01262349

8/6/2019 01262349

http://slidepdf.com/reader/full/01262349 1/7

8/6/2019 01262349


Polyimide

~ Polyimide

Si

a

I I+w2+

polyimide

I Si I

b

Fig. 1

polyimide Iuyers with (I shielding .strucfuTebetween them

a Microstrip lines on same layer of polyimideh Microstrip lines on different layers of polyimide

Cross-sectioml cuts through microstrip lines embedded in

3 Theoretical analysis

The fullwave FDTD technique [6]is used for the theoreticalcharacterisation of the forward and backward coupling, S,,and S,, respectively, between the two parallel microstriplines (Fig. 2). The E- and H-field components areimplemented in a leapfrog configuration. An adaptive gridwith neighbouring cell aspect ratio smaller than or equal to2 maintains a second-order global accuracy.

port 2

port 1

d

,P..--%3ia fence with

connecting stripport 4

Fig. 2

the coupling

Schematic of the coupled line structures used to characterise

Numerical meshes of 80 to 120 by 45 by 250 cellsterminated with 10 perfectly matched layer (PML) cells ineach direction provide accurate results for a time step ofA t = 0.99Atm,. A Gaussian pulse with fma = 60GHz is

applied vertically as a soft source close to the front end ofthe microstrip, and its values are superimposed on theFDTD calcnkated field value for all cells in the excitationregion for each time step. The via holes are modelled as

rectangular metal tubes with cross-section 23 by 20 pm, Toaccount for the coupling of even and odd modes, twosimulations are performed for each geometry exciting bothlines with equal amplitude and even or odd spacedistributions respectively. In addition, both microstrip linesare terminated with matched loads (ZO=50Q)hat arerealised as the combination of shunt resistors placed

IE E Pm.-Mioow Antems Propag., Vol. ISO,No. , Ocrober 2W3

between the microstrip and the bottom ground [7l. Twoprobes placed at the front end and at the far end of one lineare used for the combination of the results of the even andodd simulations. The application of the FFT algorithmderives the frequency-domain results from the time-domaindata (usually 25,000 time steps).

4

The circuit for experimentally characterising couplingbetween the microstrip lines is shown in Fig. 2. The four-port circuit has probe pads orientated so that each port maybe probed simultaneously. The coupling region, or thesection of parallel transmission lines labelled L in Fig. 2, is5000pm long. Note that for characterisation, the shieldingstructure extends past the transmission line right angle bendand is tapered at 45" to minimise the effects of radiationfrom the right angle bend.

The four-port microstrip circuits are fabricated on a 1 Q-

cm Si wafer. A ground plane consisting of a 3WA Ti

adhesion layer, 1.5pm of Au, and a 200A Cr cap layer isfist evaporated onto the Si wafer. This is followed byspinning on Dupont adhesion promoter and l0pm ofDupont PI-2611 polyimide, which has a permittivity of 3.12measured at 1 MHz [X I and a loss tangent of 0.002 measuredat 1 kHz [9]. After curing the polyimide at 340°C for

120min, Ni is evaporated onto the polyimide to serve as amask for the O&F4 reactive ion etching (RIE) of the viaholes. After the via holes are etched and the Ni removed,

200A of Ti and 2000A of Au are sputtered on to the waferto serve as a seed layer for the 1.3 pm of Au electroplatingthat is used to define the embedded microstrip lines and i l lthe via holes in a single step. These Au microstrip lines are

capped with 200A of Cr before applying the next, 10pmlayer of polyimide. Thus, all metal stluctures are ISpmthick. This process is repeated for each layer of polyimide.After each step, a DEKTAK surface protile is used tomeasure the polyimide and metal strip thickness. Both theDEKTAK and SEM analysis show that the surfaceroughness is low enough that it can be neglected in theanalysis.

Measurements are made on a vector network analyserfrom 2 to 50GHz. A thru/reflect/line F R L ) calibration isimplemented with MULTICAL [IO], a TRL softwareprogram, using four delay lines of 1x00, 2400, 4800, and10,000 pm and a short circuit reflect fabricated on the samesubstrate as the circuits. To improve accuracy, each circuitis measured several times and the average of thesemeasurements is presented in this paper. During themeasurement of the four-port circuits, two of the fourports are terminated in 50 R loads built into speciallydesigned picoprobes.

