ebook img

Demonstration of negative group delays in a simple electronic circuit PDF

6 Pages·0.15 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Demonstration of negative group delays in a simple electronic circuit

Demonstration of negative group delays in a simple electronic circuit T. Nakanishi, K. Sugiyama, and M. Kitano Department of Electronic Science and Engineering, Kyoto University Kyoto 606-8501, Japan (Dated: February 1, 2008) Wepresentasimpleelectroniccircuitwhichproducesnegativegroupdelaysforbase-bandpulses. When a band-limited pulse is applied as the input, a forwarded pulse appears at the output. The negative group delays in lumped systems share the same mechanism with the superluminal light propagation, which is recently demonstrated in an absorption-free, anomalous dispersive medium [Wang et al., Nature 406, 277 (2000)]. In this circuit, the advance time more than twenty percent ofthepulsewidthcaneasilybeachieved. Thetimeconstants,whichcanbeintheorderofseconds, 2 is slow enough to be observed with thenaked eye bylooking at thelamps driven by thepulses. 0 0 2 I. INTRODUCTION shift is a linear function of the frequency. n The interferencedoesnotoccuronlyinthe lightprop- a Brillouin and Sommerfeld investigated the propaga- agation, but also in other wave or oscillation dynamics. J tion of light pulse in a dispersive medium described by Signals in electrical circuits also interfere. In lumped 1 the Lorentz model. They pointed out that in the region systems, we cannot define the group velocities, because 1 of anomalous dispersion the group velocity vg could be there exists no finite length scale. Instead, we candefine v larger than c, the light velocity in a vacuum, or even be the group delay. The group delay is the time difference 1 negative [1]. The anomalous dispersion occurs near the between the input and output signal envelopes. If it is 0 center of absorption lines. In the case of superluminal negative, the output pulse precedes the input as shown 0 group velocities (v > c), the transit time of the light in Fig. 1. The negative delays are closely connected to g 1 envelope through the medium is shorter than that for a thesuperluminalornegativegroupvelocitiesinspatially 0 vacuum with the same length. In the case of negative extended systems. 2 0 group velocities (vg < 0), the envelope leaves before it In this paper, we propose a simple electronic circuit / enters into the medium. which generates negative group delays. While the dis- h Chu and Wong demonstrated experimentally that the persion relation k(ω) of dispersive media determines the p phase shifts in the light propagation, the transfer func- light in a Ga:N crystal can be propagated at such ex- - t traordinary group velocities [2]. But in these kind of ex- tion H(ω) of the circuit determines the phase shifts of n a perimentsitisinevitablethattheshapeofthelightpulse the output in reference to the input. Mitchell and Chiao u is largely distorted due to the absorption. The standard [6] demonstratednegativedelaysin a bandpass filter cir- q definition of the group velocity, v = dω/dk, which de- cuit to which the carrier signal modulated with a pulse g v: scribes the velocity of the light envelope or peak, tends is applied. They showed that if the carrier frequency i to lose its physical meaning under the strong distortion is located outside of the pass band, the envelope of the X of the waveform. A new definition of group velocities, output pulse precedes the input. In practical implemen- r considering the reshape of the spectrum caused by the tation,alargeinductorisrequiredanditisreplacedbya a absorption, was proposed recently [3]. Wang et al. realizedanomalousdispersionwithoutab- negative delay circuit sorptionusingthe gain-assistedlinearanomalousdisper- peak sion,anddemonstratedthelightpropagationattheneg- Η(ω) ative groupvelocities [4, 5]. For the light pulse propaga- IN OUT tioninanabsorption-freemedium, the normaldefinition of the group velocity describes precisely the propagating Η(ω) speed of the waveform, or the envelope. The pulse propagation with superluminal or negative IN OUT peak group velocities is very counterintuitive and is liable to Η(ω) causemany misunderstandings. However,itis the direct IN OUT results of the interference of the wave and is consistent time with the relativistic causality. The anomalous dispersion with no absorption induces FIG. 1: An electronic black box for negative delays. The phase shifts depending on the frequencies, thereby en- inputand theoutputare monitored by LEDs. When a pulse hances the front part of the pulse by constructive in- is fed to the input, the LED at the output lights up before terference and cancels the rear part by the destructive theLED at the input. interference. The pulse shape is conserved if the phase 2 simulated inductor with operational amplifiers. The use The group delay t is defined as d ofthecarriersignalalsoincreasesthecomplicationofthe circuit. dφ t = . (4) d The circuit that we present here generates negative −dω(cid:12)ω0 (cid:12) delays for baseband signals or signals with zero carrier (cid:12) ThentheenvelopeoftheoutputisobtainedfromEqs.(2) frequency. It is easy to produce negative delays as large and (3) as as a few seconds, therefore, we can observe it with the nakedeye by watching two LEDs (light emitting diodes) (t)= (t t )eiφ(ω0). (5) each of which is driven by the input and output pulses. out in d E E − It is helpful and illuminating to translate fundamental This means that, aside from the phase factor, the enve- physical concepts into a simple circuit which replicates lope ofthe output is shifted by the groupdelayt , while the essence of the phenomena [7, 8]. d maintaining the shape. For t > 0, the input precedes Inthenextsectionweexplaintherelationbetweenthe d the output (normal delay) and for t < 0, the output superluminalornegativegroupvelocityinthelightprop- d precedes the input (negative delay). We will represent agationandthe negativegroupdelay inlumped circuits. a circuit that satisfies the latter condition in the next Then, in Sec. III, we present a circuit generating nega- section. tive group delays and the pulse generator used for the Inordertotranslatetheabovediscussionintothelight input. In Sec. IV, we propose a method for larger nega- propagation through a dispersive medium with length tivedelaysbycascadingthecircuits. Itisshownthatthe L, we suppose E and E represent the field of in- negative delay time can be increased as √n without in- in out put and that of output of the medium, respectively. troducingadditionaldistortionofthepulseshape,where Whenthemonochromaticlightwithfrequencyωisprop- nisthenumberofthestages. Butforlargen,thecircuit agatedinthe medium, the phase ofthe field is shifted as becomes susceptible to the noise. φ(ω) = k(ω)L, where k(ω) is the wavenumber in the − medium. If k(ω) is linear in the bandwidth, with the II. TRANSFER FUNCTION FOR NEGATIVE help of Eq. (4), we have GROUP DELAY dk L t =L = , (6) d In this section we deal with a transfer function H(ω) dω(cid:12)ω0 vg (cid:12) in order to study negative group delays. The discus- (cid:12) where dω/dk is the group velocity v . The envelope sion can be applied to both electric circuits and light |ω0 g of the light is delayed by L/v . propagation in a dispersive medium. A band-limited g If the difference between the propagation time of the input signal E (t) can be expressed as a product of in envelopesinthedispersivemediumandthatinavacuum the carrier exp(iω t) and the envelope (t); E (t) = 0 Ein in with the same length L is negative, i.e., (t)eiω0t+c.c., where c.c. represents the complex con- in E jugate term. Our circuit deals with the baseband signal ∆t=t L/c=L(v−1 c−1)<0, (7) (ω0 = 0), but we include the carrier for the purpose of d− g − comparisonwithothercases. Thesignalisexpandedwith then the light propagation in the medium is called su- the Fourier component ˜ (u) of the envelope (t) as Ein Ein perluminal. There are two cases which satisfy this con- Ω/2 dition; vg >c andvg <0. Inthe formercase,the output E (t)= du ˜ (u)ei(ω0+u)t+c.c., (1) envelope precedes the output for the vacuum case but in in Z−Ω/2 E does not precede the input. In the latter case, the out- where Ω is the bandwidth and u is the offset frequency putprecedesthe input, orthe wholesystemprovidesthe from ω0. negative group delay td <0. The transfer function H(ω) A(ω)eiφ(ω) is defined ≡ for each frequency ω = ω +u and the output E (t) 0 out can be written as III. CIRCUITS AND EXPERIMENT Ω/2 E (t)= du ˜ (u)H(ω)ei(ω0+u)t+c.c. A. Negative delay circuit for baseband signals out in Z−Ω/2 E Ω/2 In Fig. 2, we show a negative delay circuit for base- = du ˜ (u)A(ω)ei(ω0+u)teiφ(ω)+c.c. (2) Z−Ω/2 Ein band (ω0 = 0) signals. This is basically a non-inverting (imperfect) differentiator. Its transfer function is easily Weassumethatwithinthebandwidth(u <Ω/2),the obtained as | | amplitude A(ω) is nearly unity and the phase φ(ω) can be approximated by a linear function, i.e., H(ω)=A(ω)eiφ(ω) =1+iωT, (8) A(ω) 1, φ(ω) φ(ω ) ut . (3) 0 d where T =RC . ∼ ∼ − 3 R 2.5 C V 2 out V in e d 1.5 u plit m 1 FIG. 2: Negative delay circuit for baseband signal can be a constructedwithanoperationalamplifier. Thetransferfunc- 0.5 tion is H(ω)=1+iωT. 0 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 We note thatthe correspondingactionin the time do- ωT main is 1+T(d/dt). The advancement of a pulse can be understood qualitatively. For the rising edge, which has a positive slope, the two terms interfere constructively, 80 while for the falling edge with a negative slope, they in- 60 terfere destructively. Thus the pulse is forwarded. e 40 In the low-frequency region ( ω 1/T ), H(ω) is e r | | ≪ g 20 approximated as e d A(ω)=1+O(ω2T2), (9) e / 0 s -20 φ(ω)=ωT +O(ω3T3), (10) ha -40 p which mean that the amplitude is nearly constant and -60 the phase increases linearly with frequency. Then the -80 group delay becomes negative: -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 dφ ωT t = = T <0. (11) d −dω(cid:12)ω=0 − (cid:12) As seen in Fig. 3, the a(cid:12)mplitude A(ω) and the phase FIG. 3: The amplitude of the transfer function A(ω) = |H(ω)|(upper)and thephase φ(ω)=argH(ω)(lower). φ(ω) of the transfer function are not linear except in ω T 1 owing to the higher order terms in Eqs. (9) | | ≪ and(10). Theseterms induce waveformdistortionofthe output. Inorderto keepthe distortionas smallaspossi- TABLE I: Circuit parameters. ble, the spectrum of the input signal must be restricted within the frequency region ω 1/T. For this purpose R0 6.8MΩ R 1MΩ low-pass filters are needed. | |≪ C0 0.22µF C 0.22µF Trec 1.5s R′ 10kΩ R1 2.2MΩ C′ 22nF B. Low-pass filter C1 0.22µF T 0.22s R2 10kΩ In order to prepare a band-limited, base-band pulse R3 2.2kΩ weintroducelow-passfilters. Theinitialsourceisarect- ωc 1.6Hz angular pulse from a timer IC (integrated circuit). The rectangular pulse has high-frequency components, and the negative delay circuit does not work correctly. We sothattheovershootforrectangularwavesissmall. The musteliminatethehigh-frequencycomponentswithlow- cut-off frequency is defined by ω =0.7861/T . pass filters. We introduce two 2nd-order low-pass filters c LP shown in Fig. 4. Each transfer function is α H (ω)= , (12) C. Experiment LP 1+iωT (3 α)+(iωT )2 LP LP − where T = R C and α = (1+R /R ). By changing We show the overall circuit diagram for the negative LP 1 1 3 2 α, we can tailor the characteristic of the filter. In our delay experiment in Fig. 4 and the parameters in Table experiment we choose the value α = 1.268, which corre- I. sponds to a Bessel filter [9]. The Bessel filter is designed The pulse generator in the upper section of Fig. 4 is 4 Pulse generator 2.5 timer IC 9V 7555 9V Low-pass filters R0 10k 2 output input R3 R3 TRGDIS SW 20k OREUSTTVHC C0 RR21 TL082 RR21 TL082 e / V 1.5 R1 C1 R1 C1 d C1 C1 u 1 plit m 0.5 a 0 input Negative delay circuit output Vin C’ C’ Vout -0.5 -1 0 1 2 3 4 5 6 7 LED 10k R’C R R’C R LED time / sec 100 100 20k TL082 TL082 FIG. 5: Experimental results. The oscilloscope traces show the input and output pulses. The output precedes the input owing to the negative delay. FIG. 4: Overall circuits. Upper section: pulse generator. Lower section: negative delay circuit. 0.5 subdivided into the generator of single-shot rectangular m=2 4 6 8 wave and the low-pass filters. In the first part, when 10 0.4 triggered by the switch, the timer IC generates a single pulse, whose width is determined by the time constant U. T = R C = 1.5s. The rectangular pulse is shaped A. 0.3 byrecthe tw0o-s0tagelow-passfilters. The total order of low- e / d 0.2 pass filter is m = 4. We set the cut-off frequency of u the low-pass filter as ωc = 0.35/T, so that A(ω) and mplit 0.1 φ(ω)canbeconsideredtobeconstantandlinear,respec- a tively, below the cut-off frequency (see Fig. 3). Finally, 0 the band-limited single pulse is sent out for the input of the negative delay circuit in the lower section of Fig. 4. -0.1 0 1 2 3 4 5 6 7 Two delay circuits shown in Fig. 2 are cascaded for larger advance time. The input and output terminals time ω t c are monitored by LEDs. Their turn-on voltage is about 1.1V. The variable resistor at the