EEE523: ADVANCED ANALOG IC DESIGN
PROJECT 1 CMOS -MULTIPLIER BASED CONSTANT-GM CURRENT REFERENCE Current Mirrors
NAME (LAST, FIRST, MI)
: PARAMAIAHGARI SRIKANTH R
ASU ID #
: 1206321047
Objective: 1 Objective In this project, we will design three β-multiplier based current reference circuits, using CMOS transistors in TSMC 0.25μ CMOS design library. For the current setting resistor R, we can use an ideal zero-temperature coefficient resistor from analogLib library. The resistor R and β multiplication value K sets the branch currents. We will size the transistors, K and R to optimize in such a way that the reference currents Iref1 , Iref2 match for a wide range of supply voltage Vdd for designs 1, 2 and 3 and try to be as constant as possible. Also, We would use this current to bias an NMOS transistor and prove that its gm is constant, just like the current setting resistor R across process, voltage and temperature. Design Procedure: First of all to make sure the startup circuit work properly, we need to make sure the VGS of startup transistor less than Vth after the circuit turns on completely. To make sure VGS of stratup transistor is less than Vth , we need to increase the length of the diode connected PMOS or NMOS. For all three designs, considering M1 and M2 NMOS transistors applying KVL around the loop, we can write Vgs1=Vgs2+I2R. which can only be valid if VGSI > VGS2. To ensure that this is the case, we use a larger value of ß in M2, that is, we multiply-up M1's ß in M2 so that less gate-source voltage is needed to conduct Iref. This is done by simply using a larger width in M2.The resulting circuit is called a Beta-multiplier reference circuit. 𝑉𝑔𝑠=√(2∗𝐼𝑑/(𝑘𝑛′(𝑊/𝐿)))+𝑉𝑡 Assume β2=𝑘′(𝑊/𝐿)2 = K β1 Therefore, W2 = KW1 Solving for I2 from the above two relations, we get 𝐼2=(2/(𝑅2∗𝑘′∗(𝑊/𝐿)))1∗(1−1/√𝐾)2 Using the above equation wr can solve for R for given reference current Iref=10uA. Note that we have no dependency on VDD. For k=4, solving for Iref, we have a constant gm biasing circuit. Gm= Sqrt(2*k*w/L*Iref)= 1/R. In this pro ject, “k” is chosen to be 4 in three designs. To find the correct R value at Iref=10µA, parametric analysis in Cadence is performed by varying the R(Because the calculated value of R is not exact due to ignoring body effect and channel length modulation effects).
All the transistors are sized by assuming Vdsat=125 mV. For design3, to achieve zero temperature coefficient for Vref, we need to find the correct value of resistor temperature coefficient to compensate it. The procedure to find the temperature coefficient is as follows : Plot dVth/dT of NMOS vs temperature (The Vth can be obtained from Calculator-->info-->op--> select the NMOS device and select Vth). Find the value of dVth/dT at 27C from the plot. Plot (1/Kp)*(dKp/dT) of the NMOS device vs temperature (The betaeff parameter from the op of NMOS should be divided by its (W/L) to get Kp). Find the value of (1/Kp)*(dKp/dT) at 27 C from the plot.Now use the equation (1/R)*(dR/dT) = R*Kp*(dVth/dT) – (1/Kp)*(dKp/dT) and find out (1/R)*(dR/dT) using the values calculated above. This value corresponds to the Temperature coefficient1 parameter in the resistor model used in cadence. Now we can sweep the temperature coefficient value of the resistor and select the value of temperature coefficient 1 such that dVref/dT is minimum at 27C.
Design1: Schematic:
Sizing and biasing: Parameter
Min
Nom
Max
VDD
2.0V
2.5V
3.0V
Iref1
9.043u
9.997uA
10.91u
Iref2
9.042u
9.997uA
10.91u
gm
110.6u
113.8u
117u
These are obtained from Iref vs Vdd plots attached below. And it can be observed from the below schematic that Iref1=Iref2=10uA for 2.5V.
R= 7.757 kOhms. Device
W(um)
L(um)
Vgs-Vt(m V)
Vds(mV)
Ibias(uA)
M1
2.775
0.6
136.9
594.6
10
M2
11.1
0.6
81.96
796.9
10
M3T
11.925
0.825
146
-737.4
10
M3B
7.95
0.9
151.2
-1.168
10
M4T
11.925
0.9
146
-736
10
M4B
7.95
0.6
151.3
888.5
10
MS1
0.45
45
1.554V
-2.445V
0.6
MS2
0.45
0.3
44.7m
1.169V
10.01f
MS3
0.45
0.6
132m
45.2
0.6
Iref1,Iref2 Vs Vdd: It can be observed from the below plot that Iref1 and Iref2 are with +/- 5% (<0.5uA difference for 10uA) for entire Vdd range of 2 V to 3 V.
Vref vs Vdd:
Vref vs Temperature:
Gm vs Temperature: Theoritical gm=1/R= 1/7.754K = 128.9 uA/V. which is close to practical gm=129uA.
Design2: Schematic: Current of 10 uA is flowing through both the legs as observed at the resistor.
Sizing and Biasing:
Design questions: R=7.341 kOhms(parametric analysis) Parameter Min
Nom
Max
VDD
2.0V
2.5V
3.0V
Iref1
9.54uA
10uA
10.2u
Iref2
9.51uA
10uA
10.25u
Gm
128uA/V 130uA/V 131uA/V
DESIGN QUESTIONS: Device
W
L
VGS-Vt
VDS
Ibias
M1
2.775um
.6um
136mV
594.6mV 10uA
M2
11.1um
.6um
82.2mV
796.9mV 10uA
M3T
11.925um .9um
141mV
737mV
10uA
M3B
7.95um
.9um
151.4mV 1.168V
10uA
M4B
7.95um
.9um
151.2mV 889.5V
10uA
M4T
11.925um .6um
140.6mV 736mV
10uA
M1T
7.95um
.6um
141.8mV 479.9mV 10uA
M2T
7.95um
.6um
144mV
Startup_n
0.45um
0.6um 131mV
45.2mV
0.697uA
Startup_p
0.450um
45um
1.553V
2.455V
0.697uA
NMOS_switch 0.450um
.6um
45mV
1.169V
10fA
859.5mV 10uA
Iref1,Iref2 Vs Vdd: It can be observed from the below plot that Iref1 and Iref2 are with +/- 5% (<0.5uA difference for 10uA) for entire Vdd range of 2 V to 3 V.
Vref vs Vdd:
Vref vs Temperature:
Gm vs Temperature: Theoritical gm=1/7.341K= 136 uA/V close to practical.
Design3: Schematic:
Sizing and biasing:
Design questions: R=7.4265 kOhms(parametric analysis) Parameter
Min
Nom
Max
VDD
2.0V
2.5V
3.0V
Iref1
9.9996uA
10uA
10.0006u
Iref2
9.99975uA 10uA
Gm(zero temp co R) 127.7uA/V
10.00095u
129.1uA/V 130.2uA/V
DESIGN QUESTIONS: Device
W
L
VGS-Vt
VDS
Ibias
M1
2.775
.6um
136mV
594mV
10uA
M2
11.1um
.6um
82.2mV
520mV
10uA
M3
11.925um .9um
141mV
1.905mV 10uA
M4
11.925um .9um
140.6mV 1.905mV 10uA
MA1
2.775um
.6um
135mV
1.744mV 12.27uA
MA2
2.775um
.6um
145
1.77
12.25uA
MA3
11.925um .9um
122mV
755mV
12.27uA
MA4
11.925um .9um
122mV
730mV
12.25uA
Startup_n
0.45um
0.6um 131mV
45.2mV
0.697uA
Startup_p
0.450um
45um
1.553V
2.455V
0.697uA
NMOS_switch 0.450um
.6um
45mV
1.169V
10fA
Iref1,Iref2 Vs Vdd: For entire range of Vdd Matching error between Iref1 and I ref2 is < 5%. It is very small as observed from below plot.
Vref vs Vdd:
Vth Vs Temperature:
Kpn vs Temperature:
Vref vs Temperature: With Zero temperature coefficient resistor
With 5m temp coefficient resistor
Gm vs Temperature Zero temperature coefficient resistor
With 5m temperature coefficient resistor: