e-Book | 2025

RF FILTER DIGEST
A Resource of Whitepapers and Articles
 Introduction
With the introduction of new wireless technologies, the frequency spectrum
is getting very crowded. To ensure these technologies can operate
simultaneously without impacting each other, RF Filters play a very
important role. RF Filters are crucial in optimizing signal transmission and
reception, selectively allowing desired frequencies to pass through while
attenuating unwanted signals.

everything RF has published the 2024 “The RF Filter Digest” eBook to act
as a comprehensive resource for anyone looking to learn about the
current-generation RF Filters. RF Filters not only enhance the
performance of communication systems, but they also ensure efficient
spectrum utilization while minimizing interference. This eBook discusses
the challenges involved with 5G mmWave filtering, high-rejection LTCC
filter performance, the role of MMIC filters in developing next-generation
systems, and the impact of BAW filters on 5G applications. The eBook
also includes an exciting section that discusses filter design in the
Cadence AWR design environment.

The 2024 RF Filter Digest has been created in close collaboration with
some of the leading companies in the RF Filter space. These include
Qorvo, Mini-Circuits, Marki Microwave, Knowles and Modelithics.

Content Contributed by:

2
 INDEX
The Surprising Ways MMIC Filters Are
Enabling the Development of Next
Generation Systems

Filter Idealizations 6

Filter Figures of Merit 8

Filter Technologies 8

A Reassessment of Performance Metrics 9

MMIC Filter Capabilities 10

MMIC Filter Application Considerations 11

RF Filtering for 5G millimeter wave

applications

About millimeter-wave frequencies 16

Current uses of mmwave technology 17

Mmwave applied to 5G 18

Potential uses for mmWave in 5G 18

Filters in mmwave applications 19

Filter Specifications To Consider 21

 INDEX

A New Generation of 5G Filter Technology

5G Filtering Challenges 25

Power Handling and Heat Dissipation 26

Heat Dissipation at Higher 5G Frequencies 27

Insertion Losses 27

Small Form Factor 28

Out-of-Band Attenuation with BAW Multiplexers 28

Optimizing High-Rejection LTCC Filter

Performance in Co-Planar Waveguide
Implementations

PCB Layouts in Stripline vs. Coplanar Waveguide

30
Implementations
Effects on Performance
32
The Test Setup
32
Channelization with a Silver-Plated Conducting Cover
33
Measurements
34
Discussion
37
 INDEX

Filter Design Using Discrete Part-Value

Optimization in Cadence AWR Design
Environment

Beginning with the iFilter Synthesis Module 39

Incorporating Microstrip Transmission Lines and
42
Modelithics Models
Discrete Part-Value Optimization to the Rescue
45
Measured Data Versus Simulated Results and Closing
50

Advanced BAW Filter Technology and Its

Impact on 5G

Key Properties of BAW 54

The Importance of Electro-Mechanical Coupling 56

Filter Topologies 56
High-Frequency BAW 56
Increased Bandwidth 57

Thermal Issues and Reliability 58

Reducing Size 59

Complex Filters: Multiplexers and Antennaplexers 60

Simulation 61
Published by Contributed by

The Surprising Ways MMIC

Filters Are Enabling the
Development of Next
Generation Systems
Filters are ubiquitous throughout RF systems, yet they have not seen the same degree of
advancements (e.g. size reduction and high frequency operation) that have been made to the
rest of the RF signal chain. Current filter solutions consume a large volume of a system’s overall
footprint and are susceptible to process and lot-to-lot variation. Due to the custom nature of
filter requirements, they are often designed in-house, consuming valuable engineering time and
resources through multiple iterations of cascade analysis and subsequent design.

Until recently, MMIC has not been considered a viable technology for filter development due to
misconceptions about the technology including development time, cost, suitability for custom
solutions (due in part to the aforementioned time and cost), and the focus on traditional filter
metrics, such as Q factor. In reality, MMIC design cycles are short due to fast wafer fabrication
cycles and accurate first pass design success. Combined with the high-volume production
capability of MMIC technology, this has the added benefit of reducing the cost of custom
designs at volume.

Current RF market trends indicate that next generation systems demand significant reduction
in component size. As a result, filter specifications are being reprioritized to ease loss and
rejection requirements in exchange for size reduction. This reassessment of the trade-space has
allowed MMIC planar filters to emerge as the ideal solution to these shifting market trends. Due
to the precision of MMIC lithography, when compared against other planar technologies, MMIC
filters have demonstrated unmatched scalability, repeatability and size reduction while enabling
millimeter-wave designs. This whitepaper aims to discuss what Marki Microwave contends to be
the key metrics of a modern competitive filter technology, how these are achieved through a
GaAs MMIC process, and the capabilities of MMIC filters that have been demonstrated thus far.
Together with the companion design tool that has been developed, this offers a fast, accurate
and trustworthy way of developing filters without consuming large amounts of customer and
vendor design resources.

Filter Idealizations improving system dynamic range. These

designs are typically custom and specific
RF/Microwave filters are electronic
to an application’s frequency plan.
circuits that allow only specific
Multiple filters are often needed at
frequencies to pass through a system
different points throughout a particular
while rejecting unwanted signals outside design.
of the desired passband. Filters are
commonly used to provide channel/band The fundamental building block of filter
selection and reject spurious tones, designs is a resonator. An LC resonator

6 Download this Whitepaper from Marki Microwave everythingRF.com

RF Filter eBook Marki Microwave

in a filter typically consists of an inductor get closer to an ideal transition and

and a capacitor in either a series or cutoff by increasing the order of a filter.
parallel configuration. These reactive These filters are mathematically
components along with other circuit possibleto achieve if you have perfect,
elements determine a filter’s frequency lossless resonators. A Butterworth filter,
response characteristics. In addition to as seen in Figure 1(b), features a
the series and shunt LC resonators, maximally flat passband (low ripple) with
planar designs can be implemented using a gradually sloping transition region at a
distributed elements, such as spiral rate of -20 dB/decade for a 1st order
structures or tapered traces to achieve filter. Increasing filter order increases
the desired inductance and capacitance. that steepness; a 2nd order filter would
see a rate of -40 dB/decade, and a 3rd
In the context of planar filters, the layout
order filter would see -60 dB/decade,
of an LC resonator is implemented in a
and so on. A Chebyshev filter, as shown
two-dimensional form on a flat substrate
in Figure 1(c), on the other hand can
combining resonators with transmission
achieve a steeper transition region due
lines.
to the allowable ripple in the passband
and at the passband edges. Chebyshev
filters are used when steeper roll-offs are
critical for a design.
There are many other mathematical ‘filter
prototypes’ that can be used as the
starting point for a filter design, including
Legendre, Bessel, Elliptic, and others
without formal names. However a real,
well-designed planar filter will not strictly
follow any of the filter prototypes due
nonidealities in the physical
implementation.
Figure 1. (a) Ideal filter frequency response; (b) Butterworth filter
response; (c) Chebyshev filter response. Real Resonator Filters

The Ideal (Brick-Wall) Filter Filters with real resonators will have
The ideal filter is a two-port device that insertion loss and finite Q. Real
would allow a desired range of resonators will introduce resistive and
frequencies to pass through a system parasitic losses and the filter will have
with no loss and would immediately cut- lower selectivity. The additional losses
off unwanted signals at the passband result in the widening of the filter
edge with an infinitely steep transition bandwidth. When laid out and 3D
and out-of-band rejection as shown in simulated, crosstalk and other physical
Figure 1(a). Ideal filters are impossible to effects need to be considered as well.
realize, even in digital signal processing. Many different filter topologies exist;
Butterworth and Chebyshev Filters common topologies include: lowpass,
highpass, bandpass, notch, and diplexer.
The next level of filter idealization are the The unique filter requirements of a
Butterworth and Chebyshev filter system will dictate which filter topology
approximations. While the brick-wall is the preferred solution.
response is not achievable, designers can

7 Download this Whitepaper from Marki Microwave

everythingRF.com
RF Filter eBook Marki Microwave

Filter Figures of Merit circuits are often designed using

cascaded sections. A higher order filter
typically features steeper roll-offs at the
passband edges, narrower transitions
between the passband and stopband and
high stopband suppressions.
Insertion Loss

The filter order, 𝑁, indicates the number

of zeros or poles in a filter’s transfer The
center frequency insertion loss of a filter
can be approximated by Equation 2,
Figure 2. Key performance metrics of a microwave filter. where represents the sum of the
resonator element values up to the Nth
The goal of any filter design is to resonator, and 𝑄_𝑢 is the unloaded Q
accommodate the design specification of factor.
a customer’s system. Figure 2 displays
the key performance metrics of a filter
design. Typically, this will be the
maximum insertion loss across the Equation 2

passband and the out of band rejection at

A filter with a high Q would demonstrate
some specified frequency. The resonator
low center frequency insertion loss, while
Q and filter order are key factors in
a higher order filter would present with
defining a filter’s center frequency
higher loss, but steeper rejection slopes
insertion loss and rejection skirts.
and higher selectivity. Additionally, a
Resonator Q
higher order filter would be a more
complex, larger filter design. For this
reason, the traditional approach to
The Q, or quality factor, is defined as the
improving filter performance has been to
ratio of the energy stored in a resonator
focus on improving Q.
to the energy.
Filter Technologies
Common filter technologies in use today
Equation 1 include acoustic wave filters, cavity
A filter with a high Q will have a sharp filters, lumped element filters and planar
resonant peak at the center frequency filters. No single filter technology is best
and a narrow bandwidth. Filters with a suited to fit all applications; each
high Q factor are highly selective with technology has its own tradeoff space.
respect to their passband. Cavity filters for example provide
Filter Order excellent out of band rejection, low loss
and high power handling at microwave
The filter order, 𝑁, indicates the number
frequencies at the expense of being
of zeros or poles in a filter’s transfer physically large designs that require
function. A higher order filter is achieved hand tuning. Acoustic filters on the other
by increasing the number of reactive hand offer excellent rejection in a small
components (e.g. LC or other
resonators) in the filter design. These

8 Download this Whitepaper from Marki Microwave

everythingRF.com
RF Filter eBook Marki Microwave

form factor but are limited to frequencies For traditional filter technologies, this is
below 8 GHz and with low power problematic; as a filter’s size decreases,
handling. Planar filters (e.g. MMIC, Thin there is a natural trade-off in Q and filter
Film, Laminate) fall somewhere between order. The implications of this are that
these technologies in size, but feature an the center frequency insertion loss takes
excellent tradeoff for improved a hit, passband edges are more rounded
performance, size, cost and development and the transition slopes are not as
time. Marki Microwave contends that steep. However, if insertion loss
MMIC filters are well suited to fit current requirements can ease slightly, MMIC
demands for frequency and performance filters are uniquely suited to become a
in a small form-factor. Table 1 provides a viable solution for modern filter designs.
high-level view of what can be realized With MMIC, the flexibility in filter
across different filter technologies, with 5 topology and architecture provides more
representing the strongest capabilities, tools than just Q & filter order to play
and 1 representing the worst with.
In Marki Microwave’s experience, filter
A Reassessment of specifications of modern and next gen
applications appear to be less sensitive
Performance Metrics
to insertion loss than they have been in
the past. If designs are allowed to take a
small insertion loss hit, then planar filter
sizes can shrink significantly with MMIC.
The goal of a filter design is to decrease
size as much as possible without taking
large Q penalties while still meeting the
design spec: typically maximum
passband loss and out-of-band rejection
minimum at some offset frequency. Since
lower Q is directly related to lower order
filters and increased insertion loss, the
trade-space in creating smaller filters is
to find the balance between reduced
filter size, rejection and insertion loss.
Focusing specifically on planar filter
technologies, when considering Table 1,
MMIC filter designs offer the best in
Table 1. Comparison of commonly used filter technologies.
frequency capability, size, scaling,
repeatability, integration capability, and
As next generation systems aim to take
agreement between simulation and
advantage of wider bandwidths and
measurement. While MMIC designs may
support higher channel densities, RF
have a lower Q compared to
market trends for filters are shifting to
technologies such as laminate and thin
prioritize size-weight-and-power
film, the reprioritization of design specs
(SWaP), higher frequency designs,
means that the filter Q, while important,
scalability and rapid custom filter is no longer the most critical filter metric
development over preferences of low for many applications.
insertion loss and high out-of-band
rejection.