5 Microstrip coupling results

As a first step, the measured and FDTD analysis results for

the embedded microstrip lines are compared across theentire frequency band of 2 to 50GHz. One such case forcoupled microstrip lines without any shielding structure isshown in Fig. 3. It is seen that there is excellent agreementbetween the theory and the measured results, which istypical of the other cases. Also typical of all of the resultspresented in this paper, the forward coupling increasesmonotonically with frequency, while the backward couplingis periodic. Thus, throughout the paper, the backwardcoupling results presented are the maximum coupling valueover the frequency band. Presented results are backward

Circuit fabrication and experimental procedure

345

8/6/2019 01262349


-20

-25 1 - -20 I

-70, I

0 10 20 30 40 50

frequency. GHr

Fig. 3 Comparison of meu.wied and F DTD scatrering parameters

for coupled microstrip lined on the same polyimide layer (Fig. l a )wirh no shielding .srmcfure and C=69pm

coupling defined as -20 loglS4,/ and forward couplingdefined as -20 log/S,,I. Note that the definition forforward coupling presented here differs from the definitiontypically used in the literature which is S31/S21I l l , but ifthe coupling and attenuation for the lines is very small,

IS,, 1 is approximately equal to one and the two definitionsare equivalent. Also, backward coupling is often called nearend coupling and forward coupling is often called far endcoupling in the literature [I I].

While all of the results in this paper are presented for twocoupled lines with a coupling length of 500Opn1, an FDTDanalysis of microstrip lines with shielding structures as afunction of coupling length was performed. Fig. 4 showsthat the S-parameters for two microstrip lines separated byC= 72 p n and a continuous metal filled trench. As seen inFig.4a, he maximum level of backward coupling isindependent of the coupling length, and only the periodicityof the nulls changes. The magnitude of the forwardcoupling for shielded microstrip lines with different couplinglengtbs varies with frequency in a similar manner as shownin Fig. 3 , which is for unshielded coupled microstrip lines,and to a first order, the forward coupling increases linearlywith coupling length. These results agree with the generalconclusions for coupling between weakly coupled transmis-sion lines [I I], which show that the shielding structures donot change the basic physics that cause coupling.

The effect of shielding structures on the coupling levelbetween parallel microstrip lines fabricated on the samelayer, as shown in Fig. la , is summarised in Fig. 5. For

closely spaced microstrip lines without shielding structures,the backward coupling is slightly stronger than the forwardcoupling, hut for widely spaced lines, the forward couplingis approximately 5dB stronger. It is also seen that for

closely spaced lines without shielding structures, thecoupling is very hgh (approximately 3OdB), but thecoupling decreases monotonically to approximately 45dB

as C increases.When a shielding structure is introduced in the first layer

of polyimide only (shield height equal to the microstripheight, h) , the backward coupling is reduced by 8dB forclosely spaced lines, hut there is very little improvement forlarger line spacing or in the forward coupling. However, ifthe via fence or a continuous trench is placed through bothlayers of polyimide (shield height equal to 2h ) , thebackward coupling is reduced by approximately 18dBand the forward coupling is reduced by approximately12dB at 2 5GH r for closely spaced lines. Moreover, it is

34 6

0 10 20 30 40 50

frequency,GHz

a

-25 I

-35

1 4 0~

z 5

-50

-5 5

/

4 0 I I I I ,0 10 20 30 40 50

frequency. GHz

b

Fig. 4 FDTD derived S-parameters as a f a c t i o n of coupling

length and frequ enc y fo r coupled microsrrip lines on the same

polyimide layer (Fig. l u ) with a trench shielding structure andC= 7 2 w

seen that the use of shielding structures enables microstriptransmission lines to be placed as close as 60pm whileyielding the same coupling as lines with no shieldingstructures placed 1 3 0 ~part.

It is interesting that the via fence interconnected hy ametal strip yields the same coupling level as the metal filledtrench. Also, although not shown, it was found that theresults are not dependent on the via hole spacing (DV= 60or 1OOpm). However, these results are expected when thedimensions of the rectangular mesh created by the vias andmetal strips that interconnect them is compared to awavelength. Using the largest dimensions that werecharacterised, the mesh is I O by XOpm, which at 50GHzresults in the mesh being 0.003 by 0.023 Jd where Ad is thewavelength in the dielectric. Thus, the via fence appearselectrically to be a solid wall. Another interesting result

shown in Fig. 5 is that the coupling between microstrip lineswith shielding structures with a height of 2 h is very low for

small C, it then increases as C increases, and finally reduces

again as C increases further. The small C result is believedto be true, but it is not very useful in practice because themicrostrip line is nearly touching the shielding structure forsmall C and the microstrip fields are greatly perturbed. Infact, placing the shielding structure too close to themicrostrip line results in an asymmetric surface current onthe microstrip, resulting in a higher conductor loss.

