A%202.5V,%2077-GHz,%20Automotive%20Radar%20Chipset

1
A 2.5V, 77-GHz, AutomotiveRadar Chipset
WE2B-5

• Sean T. Nicolson1, Keith A. Tang1, Kenneth H.K.
  Yau1, Pascal Chevalier2, Bernard Sautreuil2, and
  Sorin P. Voinigescu1
• 1) Edward S. Rogers, Sr. Dept. of Electrical
  Comp. Eng., University of Toronto,
  Toronto, ON M5S 3G4, Canada 2)
  STMicroelectronics, 850 rue Jean Monnet,
  F-38926 Crolles, France

2
Outline

• Motivation
• Transceiver architecture
• Circuit design layout some device insight
• Fabrication technology
• Measurements
• Conclusions

3
Applications

• W-band applications 77GHz auto radar, 94GHz
  weather radar, imaging, data communications
• All applications require a W-band radio
  transceiver.

4
The Doppler Radar Transceiver

• Doppler transceiver block diagram
• Development steps
• Design test circuit blocks optimize HBT for
  circuit performance
• Integrate circuit blocks into transceiver
• Duplicate to form arrays

5
Low-noise Amplifier

• 3-stage design, add R1 to de-Q the final stage.
• Noise impedance matching including CPAD
  Nicolson, 2006.

250mm
1pF decoupling caps
6
Power Amplifier

• Primary goal maximize PAE
• common source, class AB operation

7
Down-conversion Mixer

• Classical Gilbert cell mixer has poorlinearity
  at 2.5V
• Eliminate RF pair
• Couple to LNA using transformer
• Bias quad from center tap
• Simulations
• 9dB conversion gain
• 3dBm OP1dB (1.25VPP/side)
• 12.5mW PDC

input
8
Mixer IF Amp Layout

• Layout is critical at 77GHz.

9
Frequency Divider

• The most challenging block to operate from 2.5V.
• Given sizes of Q1-Q6, the size of Q7 Q8 can be
  optimized.
• important inductor size, swing, latch pair
  size, current density.

10
Frequency Divider

• The most challenging block to operate from 2.5V.
• Given sizes of Q1-Q6, the size of Q7 Q8 can be
  optimized.
• important inductor size, swing, latch pair
  size, current density.

11
SiGe Technology Chevalier, 2006

• 230/290GHz fT/fMAX SiGe HBT process
• Several process splits to find optimal HBT
  profile.

12
LNA Measurements

• Fabricated measured a 65nm CMOS LNA for
  comparison.
• CMOS has more power supply variation (HBT
  feedback is stronger)
• CMOS has low output resistance ? higher bandwidth

13
LNA Measurements

• S21 vs. temp. shows 6dB variation up to 125C _at_
  center band.
• Again, upper band shows greater variation (less
  feedback).

14
PA Measurements

• PAE 15.7, PSAT 14 dBm, OP1dB 11dBm

15
Mixer IF Amplifier Measurements

• DSB noise figure of 13dB is pessimistic
• harmonics from LO multiplier source, includes 3dB
  transformer loss.
• Min. NF current density at 73GHz (common base) is
  5.5mA/mm2.

16
Frequency Divider Measurements

• Operates up to 105.44GHz at 25C and 97GHz at
  100C.
• limited by power available from source.

17
Performance of Process Splits

• The best split is the reference, with the highest
  fMAX.

18
Conclusions and Future Work

• Excellent performance despite 2.5V supply.
• SiGe divider 94GHz self-oscillation, and 75mW
  power consumption.
• 77GHz power amplifier PAE of 15.7
• 5dBm OP1dB from Mixer IF amplifier
• -101.5dBc/Hz at phase noise at 1MHz offset
• Transceiver currently in the fab
• lt 500mW power consumption (180mW for receiver,
  inc. VCO)
• Contains only 33 HBTs (includes 16 in divider)
  2 MOS varactors.

19
Acknowledgements

• Ricardo Aroca and Katia Laskin for measurement
  help
• Jaro Pristupa and Eugenia Distefano for
  CAD/Network support
• STMicroelectronics CITO for fabrication and
  funding

20
Voltage-Controlled Oscillator

• Minimize phase noise, supply temp dependence
  2, 3.
• Small LB, differential tuning
• C1 CBE gtgt CVAR, C3 cancels CBC

21
VCO Measurements

• Phase noise better than -100dBc/Hz at 77GHz 2,
  3.

22
More About Process Splits

• LNA S21 for several process splits.
• Reference split looks the best.

23
More About Process Splits

• PA saturated S21 for several process splits
• Again, the reference split looks the best.

24
More About Process Splits

• PA S11 for several process splits

25
Inductor Measurements

• Accurately simulated/modeled Dickson, 2005
  passives ( 1pH).