9 Download this Whitepaper from Marki Microwave

everythingRF.com
RF Filter eBook Marki Microwave

MMIC Filter Capabilities loss, however, in exchange we gain

High Frequency Designs frequency/bandwidth and more precise
lines. The insertion loss of MMIC filters
increases slowly at higher frequencies,
whereas other technologies such as
laminate and thin film tend to struggle in
terms of insertion loss above 40 GHz.
Additionally, the insertion loss of a filter
is also directly impacted by filter order
and percent bandwidth as noted by
Equation 2. While MMIC filter order is
limited to below 9th order circuits, Marki
Figure 3. MMIC filter performance demonstrated to 110 GHz.
Microwave has found that a filter with an
order between 4 and 8 is sufficient to
The precision and fine lines of MMIC meet most rejection specs while
lithography and the relatively thin maintaining an acceptable passband
substrate used enable high frequency insertion loss.
filter designs that are not achievable
with competing technologies. Combining Rejection In addition to the insertion
unique design topologies and layout loss across the passband, a typical filter
with high volume MMIC production spec is also determined by the rejection
processes, standard frequency capability at a specified frequency outside the
has been demonstrated up to 110 GHz as passband. For most cases, Marki
shown in Figure 3. Microwave has found that a 40 dBc
minimum stopband rejection is
Insertion Loss achievable at 10% from the band edge.
At higher frequencies, stopband
rejection does take a slight hit due to
slower roll-off, however, a 30 dBc
minimum rejection above 40 GHz has
still proven to be acceptable for most
designs. Return Loss

Figure 4. Measured passband insertion loss versus frequency for

a selection of bandpass filter designs.

Typical passband losses for MMIC designs

have so far been demonstrated between 1
and 5 dB, as shown in Figure 4, with
lower frequency designs displaying
superior losses. The move to higher
frequencies and smaller filter sizes
necessitates the use of thinner substrates.
As a result, a filter takes a penalty in Q
and thus a hit in insertion Figure 5. Measured return loss for a 78 GHz - 110 GHz MMIC
bandpass filter.

10 Download this Whitepaper from Marki Microwave

everythingRF.com
RF Filter eBook Marki Microwave

For a standard reflective filter, a high wafer-to-wafer when compared against

(negative) return loss indicates that a other technologies. While the design of a
majority of the input signal is being typical planar filter is similar among the
transmitted through the filter, ensuring different technologies (laminate, thin
efficient transfer and minimal signal loss. film, MMIC), the etching tolerance of a
A high return loss indicates a well- MMIC is unmatched. The typical etching
matched filter that effectively tolerance of a laminate board is 25 um
distinguishes between the passband and whereas a typical MMIC is 0.25 um. For
stopband. MMICs have demonstrated the laminate filter, poor tolerances can
measured passband return losses result in dimensional variations that may
detune the filter, lowering yield. As for
better than 10 dB for filter designs above
thin film, the manufacturing processes
100 GHz (Figure 5) as well as typical
are not as well controlled compared to
return losses in the 15 to 25 dB range at
ICs due to the less sophisticated, low
frequencies below 40 GHz.
Tolerance & Repeatability volume tools and techniques employed
resulting in variability from lot-to-lot.
Figure 6 displays an overplot of 24
randomly selected MMIC filters from a lot
of over 8,000 filters demonstrating high
repeatability with low variation in both
passband frequency and loss.
MMIC Filter Application
Considerations
Package & Size
The heavy increase in channel count and
total bandwidth demanded by next gen
systems has driven the need for the
substantial size reduction that can be
Figure 6. Overplot of a representative sample of 24 MMIC filters
from a lot of over 8,000 filters demonstrating outstanding achieved with MMIC filters when
fabrication repeatability.
compared against other traditional
planar filter techniques. The precision of
GaAs fabrication utilizes tightly
MMIC lithography as well as the
controlled single crystals that can be
availability of extremely small, lumped
produced in high volumes while
element inductors and capacitors
maintaining repeatability. MMIC filter
enables unique circuit innovation and
production is scalable from thousands to
significant size reduction of filter circuits.
millions of units. The fine geometric
MMIC filters can be realized with die as
features available on the process allow
small as 1.5 x 1.5 mm for high frequency
for the careful tuning and optimization of
designs. Additionally, a MMIC filter can
features such as transmission zeros via
be integrated directly with another MMIC
cross-coupling of resonators. These
block or co-packaged into a surface
results have been achieved with near
mount package or multi-chip module to
perfect match with simulation, enabling
further reduce a system’s footprint.
first pass design success and a short
Standard QFN vs Thin Film
development cycle.
A tighter tolerance results in better
repeatability from unit-to-unit and
Both the standard QFN and thin film

11 Download this Whitepaper from Marki Microwave

everythingRF.com
RF Filter eBook Marki Microwave

filter are surface mountable devices that performance.

can be assembled with standard pick and
place machines and reflow ovens.
However, the GaAs MMICs have a
dielectric constant, Dk, of 12.9 (compared
to 9.8 Dk for typical thin film) allowing
much smaller designs and features
resulting in QFN package sizes 10 times
smaller than thin film filters. MMIC
lithography enables fine lines and gaps
required for operating frequencies
covering the low GHz range to beyond
100 GHz. Optimized Landing Patterns
Figure 7. Optimized landing pattern displaying an inductive
Filter performance is extremely sensitive taper.
to impedance matching and tuning to
ensure desired cutoff frequency, Power Handling
passband insertion loss and return loss, Modern communication systems require
and stopband rejection. Resonant high operating powers to transmit and
structure-based filters rely on tight receive high-bandwidth data. These
impedance tolerances to pass or to trap systems push filter capabilities well
selected frequencies. When packaging beyond 10W for continuous wave
filter die into surface mountable operation and 100W for pulsed operating
components, parasitic capacitances and conditions.
inductances from the leads and
wirebonds contribute to the filter’s tuning The maximum power handling for MMIC
network (especially at 40 GHz and above) filters depends on the topology: filters
which can detune the designed can have either a RF power handling
impedances. Therefore, die in order to and/or a max DC current specification. In
present a well-matched 50Ω input to the general, the type of power rating
filter package while maintaining the depends on if the input is AC or DC
filter’s integrity. coupled. If the resonator uses a
capacitive structure (or any gaps
Parasitic compensation can be built onto between conductors), traces are likely to
the die, but it often is not enough to short when the max power rating is
meet desired passband return loss exceeded. For DC coupled inputs, the
performance. External compensation trace width limits current throughput,
using optimized landing patterns, as seen and when there is sufficient self-heating,
in Figure 7, is necessary to tune the the conductor burns out rendering the
transmission line leading to the filter’s through path as an open circuit. For
package, typically to compensate for bandpass filters, both max RF power
parasitic capacitance from the package handling and max DC current rating are
leads landing pattern. Compensation applicable, but for filter designs that are
networks, like an inductive taper, can DC coupled, the max DC current rating
reduce stray capacitances and ensure a will typically be the first to blow. For a
better 50Ω match leading to the filter, subset of MMIC bandpass filters tested
improving passband return loss so far, typical power handling values
have been demonstrated to at least 10W
CW so far, though this may vary with

12 Download this Whitepaper from Marki Microwave

everythingRF.com
RF Filter eBook Marki Microwave

design. improved insertion loss performance. In

general, lower percent bandwidth filters
High pass filters are typically AC coupled
will also be more sensitive to
at the input therefore limited by a max
temperature change.
RF power handling specification while
The Future of MMIC Filters
low pass filters typically have a DC short
Filter specifications tend to be custom to
in the through path with a max DC
their projects. They may be planned for
current capability. Filters with larger
at the initial design phase or potentially
elements and wider traces (as seen in
towards the end to clean up unexpected
distributed element filters) enable higher
spurious signals. Regardless, it is critical
power handling capabilities, however this
that custom filters can be developed
comes with increased die dimensions.
Performance over Temperature quickly and accurately. The future of
MMIC filter design is to heavily minimize
the design burden for both the customer
Aerospace and military applications
and design engineer.
require components to withstand
extreme temperature ranges, with Marki Microwave recently demonstrated
minimum and maximum operating a new GaAs MMIC filter design tool, Filter
temperature ranging from -55°C to Prodigy, based on AI-generated finite
125°C. Over this temperature range, element models and Marki’s ever-
MMIC filters have shown very low growing library of existing filter designs.
performance variation as seen in Figure Using the tool, a customer will be able to
8. However, the magnitude in variation select the filter type and design
between insertion loss and return loss is parameters (center frequency, %BW,
dependent on filter design topology. etc.) of their project, and the tool will
Lumped element filters experience almost instantly provide accurate s-
greater temperature variation compared parameters and die dimensions. On the
to distributed element filters at very high backend, Filter Prodigy manipulates
frequencies. As element sizes shrink, the existing filter layouts in real time and
tolerance becomes a significant factor as produces a 3D model, again almost
any slight variation will cause the instantaneously, that meets the customer
resonant frequency to shift. With spec, nearly eliminating the bulk of the
temperature changes, material expansion engineering design work that would
and shrinking will also cause shift in typically go into laying out a filter. As the
frequency. Increased temperature will library of filters grows and new
generally increase insertion loss while topologies are added, the tool will only
decreasing temperature will yield become more accurate.

Figure 9 demonstrates a 10 GHz filter

design at varying percent bandwidths
simulated using the Filter Prodigy tool.
As expected by theory, as percent
bandwidth increases, insertion loss
improves while the rejection slopes
become less steep. Even then, the
insertion loss between a narrowband and
wideband filter differs by approximately
dB. The customer application will

Figure 8. MMIC filter insertion loss variation over temperature.

13 Download this Whitepaper from Marki Microwave

everythingRF.com
RF Filter eBook
Marki Microwave

determine whether this loss is feasible, but regardless, these design questions can be
determined quickly using this tool.

Figure 9. Filter Prodigy simulation of Insertion Loss vs Frequency at varying %BW

For a customer, this is a fast, accurate and trustworthy, iterative tool. Using the
provided s-parameters from each design, they themselves can quickly simulate their
filter, check and modify their requirements against their system requirements and
repeat the process before bringing Marki into the discussion. Once a design is reached
that satisfies their requirements, Marki can verify their design for accuracy. If filters
need to be packaged, some additional 3D modeling may be required, but the bulk of
the work has already been completed. This, combined with the capabilities of MMIC as
discussed, allows first-pass design success and results in shorter design cycles that are
not achievable with other filter technologies.

View Table of Contents

14 Download this Whitepaper from Marki Microwave

everythingRF.com
Published by Contributed by

RF Filtering for 5G millimeter

wave applications
This whitepaper will discuss some of the areas of Fifth Generation (5G) communications that
will be implemented in the millimeter wave (mmWave) region of the RF spectrum and will
guide the reader through some of the options available to implementing RF filtering at these
frequencies. In the first part of this whitepaper we will do a short review of mmWave
frequencies and their applications, both prior to the advent of 5G and with planned 5G systems.
In the second part we look at filtering technologies available for use at mmWave frequencies.