It is seen in Fig. 6 that for all frequencies, couplingdecreases as the line separation increases with the sameslope. Furthermore, for any line separation, the shieldingstructure reduces forward coupling by the same amount

IE E Proc-Microw. Anrenm Propuy.. Vu/. ISO,No . S,October 2003

8/6/2019 01262349


55 I

45

m4 0 -

c

a

.-

; 5 -

E

2 3 0 -P

25

r Shield = 2h

B 45

F

-

-

8 10measured noshield 1

c

a

.-

2 4 0 -

2eP

30

? FDTDvia heighkh,DTDvia heighk2h25

2040 60 60 100 120 140 160 160 200

line separation C, pm

a

50 I

-

f = 25 GHz

-Shield = 2h

V measured via heightrh

0 measuredtrench he#ghl=Zh

7 FDTDvia heighf=h

FDTDvia heiclhkzh

10 FDTDtrenchheight=Zh 120

1 I

40 60 60 100 120 140 160 160 200


b

Fig. 5 Meanrrednnd FDTD backward coi&ng ( a ) and foiwurdcoupling (h) at 25 GH z of microstrip linesfibricuted on the samelayer of polyimide (Fig. la) m ofunction of ce ntre-to-centre spacing,

C

across the entire frequency band. Lastly, as seen in Figs.3

and 6, forward coupling is very low at low frequencies, butincreases rapidly with frequency. Therefore, at lowfrequencies, backward coupling dominates, but for frequen-cies greater than 25 GHz, fonvard coupling dominates.Also, Fig. 6 shows that the results shown in Fig. 5b, and theconclusions drawn from them, can be applied to the entiremicrowave frequency spectrum.

il.......................... a............................._

/ f = 2 5 G H r9

.............

..............

........a/ / I =.... .................... ........... 50 GH z

r no Shield I.............. trench

20 I '60 60 100 120 140 160 160 200


Fonvurd coupling ofniicrostrip lines fabr icate d on the sameig.6Iqver ofpolyUn ide (Fig. a ) QY a junction o frequency

IEE Pm~Mim" nre- P r o p q . , Vo l 150,No. , October 2lW3

Coupling between unshielded microstrip lines is typicallyassumed to be dependent on the line separation C,normalised by the substrate thickness h, or coupling is

dependent on C/ h [12, 131. To investigate the effect ofshielding structues on this relationship, a set of coupledmicrostrip lines were fabricated and characterised asdescribed previously, except h = 5 wm and WI = 12m.The via hole dimensions and trench were kept at thepreviously described dimensions. This yields a 50Q micro-strip line that is an exact scaled version of the previously

reported results. Fig. 7 shows the measured and FDTDdetermined backward and forward coupling for both setsof circuits ( h = 10 and 5pm). It is seen that the resultswhen there are no via holes (filled symbols) supportsthe assumption that coupling is dependent on C/h.Furthermore, the results also show that this relationshipholds when shielding structures are used to minimisecoupling.

65 I60 -

m, 55-

6.E 50 ~

-a

fi 4 5 -e

$ 4 0 -

m

= 3 5 -

ihield /

-I

0 5 10 15 20 25 30 35 40

ratio of line separation to polyimide

thickness Clh

a

60

5 5 1 t=25GHz

25 10 5 10 15 20 25 30 35 40

ratio of line separation to polyimide

thickness Clh

b

Fig. 7

shown in Fig, l a wi th h= 10 pm and h = 5 pmAll lines are 50 R

Bac kvard andforwa rd coupling between microstrip /imsU

The measured coupling between microstrip lines fabri-cated on different layers of polyimide as shown in Fig. Ibare summarized in Fig. 8. Generally, the observationspertaining to Fig. 5 are true here as well, but the reductionin coupling with shielding structures is very small,approximately 5 dB, whch is the approximate reductionin coupling when the shielding structure height is equal tothe substrate height h, in the previous results. Comparing

347

8/6/2019 01262349


55

50 -

Q 45d

4 40-

m~

.-

0

2 35-Ir

5 30-

25

trench

measured

1no shield

200 50 100 150 200 250 300 350

line separation C, m

a

50

45-

' 0-0

.-a

a 35-

E

0

Y trench 1

0 50 100 150 200 250 300 350

line separation C, m

b

Fig. 8 Mearured backward coupling (a ) andforward coupling ( b )

at 25GHz of microstrip lines fabricated on d@erent layers of

polyimide (Fig. I b ) asa unction of centre-to-centre spacing C

the two different cases shown in Figs. la and Ib , it is seenthat microstrip lines embedded on the same layer of

polyimide with shielding structures that extend to the top ofpolyimide have approximately IO dB less coupling thanmicrostrip lines on different layers with the same shieldingstructure. Thus, while it may be necessary to placemicrostrip transmission lines on different layers of poly-imide to reduce circuit siw, this will result in highercoupling. It has also been shown that these results scale with

the polyimide thickness, which enables designen to use

these results for all 50 R microstrip lines embedded inpolyimide.

