CMOS PA provides a path to continued system enhancements through further integration

1
ISSCC 2008 Any semiconductor component which
can be implemented in CMOS will be implemented in
CMOS

• CMOS PA provides a path to continued system
  enhancements through further integration

Donald McClymont February 2008
2
Overview of Presentation

• Cellular market and associated cost pressures
• The drive to CMOS utilization
• CMOS PA
• Further integration possibilities
• Conclusion

3
Market Size
Greater than 1 billion units and growing
Source, Strategy Analytics, Axiom
4
Cost Pressures

• Increase in market size is driven to a large
  extent by the Ultra Low Cost market
• Emerging economies
• Basic telephony requirements only
• Cost of ownership is key to expanding the market

5
Drive to CMOS

• Transceiver has been converted from BiCMOS to
  CMOS in the last years
• Many technological problems were overcome
• Examples
• It was thought that 1/f noise made direct
  conversion receivers impossible
• Low phase noise VCOs were developed

6
Cost Argument

• If digital content is low, it can be argued that
  an older process technology with Bipolar
  extensions may be lower cost than a CMOS process
  of finer geometry
• But smaller geometries catch up eventually
  economies of scale
• Advantage can be taken of digital heavy
  architectures

7
CMOS PA
8
Axiom CMOS PA

• The PA had remained a bastion of specialty
  technology
• Currently represents an increasing proportion of
  the RF BOM
• CMOS PA offers a significant cost reduction
• The CMOS PA is now a reality
• Axiom AX502 quad band GPRS PA
• Shipped 10m units in first year of production
• Volume growing steadily

9
Axiom AX502 GPRS Power Amplifier
DCS/PCS Output
DCS/PCS Input

• Quad band GSM/GPRS PA
• Fully integrated power control
• Superior Reliability

RF Detector
Power Control
BS
TXEN
VRAMP
RF Detector
Full technical presentation will be made later
this week
EGSM/850 Input
EGSM/850 Output
10
Axioms Single-Chip CMOS PA Compared to GaAs PA
Modules
SINGLE DIE ADVANTAGE
Typical Module-based PA
CMOS PA
Quad-Band GSM/GPRS

• Many off-chip passive components
• Expensive laminate packaging
• Proprietary GaAs process

• No external components
• Standard 5x5mm package
• Mainstream Silicon CMOS

11
Breakdown Issues

• Voltage breakdown is a key issue to solve in ANY
  power amplifier
• Especially problematic when using standard CMOS
• Output voltage swing exceeds 20V at maximum
  output power for the GSM case
• Breakdown of a 0.13u core transistor is 3V
• An impedance transformation provides the core of
  the solution
• Other circuit techniques must be employed to
  further enhance reliability
• Design must demonstrate full compliance with
  foundry design rules

12
LC Match vs Magnetic Transformer
Resonant Match
Transformer Match
Inductor Stored Energy 49 VA
Inductor Stored Energy 7 VA
Transformer stored energy less than LC match
Transformer can have lower Q
13
Transformer Match versus Resonant Match

• High Transformation ratio is required to
  accommodate the low drain voltage of the CMOS
  transistor
• The higher the transformation ratio, the higher
  the energy stored in the matching network
• Higher stored energy in resonant match makes the
  Q extremely high
• A transformer offers a lower Q match which makes
  it possible to cover multiple bands while
  maintaining a high impedance transformation ratio

14
Output Match Integration DAT

• GSM power into 50O output
• - Transformation ratio 18
• - Drain impedance 6 O
• - Power combining 8
• - Equivalent impedance ratio
• from the drains to the output 164
• Equivalent drain impedance 0.8 O
• Passive Efficiency
  85 (0.7dB)

Active transformer topology
15
Current Consumption in a Phone
16
Further Integration Possibilities
17
Advantages of Further Integration

• Interface Optimization
• Opportunities to employ more complex systems to
  enhance performance and reduce cost
• More sensors and more actuators to feed control
  loops

18
Interface Optimization 1 PA/Transceiver

• Traditionally transceivers are single ended
  outputs to support single ended PA input
• Transceiver would prefer to be differential
  (harmonic generation, noise)
• Having PA input differential reduces current
  consumption of buffer

19
Interface Optimization 2 PA/Switch

• Customer generally drives vendors to margin stack

20
Interface Optimization 2 PA/Switch

• Antenna spec for harmonic suppression is -30dBm
• PA may be specified with -16dBm harmonic
  generation, switch module with 30dB of harmonic
  rejection gt provides -46dBm at antenna
• A combined device may be specified with -36dBm
  which is more than enough margin
• Margin stacking of filter rejection typically
  increases insertion loss, which means that the PA
  must produce more power
• gt Higher current consumption

21
Advantages of Further Integration

• Interface Optimization
• Opportunities to employ more complex systems to
  enhance performance and reduce cost
• More sensors and more actuators to feed control
  loops

22
Increased complexity 1 Sensors

• In power control having multiple measurement
  points may aid in power setting accuracy and
  immunity to load variations
• Addition of multiple measurement points is
  significantly easier in monolithic
  implementations
• Saves pins etc.

23
Increased complexity 2 Actuation points

• Control PA bias to set power levels
• Control VCO bias to save current at lower power
  levels where drive power may not be required

24
Increased Complexity why is it useful ?

• Example Polar transmit architectures
• Main goal is to linearize a saturated power
  amplifier
• Both AM/AM and AM/PM response
• Pre-distortion architectures are simple, but
  depend on characterization of PA block and
  extensive production calibration
• Feedback provides more immunity to process
  variation
• More feedback points can lead to a more accurate
  linearization
• Better RF performance
• Lower production calibration time

25
Whats our target ?

• Typical 3G Front End
• 5 PAs
• 1P9T switch
• 3 Duplexers
• 4 SAW filters
• Comes back to cost
• How do we replicate the cost curve of GSM ?

26
Conclusion

• Our challenge as an industry is to continue to
  innovate to integrate impossible blocks in CMOS

27
Thank You
28
(No Transcript)