About millimeter-wave From a dramatic increase in available

bandwidth comes an increase in
frequencies achievable data-rates, making these
The millimeter-wave part of the frequencies extremely interesting to
electromagnetic spectrum is at the high teams around the world working on 5G
end of the microwave region (which communications.
spans ~300 MHz to 300 GHz) and is
However, with this increase in frequency
usually taken to mean frequencies from
comes a set of challenges. The free
~30 GHz to 300 GHz and wavelengths in
space path loss (FSPL) equation
the range of 1mm to 1cm. The ITU
provides a way to calculate signal
designation for the band spanning 30
attenuation through free space or a
GHz to 300 GHz is EHF (Extremely High
vacuum. It tells us that signal loss
Frequency).
increases as the square of distance and
One of the characteristics that makes the square of frequency. The higher the
this part of the spectrum so interesting frequency the more the signal will
to communication applications is that degrade over distance, limiting these
the frequencies are that much higher frequencies to short range and/or high-
than those historically used in wireless so power communications.
the available spectrum in the mmWave
potentially opens up 200x the spectrum
available below 3 GHz.
In addition to losses in free space, radio
signals are attenuated both by
atmospheric constituents and by physical
objects (trees, buildings) that they
encounter. mmWave frequencies have
proven particularly susceptible to these
factors. For example, there is an oxygen
absorption band in the range of 60GHz,
signal loss due to rain and humidity is a
known engineering challenge throughout
the mmWave range, and attenuation
through materials

Table 1. ITU Band Designations

16 Download this Whitepaper from Knowles everythingRF.com

RF Filter eBook Knowles

like concrete or brick is increased for Military

high frequencies. Military applications of Microwave and
With that said though these challenges mmWave have a long heritage.
are not insurmountable and even before Table 2 shows some example military
5G began to build interest a wide range applications across frequencies ranging
of applications have utilized mmWave from 2 GHz to 40 GHz. Typically, as
frequencies. frequencies increase the applications
Current uses of mmwave move to higher resolution applications.

technology Automotive
Communications
mmWave frequencies also find
Microwaves (of which mmWaves are a
application in radar based Advanced
subset) have been widely used in point-
Driver Assistance Systems (ADAS),
to-point communications. An example of
helping drivers control vehicles and to
point to point would be in a backhaul link
assist in automated functions. These
where microwave frequencies are used
systems often use both short range (at
to transmit data in a straight line of site
24 GHz) and long range (at 77 GHz)
link between a cellular base station and a
base station controller. Point to point at
mmWave frequencies is attractive
because the reduced wavelength allows
conveniently-sized antennas to focus
them into narrow beams, which can be
aimed directly at a receiving antenna.
This prevents nearby microwave
equipment that use the same frequency
from interfering with each other, allowing
the deployment of more links in the same
area. In the US, 36 GHz to 40 GHz is
licensed for high-speed microwave data
links, and the unlicensed 60 GHz band
can be used in short range data
connections. Table 2. Some examples of Military Applications at
Microwave Frequencies.
In the US, 36 GHz to 40 GHz is licensed
for high-speed microwave data links, and radar to scan the environment around
the unlicensed 60 GHz band can be used the car. The high frequency, and thus
in short range data connections. In the short wavelengths, involved is a key
60GHz unlicensed band another advantage for mmWaves in this
application is WiGig, or IEEE 802.11ad, application. A radar system operating at
intended to support high bandwidth 77 GHz will have a wavelength of
gaming and 4k video streaming over approximately 3.8mm. As a
WiFi. The new standard adds a new 60 consequence, the size of system
GHz frequency band alongside the components such as the antennas
current 2.4 and 5 GHz offerings. required to process signals is reduced.
An additional benefit to using such short
wavelengths is the increased levels of
accuracy, since such a system will have

17 Download this Whitepaper from Knowles everythingRF.com

RF Filter eBook Knowles

the ability to detect movements that for 2020 and beyond’ including the
areas small as a fraction of a millimeter following list of frequencies: 24.25-
Mmwave applied to 5G 27.5GHz, 31.8–33.4GHz, 37–40.5GHz,
Frequencies of interest to the 5G 40.5–42.5GHz, 45.5–50.2GHz, 50.4–
community 52.6GHz, 66-76GHz, 81–86GHz.
Earlier we talked about how the use of Since then, a group of frequencies have
frequencies up in the mmWave makes come forward as the main candidates for
large amounts of additional bandwidth mmWave applications in 5G: 28 GHz, 38
available, which in turn permits an GHz, and 72 GHz. These frequencies have
increase in transmission speed. Fifth been selected because they have the
Generation (5G) communication systems following advantages:
being planned to enable a hundred-fold
increase in user data-rates – with this Reduced Oxygen absorption rates
need for huge amounts of bandwidth in compared with other mmWave
5G, mmWave has become extremely frequencies.
interesting to the 5G community.
Shown to perform well in multipath
We also talked about some of the environments, allowing them to be
challenges involved in using mmWave used in non-line-of-sight (NLOS)
frequencies, such as range limitations communications
brought about by propagation losses. Demonstrated to perform well with
However recent advances in mmWave
directional antennas, beam forming
systems have made these frequencies a
and beam tracking.
commercially attractive prospect and
have turned some of the perceived
disadvantages into key enablers in
Potential uses for mmWave in
system architecture. For example, short 5G
transmission paths and increased To include 28 GHz conceptually under
propagation loss allow for spectrum re- mmWave seems to be an industry wide
use. And the established ability of convenience – technically (according to
mmWave in point to point the ITU) 28 GHz would in fact fall in the
communications to be tightly focused centimeter-wave range. In the context of
into beams allows signal strength to be 5G mmWave is now being used to refer
directed exactly where it needs to go, to the licensed bands such as 28 GHz
reducing interference, and also allowing (US) and 38 GHz (Korea) with the
multiple beams to be combined, resulting addition for unlicensed spectrum up at
in increased range. 60 GHz.
Following the World Radio
A detailed discussion of the potential use
Communication Conference in 2015
cases is described in the 2017 paper
(WRC-15), the ITU released a resolution
‘Where, When, and How mmWave is
addressing ‘possible additional
Used in 5G and Beyond’. The applications
allocations to the mobile services on a
discussed in that paper are:
primary basis in portion(s) of the
frequency range between 24.25 and 86 28 GHz Outdoor Backhaul for Fixed
GHz for the future development of Wireless Access
International Mobile Telecommunications
The idea behind Fixed Wireless Access is
that high bandwidth connections into the
home and business can be provided by

18 Download this Whitepaper from Knowles everythingRF.com

RF Filter eBook Knowles

mmWave radio. Think about replacing major design component in radio access
cable and fiber with radio. The 2018 US technology in 5G. Next we will take a look
commercial launch will be at 28GHz, and at mmWave filter technology options.
there is talk of trials in Japan beginning Filters in mm-wave applications
in 2018 at 60GHz. In addition, there is Here we provide an overview of some of
some discussion of placing access points the common technologies that are both in
in buses, for example. use in industry today and are generally
60 GHz Indoor Access with Mobile Edge available from more than one
Computing manufacturer. This last point is a factor to
consider when looking at some of the
The paper discusses the use of 60GHz in more experimental alternatives– our own
locations such as airports in an attempt approach is to provide a solid foundation
to support data intensive applications based on known implementations that are
such as augmented reality (AR). proven to be reliable.
mmWave Mesh Networks for µ–RAN On chip or off chip?
This scenario involves the deployment of When we look at filtering options for
mmWave base stations as an overlay to mmWave RF systems, an early choice
existing LTE cells in dense urban comes from choosing between on-chip
environments to provide sufficient (integrated into the RFIC for example) and
bandwidth. It is suggested that a 6 off-chip (filtering outside the RFIC with
sector mmWave gateway would provide surface mount components or
backhaul to a number of mmWave small connectorized solutions).
cells in a mesh network approach. At these frequencies, for on-chip solutions
mmWave based V2V/V2X for the common semiconductor technologies
Automated Driving are Silicon (Si) and Gallium Arsenide
In probably the most remote use case (GaAs). It is often desirable to
described in the paper, mmWave use on-chip solutions because of not only
networking is used to support driverless the compactness of the circuits, but also
cars in Vehicle to Vehicle (V2V) and the robustness of the on-chip filters since
Vehicle to Everything (V2X) applications. the manufacturing tolerances are those of
At this point we can see that mmWave the semiconductor manufacturing process.
frequencies are going to be an integral Integrating an on-chip filter with other
part of the changes that 5G will bring devices to form a system-on-chip (SoC)
about over the next few years. The solution, as is commonly done below
different use cases that mmWave 6GHz, can dramatically reduce the
technologies enable makes up a diverse physical size of a wireless system.
set of applications – each bring their own However, the reduction in dimensions of
set of challenges when it comes to devices working in the mmWave
implementing systems in a robust and compared to devices at 3GHz can make
reliable way. interconnect between devices challenging
Something we know here at Knowles and to-date on-chip implementations have
Precision Devices is that, just as we have seen limitations
seen filtering playing a key role in some when it comes to quality factor (Q),
of the more common uses of mmWave
frequencies, filtering will continue to be a

19 Download this Whitepaper from Knowles

everythingRF.com
RF Filter eBook Knowles

losses and Noise Figure (NF). High Crystal Filters make use of a quartz
isolation and low loss can be a challenge crystal as the resonant element. The high
for on-chip filtering, and when a Q of a quartz resonator makes for a very
selfinterfering TX is in the same band as steep band-pass. These filters are usually
the RX, lumped filtering is often not implemented at IF frequencies in the
practical for on chip solutions and an off- range of 10 MHz and Q factors fall in the
chip duplexer is usually required. range of 10,000 to 100,000. SAW/BAW
(Surface Acoustic Wave and Bulk
Challenges with implementing
Acoustic Wave). Acoustic filters cover a
sufficiently highperformance filter
range of frequencies up to 6 GHz and
structures on-chip arise from various
offer a good performance/cost tradeoff,
factors, including the physical
making them the dominant off chip filter
characteristics of the semiconductor
approach in mobile devices today.
material and the cost of implementation.
In Gallium Nitride (GaN) for example, the Ceramic filters cover a range of ~100
aim is to make the circuits as thin as MHz to ~8 GHz. They offer similar
possible to encourage the heat performance to discrete lumped element
dissipation necessary for the high inductor-capacitor (LC) designs but can
powers that GaN is capable of. However, be implemented in small form factor
the Q of a filter structure is surface mount packages. Performance
proportional to the thickness of the and package thickness can be a limiting
dielectric substrate, so in GaN this means factor when comparing ceramic filters
an advantage of the material (high with SAW/BAW.
power) works in opposition to building Lumped Element. Discrete LC
filters with high Q. In addition, the space approaches provide a low-cost approach
required to implement a filter structure in to implement a filter, but the attainable
GaN takes up valuable chip real estate – Q factors are limited in such devices.
area on the wafer than can be devoted Discrete lumped element filters are
to much more lucrative active systems usually used around the 30 MHz to 300
such as amplifiers. At the time of writing, MHz range but can in principle be built
building on-chip filter structures with for applications up to 40 GHz. At
high enough Q to serve in a front-end mmWave frequencies though discrete
application has proven impractical. lumped element filters are very hard to
implement because of the dimensional
Our survey then will focus on off-chip
limitations imposed by the frequency,
filtering technologies, but it is worth
since the filter elements must be much
keeping some of the advantages of on-
smaller than the wavelength of the
chip (reduced size, repeatable and
transmission lines.
robust manufacturing) in mind as we go
forward. Cavity filters are a common approach in
A broad view of filter technology the 40 MHz to 960 MHz frequency range
and can offer high selectivity under high
power. They can achieve good
A good place to start our survey is by performance but are physically large,
looking at the frequency ranges covered and usually only seen in infrastructure
by different off-chip band-pass filter applications, such as for additional
technologies. filtering at a cell site.
We can look at seven common off-chip Planar filters are manufactured using a
approaches to building filters:

20 Download this Whitepaper from Knowles

everythingRF.com
RF Filter eBook Knowles

thin-film process, and depending on the

filter topology, can offer high Q and a
reasonable approach to achieving
performance in a small footprint when
compared with discrete lumped element
designs. In a thin film Lumped Element
approach, the filter’s transmission lines
are printed in various configurations,
depending on the required performance
and filter elements are realized through
discrete resistive, capacitive, and
inductive elements. Planar Distributed
Figure 1. Common filter fechnologies and their frequency ranges
Element filters rely on carefully
distributed transmission lines to create
resonant structures and can be designed
Filter Specifications To
to tighter tolerances than a lumped Consider
element filter. Distributed Element Percent BW: Where Bandwidth is the
designs are more practical than Lumped width of the passband of a bandpass
Element designs at increased frequencies filter and is expressed as the frequency
temperature stable ceramics to provide difference between lower and upper 3
frequency stable performance over dB points, Percent Bandwidth is a
temperature common relative merit that compares
Waveguide filters are characterized by bandwidth with carrier frequency.
high power handling capability, leading Commonly calculated as 3dBW/(Center
to their wide adoption in radar Frequency). Percent Bandwidth is likely
applications and low loss given that the to be an important consideration in 5G
waveguide itself is a low loss medium. mmWave applications. One of the
desirable attributes of the mmWave
Figure 1 summarizes the frequency spectrum is the ability to access large
ranges covered by these filter amounts of bandwidth (and) hence
technologies. Looking at the plot the increased data rates). To have a radio
main choices for filtering in mmWave access system bandwidth limited
applications (e.g. ~30GHz and up) because of available filter technology is
appear to be Cavity, Planar Thin Film and not desirable.
Waveguide. All three approaches bring Q factor: The quality factor (Q) of a
advantages of their own and are based resonator is a measure of the Selectivity
on established, readily available of a resonator and is expressed as the
technologies. Let’s take a look at some of ratio of stored versus lost energy per
the performance characteristics that are oscillation cycle. Overall losses through a
important in filtering applications, both in resonator increase as Q factor drops and
general and to 5G communications. will increase more rapidly with frequency
for lower values of resonator Q. As a
result, the edges of the passband
become more rounded and the
bandwidth narrows as the Q decreases.

21 Download this Whitepaper from Knowles

everythingRF.com
RF Filter eBook Knowles

Selectivity is another way of talking transmit power will be spread over the
about Q factor, and a measurement of individual elements in phased array
the capability of the filter to pass or antenna systems, and where the transmit
reject specific frequencies relative to the range at these frequencies is reduced,
center frequency of the filter. Selectivity filters do not need to withstand as much
is typically stated as the loss through a power.
filter that occurs at some specified
Package Type and Package Size. These
distance from the center frequency. A
factors look at how the filter is physically
filter with high selectivity exhibits high
implemented. Is the filter contained in a
slope in the transition from pass to stop
somewhat bulky connectorized package,
– Selectivity is crucial in environments
or can it be implemented in a compact
where adjacent channels are close surface mount or flip-chip configuration?
together and high selectivity enables How consistent are the interconnects at
designers to make good use of available mmWave frequencies – that is, how does
bandwidth. the packaging impact the amount of
Insertion Loss: Loss of signal power post assembly tuning that a system will
through the filter. Important to consider need to undergo before shipment?
on the Tx side because power is a Component size will be a key enabling
system cost driver, and on the Rx side (or restricting) factor in practical
because loss impacts the overall Noise mmWave front end designs. In phased
Figure of the receiver. array antennas the elements must be
Temp Stability: Not often specified on sufficiently close together to avoid
generating grating
filter datasheets the ability of the filter to
lobes – and half wavelength spacing for
hold its specifications (not just center
mmWave frequencies amounts to a few
frequency but percent BW and
millimeters. The resulting compact arrays
Selectivity) over temperature is an
will need to find a way to integrate the
important factor to consider in systems
necessary filtering and depending where
that want to make good use of available
in the architecture RF filtering is
bandwidth. The near term mmWave
deployed, space for filtering will come at
applications in 5G, such as FWA, will
a premium and compact packaging will
entail small scale systems (e.g. pole
be desirable.
mounted) in exposed environments just Weighing up available filter
as with today’s Small Cells. Such technologies
deployment scenarios have raised the
bar when it comes to a front-ends ability
to withstand both extremes in Viewing the available technologies at
temperature and temperature variation. mmWave, we can begin to make some
Further, overall size reduction in systems general statements about what
lead to densely populated boards o heat approaches are likely to be useful in a 5G
from the sourrounding components can system. An approach to doing this is to
impact the stability of the filter. look at the factors identified above in
Power Handling: The ability of the filter Table 3.
to withstand large amounts of transmit
When we look at some of the driving
RF power, this is mostly a concern in
factors that will influence 5G mmWave
traditional macro-cell applications below
front end designs, Planar Thin Film
3GHz. In mmWave 5G applications the
implementations emerge as a desirable
approach from the standpoint of size,

22 Download this Whitepaper from Knowles

everythingRF.com
RF Filter eBook Knowles

cost and acceptable performance. Filtering in 5G applications will be just as

important at mmWave frequencies as it is
below 3GHz, but with a new set of challenges
brought about by the available filtering
technology and the characteristics of mmWave
systems.
5G systems will require filters with high percent
bandwidth, good selectivity, and excellent
temperature stability in a . compact package
but will not require the highest power handling
capability.

Based on these criteria, the Planar Thin Film

filter implementation becomes a desirable
approach. Successfully implementing filters
using Thin Film manufacturing requires the right
choice of substrate (to ensure suitably high Q)
and sufficiently high thin film tolerances (to
ensure repeatable performance at the scales
involved in mmWave applications).
Approaching this design challenge will be the
subject of a future whitepaper

Table 3. Factors influencing off-chip mmWave filter technology

choice

Conclusion
Frequencies in the mmWave spectrum will play
a key role in 5G communications. RF technology
that was developed around existing mmWave
applications have evolved to encompass the
needs of 5G wireless access.

View Table of Contents

23 Download this Whitepaper from Knowles

everythingRF.com
Published by Contributed by

A New Generation of
5G Filter Technology
5G implementation is accelerating worldwide, creating new challenges for wireless technology
in mobile devices. The drive to deliver higher data rates with 5G has resulted in an enormous
increase in RF complexity: the use of new higher-frequency bands, increasing complexity of
modulation schemes, increasing number of antennas, carrier aggregation (CA) and E-UTRA
New Radio Dual Connecticity (EN-DC) are all being used extensively to increase bandwidth.
These trends increase the probability of interference in mobile devices, to mitigate this,
advances in RF filtering technology are required. Space available for the RF front end (RFFE) is
shrinking, which means filters must also become smaller and more highly integrated. If that
wasn’t enough, filters also need to handle higher power to maintain adequate design margins
and maximize handset operating range.

Bulk acoustic wave (BAW) filters are evolving to address these challenges, along with other
approaches such as antenna routing and antenna tuning. BAW technology is advancing toward
handling higher frequencies up to 7 GHz – the upper limit of 5G frequency range 1 (FR1). BAW
filters provide low insertion loss that helps maintain RF output power and steep skirts to avoid
interference, which makes them extremely efficient at dissipating heat at high power levels. A
new generation of BAW technology, called micro-BAW (µBAW), saves board space by making
filters smaller and easier to integrate into RFFE modules that support 5G requirements.

5G Filtering Challenges
Table 1 illustrates how RF complexity in mobile infrastructure and devices has
increased with each generation of wireless technology. 5G adds yet another level of
complexity, intensifying the challenges for mobile device design engineers. RFFEs in
5G devices must support many more RF paths, higher power output and much greater
bandwidth, all while occupying less space – minimizing link losses and efficiently
removing heat are vital. Advances in filter technology are critical to achieving all these
goals.

Table 1. Band and feature developments over time.

25 Download this Whitepaper from Qorvo everythingRF.com

RF Filter eBook Qorvo

Power Handling and Heat Dissipation

Rising RF power requirements have increased overall system heat. To maintain
handset reliability and performance, filters must be able to handle higher power levels
and efficiently dissipate the additional heat.
RF power output in mobile devices has increased due to requirements such as Power
Class 2, a specification introduced by 3GPP to improve 2.5 GHz TDD-LTE coverage
worldwide. Power Class 2 raises power output at the handset antenna by 3 dB, as
shown in Figure 1, to increase uplink (UL) range. This specification is also being usedto
support new higher frequency 5G bands, such as n77, n78 and n79.

Furthermore, handset manufacturers require RFFE components that handle higher

power in order to maintain adequate design margins as system losses increase due to
growing RF complexity. In today's mobile and infrastructure devices, filters must
handle power levels as high as 33 dBm or more. By effectively dissipating the heat
caused by these higher input power levels, filters can maintain performance and avoid
lifetime degradation.

Figure 1. Power Class 2 increases handset uplink power and range compared to Power Class 3.

As shown in Figure 2, the structure of solidly mounted resonator (SMR) BAW filters
removes this heat much more efficiently than other filter technologies, like film bulk
acoustic resonator (FBAR). SMR BAW uses an acoustic reflector comprised of solid
layers below the piezoelectric resonator. These solid layers provide a direct link to the
silicon substrate below, allowing the heat generated within the filter to efficiently move
away from the piezo resonator through the reflector layers to the substrate. In contrast,
FBAR devices use an air cavity as the acoustic reflector. Because air is a poor heat
conductor, heat must travel around the cavity rather than through it. Therefore, SMR
devices are more efficient at moving heat into the silicon substrate, as shown in Figure
3. Increased system heat only has a minimal effect on SMR BAW filter and system
performance, in comparison to FBAR, because of these structural differences.

26 Download this Whitepaper from Qorvo everythingRF.com

RF Filter eBook Qorvo

Figure 2. SMR BAW power and heat handling

Heat Dissipation at Higher 5G Frequencies

The advantages of SMR BAW over FBAR are even more pronounced at the higher
frequencies used for 5G. Power handling in filters becomes more challenging at these
frequencies due to the smaller resonator size. However, with SMR BAW, the reflector
layers also become thinner, which further improves heat extraction (Figure 3). In
contrast, with FBAR, the membrane above the air cavity becomes thinner, reducing its
ability to move heat away from the resonator. As a result, with SMR BAW the
temperature rises 20⁰C per Watt of transmit power, compared to 70° C rise per watt
with FBAR. Less temperature variation means better performance, which enables SMR
BAW filters to meet system requirements for insertion loss and out-of-band (OOB)
attenuation under high-power and temperature conditions

Figure 3. SMR BAW versus FBAR filter power and heat handling

SMR BAW filters also have high quality (Q) factors due to their low insertion loss and
steep skirts, which also helps to reduce dissipated power and associated thermal
issues.

