 hAt (4)

h 1

given that A t 1/2 (fB-fH). A f (fe- -fH) / m and fH / (fe-fH) K, follows

Gx (fH + n A f) 4, 5Kx (0) +

+ .2 MhAt) (- 1), (5)

m

J

whence the expressions for the weighting factors (1) and (2)

From the above reasoning and the type of formulas (4), (5) it can be concluded that whatever the number of subband I,., I, if the boundaries of subbands are chosen from the relations fHi / (fBi-fni) K; fHi + i fBi, ... I, where K is any positive integer and (fBi-fHi), then to obtain estimates of the spectral power density at m discrete points with step Afi (fBi-fHi) / m in any of the I channels, DFT device with the same set of function values (-1) nxoz (lgp / t)

ex (fHi + n AfiH, 5 Kx (O) +

J / x ChAt.X-ircos, (6)

1 1 ... I

 ... m

where Kx () is the estimate of the correlation function obtained in the ith channel.

The drawing shows a functional diagram of the device that implements a method for rapid spectral analysis of broadband processes.

The device contains four channels

analysis 1, each of which includes an input filter 2 connected in series, a multi-correlator 3 for six values of the multi-correlation function,

switch 4, resistor divider 5, analog-to-digital converter (ADC) b, digital multiplier 7 and accumulating adder 8, the outputs of all channels are connected to the inputs of the consumption unit

information 9, which includes a random access memory (RAM) 10, as well as a control unit 11 containing a clock pulse generator 12, a counter-divider of six 13, two counters

14, 15, two RAMs 16, 17.

The device works as follows. From the clock generators (GTI) 12 to the counters 13-15, the pulses are successively received. With the arrival of the next pulse at the output of the counter 13, a digital code is established that simultaneously enters the control inputs of the switches 4 of each channel, causing a connection to the output of each channel switch 4 of one of the outputs of the channel multi-correlator 3. From the output of switch 4, a voltage divider 5 to the ADC 6, from which it is digitally fed to one of the inputs

multiplier 7. The mentioned clock pulse goes to counters 14 and 15 that form an eight-bit counter. The output code from these counters goes to address inputs of RAM 16, 17, which make up the device

memory (prior to the beginning of the work, the values of the weighting factors Bk (n)

 (-1) Khcos, .., 5, ... 5,,).

From the outputs of RAM 16, 17 simultaneously the digital code of the weight coefficient B (h) corresponding to the current values of n and h enters the inputs of the channel multipliers 7 of each channel.

Thus, on the two inputs of the multiplier 7 of each of the channels, digital values of the correlation function K (hAti) and weight coefficient Bk (n) respectively appear at the current values of n and h. The result of the multiplication enters the digital accumulator adder 8. The described procedure is performed six times for each n, then at the output of the adder 8

For each of the channels i 4, the spectral density G (Ati) appears.

Under the action of a pulse coming from the divider counter to six 13, the indicated spectral density estimates from all channels simultaneously ... 4 are recorded in the information consumption block 9, where each channel has its own RAM 10.

After this, from GTI 12 a channel impulse goes to channel adders 8 of all channels, and the whole described process repeats for the next value of n. After repeating the procedure for all ... 5, all I m 20 spectral density values are in the information consumption block 9 .

Claims (1)

Claims of the Invention A method of fast spectral analysis of broadband processes in which the estimate of the correlation function of the process Kx (h At) at m + 1 points, 1, ... is determined using electrical signals using the discrete transformation of the hfury of the correlation function Kx (h At) determine the estimated power spectral density Gx (f) at m points of frequency, characterized in that, in order to increase the speed of the analysis, the original signal in the test The frequency range (Cm, max) is divided by smart filters into I signals in the f, i, fail subbands. M ... I, whose boundaries are found from the relations JMI fel - fnl K, fHr Gmin, (H-l); . .. I, (where K is any positive integer), parallel to all I filtered signals are sampled with Ati-1/2 times (fBi-fHi): Y, ... v are determined by their estimates of the correlation function Kx (hAti) ( where, 1 ... m) is parallel in all ... I channels, the received signals are weighed with coefficients BK (h Al), not dependent on the channel number, equal to five VK (PDP) 0.5 (-1) .... t, l g t i t and then summarize, and the resulting total voltages in all ... I channels are weighed again with a factor of 4 Ati, which is proportional to the spectral density estimates Gx (fHi + Afi An), ... m.