Professional Documents
Culture Documents
Ingo Doerr1, Lih-Tyng Hwang2, Grit Sommer1, Hermann Oppermann1, Li Li2, Michael
Petras2, Sabine Korf1, Faical Sahli1, Tom Myers2, Mel Miller2 and Werner John1
1
Fraunhofer IZM
Gustav-Meyer-Allee 25, D-13355 Berlin
2
Motorola, SPS
2100 E. Elliot Road, Tempe, AZ 85284
S
Span of Loop 1mm – 3mm
Relative 1 (no Figure 2: Definition of design of experiment
εr Permittivity of encapsulation) –
encapsulation 3.5 Coupled Wire (Ball-Wedge)
According to the angle between the two adjacent wires,
DH Die Height 0 µm (recess) – the coupled wire bond simulations were divided into small
350µm angle (angle less than 20 degrees) and large angle (angles of
Electric Loss 45°, 90°, 135° and 180°) coupling simulations.
Tanδ Tangent of up to 0.01
encapsulation Case 1 – Small Angle (Ball-Wedge)
Table 1: Typical RF relevant parameter of single wire bond The main independent parameters for the small angle case
interconnection are the pitch of the wires on the die side (CH_PP), the angle
between the two wires (alpha) and the spans of the two wires
Furthermore, a two-level full factorial design with center (Loop A, Loop B).
point was chosen for the wire bonds. For each parameter, two Figure 3 illustrates the parameters of a coupled ball-wedge
levels within the range interval (see Table 2) were selected wire bond.
and combined with each other.
Since the glob top affects the capacitive coupling between
Figure 2 gives an idea how the design space appears the two wires, it is also included in the DOE definition as
geometrically for the inductance L, which depends on MH, S critical parameter.
and DH. The star markers (green) represent all combinations
of the chosen levels (Num_Levels Num_Parameters = 23 = 8) and
form one part of the design space (green box). The dot Loop A
marker (blue) is the center point and lies in the geometrical Chip Loop B
middle of the lower and upper range limit. Port 3 Port 2
Now, to enlarge the design space, two additional levels CH _PP
(the upper and lower range limit) for each parameter were
chosen and combined with the corresponding values of the
center point. They are represented by the rhomb markers
(red) and form the outer sphere of the design space.
Figure 2 on the whole shows very demonstratively the Port 4
geometry space, which closely related to a parameterized
function. Each of the markers describes the geometry for one Port 1
simulation. In fact, the parameterized function for the single Substrate
wire bond was based on 8+1+6 = 15 combinations
(simulation runs). Substrate
Sa Span of 0.75 mm – 2 mm
Loop a
Sb Span of 0.75 mm – 2 mm Figure 4: Cross-section of wire bond between die and
Loop b substrate, loop form
Relative 1 (without
εr Permittivity of encapsulation) – Main parasitic characteristic of wire bond is inductance,
encapsulation 3.5 as it is shown in Figure 9 with frequency dependent
α Angle between 0° - 180° magnitude of transmission coefficient. With increasing of
wire a and b span inductance increases as well. Second order geometry
Tanδ Electric Loss parameters are loop height and die height. The encapsulation
Tangent of up to 0.01 material influences capacitance of wire to ground.
encapsulation
CH_PP Die Pad Pitch 75µm – 300µm port1
Table 2: Typical RF relevant parameter of coupled wire bond
interconnection
Table 2 shows the varied parameter for both, the coupled
small angle and the coupled large angle wire bonds with port2
respective ranges. Again, a full factorial design with center
points was chosen for small angle coupled wire bond
simulation. Figure 5: EM simulation model, loop form
3D FEM Simulation
De-embedding
Ansoft High Frequency Structure Simulator (HFSS), using
finite element method, was employed to obtain the full wave The goal of de-embedding is to obtain RF interconnect
characteristics of wire bonds. Since the full wave simulator model without the pad parasitics. This leads to an
requires a uniform traveling wave to obtain the scattering interconnect model independent from chip and substrate
parameters (S-parameters) of the physical structures, a section technology. EM simulation and RF measurement takes place
of well developed transmission line section was added to each for the whole structure including feeding pads. Subtraction of
port. The effect of these transmission line feeds can be later pad parasitic can be carried out as explained in Figure 6.
removed from the total effects.