Insertion Losses
As Table 1 shows, the number of RF and filter band paths has increased dramatically in
LTE-Advanced Pro and 5G smartphones. These additional paths increase insertion loss
as signals pass through the system. Any increase in insertion loss reduces RF output,
especially in the higher 5G FR1 ranges above 3 GHz. PA output can be increased to

27 Download this Whitepaper from Qorvo everythingRF.com

RF Filter eBook Qorvo

compensate for these incurred losses, but it comes at the cost of higher system
current consumption. A better way to mitigate the system link budget increase is to
use lower-loss switches and filters, integrate individual components into modules and
reduce PC board line lengths. New SMR BAW generations offer reduced insertion loss
to help design engineers meet link budget specifications in today’s 5G devices.
Small Form Factor
As illustrated in Figure 4, PC board area in 5G smartphones is shrinking due to several
factors. Phone manufacturers are increasing battery size to support 5G and other new
features. To support broad range frequencies and new wireless standards including
Wi-Fi, low-band, mid-band, high-band, ultra-high band and millimeter wave, more
antennas are also being added. Therefore, system designers need the RFFE to occupy
less area on the PCB. As a result, this drives semiconductor manufacturers to make
filters and other components smaller and integrated into increasingly dense, complex
RFFE modules.

Figure 4. Larger batteries in 5G phones reduce the available RF related PCB space.

The development of µBAW technology is a key step toward addressing these

challenges in 5G smartphones. Like previous BAW generations, µBAW uses wafer level
packaging (WLP) with Cu-Sn-pillars. µBAW reduces the die peripheral area by placing
the device input and output terminals on top of the WLP-roof, overlapping with the
active area. As shown in Figure 5, µBAW shrinks the vias needed to connect the input
and output pads to the BAW filter. This also means the distance between the active
resonators can be further reduced when compared to previous BAW generations.

Out-of-Band Attenuation with BAW Multiplexers

Figure 5. Comparing previous WLP BAW generations with µBAW

28 Download this Whitepaper from Qorvo

everythingRF.com
RF Filter eBook Qorvo

In the 5G ecosystem, attenuating adjacent signals is becoming more difficult as

technologies such as CA are used to increase device data rates. CA, first introduced
with LTE Advanced, enables mobile network operators (MNOs) to maximize the use of
their existing mobile spectrum and bandwidth by combining multiple component
carriers into one data channel. By increasing uplink and downlink data rates, CA
enhances network performance and enables MNOs to ensure a high-quality user
experience.
5G allows many more carrier combinations than LTE, which increases the out-of-band
(OOB) attenuation challenges. Enabling CA requires simultaneous communication on
multiple carriers, which may be different frequency bands. The system must include
mechanisms to help achieve isolation between these RF paths. BAW multiplexers
achieve the required cross-isolation between the aggregated RF pathways, allowing
communication to occur simultaneously on all the aggregated carriers while
attenuating OOB signals for each pathway. BAW filters are well suited to attenuating
OOB signals thanks to their steep skirts.
Advancements in filter technology were critical to enabling the adoption of CA with
LTE Advanced Pro, and are equally critical for 5G. Filter material advancements in
coupling, aluminum nitride enhancement and material doping were important
milestones to achieving high attenuation, low return loss and the cross isolation
required for enabling CA.

Advances such as µBAW allow for the development of more complex integrated
multiplexers, as shown in Figure 6. These highly integrated multiplexers reduce board
space requirements and have the added benefit of reducing losses by eliminating the
need for additional matching elements.

Figure 6. Tackling CA with advanced BAW multiplexer filter design

Conclusion
SMR BAW filter technology has been critical to enabling successive waves of wireless
technology and is playing equally important roles in 5G mobile devices. New advances
in BAW technology are mitigating thermal issues, reducing board space requirements,
enabling greater integration, supporting complex CA requirements and handling the
higher frequencies used for 5G. Future BAW technology developments will continue to
provide system designers with the tools to further enhance our mobile devices and
networks.

29 Download this Whitepaper from Qorvo everythingRF.com

Published by Contributed by

Optimizing High-
Rejection LTCC Filter
Performance in Co-Planar
Waveguide
Implementations
Mini-Circuits’ BFHK-series of high-rejection LTCC filters have been characterized with stopband
rejection floors on the order of 90 dB and higher with a combination of size, reliability, and cost
currently unmatched by other filter technologies with comparable performance. Performance in
the customer’s system, however, can vary depending on the specific implementation. The
unique design of these devices features coaxial input and output pins on the bottom surface of
the structure requiring blind vias to the conductive layer of a stripline circuit board. While many
PCB manufacturers have developed the capability to build surface-mount assemblies with blind
vias reliably, some designers still prefer to use coplanar waveguide (CPW) wiring boards where
the contact between the conductive trace and the device ports is exposed on the top layer. In
addition to avoiding any concerns about blind vias, in certain use cases, CPW allows soldering
of other surface-mount components in shunt or series to the signal trace as well as fine-tuning
of the trace width and characteristic impedance for optimal matching conditions. To this end,
Mini-Circuits has developed the BFHKI series of CPW-compatible filters by on a pick-and-place-
ready platform, consisting of a sub-assembly with the LTCC component and an interposer
substrate, which converts the coaxial launch of the LTCC into a CPW interphase.

This paper describes the physical differences between stripline and coplanar waveguide
implementations of Mini-Circuits’ high-rejection LTCC filters and related effects on
performance. Channelization is proposed as an effective technique to achieve comparable
performance to stripline implementations in CPW environments. Real test data from a leading
customer’s evaluation of the new BFHKI series of these filters on interposer boards in a
channelized housing is then presented as proof of concept.

PCB Layouts in Stripline vs. Coplanar Waveguide Implementations

Because the land pattern of the BFHK-series filter package features conductive plating
across the bottom surface as shown in Figure 1, simply soldering the unit to the
exposed trace of a CPW board would cause functional issues due to shorting between
the conductive metallizations on the PCB and the bottom surface of the filter.

30 Download this Whitepaper from Mini-Circuits everythingRF.com

RF Filter eBook Mini-Circuits

Figure 1: Stripline PCB layout for a BFHK-series high-rejection LTCC filter

Mini-Circuits developed the BFHKI series of filters incorporating a novel interposer

board to allow drop-in use on CPW launches to facilitate integration on the user’s end
(Figure 2). The LTCC filter is pre-mounted on a stripline base designed for coaxial
mating between the top-layer metallizations on the customer’s PCB and the filter input
and output ports as shown in Figure 3.
The interposer board enables surface mount assembly of Mini-Circuits’ high-rejection
LTCC filters on coplanar waveguide substrate, but as with everything in the RF world,
it comes with caveats and requires due consideration in the design and assembly
process.

Figure 2: BFHKI-series LTCC filters feature an interposer between the filter and the
customer’s PCB, allowing easy use with top-layer transmission line.

Figure 3: Suggested PCB layout for BFHKI-series LTCC high-rejection band

pass filter on coplanar waveguide launches.

31 Download this Whitepaper from Mini-Circuits

everythingRF.com
RF Filter eBook Mini-Circuits

Effects on Performance
The specified rejection performance of BFHK-series filters is characterized on stripline
test boards where the line to and from the filter is shielded from cross-coupling, buried
in the PCB stackup. By contrast, in a CPW implementation, the exposed launch from
the PCB to the device creates leakage, which affects rejection performance. For
illustration, consider Figure 4. The filter response for BFHK-8501+ is shown on the left
exhibiting a rejection floor of about 100 dB. The response for BFHKI-8501+, the same
filter mounted on the interposer, exhibits typical rejection of about 70 dB in the lower
stopband and 50 dB in the upper stopband.

Figure 4: Characteristic response of BFHK-8501+ (left) vs. BFHKI-8501+ (right).

Note that in the test fixture for the BFHK-series model, the filter is mounted directly to
stripline substrate, while the BFHKI-series model is characterized with the interposer
mounted on an open CPW test board. It’s important to qualify that the variation in
rejection performance between the two models is a function of the quality of the
launch from the PCB to the filter rather than an intrinsic property of the filter itself.
An ideal solution would replicate the insulation provided by the top layer of a stripline
PCB in a CPW environment. Fortunately, this can be achieved by making use of the
LTCC’s conformal metallic coating and industry-standard channelization techniques,
yielding a response similar to that of the BFHK-series mounted directly on stripline.
What follows will present data measured by a leading customer in the test and
measurement field evaluating the performance of Mini-Circuits’ BFHKI-series filters on
CPW substrate in channelized housings. The results will demonstrate performance
parity between the baseline case of a filter mounted directly to stripline and the CPW
use case with the interposer board and channelization.

The Test Setup

The customer used a 20 mil test board with four positions for BFHKI filters of identical
layout using PC 2.92mm connectors. Input and output lines were grounded CPW
types. Ground vias were made through the whole PCB stackup. Additionally, there was
a TRL Calibration Board not shown here, which allowed de-embedding of the
measurements excluding effects from PC 2.92mm connector and CPW line to the
BFHKI unit. The test board and PCB layout are shown in Figure 5:

32 Download this Whitepaper from Mini-Circuits everythingRF.com

RF Filter eBook Mini-Circuits

Figure 5: Test board and PCB layout for the customer’s measurement of BFHKI filters.

Additionally, as an experiment, a conductive die attach adhesive (Ablebond 84-1LMI)

was applied around the interposer except at contacts with RF lines. The adhesive
application is pictured in Figure 6:

Figure 6: Conductive die attach adhesive was applied around the edge of the
interposer board as a measure to maximize rejection.

Channelization with a Silver-Plated Conducting Cover

To test the filters in a channelized environment, a silver-plated conductive cover was
used. The cover (pictured in Figure 7) contained single chambers for each of the four
BFHKI filter positions on the test board. Note that the dimensions of the chambers
were not optimized for the BFHKI filters specifically because the cover was built for
use with devices of various sizes.

Figure 7: Silver-plated conductive cover used with the test board for 4x BFHKI filters.

33 Download this Whitepaper from Mini-Circuits everythingRF.com

RF Filter eBook Mini-Circuits

The top cover contained a conductive sealing cord, and the device chambers were
fitted with a compressible material to create a firm electrical contact with the top face
of the BFHKI filter package. This so-called “EMI D-Profile” is built with a foam core
covered with metallic woven fabric (2.0 x 6.4mm, LAIRD 4202-AE-221-07900), and
must be mounted transverse to the channel. Otherwise, there may be a crosstalk path
through the non-conductive foam core. 3x3mm pieces of damping material (0.76mm
thick LAIRD Eccosorb GDS) were glued into the input and output channels to further
suppress any crosstalk (Figure 8).

Figure 8: Cover pictured with compressible EMI D-Profile mounted in device chambers and damping material
glued to input and output channels.

The strong imprints from the BFHKI units in Figure 8, indicate a good electrical contact
to the top of the filter package, although the chamber depth may be a little low as
dimensions were not optimized for these filters specifically.
Measurements

Multiple BFHKI models were measured in the test setup described above using a
Rohde & Schwarz ZNA43 2-port vector network analyzer (Figure 9). Calibration was
performed, and the DUT was de-embedded to the outer edge of the component
footprint. The taper from the CPW line width to the pad’s width is therefore still
included in the measurement.

