Transcript V t

Lecture 20
Current Source Biasing and MOS
Amplifier Design
• Temperature and supply-independent biasing
• OPAMP design
• Summary
Michael L. Bushnell
CAIP Center and WINLAB
ECE Dept.,Analog
Rutgers
U., Piscataway, NJ
and Low-Power Design
Lecture 20 (c) 2003
1
Temperature and SupplyIndependent Biasing
• Biasing must also be independent of process variations
– Critical for MOS circuits
• Bias I fluctuations with T, VDD, and process cause
wasted energy
• Supply-independent biasing is important
– Avoid injecting high-f noise on power lines into circuit
signal path
Analog and Low-Power Design
Lecture 20 (c) 2003
2
Supply-Independent Biasing
• Refer bias circuit to potential other than VDD:






Vt – threshold voltage of MOSFET
Vbg – band-gap voltage
DVt of dissimilar devices
VBE of parasitic bipolar transistor in CMOS
VT – thermal voltage (k T / q)
Zener diode breakdown V – too high breakdown V
• Self-biasing requires a start-up circuit
– Must force circuit into equilibrium in desired stable state
Analog and Low-Power Design
Lecture 20 (c) 2003
3
Vt Referenced Self-Biased Circuit
• Fig 12.25a (old book)
Analog and Low-Power Design
Lecture 20 (c) 2003
4
Analysis
• Feedback from M2, M3 & M4 forces same current I to flow in
•
•
•
•
M1 and R
Operating point must satisfy:
I R = VGS1 = V t1 +
2I
mn Cox (W / L)1
Neglect channel length modulation & body effect
Make 2nd term small compared to Vt
1. Use low bias current
2. Use large W/L
• I  Vt / R
Analog and Low-Power Design
Lecture 20 (c) 2003
5
Analysis (continued)
• If 2nd term included, O/P current is slightly reduced but T and
•
VDD dependence is the same
Must ensure stability – need to verify that feedback loop gain is
< 1 at operating point
– Do by breaking the loop, injecting a signal, & checking the gain
• Must determine degree of supply independence
– Channel length modulation in M2 & M1 causes bias current
variation
• Reduce variation with cascode current source
Analog and Low-Power Design
Lecture 20 (c) 2003
6
Problem
• In typical MOS process, Vt not well controlled
 0.5 V  Vt  0.8 V
• Vtn has TC (temperature coefficient) of –2 mV / oC, but
diffused R’s have a large positive TC
• Results in O/P current with large negative TC
Analog and Low-Power Design
Lecture 20 (c) 2003
7
Delta Vt Temperature
Independence
• Use differences in Vt of two devices of same polarity
but with different channel implants
• Advantage: TC’s of two devices cancel to first order
– Can get O/P Voltage TC as low as 20 ppm / oC
Analog and Low-Power Design
Lecture 20 (c) 2003
8
DVT Referenced Biasing
• Fig 12.25b (old book)
Analog and Low-Power Design
Lecture 20 (c) 2003
9
Disadvantages
• Large initial tolerance in O/P voltage value
– Threshold voltages have large tolerance
• Extensively used for precision voltage references in
nMOS and CMOS
– Need to trim a resistor to adjust absolute O/P voltage
Analog and Low-Power Design
Lecture 20 (c) 2003
10
VBE Referenced Biasing
• Fig 12.26 (old book)
Analog and Low-Power Design
Lecture 20 (c) 2003
11
Analysis
• pnp is a parasitic bipolar device in p-substrate CMOS
• Can also use a parasitic npn transistor
• Feedback involving M1, M2, M3, M4 forces emitter current in Q1
•
•
to match R current
I R = VT ln I or I = VBE1
IS
R
Advantage: VBE is well-controlled, with 5% variation
Disadvantage: VBE has a negative TC of –2 mV / oC
– R has a strong positive TC
– Leads to a strong negative TC in bias current
– Can reduce reference current variation with a cascode or Wilson
current source
Analog and Low-Power Design
Lecture 20 (c) 2003
12
VT-Referenced Biasing (Thermal
Voltage)
• Fig 12.27 (old book)
Analog and Low-Power Design
Lecture 20 (c) 2003
13
Analysis
• Q1 & Q2 transistor areas differ by n factor
• Feedback circuit makes them operate at same bias current
 Difference between two VBE’s appears across resistor R
 VBE = VT ln I
VBE / VT

IS
I = IS e
( )
• Get: I R = VBE1 – VBE2
[
]
( ) ( )
( )
Or
= VT ln I - VT ln I
IS
n IS
= VT ln I n IS
IS I
I = VT ln (n)
R Analog and Low-Power Design
Lecture 20 (c) 2003
14
Discussion
• Advantage: VT has positive temperature coefficient
(VT = kT / q)
• R has a positive TC, so current output is relatively T
independent
Analog and Low-Power Design
Lecture 20 (c) 2003
15
VT Referenced Self-Biased
Reference Circuit
• With cascoded devices:
– Improves power-supply rejection and initial accuracy
• DV across R is ~ 100 mV
• Small differences in VGS for M1 & M2 cause large O/P
current (IOUT) changes
– Result from device mismatches or from channel length
modulation in M1 & M2 (with different drain voltages)
Analog and Low-Power Design
Lecture 20 (c) 2003
16
Self-Biased Reference Circuit
• Fig 12.28 (old book)
Analog and Low-Power Design
Lecture 20 (c) 2003
17
Band-Gap Referenced Biasing
• Fig 12.29 (old book)
Analog and Low-Power Design
Lecture 20 (c) 2003
18
Analysis
• IM8 drain = VT ln (n)
•
•
•
R
V0 = VBE + VT ln (n) x R
R
= VBE + x VT ln (n)
OPAMP maintains V0 at both + and – terminals due to feedback
IOUT = V0 / R2
Advantage: By weighting VBE and VT components, one gets a
voltage of any desired TC
(
)
– Can exactly cancel an R TC
Analog and Low-Power Design
Lecture 20 (c) 2003
19
Weighting
• Parameter x determines weighting of VT-dependent
portion:
1. Can use only common collector transistors
2. OPAMP has MOS transistors, so their input offset
voltage and input offset voltage temperature drift
influence O/P voltage of the reference
• Must remove offset with analog storage and cancellation
Analog and Low-Power Design
Lecture 20 (c) 2003
20
Summary
• Temperature and supply-independent
biasing
• OPAMP design
Analog and Low-Power Design
Lecture 20 (c) 2003
21