The electrical modeling of the wire bonds is based on Deembedding of the Wire from Pad
electromagnetic simulation results. Numerical simulation is
able to capture more physical configurations than assembly
Other advantages include less sample preparation times,
quicker to obtain results. It is, however, necessary to e γ l - Deembedding
experimentally validate the simulation-based parameterized
models.
ne of challenges was to generate a relevant loop form. e γ l - Deembedding
With definition of Span, Loop height and Die height the loop
and the effective line length is defined (see Figure 1). By
conducting a detailed analysis of bonding machine fabricated Square Pads: Chip = 125 x 125 µm
loop forms, a macro for automated wire bond model Substrate = 225 x 225 µm
generation was developed. Figure 5 shows a three [Y deemb ]= [Y Wire and Square Pads ]− [Y OPEN Square Pads ]
dimensional loop form in Ansoft’s HFSS simulation. A
geometrical analysis illustrates very good agreement between Figure 6: De-embedding of wire bond from pads
the assembled wire bond and the geometrical model (Figure 4
and Figure 5).
De-embedded Up to De-embedded Up to
the Square Pads the Square Pads
2.5
Mutual Inductance in
Span = 2000µm
Span = 750µm
2
L_W ire in nH
0.25
1.5 M H=250,DH=262.5
M H=200,DH=175
M H=150,DH=262.5
1
M H=250,DH=87.5
M H=150,DH=87.5
0.5
0 0.5 1 1.5 2 2.5 3 3.5 0.00
S in mm 0 30 60 90
A ngelin °
Figure 10: Inductance of wire as function of loop span, loop
Figure 12: Mutual inductance as a function of angle between
height and die height
the wires and the span after compact modeling
Coupled Wire Bonds
Pitch Angle Span a Span b Coupling
Figure 11 shows the compact model for coupled wire in µm in mm in mm at 1GHz / dB
bonds. The schematic consists of two symmetrical T models 300 0 0.75 0.75 -31
and mutual inductance M12 and coupling capacitance CC 300 20 0.75 0.75 -33
between. It could be shown that for the transmission paths 300 45 0.75 0.75 -35
between ports 1 and 4 and/or ports 2 and 3, the results of 300 90 0.75 0.75 -38
300 135 0.75 0.75 -46
single wire modeling can be used. Moreover, this modeling 300 180 0.75 0.75 -65
procedure gives the opportunity to use the single wire results 300 0 2 2 -26
and switch on or off the coupling elements, dependent from 300 20 2 2 -31
application requirements. This model describes the 300 45 2 2 -35
electromagnetic simulation results in defined design space up 300 90 2 2 -41
300 135 2 2 -42
to 10 GHz with an error less than 1%. 300 180 2 2 -61
The dominant coupling effect is the mutual inductance Table 3: Coupling scattering parameter S13 as a function of
(Figure 12). Fore angle more than 90° the coupling reaches angle between the wires
values in order of – 40dB and below. In typical system
applications below – 40dB can be neglected (Table 3). These
values are very difficult to determine. Parameterization of Model
After the R, L, and C values of the single wire bond were
obtained for the entire DOE, as shown in Table 4, the R, L,
and C can be expressed as functions of the corresponding
independent factors (geometry parameters, see Table 1) by
applying appropriate optimization methods.
For the example of the single wire bond, following
functional dependencies are required:
C = f(MH,DH,S,εr)
L = f(MH,DH,S)
R = f(MH,DH,S)
+ a 5 MH S + a 6 DH S + a 7 MH DH S