Figure 9: Measurement was performed using a Rohde & Schwarz ZNA43 2-port vector network analyzer

For each model of interest, S11 and S21 were swept over the full operating bandwidth
under the following conditions:
• Open CPW test board without cover
• DUTs channelized with silver-plated conductive cover

34 Download this Whitepaper from Mini-Circuits

everythingRF.com
RF Filter eBook Mini-Circuits

• Channelized with cover and conductive die attach adhesive on outer edge of
interposer
These measurements were compared to S-Parameters for the BFHK-series counterpart
of each model as a baseline reference to study any deviations in each implementation
of the BFHKI model on CPW. The S-Parameters were taken from the Mini-Circuits
website and were measured on stripline test boards. The data presented below are
compiled from the different measurements and superimposed on the same set of axes
to easily compare the filter response under the different conditions tested.
Color Key:

• Red = Open
• Blue = With Cover
• Green = Cover + Adhesive
• Gray = BFHK on stripline (Mini-Circuits S-Parameters)

Figure 10: S11 and S21 plots for BFHKI-6751+ and BFHK-6751+ under various test conditions

Figure 11: S11 and S21 plots for BFHKI-1072+ and BFHK-1072+ under various test conditions.

35 Download this Whitepaper from Mini-Circuits everythingRF.com

RF Filter eBook Mini-Circuits

Figure 12: S11 and S21 plots for BFHKI-1572+ and BFHK-1572+ under various test conditions.

Figure 13: S11 and S21 plots for BFHKI-2492+ and BFHK-2492+ under various test conditions. The dip in
the passband at 28 GHz is caused by the cover.

Figure 14: S21 plots for all BFHKI models tested on open CPW.

36 Download this Whitepaper from Mini-Circuits everythingRF.com

RF Filter eBook Mini-Circuits

Figure 15: S21 plots for all BFHKI models tested with cover. Observe the dip at 28 GHz caused by
the cover.

Discussion
In all cases, the addition of a channelized cover resulted in significant improvement in
stopband rejection compared to the open CPW test case. Measurements of the BFHKI-
series filters with the cover exhibited a response more closely approximating the
reference case of the BFHK filter mounted directly on stripline with 90+ dB rejection
floor in the lower stopband and upper stopband rejection ranging approximately from
60 to 90 dB depending on the model and the frequency of the upper stopband. More
deviation was evident in the upper stopband, particularly at higher frequencies, but
the channelized implementation still exhibited 60 to 70 dB rejection even at
millimeter-wave range.
Application of conductive die attach adhesive achieved additional improvement in
rejection in some cases, but caused resonances in others (e.g. BFHKI-1572+) resulting
in poorer rejection. The cover may also cause resonances for higher passband
frequencies, and a cover with chambers optimized to the dimensions of the BFHKI
filter would presumably yield even better results.

Lastly, the measurements presented here exhibit poor return loss for passbands with
higher center frequencies. A 10 mil test board rather than the 20 mil test board used
here for filters with Fc higher than 18 GHz would likely correct this effect but was not
included in this evaluation.
Conclusion
Mini-Circuits’ BFHKI-series high-rejection LTCC filters with an interposer board were
developed to extend the revolutionary capability of the BFHK series to CPW
implementations with top-layer RF traces. While specifications on the datasheet exhibit
degradation in rejection performance due to characterization on an open CPW test
board, customers have demonstrated performance comparable to that of the BFHK
filter on stripline PCB with channelization.
The channelization techniques presented in this paper are well-understood and widely
used in the industry, making this a practical solution for customers using high-rejection
LTCC filter technology in coplanar waveguide implementations.

37 Download this Whitepaper from Mini-Circuits everythingRF.com

Published by Contributed by

Filter Design Using Discrete

Part-Value Optimization in
Cadence AWR Design
Environment
Optimization is often required to achieve desired performance when designing RF filters and
other highfrequency circuits with today’s simulation software tools. As an example, optimizing a
lumped-element filter involves adjusting the values of its lumped components until the filter
achieves an optimal frequency response. But once the component values have been determined
via optimization, they may still need to be adjusted to the closest discrete, or “real-life,”
manufacturer part values. Depending on the design’s complexity, this extra step can create a bit
of extra legwork for the designer.

Of course, additional simulations must be performed after changing the optimized part values
to the nearest available manufacturer part values. And if an optimized value falls roughly
halfway between the two closest available manufacturer part values, one may need to
experiment to determine which of the two values allows for better performance. In such cases,
designers may need to carry out further refinements by adjusting interconnect dimensions to
fine-tune the filter performance after setting the component values to the closest available
manufacturer part values.

Fortunately, proper tools make it possible to perform discrete part-value optimizations in which
component values are directly adjusted to optimal manufacturer part values. This optimization
method eliminates the need for designers to manually adjust optimized component values to
the closest available real-life part values, thereby cutting one step from the overall design
process.

This application note explains how a discrete part-value optimization method can be leveraged
to design a lumped-element bandpass filter designed using the Cadence AWR Design
Environment (AWRDE). The design includes Modelithics measurement-based passive-
component models, which enable simulations to accurately predict the filter’s real performance.
After completing an initial filter design, a discrete part-value optimization is carried out to
achieve the desired frequency response. A yield analysis is also performed to determine the
effect of component tolerances. At the conclusion, measured data is compared to the simulated
results.

Beginning with the iFilter Synthesis Module

The AWRDE consists of several software tools used for developing RF/microwave
products. Among them are Microwave Office for RF/microwave circuit design and the
AXIEM 3D planar electromagnetic (EM) simulator. For this design example, both of
these tools are used.

39 Download this Whitepaper from Modelithics everythingRF.com

RF Filter eBook Modelithics

Various add-on modules can also be utilized within the AWRDE platform. One of them
is iFilter, an integrated synthesis wizard used to develop RF/microwave filters. With
iFilter, designers can synthesize lumped-element and distributed filters and then
directly export them to Microwave Office for further analysis.
For this example, the design process begins with iFilter. Users open the iFilter module
within AWRDE by double-clicking “iFilter Filter Synthesis,” which is located under
“Wizards” in the “Project” browser. Starting a new filter design with iFilter prompts
users to specify the “Passband” and “Realization.” Subsequently, “Main Filter Type”
and “Options” must be specified (Fig. 1).

Figure 1. User interface when starting a design with iFilter.

For this example, “Passband” and “Realization” are specified as “Bandpass” and
“Lumped,” respectively. Next, “Narrowband Lumped Filter” is chosen as the Main Filter
Type. Subsequently, “Inductive (identical shunt C)” is chosen from the list of options.
Specifying the type of filter leads to the next user interface, which contains the filter
schematic, frequency response, and several user-defined parameters (Fig. 2).

40 Download this Whitepaper from Modelithics everythingRF.com

RF Filter eBook Modelithics

Figure 2. User-defined parameters, frequency response, and filter schematic.

For this design example, a Chebyshev filter with a center frequency of 950 MHz is
specified. In addition, the bandwidth is set to 300 MHz.
Clicking “Generate Design” automatically creates a new schematic of the filter in
Microwave Office (Fig. 3)

Figure 3. Generated Microwave Office schematic. Bold reference designations that correspond to the
variables are shown for illustration.

Note that the tool automatically set several variables. The bold reference designations
that correspond to the variables were added to the schematic for illustration. In this
case, the variable C_v1 was created for all the capacitor values, while variables L_v1,
L_v2, L_v3, and L_v4 were created for all inductor values. Table 1 lists these variables

41 Download this Whitepaper from Modelithics

everythingRF.com
RF Filter eBook Modelithics

and their corresponding values; Figure 4 shows the filter’s simulated frequency
response.
Table 1

Variables generated by iFilter that represent the part values of the inductors and capacitors.

Figure 4. Simulated frequency response shows a center frequency of 950 MHz.

Incorporating Microstrip Transmission Lines and Modelithics

Models
While the schematic shown in Figure 3 represents a good starting point for the design,
the overall design process is far from complete. For one, to properly model a real filter,
one must add microstrip interconnects to the schematic. Adding the microstrip
interconnects makes it possible to generate a layout that can later be sent to a
printed-circuit-board (PCB) manufacturer for fabrication.
One must also consider component modeling. In the schematic shown in Figure 3, the
inductor and capacitor models are ideal closed-form elements with user-defined
quality-factor (Q)

42 Download this Whitepaper from Modelithics everythingRF.com

RF Filter eBook Modelithics

parameters. Alternatively, utilizing Modelithics measurement-based inductor and

capacitor models makes it possible to predict the filter’s performance more accurately.
Modelithics models offer the benefit of both part-value and substrate scalability. These
models accurately capture substrate-dependent parasitic behavior and offer advanced
pad features for accurate EM/circuit co-simulations. Hence, Modelithics inductor and
capacitor models will be used in place of the ideal elements seen in Figure 3.
Of course, one could simply modify the original schematic by adding the microstrip
transmission lines and the Modelithics models. Another approach-and the one used in
this case—is to duplicate the original schematic, allowing it to be left as a reference if
needed. Figure 5 shows a new filter schematic created by duplicating the original one.

Figure 5. New filter schematic with microstrip interconnects and vias along with Modelithics passive-component models.

This new schematic contains all the necessary microstrip interconnects and vias. In
addition, the ideal inductor and capacitor models are replaced with Modelithics
models. For this example, the Würth Elektronik WE-KI 0402-size inductors are chosen
for all the inductors. For the capacitors, Kemet’s CBR04C 0402-size capacitors are
used. The WE-KI 0402-size series covers an inductance range of 1 to 120 nH, while the
CBR04C 0402-size series covers a capacitance range of 0.1 to 100 pF. Furthermore,
the substrate used for this filter is 10-mil-thick Rogers RO4350B, which was
implemented by placing the corresponding Modelithics substrate definition on the
schematic.
The schematic shown in Figure 5 also includes an “EXTRACT” block and a “STACKUP”
multi-layer substrate definition. These elements were added to the schematic via the
“Create_Stackup” script, which can be accessed by selecting “Scripts” from the
toolbar followed by “EM” and then “Create_Stackup.” These two elements together
facilitate EM/circuit co-simulation by creating and configuring a metal geometry from
the PCB layout of selected schematic elements. In this case, all the microstrip
interconnects and vias are selected for extraction (Fig. 6).

43 Download this Whitepaper from Modelithics everythingRF.com

RF Filter eBook Modelithics

Figure 6. Microstrip element specified for EM extraction.

Therefore, these elements will be analyzed with the AXIEM EM simulator. The
inductors and capacitors are the only elements from the schematic that are not
extracted for EM analysis, because they will be analyzed within the Microwave Office
circuit simulator. Figure 7 shows the extracted EM structure comprising of all the
interconnects and vias.

Figure 7. EM structure extracted using the “Create_Stackup” script.

Now that the EM extraction is properly configured, an EM/circuit co-simulation can be

performed by simulating the schematic shown in Figure 5. As explained above, the
microstrip interconnects and vias will be analyzed with the AXIEM EM simulator, while
the components will be analyzed within the Microwave Office circuit simulator. Also,
this schematic includes the same variables found in the original schematic (Table 1,
again). The inductor and capacitor values of the Modelithics component models are
set to the same variables as the corresponding components in the original schematic,
meaning the component values are unchanged.
Some designers may want to see how the microstrip interconnects alone affect the
filter’s performance. To do so, set the “Sim_mode” parameter of all Modelithics models
to 1. Setting a model’s “Sim_mode” parameter to 1 enables it to simply behave as an
ideal element, meaning that real-world parasitic, pad, and substrate effects are not
considered. Thus, utilizing this setting for all models makes it possible to determine
how the microstrip transmission lines alone affect performance, since the design is still
employing ideal component models like the original schematic.

44 Download this Whitepaper from Modelithics everythingRF.com

RF Filter eBook Modelithics

Figure 8 shows the simulated performance of the filter with the “Sim_mode”
parameter set to 1 for all models. Notice how the response is shifted downward in
frequency in comparison to the original frequency response (Fig. 4, again).