input is adjusted so FIG. 6: Responses of the Bessel filter to a rectangular wave. that the input and the output have the same height. The orders of filters are m=2, 4, 6, 8, and10. The experimental result is shown in Fig. 5. The input and output waveforms are recorded with a oscilloscope. The origin of the time (t = 0) is the moment when the switch in Fig. 4 is turned on. We see that the output precedes the input considerably (more than 20% of the pulse width). The slight distortion of the output wave- waveforms. We could dispense with an oscilloscope. form is causedby the non-idealfrequency dependence of A(ω) and φ(ω), as mentioned in Sec. IIIA. The actualnegative delay circuit in Fig. 4 differs from ′ ′ The expected negative delay derived from Eq. (11) is thatshowninFig.2. TheresistorR andthecapacitorC 2T = 0.44s; we have connected two circuits (each time are supplemented for the suppression of high-frequency constantT =0.22s)inseriesforlargereffect. Thisquan- noises. As shown in Fig. 3, the gain at ω T > 1 is | | titatively agrees with the experimental result, where the large. Although the high-frequency components of the time difference between the output and the input peaks input signal are suppressed by the low-pass filters, the is about 0.5s. The time scale is chosen so that one can internal and external noises with high frequency are un- ′ ′ directly observe the negative delay with two LEDs con- avoidable. With R and C , the high-frequency gain is ′ ′ nected at the input and the output terminals. We could limited. The parameters are chosen as RC,C R T so ≪ alsousetwovoltmeters(orcircuittesters)tomonitorthe that the phase φ(ω) for ω T <1 is not affected. | | 5 IV. REALIZATION OF LARGER NEGATIVE 1.2 DELAY e 1 d Theobtainednegativedelayintheaboveexperimentis u about20%ofthepulsewidth. Thisislargerthantheval- plit 0.8 m output input ues obtained in other superluminal-velocity or negative- a 0.6 delay experiments. We consider the way to make even d e larger negative delays. We assume noise-free environ- z 0.4 ment for a simplicity. ali m Cascadingthenegativedelaycircuits,thetimeadvance r 0.2 o can be increased. One might simply expect that, by in- n 0 creasing the number of stages n, the total time advance canbeincreasedlinearlywithn. But,unfortunately,this 0 1 2 3 4 5 6 7 is not the case. It is obvious from Eqs.(9) and (10) that time /sec the distortion of the waveform of the output is also in- creased. When n circuits are connected in cascade, the total transfer function can be written as Hn(ω). Corre- FIG. 7: A simulation result for multi-stage negative delay circuit (n=10). spondingly, the amplitude and the phase are given as n(ωT)2 An(ω) 1+ , (13) ∼ 2 butthe higherthe orderoffilterthe morethe risingpart nφ(ω) nωT. (14) is delayed. Hence we need a high order filter to attain ∼ large time advance. In other words, the pulse must be Inordertokeepthewavedistortionbelowacertainlevel, delayed appropriately in advance in order to get a large we have to limit the excess gain An(ω) 1 within the negative delay. − bandwidth by some value γ; Moreover the short-time behavior of the total circuit including the low-pass filters and the negative delay cir- n(ωT)2 n(ω T)2 c cuitsisdeterminedbythecompositetransferfunctionat =γ. (15) 2 ≤ 2 highfrequency(ω ). Theorderofthelow-passfilter →∞ mshouldnotbe smallerthanthenumberofthe stagesn Then the advance time per circuit should satisfy of the negative delay circuits. Otherwise the total trans- fer function would diverge at ω and the derivative T = 2γω−1 = 2γT , (16) of the rectangular pulse would a→ppe∞ar at the output. r n c r n w Figure 7 represents a result of simulation for the re- sponse of ten negative delay circuits with the input of where T is the pulse width. It is determined by the w a tenth order Bessel filter. The used parameters are cut-off frequency of the low-pass filter. T = 1s and γ = 0.2. The total time advance is esti- If we want to increase the time advance in conserving w mated 2s from Eq. (17). This estimation is consistent the pulse widthandthe distortionofthe signal,wemust with the result of the simulation. The advance time is reducethetimeadvanceT percircuitbythefactor1/√n. comparable to the pulse width. It is so large that the Therefore, the total time advance T scales as total input starts to rise when the output begins to fall. T =nT = 2nγT , (17) total w p V. DISCUSSION AND CONCLUSION which is a slowly increasing function of n. Inaddition,thereisanotherfactortobeconsidered. A causal transfer function cannot generate negative delays Letusconsiderapositivedelaycircuit. Acircuitcalled unconditionally. It is impossible to advance the signal all-pass filter has the transfer function [9]: beyond the time when the switch is turned on in the 1 iωT rectangular pulse generator in Fig. 4. The reason of the H (ω)= − . P 1+iωT advancement is that the slow rising part of the pulse, whichhasbeensuppressedbythelow-passfilter,isreem- The all-pass filter can be build with an operational am- phasized by the negative delay circuit. The slowness of plifier, three resistors, and a capacitor. It is very conve- rising part of the pulse is determined by the order m of nientfor generatingpositive delaysforbase-bandsignals the low-pass filter. Figure 6 represents responsesof vari- because the amplitude and phase of the transfer func- ousorderBesselfilterswiththesamecut-offω toarect- tions are given as A(ω)=1, and φ(ω)= 2tan−1ωT c angular wave with a unit height and a pulse width ω−1. 2ωT, respectively. The flat amplitude r−esponse allow∼s c − Thepulseshapes,especiallythe widthsofthepulses,are ustocascadethencircuitswithoutintroducingpulsedis- similartoeachotherowingtothesamecut-offfrequency, tortionandresultsinthe delayt =2nT. Unfortunately d 6 the negative version (T T) of all-pass filter negativedelay(0.44s)couldeasilybeachieved. Itisslow →− enoughtobeobservedwiththe LEDsorvoltmeterswith 1+iωT H(ω) the naked eye. The negative delay amounts to 20% of H (ω)= = , N 1 iωT 1 iωT the pulse width. In the light experiment [4], t L/c is d − − − 62ns and is only a few percents of the pulse width, 4µs. will not work because it is not causal. The pole of Includingthepulsegeneratorfortheinputofthenega- H ( is), s = 1/T, is located in the right half plain. N − tivedelaycircuit,theapparatusconsistsofcommonparts If one makes this circuit, it will be unstable owing to availableinanylaboratories. Itissosimplethatabegin- the time response function exp(t/T). This example tells ner can build it in an hour. The setup can be operated us aboutthe asymmetry betweenthe negative delay and stand-aloneandnoexpensiveinstrumentssuchasoscillo- the positive delay. The former is much more difficult to scopes and function generators are required. It is useful achieve than the latter. for understanding the physics of superluminal propaga- The group velocity has no direct connection with the tion as well as the negative group delays. relativistic causality, therefore, it can exceed the speed of light c in a vacuum. But the front velocity v (or the f wavefrontvelocity)is constrainedby the causalityandis equal to c, namely, v = L/t = c. In lumped systems | f| | f| Acknowledgments (L = 0), the wavefront delay t must vanish. Actually f all of the pulses in our system have their wavefronts at t = 0, the moment when the original rectangular pulse This research was supported by the Ministry of Ed- rises or the switch is turned on. ucation, Culture, Sports, Science and Technology in To conclude, we demonstrated the negative group de- Japan under a Grant-in-Aid for Scientific Research lay in a simple electronic circuit. A considerably large No. 11216203and No. 11650043. [1] L.Brillouin,WavePropagationandGroupVelocity(Aca- 053806-1 – 053806-11 (2001). demic Press, New York,1960) pp.113 – 137. [6] M.W.MitchellandR.Y.Chiao,“CausalityandNegative [2] S.ChuandS.Wong,“LinearPulsePropagationinanAb- Group Delays in a Simple Bandpass Amplifier,” Am. J. sorbing Medium,” Phys. Rev.Lett. 48, 738 – 741 (1982). Phys. 66 (1), 14–19 (1998). [3] M.Tanaka,M.Fujiwara,andH.Ikegami,“Propagationof [7] J. L. Rosner, “Tabletop time-reversal violation,” Am. J. a Gaussian wave packet in an absorbing medium” Phys. Phys. 64 (8), 982–985 (1996). Rev.A 34, 4851 – 4858 (1986). [8] W.FrankandP.Brentano,“Classicalanalogytoquantum [4] L. J. Wang, A. Kuzmich, and A. Dogariu, “Gain-assisted mechanicallevelrepulsion,”Am.J.Phys.62(8),706–709 superluminal light propagation,” Nature 406, 277 – 279 (1994). (2000). [9] U.Tietze and Ch. Schenk, Electronic Circuit (Springer- [5] A. Dogariu, A. Kuzmich, and L. J. Wang, “Transparent Verlag, Berlin, 1991) pp. 350–408. anomalous dispersion and superluminal light-pulse prop- agation at a negative group velocity,” Phys. Rev. A 63,

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.