Current-Sensing Techniques for DC-DC Converters

Hassan Pooya Forghani-zadeh, Student Member, IEEE, and Gabriel A. Rincón-Mora, Senior Member, IEEE
Georgia Tech Analog Consortium
School of Electrical and Computer Engineering
Georgia Institute of Technology
Atlanta, GA 30332-0250
(gtg054c@mail.gatech.edu, rincon-mora@ece.gatech.edu)

ABSTRACT Vin

Current sensing is one of the most important functions on a smart

+Vsense-
power chip. Conventional current-sensing methods insert a
resistor in the path of the current to be sensed. This method Vout
incurs significant power losses, especially when the current to be
sensed is high. Lossless current-sensing methods address this

Rload
L Rsense
issue by sensing the current without dissipating the power that Cf
passive resistors do. Six available lossless current sensing
techniques are reviewed. A new scheme for increasing the
accuracy of current sensing when the discrete elements are not
known is introduced. The new scheme measures the inductor Figure 1. Series-Sense Resistor.
value during the DC-DC controller startup.
This method obviously incurs a power loss in Rsense, and
therefore reduces the efficiency of the DC-DC converter. For
1.INTRODUCTION accuracy, the voltage across the sense resistor should be roughly
more than 100mV range at full load because of input-inferred
Regardless of the type of feedback control, almost all DC-DC offsets and other practical limitations. If full-load current is 1A,
converters and linear regulators sense the inductor current for 0.1W is dissipated in the sense resistor. For an output voltage of
over-current (over-load) protection. Additionally, the sensed 3.3V, the output power is 3.3W at full-load and hence the sense
current is used in current-mode control DC-DC converters for resistor reduces the system efficiency by 3.3%. In lower output
loop control [1]. Since instantaneous changes in the input voltage voltages, the percentage of power lost in the sense resistor
are immediately reflected in the inductor current, current-mode increases, which degrades efficiency further.
control provides excellent line transient response. Another
application for current sensing in DC-DC converters is also 2.2 RDS Sensing
reported [2,3], where the sensed current is used to determine
Vin
when to switch between continuous-conduction mode (CCM) and
+ Vsense

discontinuous-conduction mode (DCM), which results in an

overall increase of power efficiency in the DC-DC converters.
This paper gives an overview of current sensing techniques in
-

DC-DC converters and their impacts on power losses. Section 2

discusses the traditional way of sensing current, which is not
really lossless, as well as nontraditional lossless techniques. In
section 3, the advantages and disadvantages of the different
current sensing techniques are compared. Section 4 discusses the
Figure 2. MOSFET RDS Current-Sensing.
concerns of sensing current accurately when no information
about the discrete elements is available, which is a reasonable MOSFETs act as resistors when they are “on” and they are
assumption in integrated solutions, and introduces a new scheme biased in the ohmic (non-saturated) region. Assuming small
for surviving in such an environment. drain-source voltages, as is the case for MOSFETs when used as
switches, the equivalent resistance of the device is
2. REVIEW L
RDS = , (1)
WµCox (VGS − VT )
2.1 Series-Sense Resistor where µ is the mobility, Cox is the oxide capacitance per unit area,
and VT is the threshold voltage [4]. Consequently, the switch
current is determined by sensing the voltage across the drain-
This technique is the conventional way of sensing current. It source of the MOSFET, provided that RDS of the MOSFET is
simply inserts a sense resistor in series with the inductor (Fig. 1). known (Fig. 2). The main drawback of this technique is low
If the value of the resistor is known, the current flowing through accuracy. The RDS of the MOSFET is inherently nonlinear.
the inductor is determined by sensing the voltage across it. Additionally, The RDS (for on-chip or discrete MOSFET) has
significant process variation because of µCox and VT, not to
mention how it varies across temperature, which can yield a total
variance of –50% to 100%. The RDS depends on temperature

1
exponentially (35% variation from 27°C to 100°C) [5]. In spite of This current-sensing technique, referring to Fig. 5, uses an RC
low accuracy, this method enjoys commercial use because of its low-pass filter at the junction of the switches of the converter.
power efficiency (no additional resistor is added, lossless Since the average current through resistor R is zero, the output
technique). averaged-current is derived as
Vout − VC
2.3 Filter-Sense the Inductor Io = IL = , (4)
RL
This technique, reported in [5,6], uses a simple low-pass RC where VC is the average capacitor voltage.
network to filter the voltage across the inductor and sense the
current through the equivalent series resistance (ESR) of the V(1) IL
L RL
Vin Vout
inductor (Fig. 3).

-
+

Rload
+ <---VL--->
+Vc-
- R

Vc
C
Vin
Rf Cf
Vout
L RL Controller
RLoad

Figure 3.Filtering the voltage across inductor to sense the Figure 5. Average current-sensing technique.
current. If RL is known, which is not the case for IC designers, the output
The voltage across the inductor is current can be determined. Sensing the current in this method
vL=(RL+sL)IL , (2) depends only on RL , and not on the parasitic switch resistor or the
where L is the inductor and RL is its ESR. The voltage across values of R and C. This scheme is used mainly for load sharing in
capacitor Cf is multiphase DC-DC converters [8].

vc =
vL
= L
(R + sL )I L
1 + sR f C f 1 + sR f C f 2.6 Current transformers
(3)
⎛ 1 + s ( L / RL ) ⎞
⎟ I L = RL ⎜ 1 + sT ⎟ I L ,
⎛ ⎞ The use of current-sensing transformers is common in high
= RL ⎜ ⎜ 1 + sT ⎟
⎜ 1 + sR f C f ⎟ power systems. The idea is to sense a fraction of the high
⎝ ⎠ ⎝ 1 ⎠
inductor current by using the mutual inductor properties of a
where T=L/RL and T1=RfCf. Forcing T = T1 yields vc=RL iL and,
hence, vc would be directly proportional to iL. transformer. The major drawbacks are increased cost and size,
To use this technique, the values of L and RL must be known and non-integrablity. Furthermore, the transformer cannot
(at least their ration), and then R and C are chosen accordingly. transfer the DC portion of the current, which makes this method
This technique is not appropriate for integrated circuits because inappropriate for over-current protection.
of the tolerance of the components required. It is, however, a
proper design for a discrete, custom solution where the type and
value of the inductor is known. 2.7 SENSEFETs

2.4 Sensorless (Observer) Approach This method is a practical technique for current sensing in
many new power MOSFET applications [9,10,11,12]. The idea is
This method is introduced by Midya [7]. It uses the inductor to build a current sensing FET in parallel with the power
voltage to measure the inductor current (Fig. 4). Since the MOSFET (Fig. 6) and use its “measuring” capabilities for
voltage-current relation of the inductor is v=Ldi/dt, the inductor sensing the current.
Vin
current can be calculated by integrating the voltage over time.
The value of L also must be known for this technique.
Integrator
Vin IL W/L=1 W/L=N
1/L
SENSEFET MOSFET
L Vout
Is SENSEFET Im
RLoad

IL
S M

Figure 6. SENSEFET.
Figure 4. Observer current sensing technique.
The effective width (W) of the sense MOSFET (SENSEFET) is
2.5 Average Current significantly smaller than the power FET. The width of the power
MOSFET should be at least 100 times the width of the

2
SENSEFET to guarantee that the consumed power in the Table 1. Comparative overview of current-sensing
SENSEFET is low and quasi-lossless. The voltages of nodes M techniques.
and S should be equal to eliminate the current mirror non-
idealities resulting from channel-length modulation. By using Technique Advantages Disadvantage
current conveyors [13] and other circuits, these effects can be
mitigated. A complete current-sensing circuit using a SENSEFET A. RSENSE Good accuracy High power dissipation
is shown in Figure 7. M1 is the power MOSFET and M3 is the B. RDS Lossless Low accuracy
SENSEFET. The op-amp is used to force the drain voltages of
M1 and M3 to be equal. C. LFilter Lossless Knowledge of L
High number of discrete
elements

M1 D. Observer Lossless Knowledge of L

M3

E. IAverage Lossless Known inductor ESR

Average inductor current
--

M2
only
+

F. Transformer Lossless Cost

Size
+Vsense-

Rsense Not integrable

No IDC information
Not practical
Figure 7. Sample circuit to increase the accuracy of the G. SENSEFET Lossless Special MOSFETs
SENSEFET method. Integrable Matching issues
Practical Low bandwidth
As the width ratio of the main MOSFET and SENSEFET Moderate accuracy
increases, the accuracy of the circuit decreases (matching
accuracy of the FETs degrades). The bandwidth of this technique 4. PROPOSED SCHEME
is reported unexpectedly low [14,15]. The low bandwidth cannot
be estimated by RC pole analysis and is due to another
phenomenon called transformer effect [14,15]. Since the current All of the discussed current-sensing methods, except for the
ratio in SENSEFET circuits is in the order of 100:1, even a low SENSEFET technique, depend on knowing the value of discrete
degree of coupling between the power MOSFET and the elements, such as the inductor, sense resistor, or MOSFET’s RDS.
SENSEFET circuits can induce a significant error, and large In a lossless current-sensing technique, only node voltages are
spikes are injected in the sense signal during periods of high sensed and the value of the current in a branch is estimated using
di/dt. Therefore, proper layout schemes should be chosen to the values of passive elements (i.e., v=Ri, i=Cdv/dt and v=(∫
minimize the mutual inductance between the SENSEFET and the idt)/L).
power MOSFET. In a custom discrete design, the values of external components
are known and the current-sensing technique can be adjusted
before mass production. On the other hand, if a current sensing
3. COMPARITIVE OVERVIEW and scheme is designed for a general-purpose controller, where the
CHALLENGES end-user can select the inductor, capacitor, and switches from a
specified range, the IC designer is incognizant of the values of
Selecting a proper current-sensing method depends on the DC- the external components. Hence, current-sensing techniques are
DC converter control scheme. If voltage-mode control is used, best designed if they are independent of external component
current sensing is only needed for over-load current protection. values. To solve this problem, the circuit shown in Figure 8 is
Therefore, a simple, not very accurate but lossless method, like proposed, where the value of the inductor is measured and stored
RDS current sensing, is sufficient. If current-mode control or mode during startup and the voltages are sensed during normal
hopping (for high efficiency) is used, more accurate algorithms operation to determine the current. Just before startup, the power
may be necessary. The traditional series resistor technique can be MOSFETs are forced off and switches S1 and S2 are “on”.
used when power dissipation is not critical, which is almost never
the case for portable applications. For applications like desktop Vcc
computers, decreasing the power efficiency by about 5% may not Vin Current mirror
as critical. The majority of commercial current-mode controller S1 Iref
solutions for desktop computers use a sense resistor. RDS sensing +
Controller

is the other dominant technique, which is used even in current- -

mode controllers in commercial products, but its accuracy is L
1 R
poor. The other techniques are not appropriate for integrated C
solutions. Table 1 summarizes the advantages and disadvantages
of the different current-sensing techniques explored. S2
Memory A/D

Figure 8. Inductor measuring circuit

3
Constant current source Iref charges capacitor C during that time, 7. REFERENCES
creating a linear voltage ramp, which is turned into a linear
current ramp by the op-amp. [1] W. Kester and B. Erisman, “Switching Regulators”,Analog
The voltage at the positive port of the op-amp is Devices Technical Library on Power Management, 1999.
1 I ref t [2] A. Prodic and D. Maksimovic, “Digital PWM Controller
Vc (t ) = I ref dt =
∫ , (5)
C C and Current Estimator for a Low-Power Switching
where t is time. The op-amp, along with the negative feedback, Converter”, in Proc. The 7th workshop on Computers in
forces the negative port of the op-amp to be equal to the positive Power Electronics, pp.123 –128, 2000.
port; thus, the current forced is a ramp (IL(t)=Vc/R=Ireft/(RC)) . [3] T. Wang, X. Zhou and F Lee, “ A Low Voltage High
Therefore, the voltage across the inductor is Efficiency and High Power Density DC/DC Converter”, in
dI L I ref Proc. 28th Annual IEEE Power Electronics Specialists
VL (t ) = L =L , (6) Conference, pp. 240-245, 1997.
dt RC [4] P. Gray, P. Hurst, S. Lewis and R. Meyer, Analysis and
where Iref is a function of a constant reference voltage and Design of Analog Integrated Circuits, Wiley, New York,
another integrated resistor. The voltage given by (6) is a constant 2001.
voltage inductor across a 1uH inductor. Fig. 9 shows the [5] R. Lenk, “Application Bulletin AB-20 Optimum Current-
simulation results for the voltage across the inductor. Capacitor Sensing Techniques in CPU Converters”, Fairchild
C, resistor R and current source Iref are 1pF, 100Ω, and 10uA, Semiconductor Application Notes, 1999.
respectively. This measurement technique boosts the observer [6] E. Dallago, M. Passoni and G. Sassone, “Lossless
technique accuracy, especially when the value of the inductor is Current Sensing in Low Voltage High Current DC/DC
not known. Modular Supplies”, IEEET Trans. Industrial Electronics,
vol. 47, pp. 1249-1252, Dec. 2000.
5V
[7] P. Midya, M Greuel and P. Krein, “Sensorless Current Mode
Control-An Observer Technique for DC- DC Converters”,
IEEE Trans. Power Electronics, vol. 16, pp. 522 –526, July
0V
0s 200ns 400ns
2001.
(a) [8] X. Zho and P. Xu, “A Novel Current-Sharing Control
Technique for Low-Voltage High-Current Voltage
Regulator Module Applications”, IEEE Trans. Power
100mV
Electronics, vol. 15, pp. 1153-1162, Nov.2000.
[9] W. Schultz, “Lossless Current Sensing with SENSEFETs
Enhances the Motor Drive”, Motorola technical report,
0V
1988.
0s 200ns 400ns [10] S.Yuvarajan, “Performance Analysis and Signal
(b)
()
Processing in a Current Sensing MOSFET
(SENSEFET)”, in Proc. Industry Applications Society
Annual Meeting, vol. 2, pp. 1445 – 1450, 1990.
Figure 9. Simulation of transient voltage across (a.) Capacitor [11] P. Givelin, M. Bafleur, “On-Chip Over Current and Open
C and (b.) Inductor L. Load Detection for a Power MOS High Side Switch: A
CMOS Current Source Approach”, in Proc. Fifth European
5. CONCLUSION Conference on Power Electronics and Applications, vol. 2,
pp.197-200, 1993.
Six lossless current-sensing techniques were reviewed. Among [12] S. Yuvarajan and L. Wang, “Power Conversion and
these techniques, only the SENSEFET method is independent of Control using a Current Sensing MOSFET”, in Proc. 34th
external component values. A novel approach for sensing current, Midwest Symposium on Circuits and Systems, vol. 1,
when values of external components are not known, is proposed pp.166–169, 1992.
and simulation results were presented. The new approach enjoys [13] A. Sedra, G. Roberts and F. Gohh, “The Current Conveyer:
the benefits of being integratable, accurate, and lossless (low History, Progress and New Results”, IEEE International
power loss during startup). Symposium on Circuits and Systems, pp. 78-87, 1989.
[14] D. Grant and R. Pearce, “Dynamic Performance of Current-
6. ACKNOWLEDGEMENTS Sensing Power MOSFETs”, Electronic Letters, pp. 1129-
1131, Sept. 1988.
[15] D. Grant and R. Williams, “Current Sensing MOSFETs
This research was funded by Intersil Corporation through the for Protection and Control”, IEE Colloquium on
Georgia Tech Analog Consortium (GTAC) project. The authors Measurement Techniques for Power Electronics, pp. 8/1-
thank Mr. Todd Harrison from Intersil Corporation for his 8/5, 1992.
valuable suggestions.