Figure 8. Simulated S21 and S11 with the “Sim_mode” parameter of all the Figure 9. Simulation results with “Sim_mode” set to 0 for all Modelithics
Modelithics models set to 1. Adding the microstrip elements resulted in a models. The filter shifts further downward in frequency and exhibits
downward shift in frequency in comparison to the original frequency greater passband insertion loss.
response.

The “Sim_mode” parameter of all models can now be set to 0. This setting enables a
model to behave as a full parasitic model, thus accounting for all real-world parasitic,
pad, and substrate effects. Figure 9 shows the frequency response after simulating the
filter with “Sim_mode” set to 0 for all models.
It is evident that the response shifted further downward in frequency and is more lossy
compared to the results shown in Figure 8. Specifically, the center frequency of the
passband is approximately 800 MHz, a 150-MHz shift from the desired center
frequency of 950 MHz. The filter also falls short of achieving the desired 300-MHz
bandwidth.

Discrete Part-Value Optimization to the Rescue

Because the filter does not currently meet the design goals, the next step is to perform
a discrete part-value optimization. This optimization technique will directly adjust the
inductor and capacitor values to the optimal part values in the Würth Elektronik WE-KI
and Kemet CBR04C part families, respectively.

To facilitate discrete part-value optimizations in the AWRDE, Modelithics offers .txt

files for all passive component models. Each file contains a list of all manufacturer part
values in the part family represented by the model. These files are located in the
\Modelithics\Data folder, where is the installation folder (C:\Program Files
(x86)\AWR\Foundry by default). The IND_WTH_0402_002.txt file contains all part
values in the Würth Elektronik WE-KI family of inductors, while the
CAP_KMT_0402_004.txt file contains the same information for the Kemet CBR04C
capacitor family. Figure 10 displays the equations that must be added to the
Microwave Office schematic (Fig. 5, again) to utilize these files for a discrete part-
value optimization. The equations shown are based on the default file locations.

45 Download this Whitepaper from Modelithics everythingRF.com

RF Filter eBook Modelithics

Figure 10. Equations added to the Microwave Office schematic to enable a discrete part-value optimization.

Note that the IND_WTH_0402_002.txt and CAP_KMT_0402_004.txt files generate

two vectors, “L_vector” and “C_vector,” respectively. As a result, the “L_vector”
element contains each of the part values in the Würth Elektronik WE-KI part family,
while the “C_vector” element contains each of the part values in the Kemet CBR04C
part family. To explain further, the Würth Elektronik WE-KI model encompasses a total
of 48 inductance values that range from 1 to 120 nH. Thus, the first vector index in the
“L_vector” element corresponds to 1 nH, while the 48th and final vector index
corresponds to 120 nH. The Kemet CBR04C model covers a capacitance range of 0.1
to 100 pF, with a total of 124 capacitance values. Therefore, the first vector index in
the “C_vector” element corresponds to 0.1 pF, while the 124th and final vector index
corresponds to 100 pF.
From the values shown in the bottom portion of Figure 10, one can see that the
variables are no longer equal to the initial values of Table 1. Rather, they are now set to
the vector indices that correspond to the manufacturer part values closest to the initial
values. Users may easily determine these indices by copying the part values from the
.txt files and pasting them into a spreadsheet (Fig 11).

View Table of Contents

46 Download this Whitepaper from Modelithics everythingRF.com

RF Filter eBook Modelithics

Figure 11. The part values from the .txt files can be pasted into a
spreadsheet (columns A and D) to determine the vector indices.

Simulating the filter produces the frequency response shown in Figure 12. The results
are similar—but not quite identical—to the simulated results of the filter with the initial
part values (Fig. 9, again). Table 2 lists the manufacturer part values used for this initial
simulation.

47 Download this Whitepaper from Modelithics everythingRF.com

RF Filter eBook Modelithics

Figure 12. Simulated results after changing the filter’s initial part values to the closest manufacturer part values

Table 2

Manufacturer part values closest to the initial values.

It is now time to perform a discrete part-value optimization. To do so, the variables

L_v1, L_v2, L_v3, L_v4, and C_v1 must be specified for optimization in the “Variable
Browser” (Fig. 13)

Figure 13. To execute a discrete part-value optimization, the variables that represent the part values of the inductors and
capacitors must be specified for optimization in the “Variable Browser.”

Next, the optimization goals must be set. Over the passband of 800 to 1,100 MHz, the
optimization goals are an S21 value greater than −2 dB and an S11 value less than −12
dB. For the lower and upper rejection bands, the goal is an S21 value of less than −20

48 Download this Whitepaper from Modelithics everythingRF.com

RF Filter eBook Modelithics

dB. Finally, upon opening the optimization user interface, “Discrete Local Search” must
be selected from the list of optimization methods.
Figure 14 shows the filter’s simulated frequency response after performing the discrete
part-value optimization, while Table 3 lists the optimized manufacturer part values.

Figure 14. Post-optimization frequency response. The response reveals a center

frequency of 950 MHz along with the desired bandwidth.

Table 3

Final optimized manufacturer part values.

The filter now achieves the desired performance. Keep in mind that one may fine-tune
interconnect dimensions to tweak a design so as to achieve the desired performance.
In this case, that step wasn’t necessary.
A yield analysis is the final step in the design process. This analysis is performed to
determine the sensitivity of the design with respect to part-value changes due to
component tolerances. The tolerances of the components in the optimized design can
be obtained from the Würth Elektronik and Kemet websites. L_v1 (3.6 nH) is specified
for a tolerance of ±5%, while the tolerances of L_v2 (9 nH) and L_v3 (5.6 nH) are each
specified for ±0.2 nH. L_v4 (11 nH) is specified for a tolerance of ±2%. Finally, C_v1 (7.0
pF) is specified for a tolerance of ±0.1 pF.

49 Download this Whitepaper from Modelithics

everythingRF.com
RF Filter eBook Modelithics

The Modelithics models include a “Tolerance” parameter that can be utilized for yield
simulations. To perform a yield analysis, users must check the “Use Statistics”
checkbox next to the “Tolerance” parameter and then specify the tolerance in
percentage as well as the distribution (Fig. 15).

Figure 15. Model parameters with the “Tolerance” parameter activated.

In this case, a normal (Gaussian) distribution was specified for all components. Figure
16 shows the results of the yield analysis.

Figure 16. Yield analysis simulation results showing S21 and S11.

It is clear that part-value tolerances have a greater effect on the passband, as the
results show a noticeable variation in S11 and a slight variation in S21. In contrast, part-
value tolerances have little effect on out-of-band performance.
Measured Data Versus Simulated Results and Closing

The final step is to validate the design by building and measuring the filter. Two filters
were assembled using the same Würth Elektronik WE-KI inductors and Kemet

50 Download this Whitepaper from Modelithics everythingRF.com

RF Filter eBook Modelithics

CBR04C capacitors from the final simulation (Fig. 17). Figure 18 shows the measured
data of both filters.

Figure 17. One of two filters built.

Figure 18. Measured S21 and S11 of both filters.

As a comparative analysis, the “Sim_mode” parameter of all the Modelithics models

can be set to 1 to see how the performance changes when using ideal models with the
final optimized part values. Figure 19 shows the simulated S21 with every model still
set to the optimized part value but with the “Sim_mode” parameter of each now set to
1. The measured data of both filters is also shown. The simulated results clearly do not
agree with the measured data, as the simulated frequency response is shifted higher in
frequency. The simulated bandwidth is also wider.

51 Download this Whitepaper from Modelithics

everythingRF.com
RF Filter eBook Modelithics

Figure 19. Measured data (dashed traces) and simulation results (solid lines) when models are set to ideal mode with
the optimized part values.

Figures 20 and 21 show the measured data along with the final simulated results when
the “Sim_mode” parameter is set to 0 for all models. The measured data corresponds
to the simulated results, thereby illustrating the benefit of using real-world parasitic
models. The design process is therefore validated.

Figure 20. Wideband measured data for S21 and S11 (dashed traces) along with the final simulated results (solid traces) when the
“Sim_mode” parameter of all models is set to 0.

52 Download this Whitepaper from Modelithics everythingRF.com

RF Filter eBook Modelithics

Figure 21. Narrowband measured data for S21 and S11 (dashed traces) along with the final simulated results (solid traces)
when the “Sim_mode” parameter of all models is set to 0.

In closing, utilizing Modelithics measurement-based models combined with a discrete

part-value optimization method within the Cadence AWR Design Environment can be
an effective way to design filters. The approach pinpoints exactly which manufacturer
part values are best suited for a design. A yield analysis can also be performed to
determine the sensitivity with respect to component tolerances.

53 Download this Whitepaper from Modelithics everythingRF.com

Published by Contributed by

Advanced BAW Filter

Technology and Its Impact on 5G
Abstract
The global rollout of 5G, and the rapid expansion of the Internet of Things, create significant new
RF filtering challenges. Qorvo’s bulk acoustic wave (BAW) filter technology is advancing to
overcome these challenges. Among the chief developments: BAW filters are evolving to support
higher frequencies and increased bandwidth for the new 5G and Wi-Fi band expansion. Complex
multi-filter modules (multiplexers and antennaplexers) are being used to address RF system
challenges, particularly with 5G. Additionally, a smaller uBAW (micro-BAW) form factor helps
squeeze complex RF front end (RFFE) architectures into the limited space available in handsets
and Internet of Things (IoT) devices. At the same time, Qorvo’s BAW technology helps to
mitigate the problem of heat dissipation associated with higher frequencies and smaller form
factors.
Introduction
Qorvo’s BAW RF filters play indispensable roles in wireless devices and infrastructure. They
ensure isolation between the ever-growing number of frequency bands and offer low insertion
loss to support device performance requirements. However, the global rollout of 5G, and the
evolution of other wireless standards such as Wi-Fi, create new filtering challenges that are
beyond the capabilities of traditional acoustic filters. Significant advances in BAW technology
were required to solve these new challenges.

Key Properties of BAW

The basic structure of BAW resonators consists of a thin-film layer of piezoelectric
material sandwiched between two metal thin-film electrodes. The voltage or electrical
field between the two electrodes excites an acoustic wave (see Figure 1). The wave
bounces between the surfaces of the two electrodes, forming an acoustic cavity

54 Download this Whitepaper from Qorvo everythingRF.com

RF Filter eBook Qorvo

between the top surface of the upper electrode and the bottom surface of the lower
electrode. The resonant frequency is inversely proportional to film thickness and
material properties; it is determined both by the thickness of the piezolayer and by the
thickness and material properties of the electrodes.
In BAW solidly mounted resonators (SMR), alternating solid layers of low and high
acoustic impedance below the piezoelectric and electrode layers act as an acoustic
Bragg reflector to confine energy in the vertical direction. The Bragg layers provide
reflectance that approaches the performance of a free surface. The perimeter of the
resonator is engineered to confine acoustic energy laterally, minimize acoustic leakage
and suppress unwanted modes.
The RF properties of a BAW filter are directly related to the thickness and material
properties of the piezoelectric, dielectric and metal films. Precise thicknesses are
achieved by optimized thin-film deposition and ion milling based on metrology and
electrical measurements. A higher resonant frequency can be achieved by removing
small amounts of the top layer film using ion milling.

The most common piezoelectric material used for BAW is Aluminum Nitride (AlN). AIN
provides the best balance of performance, reliability and manufacturability. The AlN
piezoelectric layer must have a high level of crystal orientation to achieve optimum
electromechanical coupling.

