DE2014529B2 - Digital correlator - Google Patents

Description

The invention relates to a digital correlator for correlating electrical digital signals, with series-connected, each causing a delay of the supplied digital signals by one clock interval Delay circuits with several coincidence gates, one of which has a first input one of the delay devices is connected to the output and a second input to the Digital signals is applied, and with several counters, each with an output of the coincidence gate are connected.

For many purposes in science and technology it is necessary to determine the cross correlation function d. H. the existence of some kind of connection between two or more random variables, which are generally available as an analog signal, the limit being cross-correlation represents the autocorrelation when the two signals to be combined are up to a phase shift are the same.

With an already known digital correlator of the type mentioned at the beginning (see IBM Technical ίο Disclosure Bulletin, Vol. 4, No. 7, December 1961. P. 51) nothing further is said about the origin of the electrical digital signals, while another is known similar digital correlator (cf. German patent specification 1 184 533), which however does not have a counter having, is provided in a device for machine character recognition, the digital signals can be obtained by adding a recognizable character called a detent; with light and dark egg mentarquadralen is present, from a light point scanner scanned under control of a vertical and a horizontal tilt generator and that light reflected by the character to be recognized is picked up by a photocell, its electrical Output signal represent the electrical digital signals corresponding to the light and dark, Elementary squares correspond to characters that are already digitally available.

In contrast, the invention deals with the correlation of electrical digital signals that arise through digital conversion of analog signals.

Correlators that have already been developed obtain electrical digital signals from the analog signals by using the Analog signals are fed to a cut-off circuit, which when a cut-off level is exceeded or undershot the analog signal produces an electrical digital signal with individual signals 1 or O am Output supplies, see for example F 1 g. 2a and 2, the horizontal line in FIG. 2 a the clipping level represents. F i g. However, FIG. 2 also shows that the cut signal (cf. FIG. 2c) is only a very rough one Approximation of the analog signal (see. Fig. 2a), i. H. information is lost when the analog signal is cut off.

The already developed correlators specifically perform the autocorrelation of the already cut off Analog signal through, d. i.e. the cut analog signal or digital signal is split into two channels, one of which feeds the signal delayed into one input of the coincidence gate, while it gets to their other input directly via the other channel. This way it becomes autocorrelation an already cut analog signal is linked to itself with a delay. The combination of a clipped analog signal with a delayed clipped analog signal then results in a correlation function that includes both has become inaccurate by truncation as well as by multiplying.

In contrast, it is the object of the invention to provide a digital correlator of the type mentioned at the beginning creating a more accurate correlation function.

According to the invention, this object is thereby achieved solves that the incoming, identified in each clock interval by a variable number of pulses Data signals to the / wide inputs of the coincidence gates directly and the first of the

Delay devices connected to each other (stages of the shift register) are fed via a cut-off circuit (bistable multivibrator), which supplies a signal 1 or O to the output for each clock interval according to the exceeding or non-exceeding of the number of pulses via a reference value.

According to the invention, the truncation is therefore only carried out in one channel, since the data signal are fed directly to the second inputs of the coincidence gates, so that a truncated Signal with the clipped and delayed Signa! what is linked in the coincidence gates a reduction in the information loss associated with the clipping and thus a more accurate one Correlation function causes.

The invention is further developed in that the reference value corresponds to the mean pulse count rate is set.

For this case of setting the reference value, the value obtained by the correlator is approximated (cut off) autocorrection function equal to the real one.

When the mean pulse counting rate approaches zero, the reference value also expediently becomes chosen equal to zero.

In the latter case, the cut-off circuit is simply a bistable multivibrator whose set input is the incoming digital signals and their reset input can be supplied with the clock signal.

A particularly important application of the digital correlator according to the invention is the determination of the Diffusion coefficient of a material or the determination of its molecular weight.

Determining the molecular weight of large molecules such as protein molecules has so far been difficult and time consuming. According to the Svedberg formula, the molecular weight is proportional to the sedimentation weight divided by the diffusion coefficient. While the sedimentation weight can be determined relatively easily in an ultracentrifuge, the determination of the diffusion coefficient has been very time-consuming up to now.

The digital correlator according to the invention for determining the diffusion coefficient of a material is now characterized in that the digital signals are the output of a single photon detecting photon detectors that emanate from a laser and from the suspended Material scattered light is supplied.

With this application of the correlator according to the invention, the measurement of the diffusion coefficient is now possible previously possible within minutes as opposed to several days.

The invention is explained in more detail with reference to the drawing. Hs shows

F i g. 1 shows the block diagram of a preferred application of the correlator according to the invention, namely to determine the diffusion coefficient,

F i g. 2 several wave trains over time to explain the theoretical basis of the invention, and

F i g. 3 shows the block diagram of a detector in an exemplary embodiment of the one according to the invention Correlators.

According to FIG. 1 contains a transparent container 1 suspended material, the molecular weight of which is to be determined. The material in the container 1 is irradiated with strong radiation from a laser 3. Light scattered by the material in the container 1 is collected and detected by a photon detector 5, the axis of which is inclined at an angle (-) to the axis of the laser and the output signal of which is fed into a digital correlator 7 to form the autocorrelation function. This arrangement works as follows:
The frequency of the light scattered by the material in the container 1 is compared to the frequency of the incoming

Falling light from the laser 3 shifted by frequency shifts that are characteristic of Brownian Molecular movement of the material in container 1 are. It follows that the mean frequency shift characteristic of the diffusion coefficient

s cient of the material in the container 1, so that the diffusion coefficient of the material in the container I can be determined by evaluating the spectrum of the scattered light. The digital correlator 7 for forming the autocorrelation function is used to obtain sufficient information from the spectrum of the scattered light in order to calculate the diffusion coefficient. This is done by the so-called autocorrelation, which is now based on FIG. 2 is to be explained, where several wave trains are plotted over time in order to explain the theoretical basis of the device according to the invention.

We assume a wave train £ (f), which has a complicated frequency spectrum,

z. B. a wave train (α) from F i g. 2. When such a wave train is shifted in time by an interval the wave train E (i + τ) results. The autocorrection function G (r) is now defined as

X.

G (t) = Γ E (t) E (t + T) di,

and its physical meaning is that it represents the area under the curve Γ (ί, τ), the value of which at any point in time t is given by

F (Ur) = E (t) E (t + τ).

C (τ; therefore depends on the value of the interval τ .

Since it is impossible to measure every wave train over an infinitely long time, it becomes practical Reasons G (t) as mean <F (r, r)> of F (r, r) Are defined

G (t) = < E (t) E (t + τ)>.

;, o The calculation of such a function, however, requires the calculation of F (J ₀₅ T ₀ ) for a large number of values t ₀ , T ₀ .

It has already been proven that if E (t) is replaced by the statistical telegraph function E _k {t) :

E _k (t) = 1 for E (O> k ,
E _k (t) - = 0 for E (t) < k
for any value of k, it follows:

ü _k (r) = <E _k (t) E _k (t

= -sin "'G (T),

provided that k is the mean E of E (f). The wave train (r) of FIG. 2 shows the function E _k (t) for k = 0. The replacement of E (O by E _k {t) is called clipping.
In practice this means that if the

Concentration of the material in container 1 in FIG. I is low, little light is scattered by the material and the photon detector 5 must be very sensitive, namely respond to the individual photons got to. which are emitted in the direction of the photon detector.

This means that the wave train that is to be subjected to an autocorrelation has only integer values and n (l) can be written, where n is equal to the number of photons that are in the photon selector 5 in the interval between (ι - -, TJ and

(r + j 7 ") occur, where 7 'is a small time interval. Under these conditions, a normalized autocorrelation function g' ² '(r) can be defined which is given by

g ^l2l ( _T ) = <n (f) / i (f + τ) > it ² , with H - ·> η ■.

The truncated function n _t (t) can be defined as

H ₁ (O = I for n (t)> k.
H ₁ (O = 0 for / i (() sr k.

n _k {t) can be measured with an execution easier than »(/).

Two truncated forms of the normalized autocorrelation functions appear useful. The first form j »'?' (T) is delineated as

g'r'fr) = <H ₁ (IIn ₁₁ H 4 r)><W ₀ > \

where hU) and (f + τ) are both truncated at 0 It can be shown that

for η ■■ < 1 applies.

The second form jj'fV) is defined as

g'ifV) = < n _k (t \ n (t -t 7) - ν N ₁ ■ 11
It can be shown that

^{Sl <T +} 1 -ι η ^{s tT} "

where 1, '"V) is the Fourier transform of the function H (<nl is:

¹¹¹ Ui =

n (...> exp (-; 2.

and mi. ·)) is the spectrum of the phonon counting nation: the corresponding relationship for g ^a V) is

So if k = η , ie the signal is clipped at the mean count rate, the clipped autocorrelation function is equal to the correct option, if a channel is clipped at I- = 0, the clipped autocorrelation function approaches the correct one. when ssch Γι Ο approaches.

F i g. 3 shows the block diagram of a detector and an exemplary embodiment of the digital correlator according to the invention with a cutoff circuit that delivers a signal 1 or O at the output for each clock interval Γ in accordance with the exceeding or not exceeding of the number of pulses via a reference value k = 0 in order to achieve the normalized autocorrelation function g _t ^u V) (as defined above).

A photomultiplier II feeds a pulse shaper 13. The photomultiplier 11 and the pulse shaper 13 together form the photon detector 5 of FIG. I.

ίο The output signal of the pulse generator 13 is sent through an inhibition gate 15 to be split into two channels.

One channel feeds a cut-off circuit in the form of a bistable multivibrator 17 (RS flip-flop), which outputs an output signal I or O to a first stage 25, a shift register 25, which is passed on by the clock signals from a clock unit 23. The clock signals from the clock unit 23 also reach the inhibition input of the inhibition gate 15 and the reset input of the bistable multivibrator 17. The shift register 25 has seventeen stages 25 ₁ , 25 2. 25, ... 25 _r , whose output signals are connected to a first input of coincidence.gates 21, .21, .2I ₁ . . . or 21 _{r are} supplied.

js The other channel from the inhibition gate 15 is fed in parallel into a counter 19 ″ and into a second input of the coincidence gates 2I ₁ to 2I _17. The output signals of the coincidence gates 2I ₁ to 21 _{| 7} reach further counters 19, 19 ₂ , 19, .... or 19 _n .

In operation, each photon received by the photomultiplier 11 is converted into a pulse which is processed by the pulse shaper 13 to give a pulse of a certain or constant height, direction. To give rise and fall rates. At the end of each clock interval T , the clock unit 23 emits a pulse which temporarily blocks the passage of signals through the inhibition gate 15, resets or clears the bistable flip-flop 17 and the shift register 25 in such a way that the first stage 25, the State of the bistable multivibrator 17 is entered before the clock pulse. A pulse fed into the bistable flip-flop 17 by the pulse shaper 13 during a Taktinlcrvalls T sets the bistable flip-flop 17. while further pulses that are received during the same Taktinlervalls do not change its state. The bistable flip-flop 17 cuts the signals from the pulse shaper 13 at 0 . that is, it delivers 7 "for each clock interval according to whether it is exceeded or not exceeded

so the reference value O by the pulse number from the pulse shaper 13 a signal 1 or O at the output.

The arrangement shown in Fig. 3 is therefore for weakly dispersing liquids used when the Number of photons received or the pulse

ss number is small. d. h for η ^ 1; k = 0. The clock interval T is chosen so that it is significantly shorter a!> the characteristic time of the signal to be examined.

The pulse number wire generated by the pulse shaper

ho to a simply or once truncated correlation function processed.

At the beginning of the measurements, it is ensured (in a customary fin direction not shown in FIG. 3) that the counters 19 ₀ to 19, _{7 are} cleared

(If all stages of shift register 25 are set to "0", sirui The first clock pulse from the clock unit 23 set! di < bistable multivibrator 17 to "0". any of the pulse formcr 13 during the first clock interva'-K /

cmr ^r pulses are stored in the counter 19 ,, and nowhere else, since all coincidence gates 21 to 2I _{n are} held blocked because all stages of the shift register 25 are "()" If any pulses are present in the first clock interval 7, the bistable multivibrator 17 is set to "I". At the linde of the clock interval 7 of the 'laklcinhcil 23, the shift register 25 resounds by the content of the bistable flip-flop 17 being fed into the first stage 25 at the end of the first clock interval 7

The clock unit 23 also resets the bistable multivibrator IV to M) << and prevents the arrival of pulses from the pulse shaper 13 while the shift register 25 is running (by blocking the inhibition gate 15

During the second Ί .Λ t inier valls the number of pulses from the pulse shaper 13 in the counter 1 *) ,, / uaddicri, and, if the first stage 25, the shift

7 _{) lr}

I ¹ J ₁₁
/1,

nji, register 25 is at "1", the number of pulses received by the pulse shaper 13 during the / wide clock interval 7 'is stored in the counter 19 via the coincidence gate 21. This process continues for very many Ί actinlei valle Γ fort. The content of the counters 19 "to 19, ₇ , which is generated in this way for a certain time as IK / after the start of the measurement, can be seen from the long table in which the columns in the individual counters are equivalent Specify times according to each line of stored rumble. Therefore, the sums of all values in the individual columns result in the total number stored in each individual counter. In the table, n _{x determines} the number of pulses in the time interval 7 _T crfaUtc and H _x the Boolean limitation. which is defined by

/ Miller

i Γ ».

