R&S 5G FR1 Power Amplifier Seminar
5G Doherty
PA Design and Measurement
Clement POTIER – Staff MMIC Designer
29/06/2023
CONFIDENTIAL
Outline
Introduction Doherty design examples
• New challenges in PA design • Ideal structures and product variants
• iCana’s profile • Design tips
Different PA topologies for complex telecom Back in the lab
signals • iCana’s setup in today’s booth
• Review on amplification classes Doherty PA FR1 catalog of iCana
• Efficiency VS linearity
• AB class PAs, LMBA, Outphasing, Doherty,… Conclusion
Basics of Doherty
• Load modulation
• Matching and lambda/4
• Importance of phase distribution
• sizing
CONFIDENTIAL 2
A World of Data
IOT, augmented reality, connected cars, … => Need for more
bandwidth, higher data rate, higher output power, higher
integration, better efficiency… But always with the best linearity!
5G Radio Access Network (RAN): a new architecture enabling us to
achieve these goals, but chips’ architecture also need to re-invent
themselves while staying cost competitive
PA designers will be able to play with different leverages:
• Active components semiconductor process
• Biasing techniques
• Exotic amplification classes
• Load modulation
• Packaging co integration
CONFIDENTIAL 3
iCana’s Profile
iCana is a fabless semiconductor component supplier
specializing in design and manufacturing of RF components for
wireless communication. Our primary markets are 5G NR FR1
and FR2 infrastructure together with automotive connectivity
CONFIDENTIAL 4
PA Design Must Comply With New Standards And Modulations Signals
CAD design tools to help achieve best performances
and multi-chip integration
• Small signal • Packaging
• High power • 3D EM
• Linearity • LVS / DRC
• Multi technology • Sensitivity / Montecarlo
Higher order modulation
• ➔ 802.11be supports up to 4096-QAM!
(WiFi 5 “only” 256-QAM”)
• Circuit envelope simulator
Higher bandwidth but avoid signal
Higher Peek to Average Ratio (PAR) coexistence and spectrum sharing
• Design for higher Psat • Strong gain flatness
• Design for higher backoff efficiency • Sharp gain profile
• Harmonic balance / Intermod • Harmonics rejection and dedicated filtering
CONFIDENTIAL 5
Designers’ Choice – Technology / Process
Depending on the targeted power, voltage biasing
availability, integration and cost several options can
be picked
GaN/Si a game changer?
CONFIDENTIAL 6
Designers’ Choice – Amplification Class
“Conventional” classes or high efficiency classes using
harmonics and waveform shaping (Class F, F-1,..),
switching devices (D, E), compensation network (J)
CONFIDENTIAL 7
By The Way How To Define Efficiency ?
No overlap No power
between delivered at
voltage and harmonics
current
CONFIDENTIAL 8
… And Linearity ?
Biasing, physical effects in the semiconductor, gain
compression, intermodulation,… are sources of linearities
In band distortion
• Gain compression
• Trapping effects
➔ EVM, OP1dB, AM/AM, AM/PM
Out of band distortion Example of 2 tones response
• Harmonics
• Intermodulation
➔ OIP3, ACLR
CONFIDENTIAL 9
Nonlinear Behaviour Caused By Memory Effects
Description
The memory effects are defined as changes in
the amplitude and phase of the distortion
components, due to the previous signals. Time
responses are not instantaneous anymore but
will be convolved by the impulse response of
the system
They appear as:
• Asymmetries in the IMD products (frequency
domain)
• Dispersion in the decision points of the
constellation (time domain)
CONFIDENTIAL 10
Efficiency vs. Linearity
CONFIDENTIAL 11
Designers’ Choice – PA Topology
5G modulated signals present non constant envelope
with high PAPR ratios
➔ We need good efficiency at high power and backoff
Several PA topologies were investigated since several
years to address these trade-offs
• Load Modulated (Doherty, Chireix, Balanced,…), Envelope
Tracking, Harmonic Tuning, Switch mode, …
• Trade-off between performances and PA complexity
CONFIDENTIAL 12
Designers’ Choice – PA Topology
Several PA topologies were investigated since several years to address these trade-offs
• Load Modulated (Doherty, Chiriex, Balanced,…), Envelope Tracking, Harmonic Tuning, Switch mode, …
• Trade-off between performances and PA complexity
CONFIDENTIAL 13
How To Choose? With The Specs!
Standards define the appropriate levels
• 3GPP, small cells consortiums, ..
• Define the modulated schemes and signals
• Define the required linearity and output power for each frequency bands
Market
• System line-ups
• Overall costs, footprint, design time
• Regions specificities
CONFIDENTIAL 14
Doherty Seems Interesting, Let’s Have A Look!
Invented by W. H. Doherty at the Bell Labs in 1936
The Doherty architectures propose to use 2 parallel
power cells, to try to maintain a constant saturation
level of the total PA, thus a high efficiency
This can be achieved by several key parameters :
• One power cell (Carrier / Main) acts as normal PA
with high efficiency at saturation, then needs to see
a different load with the increasing Pin/Pout
➔ Use another amplifier like in active load pull
• The other power cell (peaking / auxiliary) will not be
“active” at low power ➔ Class C
• Keep voltage swing constant to keep efficiency
constant
➔ Impedance Inverter Network in between
CONFIDENTIAL 15
The Doherty Amplifier Basics (1/7)
B, AB class
C class
CONFIDENTIAL 16
The Doherty Amplifier Basics (2/7) – Keep Constant VoutM
If we consider a class B amplifier for the Main, VoutMax = Vdd – Vk
π Vout
Thus efficiency is η = ∙ , if the load is constant, efficiency will
4 Vdd
change as collector Voltage will change with current increase
If we can modulate the load such as Vout stays constant and only Ic
increases with Pin/Pout then the efficiency shall remain constant
CONFIDENTIAL 17
The Doherty Amplifier Basics (3/7) – Lambda/4 Section
But then the 2 power cell must operate with independent voltage swings!
➔ Need to design an Impedance Inverter Network to isolate both of them
One of the bandwidth limitation…
CONFIDENTIAL 18
The Doherty Amplifier Basics (4/7) – OBO Definition
We define :
Saturation
If OBO increases:
➔ Aux power cell size increases vs. Main size
➔ Aux bias point shift from deep class C
towards class B
CONFIDENTIAL 19
The Doherty Amplifier Basics (5/7) – Power Cells Sizing
Starting from the required DPA output power Pout,DPA, and OBO,
the Main device can be sized
The parameter identifying the Doherty region xbreak can be inferred
by solving the equation
The Aux device can be sized
The “virtual” bias point for the Aux device can be estimated
CONFIDENTIAL 20
The Doherty Amplifier Basics (6/7) – Defining The Loads
The DPA output design parameter can be inferred
CONFIDENTIAL 21
The Doherty Amplifier Basics (7/7) – Trade-Offs
< 6 GHz > 6 GHz
Parallel combination of transistor fingers Parallel combination of transistor fingers
(power bars) (power bars)
• Negligible de-phasing • Significant de-phasing
Series combination of transistors Series combination of transistors
(stacking) (stacking)
• Little interstage matching needed • Complex interstage matching needed
• Stability • Stability
Combination at PA level Combination at PA level
• Off-chip • Off-chip
• On-chip • On-chip
CONFIDENTIAL 22
The Doherty Amplifier – Ideal Example
CONFIDENTIAL 23
The Doherty Amplifier – Ideal Example Driver And Hybrid
GaN power cells
CONFIDENTIAL 24
The Doherty Amplifier – Ideal Example Driver And Hybrid
CONFIDENTIAL 25
MMIC Solutions For Higher Frequencies
CONFIDENTIAL 26
The Doherty Amplifier – Design For Real 5G FR1 Product
The PA must be cost competitive so hybrid solutions in over-molded packages are the best
approach
Several topologies with one driver followed by one splitter or one splitter followed on
each branch by a driver then the Doherty power cell, uneven divider, uneven power cells,
die bar + matching, pre-matched cells,…
Variants of Doherty, inverted, multi-way
All the parasitic must be taken in account as the phase is critical ➔ need for complex
simulation benches and cosimulations with packages and different process in one module
Bond wires need to be included
Electro thermal simulations and accurate thermal / trapping models are needed
CONFIDENTIAL 27
The Doherty Amplifier – Design For Real 5G FR1 Product
Active biasing technics aligned with process variations are mandatory,
especially for HBTs
IMD3 simulations with different tone spacings
Envelope simulations with 3gpp standard input signals
Raw ACLR < -35dB to achieve -50dB after DPD
All of that to … tune in the lab after manufacturing!
CONFIDENTIAL 28
Back In The Lab – iCana’s Booth Setup
Large Signal Test Setup Diagram and non-overmolded DUT
Modulated Signal: 5G NR, FDD, TM3.1, scs = 30kHz,
Bandwidth = 1 × 100 MHz, PAR = 8.5 dB
Frequencies: 3.45, 3.55, 3.65 GHz
Vcc1 = Vcc2 = Vcc3 = VBIAS = 5.0 V, VPAEN = 2.0 V
NGP800×2
CONFIDENTIAL 29
Back In The Lab – Measured Result
CONFIDENTIAL 30
Back In The Lab – Measured Result
ACLR-L1 & ACLR-U1 vs. Output Power with / without DPD
H ACL L vs t t o er H ACL vs t t o er
H H
H H
H H
d c
d c
Raw ACLR
ACL
ACL
t t o er d m t t o er d m
H D D ACL L vs o er S ee H D D ACL vs o er S ee
H H
H H
H H
d c
d c
ACLR after DPD
ACL
ACL
t t o er d m t t o er d m
CONFIDENTIAL 31
iCana’s Doherty PA Portfolio For 5G FR1
4 W High-Efficiency Power Amplifiers 8 W High-Efficiency Power Amplifiers
Part Number Frequency Range (GHz) Part Number Frequency Range (GHz)
ARQSP1819-4 1.8 – 1.9 ICASP3338-8 3.3 – 3.8
ARQSP2122-4 2.1 – 2.2 ICASP3742-8 3.7 – 4.2
ARQSP2324-4 2.3 – 2.4 ICASP4450-8 4.4 – 5.0
ARQSP2527-4 2.5 – 2.7
20 W High-Efficiency Power Amplifiers
ARQSP3336-4 3.3 - 3.6
Part Number Frequency Range (GHz)
ARQSP3437-4 3.4 – 3.7
ICASP3338-20 3.3 – 3.8
ARQSP3742-4 3.7 – 4.2
ICASP3742-20 3.7 – 4.2
ARQSP4450-4 4.4 – 5.0
ICASP4450-20 4.4 – 5.0
CONFIDENTIAL 32
Conclusions
5G and new communications are always pushing for more
power/efficiency/linearity
There will always be trade-offs, pushing the efforts towards designers, foundries
to develop innovative structures who need back strong support and innovative
tools and instruments from testing and automation industry
Doherty PAs are one of the topologies to assess these issues but as we explained
as also its limits. Is Doherty the answer to future 5G/6G requirements? Variations
and innovation around the DPA are flourishing thanks to the universities and
industry works to push these limits.
Meet us our booth for more discussions ☺
CONFIDENTIAL 33
Sources
The Doherty Power amplifier, Paolo Colantonio SC1 EuMIC, EuMW22
Load Modulated RFPAs, Steve Cripps, SC1 EuMIC, EuMW22
Linearization techniques Overview, Pere L. Gilabert, SC1 EuMIC, EuMW22
PA Theoretical Foundation, Franco Giannini, SC1 EuMIC, EuMW22
Doherty Power Amplifiers for Sub-6 GHz and Beyond-6 GHz: Challenges and Design Approaches,
A. Piacibello, V. Camarchia, EuMC31, EuMW22
Improving RG GaN Success, Wolfspee, Mircowave Journal ebook, March 2022
The Doherty Power Amplifier, P. Colantonio, F. Giannini, R. Giofre, L. Piazzon, Class Chapter of
University of Roma Tor Vergata, Italy
Compact N-Way Doherty Power Combiners for mm-wave 5G Transmitter, A. K. Kumaran, H. M.
Nemati, L. C. N. de Vreede, M. S. Alavi, Benelux RF conf. May 2023
CONFIDENTIAL 34
[Link]