Figure 1. Cross section of BAW SMR filter.

A filter’s insertion loss is defined by several factors, including the losses in its individual
resonators. There are several possible loss mechanisms, but the most frequently
discussed is the loss of energy due to acoustic leakage, as shown in Figure 2. However,
most of the losses come from viscous losses in the materials.

Figure 2. Possible acoustic leakage mechanisms in SMR BAW.

55 Download this Whitepaper from Qorvo everythingRF.com

RF Filter eBook Qorvo

The Importance of Electro-Mechanical Coupling

Superior coupling is imperative to meet high-frequency passband requirements for 5G
and higher frequency Wi-Fi. The achievable width of the filter passband is determined
by the electromechanical coupling coefficient, k2eff. Maximizing is one of the
biggest challenges in thin-film BAW design.

The of a BAW resonator can be obtained from the resonance and anti-resonance
frequencies measured from its impedance response. The design of the resonator layers
can greatly enhance or degrade the coupling.

Filter Topologies
Networks of resonators can be designed to achieve various filter characteristics. BAW
filters can be classified into two major topology categories: ladder and lattice. A lattice
is a balanced topology, while a ladder is an unbalanced topology (see Figure 3).
Each topology has its advantages and disadvantages; however, the ladder topology is
the most common. To achieve the required bandwidth, the shunt elements are tuned
to a lower frequency than the series elements. The out-of-band rejection is determined
by the number of elements and the net capacitor divider. To meet a desired rejection
there is typically an optimum topology.

Figure 3. Ladder, lattice and ladder-lattice configurations.

High-Frequency BAW
5G uses new frequency bands that are well above those traditionally used for 4G LTE,
and above the range customarily supported by BAW filters. These include new 5G
Frequency Range 1 (FR1) and Frequency Range 2 (FR2) bands as shown in Figure 4
below. The upper limit of the Wi-Fi spectrum is also being extended above 7 GHz.
Evolving BAW technology to support these considerably higher frequencies requires
solving several technology challenge. Such as:

56 Download this Whitepaper from Qorvo everythingRF.com

RF Filter eBook Qorvo

Figure 4. Frequency bands for major wireless standards.

Creating Thinner Layers

The ability to create thinner layers, with accurate control over thickness and uniformity,
is critical to support higher frequencies. The layer thickness determines the frequency
response of the filter: thickness is inversely proportional to frequency. At high
frequencies, the layers must be extremely thin. Small variations in thickness have a
major impact on the filter’s frequency response; variations of just a few nanometers will
create an unacceptable deviation in the response. For example, one nanometer of
tungsten on a 5 GHz filter results in a frequency shift of approximately 10 MHz. Qorvo’s
BAW engineers were able to solve these challenges, capitalizing on the experience from
producing and selling more than 20 billion BAW filters over the last 10 years.

Minimizing Losses
Maintaining low resonator loss in acoustic filters becomes inherently more difficult at
higher frequencies, for a variety of different reasons. First, material losses generally
increase with higher frequencies. In addition, as the layers within the resonators
become thinner to support higher-frequency operation, the resonators must become
smaller in area to maintain the same capacitance. Unfortunately, this means that the
ratio of resonator perimeter to area increases, which makes the resonator more prone
to loss lateral through acoustic leakage. Mitigating above challenges requires careful
engineering of the resonator structure and a clear understanding of the underlying
physics. Qorvo’s BAW engineers have addressed these hurdles, resulting in lower
insertion losses and steeper filter skirts.
Increased Bandwidth
A key advantage of the new higher-frequency bands allocated for 5G is that they offer
much more bandwidth than is available in lower-frequency spectrum. 5G uses that
additional bandwidth to deliver higher data rates and increase network capacity. As a
result, 5G BAW filters must not only be able to operate at higher frequencies than
previous filter generations, they must also be able to support those frequencies with
much greater bandwidth.
Increasing filter bandwidth requires enhanced piezo-electric coupling. The use of
Scandium-doped AlN piezo layers has been key to overcoming this problem –
however, filter manufacturing becomes more challenging in comparison to regular AlN.

57 Download this Whitepaper from Qorvo everythingRF.com

RF Filter eBook Qorvo

Key factors that influence piezoelectric coupling in BAW resonators are the piezo
material, piezo layer quality, electrode configuration, acoustic reflector and parasitics.

Attaining perfect deposition for the piezo layer is challenging. A highly tuned pulsed
DC magnetron sputtering process has proven to be the best choice in attaining a
suitable layer for quality high-volume manufacturing.

The electrode configuration also significantly affects coupling. The thickness of the
electrodes relative to the piezo layer must be optimized to meet the filter’s frequency
and performance requirements. For each electrode material, there is an optimum
thickness ratio that delivers the highest coupling coefficient (k2eff).

Filters for 5G bands n77 (3.3-4.2 GHz), n78 (3.3-3.8 GHz) and n79 (4.4-5 GHz) require
wide bandwidth and benefit greatly from optimized coupling. Enhanced coupling is
also required when designing multiplexers for CA applications.

By using Sc-doped AlN, filter designers can increase coupling while keeping losses
constant, as shown in Figure 5 below. Choosing the right percentage of Sc allows for, a
good compromise between filter skirt steepness and bandwidth.

Figure 5. Measured coupling and relative acoustic loss of BAW resonators with various Sc%.

Thermal Issues and Reliability

5G is driving the need for smaller and sleeker devices. To maintain reliability in these
devices, filters must be able to effectively dissipate the heat generated during
operation. As BAW filters evolve to address higher frequencies, this becomes more
challenging. Resonator areas shrink and layers become thinner, thus the overall
resonator volume decreases. As a result, there is an increase in the power density in
the device; the filter must dissipate heat more efficiently to maintain an acceptable
operating temperature.

Fortunately, the structure of BAW helps to mitigate this problem, as shown in Figure 6.
BAW filters dissipate heat to the substrate through the solid reflector layers. Because
the layers are thinner at higher frequencies, this heat path becomes shorter, increasing
the filter’s ability to dissipate heat.

58 Download this Whitepaper from Qorvo everythingRF.com

RF Filter eBook Qorvo

Figure 6. Heat dissipation path in BAW filters.

Reducing Size
RFFE complexity continues to increase rapidly, as shown in Figure 7 – 5G accelerates
that process, due to the addition of new bands as well as new requirements such as
4x4 MIMO. At the same time, space allocated to the RFFE in handsets is actually
shrinking as smartphone makers cram in more features that are desirable to
consumers, such as bigger batteries and more-sophisticated cameras.
These trends mean there is constant pressure to make all RF components, including
filters, smaller and more highly integrated. The growth of the IoT also adds to demand
for very small filters and other RF components that can fit into tiny IoT devices.

Figure 7. Mobile device front-end module with CA filter technology.

59 Download this Whitepaper from Qorvo everythingRF.com

RF Filter eBook Qorvo

Qorvo’s uBAW, shown in Figure 8, is a key step in a continuing strategy to reduce filter
size. In traditional wafer level packaged (WLP) BAW filters, less than half of the total
area of the filter is occupied by the resonators. The periphery of the filter, which
includes the Cu-Sn input-output pillars, accounts for about 40% of the total area.
uBAW almost eliminates this peripheral area entirely by placing the pillars on top of
the WLP roof. This reduces the total size of the filter by an average of 30%. An
example filter based on uBAW to reduce size is the Qorvo QPQ2200Q, a fully 50 Ohm
matched automotive tested Wi-Fi and LTE coexistence filter with a footprint of
1.1x0.9x0.585 mm.

Figure 8. Simplified cross-section of legacy WLP BAW versus uBAW.

Complex Filters: Multiplexers and Antennaplexers

Complex modules that combine several filters have become important RFFE elements in
recent years and play even bigger roles in 5G.
Multiplexers are employed to support carrier aggregation (CA). CA is used to deliver
higher data rates by combining fragmented slivers of spectrum, often located in
different frequency bands. Multiplexers include all the filters required to support uplink
and downlink communications on each of the aggregated carriers, allowing concurrent
operation on all of them while preventing interference between them. Considerable
care and knowledge are required to design filters that work well together in a
multiplexer, since the performance of each filter is impacted by other filters in the
module. Unless the filters are carefully matched, a multiplexer may not perform
adequately in terms of insertion loss and isolation. For example, signals could leak
from a transmit to a receive path, desensitizing the receiver.

Figure 9. Band 1, 3, 7 multiplexer device.

60 Download this Whitepaper from Qorvo everythingRF.com

RF Filter eBook Qorvo

Antennaplexers address another growing challenge: the limitation on the number of

antennas that handsets can support. Trends such as the growing number of bands in a
small package or device, 4x4 MIMO and the addition of standards like ultra-wide band
(UWB) add to the pressure to increase the number of antennas in each handset. The
problem is that handset size and space limitations make it hard to add yet more
antennas while maintaining adequate performance. Antennaplexers reduce the need
to add antennas by allowing more frequency bands and standards (such as 5G cellular
bands, Wi-Fi, and GPS) to share each antenna. This frees more space for
manufacturers to add features that are attractive to consumers, such as additional
cameras and edge-to-edge screens. BAW antennaplexers offer low insertion loss, as
well as excellent isolation and power handling, maximizing radiated power and
reducing the risk of system de-sense.

Figure 10. Examples of Qorvo antennaplexer devices.

Simulation
Optimization tools are ubiquitous in electronic design. BAW filter design optimization
requires a scalable model that precisely predicts performance over a wide frequency
range. The rapidly increasing complexity of RFFE architectures requires an
unprecedented level of accuracy in simulation and modeling. The time allotted to

Figure 11. Comparison between simulation and measured data for SMR BAW.

develop a complex integrated RF module, from defining the required functionality to

producing a final qualified sample, is very short and leaves no room for trial and error.
Therefore, it is vital to be able to have extremely accurate methods for simulating the
performance of each functional block during design, to provide confidence that the

61 Download this Whitepaper from Qorvo everythingRF.com

RF Filter eBook Qorvo

final product will match the performance of the model.

Conclusion
Qorvo’s BAW filter technology has been advancing to provide a formidable solution to
the filtering challenges created by the global rollout of 5G. BAW filters are evolving to
support higher frequencies and increased bandwidth with ever increasing
performance. They are becoming smaller, more highly integrated and they are being
incorporated into a growing range of complex multiplexers and antennaplexers. Highly
accurate simulation methods are facilitating the rapid development of new filters and
modules to meet stringent requirements. BAW technology will continue to evolve to
address the future challenges presented by 5G and other wireless standards.

View Table of Contents

62 Download this Whitepaper from Qorvo everythingRF.com

 Conclusion
This concludes the ‘2024 RF Filter Digest’ eBook from everything RF. This
comprehensive resource has been developed for those looking to learn
about the current-generation RF Filters. This eBook discusses the
challenges involved with 5G mmWave filtering, high-rejection LTCC filter
performance, the role of MMIC filters in developing next-generation
systems, and the impact of BAW filters on 5G applications. It also discusses
filter design in the Cadence AWR design environment.

With the introduction of new wireless technologies, the frequency spectrum

is getting very crowded. To ensure these technologies can operate
simultaneously without impacting each other, RF Filters play a very
important role. RF Filters are crucial in optimizing signal transmission and
reception, selectively allowing desired frequencies to pass through while
attenuating unwanted signals.

The ‘2024 RF Filter Digest’ eBook would not have been possible without
contributions from industry leaders like Qorvo, Mini-Circuits, Marki
Microwave, Knowles and Modelithics. Stay tuned for more eBooks from
everything RF exploring cutting-edge technologies.

Content Contributed by: