The principle:
VBE
VREF = KVt + VBE
Σ
KVt
kT Vt
Vt = K
q
Generate an inverse PTAT (Proportional To Absolute Temperature) and a PTAT and sum
them appropriately.
• Vt (Vt = kT/q) is PTAT that has a temperature coefficient of +0.085 mV/°C at room
temperature.
VBE+VPTAT
VBG
VBE
VPTAT
0K 300 K
Temperature
VBE  VG0
JC = ATγ exp (2a)
Vt
γ kT T0 kT JC
VBE = VGO 1 − + VBE0 +
T T
q ln T + q ln JC0 (2b)
T 0 T0
By taking the derivative of Eq. (2) at T0 with respect to temperature, (assuming that JC
has a temperature dependence of Tα ), the dependence of VBE on temperature is clearly
seen to be
At 300 °K the change of VBE with respect to temperature is approximately −2.2 mV/°C.
Now, it is also necessary to develop the relationship for ∆VBE (PTAT) for two bipolar
transistors having different current densities. A relationship for ∆VBE can be given as
Allen/Holberg : Chapter 10 : 4/12/00 3
∆VBE = q ln JC1
kT J
(4)
C2
Therefore
∂∆VBE
= Tt ln JC1
V J
(5)
∂T C2
In order to achieve zero temperature coefficient at T0, the variations of VBE and ∆VBE as
given in Eqs. (3) and (5) must add up to zero.
V −V (α − γ)Vt0
0 = K Tt0 + BE0T GO +
V
(7)
0 0 T0
V − VBE0 + (γ − α )Vt0
K = GO (8)
V t0
The term K (K = K'' ln [JC1/JC2]) is under the designer's control, so that it can be
designed to achieve zero temperature coefficient. Rearranging Eq. (8) yields
Noting that K in Eq. (9) is the same as that in Eq. (1), as both are constants required to
achieve a zero temperature coefficient, then substituting of Eq. (9) into Eq. (1) gives
For typical values of γ = 3.2 and α = 1, VREF = 1.262 at 300 °K. A typical family of
referencevoltage variations as a function of T for various values of T0 is shown below.
Allen/Holberg : Chapter 10 : 4/12/00 4
∂VREF ∂VREF
=0 =0
∂T ∂T
1.290
T0=400 K
VREF 1.280
T0=300 K
1.260
∂VREF 1.250
=0
∂T T0=200 K
1.240
Temperature (ºC)
R3 R2
VOS

+ +
VR1 R1

Q2 Q1 VREF
Transistors Q1 and Q2 are assumed to have emitterbase areas of AE1 and AE2,
respectively.
K = 2 ln 2 E1
R R A
(15)
R1 R3AE2
Thus, the constant K is defined in terms of resistor and emitterbase area ratios. It can be
shown that if the inputoffset voltage is not zero, that Eq. (14) becomes