Transcript ON-CHIP CURRENT SENSING TECHNIQUE FOR CMOS …

ON-CHIP CURRENT SENSING TECHNIQUE
FOR CMOS MONOLITHIC SWITCH-MODE
POWER CONVERTERS
Problems In Circuits Design
Ronen Gabai
June-2006
Presentation Outlines
 The Need for Current Sensing
 Current Sensing Techniques – Overview
 The Chosen Technique – SENSEFET




Article Overview
Measurements Results
Circuit Efficiency and Performance
Conclusions & Summary
2
The Need for Current-Sensing Techniques
 In Switch-mode power converters (SMPC), current-mode
pulsewidth-modulation (PWM) control and current-limited
pulse-frequency-modulation (PFM) control schemes are widely
used in industries due to their fast dynamic response and
automatic over-current protection.
 Both control make use of the Inductor current (of the
Buck/Boost power stage) to modify the pulse width in PWM or
oscillation frequency in PFM for voltage regulation.
 The Inductor current is particularly important for PWM, as the
signal sensed from the inductor current is mandatory to combine
with the artificial ramp signal in order to avoid sub harmonic
oscillation in current-mode control PWM converter.
3
PWM Converters with Switch Model Inserted
 (a) Buck Converter 
 (b) Boost Converter 
 (c) Flyback Converter 
4
Current-Sensing Techniques for DC-DC Converters
 Current sensing is one of the most important functions on a smart





power chip.
Regardless of the type of feedback control, almost all DC-DC
converters and linear regulators sense the inductor current for
over-current (over-load) protection.
Additionally, the sensed current is used in current-mode control
DC-DC converters for loop control.
Conventional current sensing methods insert a resistor in the path of
the current to be sensed; This method incurs significant power losses,
especially when the current to be sensed is high.
Lossless current-sensing methods address this issue by sensing the
current without dissipating the power that passive resistors do.
We’ll now review Six available lossless current sensing techniques.5
1. Series Sense Resistor
 This technique is the conventional way of






sensing current.
It simply inserts a Sense Resistor in series with
the inductor.
If the value of the resistor is known, the current
flowing through the inductor is determined by sensing the voltage across it.
This method obviously incurs a power loss in Rsense, and therefore reduces
the efficiency of the DC-DC converter.
For accuracy, the voltage across the sense resistor should be roughly 100mV at
full load because of input-inferred offsets and other practical limitations.
If full-load current is 1A, 0.1W is dissipated in the sense resistor.
Main Disadvantage: For an output voltage of 3.3V, the output power
is 3.3W at full-load and hence the Sense Resistor reduces the system
6
efficiency by 3.3%.
2. MOSFET RDS Current Sensing
 MOSFETs act as resistors when they
are “ON” and they are biased in the ohmic
(non-saturated) region.
 Assuming small VDS, as is the case for
MOSFETs when used as switches:
I D    COX 
W
 (VGS  VT )  VDS
L
;
for : VDS  (VGS  VT )
 The equivalent resistance of the device is:
RDS 
L
W    COX  (VGS  VT )
 RDS in this case should be known ; VDS is measured.
 Main Disadvantage: The RDS of the MOSFET is inherently
nonlinear: It usually has significant variation because of:   COX ,
VT, and exponential variations across temperature
(35% variation from 27°C to 100°C).
7
3. Filter-Sense the Inductor
 A simple low-pass RC network filter the voltage across the
Inductor and sense the current through the equivalent series
resistance (ESR) of the inductor.
vL  ( RL  sL) I L
 The Voltage across the Inductor is:
 The Voltage across the feedback Capacitor is:
 1  s L / RL  
vL
( RL  sL ) I L
 I L  RL 1  sT I L
vc 

 RL 
 1  sR C 
1  sR f C f
1  sR f C f
1  sT1
f
f 

 Forcing:
T  T1    vc  RLiL    Vc  iL
 Main Disadvantage:
L and RL need to be known 
Not appropriate for IC, but it is a proper
design for a discrete/custom solution
8
4. Sensorless (Observer) Approach
 This method uses the Inductor voltage to measure the Inductor
current.
di
 Since the Voltage-Current relation of the Inductor is: v L  L , the
dt
Inductor current can be calculated by integrating the voltage over
time.
 Main Disadvantage: As at the previous method, The value of L
also should be known for this technique.
9
5. Average Current
 This technique uses RC LPF at the junction of the switches of the
converter.
 Since the average current through the resistor is zero, the Output
averaged–current is: I o  I L  Vout  Vc
RL
 If RL is known (not the case of IC designers), the output Average
current can be determined.
 Main Disadvantage: Only average current can be measured.
10
6. Current Transformers
 The use of this technique is common in high power systems.
 The idea is to sense a fraction of the high Inductor current by
using the mutual inductor properties of a transformer.
 Main Disadvantages:
 Increased cost and size and non-integrablity.
 The transformer also cannot transfer the DC portion of
current, which make this method inappropriate for
over current protection.
11
7. SENSEFETs
 This method is the practical technique for Current sensing in new
power MOSFET applications.
 The idea is to build a current sensing FET in parallel with the
power MOSFET (Current Mirror).
 The effective width (W) of the sense MOSFET (SENSEFET) is
significantly smaller than the power FET, and therefore linearly
reduces ID per the MOSFET known equation: I    C  W  (V  V ) V
D
OX
L
GS
T
The voltage of nodes M and S
should be equal to eliminate the
Current-Mirror non-ideality
resulting from channel length modulation
 The width (W) of the Power MOSFET is X100-1000 times the
12
width of SENSEFET to guarantee it’s low power consumption.
DS
Complete Current Sensing SENSEFETs circuit
 The Op amplifier is used to force VDS
of M1 and M3 to be equal.
 As the width of the main MOSFET and
SENSEFET increases, the accuracy of
the circuit decreases.
 Main Disadvantage:


Relatively low Bandwidth
Proper layout scheme should be designed to minimize coupling between
the transistor (which can induce significant error).
 Advantages:




Lossless
Integrable
Practical
Relatively good accuracy
13
Comparative overview of Current-Sensing Techniques
 For Voltage mode control, which current sensing is only needed
for over-load current protection, RDS method can be used.
 In current-mode control for
desktops (no power
dissipation constraint), we
can use RSENSE technique.
 RDS and SENSEFET are
the dominant techniques
were power consumption is
critical (portable applications);
SENSEFET is much more
accurate.
14
The Current-mode Buck Converter
= Vg*DON
 The next few slides
focus on few design issues
of the Current–Mode
Buck converter:
 Pole-Zero
cancellation.
 On-Chip Inductor
current sensing.
 Subharmonic
oscillations.
 Pulsewidth Generator
4
1
3
2
 The output of the Compensator, Compensation ramp, and Sensed Inductor
current pass through the modulator and the digital control block to define15d(t).
1. Pole-Zero Cancellation Compensator
 The Power stage of Current-mode



Converters has Control-to-output
Transfer function of two separated
real poles.
The Pole from the output filtering
capacitor is heavily dependent
on the equivalent resistance of the output load – RL.
For dynamic response consideration, Pole-Zero cancellation is
preferable as the bandwidth can be extended with Pole-Zero
cancellation  Speed up the response time.
v
1  sC R
The Transfer Function of the Compensator is: A(s)  bv  g R 1  sC R ; R  R
R0 is the Output resistance of the Operational Transconductance Amp
a
m
0

c
z
c
0
0
0
16
z
Calculation of the two frequency
compensation components – RZ , CC
 The general purpose of introducing Zeros and Poles in the
compensator is to cancel Poles and Zeros in the control-to-output
function, respectively.
 This will yield an average -20 dB/decade closed-loop gain
response with sufficient phase margin below the unity gain freq.
 When determining the unity gain frequency, it should not be too
close to the converter’s switching frequency as the amplifier
would amplify the output ripple voltage; A safe value of unity
gain frequency is below 20% of the switching frequency.
 Since the dominant pole shifts inversely proportional to the load
resistance, the lowest frequency occurs at the highest load
resistance, two frequency compensation components - RZ and CC
can be calculated using the corresponding Transfer Function.
17
2. Subharmonic oscillation are eliminated
by Compensation Ramp - mc
 Subharmonic oscillation is a well-known problem for current-
mode switching converters with the duty ratio – D > 0.5.
 To avoid Subharmonic oscillation, the slope of the compensation
ramp must be larger than half of the slope of inductor current
during the second subinterval D'T.
 The Compensation ramp (Ramp signal) and the Inductor current
signal (Sensed signal) are summed together.
18
3. On-Chip Current Sensing Circuit
PMOS
 Op Amp enforces: VA=VB
 L & C are off-chip
 I1,I2 are small and equal,
pull current from VA,VB
 'ON' state: M1 – ON ,
IL is mirrored to M2
 VDS and current density of
M1, M2 are almost the same.
 Due to different Aspect ratios
of M1, M2 (WM2:WM1 = 1:1000)  IS is much smaller and proportional to IL:
VSENSE  I SENSE RSENSE 
IL
RSENSE
1000
19
On-Chip Current Sensing Circuit - Continue
 I2 (Small Biasing current) << IS
PMOS
ISENSE  IS  I L
 For Current mode DC-DC
Converters, only VSENSE is
needed in the control feedback
loop during the ON-State
(ramp up of the inductor current)
 MS2 tie VA to Vg during the
OFF-State  ISENSE~0.
 The output of the Op Amp should be able to go up to Vg in order to make the
Transistors – Mrs & MCS5 operate in Saturation region.
20
Current Sensing - Design Issues
 The accuracy of the sensed Inductor current depends on the




Current Mirror of transistors M1 and M2 and in the on-chip
Poly Resistor - RSENSE.
The matching of transistors M1 and M2 depends on the process
parameters such as Mobility, Oxide Capacitance (COX) and
Threshold voltage (VT).
Therefore, proper layout technique should be well considered,
especially the location of the transistor M2, to minimize error.
In the suggested design, M2 is surrounded by 500 fingers of M1.
Of course, This on-chip current-sensing circuit can be extended
to sense power NMOS transistor by simply building a
complement circuit for other topologies (as Boost, Buck-Boost).
21
Pulsewidth Generator
 This Implementation deals with the Startup situation in which
both inputs are high  In this situation, the Latch is SET.
22
Measurements Results - 1
 The Converter is supplied with an Input voltage of 3.6V,
and Switching frequency of 500KHz.
 Attached Steady-state measurements with:
Maximum Loading current = 300mA ; RSENSE = 400 Ohm
Output voltage = 2.1V and Duty Ratio > 0.5 (Subharmonic oscillation zone)
Inductor Current
Sensing Voltage
Inductor Current
Inductor Voltage (Vx)
23
Measurements Results - 1
DC Output Voltage = 2.12V
Output Ripple Voltage = 6.4mV
24
Measurements Results - 2
Output voltage = 1.4V and Duty Ratio < 0.5
Inductor Current
Sensing Voltage
Inductor Current
Inductor Voltage (Vx)
25
Measurements Results - 2
DC Output Voltage = 1.4V
Output Ripple Voltage = 3.17mV
26
Circuit Efficiency and Performance
 The Current Sensing circuit performs accurately and the absolute




error between the sensing signal and the scaled inductor current
is less than 4% (10mA with load current of 300 mA);
This absolute error is mainly due to
the mismatch of transistors M1 and M2
in the sensing circuit.
The efficiency is shown with the
Conduction Loss
Input voltage of 3.6 V and the
Output voltage of 2.0 V.
The maximum efficiency is 89.5%
at loading current 300 mA.
There are two major power dissipations:
27
Conduction loss and switching loss
Conclusions & Summary
 Experimental results show that the converter regulates properly




with duty ratio – D larger and smaller than 0.5.
Using the internal Current-Sensing technique, it not only reduces
the external pins for the monolithic controller, but also reduces
the complexity of the design.
Due to the accurate Sensing performance, a compensation ramp can be added
to the sensing signal without any consideration
on the variation of the sensing performance.
The accurately sensed Inductor current can
also be used for over-current protection and
Load-dependent mode-hopping schemes for
optimizing power efficiency.
In addition, this current-mode DC–DC buck
converter with internal current sensor can
operate from 300 kHz to 1MHz with the input voltage range from 3 to 5.2 V,
which is suitable for lithium-ion battery supply applications.
28
References
 A Monolithic Current-Mode CMOS DC–DC Converter With
On-Chip Current-Sensing Technique.
Cheung Fai Lee and Philip K. T. Mok, Senior Member, IEEE,
IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 39, NO.
1, JANUARY 2004.
 On-Chip Current Sensing Technique for CMOS Monolithic
Switch-Mode Power Converters.
Cheung Fai Lee and Philip K. T. Mok,
In IEEE Int. Symp. Circuits and Systems, vol. 5, Scottsdale, AZ,
May 2002, pp. 265–268.
 Current-Sensing Techniques for DC-DC Converters.
Hassan Pooya Forghani-zadeh, Student member, IEEE, and
Gabriel A. Rincón-Mora, Senior member, IEEE,
29
Georgia Tech Analog Consortium.
30