6 Physicsof coupling between microstrip lines

Up to this point, it has been shown that shielding structuresplaced between adjacent microstrip lines can reduce the

coupling between them, but the nature of the coupling andhow the shielding structures work has not been explored.The FDTD analysis is useful for this discussion since it cangenerate electric and magnetic field plots, which providevaluable insights.

Fig. 9 shows the electric and magnetic field plots of twomicrostrip lines on the same layer of polyimide (Fig. la)

with no shielding structure. The line on the left is the excitedline, while the line on the right is the coupled line.Comparing the electric (Fig. 9 4 and magnetic (Fig. 96)field distributions, it is seen that the coupling is primarilydue to the magnetic field. In fact, the maximum field

348

a

I 0

-10

-20

-30

4 0

-50

4 0

-70

4 0

b

Fig. 9 Electric fel d (a ) and magnetic field (b ) di.rtri!mtionfor

microstrip lines (C=69pm) on the same poly imde layer without

shielding structure (Fig. l a )

magnitude under the centre of the coupled line is -34.28 dBfor the electric field and -32.83dB for the magnetic field.Fig. I O shows the electric and magnetic field distributions

for the same microstrip lines when a shielding structureextends to the top of the polyimide. Examining the electricfield distribution in Fig. l o a first, it is seen that the shield is

very effective and reduces the maximum electric field underthe centre of the coupled line to -45.72dB. However, themagnetic fields shown in Fig. lob are able to encircle thecoupled microstrip line and still excite the microstrip mode.The maximum magnetic field under the centre of thecoupled line is now 4 1. 57 dB. When the coupled microstriplines are on different layers of polyimide, similar results areobserved with the shield reducing the electric field by 7.3 dBand the magnetic field by 5.1 dB. Thus, shielding structuresare effective in reducing the electric field coupling, but the

magnetic field coupling remains the dominant couplingmechanism and it is only moderately reduced by shielding

structures.To derive general conclusions about the effectiveness ofshielding structures in reducing electric and magnetic fields,the F DTD analysis was used to determine the excited fieldsat different points with and without a shielding structure.The analysis was first completed as a function of frequency,but it is found that the ratio of fields shielded to unshieldedis not frequency dependent, which confirms the conclusionsdrawn from Fig. 6. Fig. I l a shows a schematic of the singlemicrostrip line that was analysed with and without theshielding structure, whch was a trench, and Fig. 1 I b showsthe ratio of shielded to unshielded electric and magnetic

IE E Proc.-Microw. A n f e m P r o p g . . Vol. 150, No . 1 Oclobrr ZW3

8/6/2019 01262349


8/6/2019 01262349


6 Taflove. A. : 'Comoutational ele~uodvnamics' Artech House. Ded- 10 Marks, R.B.: 'A multilane method of network analyser calibration',ham. MA, 1995) ' IEEE Trm. Microw. Theory Tech., IWI, 39, pp. 1205-1215Roselli, L., Tentzris, E., and Kalehi, L.P.B.: 'Nonlinear circilit Schelkunoff, S.A., and Odarenko, T.M.: 'Crosstalk between coaxialcharacterisation using a multiresolution time domain technique transmission lines', Bell Sysl. Tech J , 1937, 16, pp . 144-164(MRTD)'. Proc. of the 1998 MTT-S Conference, Baltlmorc, MU, Kal, S. , Bhattachalya. D., and Chakraborti, N.B.: 'Empirical relalionspp. 1387-1390 for wpacitive and inductive coup ling coefficients of coupled micro-LEU, I. , Ho, H:M., Lee. .K. , KaXhu"rangan. I . , Liao, C.N.. and sinp lines', IEEE Trans. Microw. Theory Te d% . ,1981, 29, (4).Ho, P.S.: 'The evaluationof low dielectric constant materkals for deCp pp. 38 63 88submicron interconnect applications'. Proc. 6th Meeting Dup ont Rainal, A.J.: 'Impedance and crosstalk of stripline and micrasvlpSvmo. PolyimideMicroelectronics. May 1-3. 1995 transmission lines', IEEE T rm . Conrpm Paekuq.. Mmuf Techol.

7 I

12

8

13

9 Dutoont Cbmpa ny Pyalin LX data she& B, Ado. Packnq., 1997, 20, (3), pp. 217-224

350

Documents

01262349