! Il JK ι i lh η'Κ

ί ". JK ·, Ii, 0 for n _x
H, I for n _v

Γ ',

0, 0

i'JK ■ 'K ₁₁ IK

Each / ihlei ', K ^- H ". ·! ■"'^{<ll (} 'Diflei.Mi / < /wiM.hcn two Zcii.nicrv., 11.,. «La. .1 h the content of the payer 19, deriving from the Vc-r / og. ^- i ung , ι I is equal to η Ii ,,,, ν.a ·. pioix '"""' · '' ^ ¹ " ¹ Aulocorrelations-

function «. '" (Ο im 1 ¹ ^ "-' ^{slll | in llu} '"'^{<lcn/; ihlcr} " ^H 'to 19 _{,, ges rl} .hc— /ahl.·.

r, .nklion, r " ^lJ Iv, dr / cn m S, hntten von / da ,.

....., J-, "<! <■ ·>. · Μ .-, Im.-id .-- HeZUgsWerl O very close

comes', also d .. -. · η Aiilokonvlatinnslunkt.on

viii come close. I>. · ■ Ar./ahl der Wi.dr, liolungcn der follow the sicb /.- hn S.> .. ϋ · «hcstimmi die (, cnauigke.t the knrr «ilatioii ', f =! iiUion

De, content «» - A.hU- 1% In 19, ₇ Rr e.nc graph.sche An /. - .. ■ · - '"'» "« «"' ^] f ^,.', ^ .Ez / Mgl bcnui.'i wr-du.>', E cr /.- .. glc Autokoirc-

iauonsnmk.ion? ■ «" all, ememcn, men exponeiv

ι. ₍ · ||. · Η waste | ..>! Hi. · 1 <: γ in turn the louiici-

ΙΪ, γΑ ".« ■■ "month fluctuation spectrum

of the Ma ,,.,.! ■■ ·. Contents, 1 (cf. Mg. Il ge - K-h »s - ■ 'Di-- Fall-off time of the Lxponential-

^^! srm.t.l.n ^ M ^ aou.ken

,,,! -, imHehidu, I ^- - '' »■ 'I m l-logically ve,

'iic watch, ·.> ..... k .., rolaMonsf »nkt.on ha. die lorm

2 /), K ' ² -I autocorrelation function is characterized by a Λ' · 1 D ₁ K ' , where /) ·, tier Di! coefficient is De, DifTusionskoefliz.icnt can keep grazing by the Dalcn in the above '. ti are used. The mean value ή of iit'l ^K corresponds to the total counter reading in counter I9 _O ; The mean value n "of" (i) is calculated from the values in the counter 19. The value v> given dun h

Y4i / Vsin ₂ (-λ /

I ι r,
case ·, the scattering .in random, process isl this

1 sheet of drawings of the refractive index of the solution, in <' Material is suspended, Strcuwinkel (i.e. the angle 'unit' the laser beam is deflected). Wavelength of the laser radiation in Va '

The low 11 approaches - ~ n ,, · the ^ ' ^: for higher values of ή a correction dm · ^! > which depends on the statistics of the gostii'i light

By linking the diffuse thus determined Coefiicient with the sedimentation rate, the m cm I llllra centrifuge in about an hour wc,>! can, the molecular weight of the material container 1 of Fig. I can be determined using dti GI cluing from Svedberg as stated above be reckoned

Claims (5)

Patent claims: 1.Digital correlator for the correlation of electrical digital signals, with one behind the other, each delaying the supplied digital signals by one clock interval causing delay circuits, with several coincidence gates, of which a first input is connected to the output of one of the delay devices and a second input is applied to the digital signals is, and with several counters which are each connected to an output of the coincidence gates, characterized in that the incoming data signals, identified in each clock interval (T) by a variable number [n (t)] of pulses, to the second inputs of the coincidence gates ( 2I ₁ ... 2I ₁₇ ) are fed directly and the first of the series-connected delay devices (stages 25, ... 25 _{17 of} the shift register 25) via a cut-off circuit (bistable multivibrator 17), which exceed or exceed for each clock interval (T) Not overc the pulse number [n (r)] rises above a reference value (fc; in Fig. 3 k = 0) delivers a signal 1 or O at the output. 2. Digital correlator according to claim 1, characterized in that the reference value (k) is set according to the mean pulse count rate (Ti). 3. Digital correlator according to claim 1, characterized in that the reference value (k) is equal to zero. 4. Digital correlator according to claim 3, characterized in that the clipping circuit is a bistable multivibrator (17) whose set input is the incoming digital signal and whose Reset input, the clock signal can be fed. 5. Digital correlator according to one of claims 1 to 4 for determining the diffusion coefficient of a material, characterized in that the digital signals are the output signal of a single photon detecting photon detector (5), which is that of a laser (3) light emanating from and scattered by the suspended material is applied.