Micronote 1302 - (AN-7) A Simple Current-Sense Technique Eliminating a Sense Resistor (154 kB)

LINFINITY Application Note
AN-7
A Simple Current-Sense
Technique Eliminating a
Sense Resistor
Copyright © 1998
Rev 1.1 07/98
AN-7
Loss-Less Current Sense Technique
A SIMPLE CURRENT-SENSE TECHNIQUE ELIMINATING A SENSE RESISTOR
INTRODUCTION
A sense resistor RS, as shown in Figure 1, is often used for
the purpose of over-current protection (OCP). RS is usually
a power device because it needs to handle the large current flowing through it. Therefore, it can be costly, bulky,
and inefficient. Even though the RDS(ON) of the MOSFET or
monitoring the output voltage are also used in OCP, those
two techniques have very poor accuracy and are good only
for output short-circuit protection. They can not protect the
power supply when there is a weak short circuit at the
output or the output is over current.
This application note introduces a simple current-sense
technique that eliminates that sense resistor, resulting in
system-cost reduction, PCB space saving, and powerefficiency improvement. Furthermore, the new currentsensing mechanism allows higher dynamic tripping current
than the static one (built-in low-pass filtering) to improve
current-sense noise immunity.
Q
L
RL
with the inductor. The voltage VCs across CS is the sensor's
output.
The operation of the new sensor can be understood by
examining the inductor current and the capacitor CS voltage. When the buck converter is operating, the voltage on
the left side of the inductor is a chopped voltage while the
one on the right side is constant. The equivalent voltage
across the inductor and the RC sensor is a square wave, as
shown in Figure 3 (a) and (c).
(b)
(a)
v
Rs
(c)
(d)
D
Figure 3 – The equivalent circuits and waveforms for the
inductor and the sensing circuitry.
Figure 1 – A buck converter with a resistor Rs for sensing the
inductor current. The resistor RL is the parasitic resistance of the
inductor.
When L/RL is much greater than the switching period TS,
i.e.
L / RL >> Ts ,
THE NEW SENSING TECHNIQUE
This new sense technique utilizes the inductor parasitic
resistance, RL, to sense the inductor current. Figure 2
shows the new sensor circuitry with the original sensing
resistor RS eliminated. The new sensor consists of a resistor R and a small ceramic capacitor CS, in parallel directly
+ V_Cs R
Q
the inductor current is a triangular wave, as shown in Figure 3 (b). If values are selected such that
R ⋅ Cs = L / RL ,
(2)
one can find that the capacitor voltage is directly proportional to the inductor current, as shown in Figure 3 (d). In
fact,
Cs
L
(1)
V _ Cs = i LRL
(3)
RL
D
where iL is the inductor current. Therefore, one can use the
capacitor voltage for OCP.
C
Figure 2 - The new current sense technique
Copyright © 1998
Rev 1.1 07/98
LinFinity Application Note
Page 2
AN-7
Loss-Less Current Sense Technique
sL + R L
sL / R L + 1 .
V _ Cs (s )
=
= RL
i L (s )
sRC S + 1
sRC S + 1
(4)
PERFORMANCE ANALYSIS
T (s ) =
Equation (3) is valid under the condition that equation (1)
and (2) are valid. Therefore, the performance of the new
current-sense technique relies on whether or not (1) and
(2) are valid and what the impact is if they are not valid.
If equation (2) holds, (4) can be simplified as RL, as shown
in Figure 4 (a); otherwise, the frequency-dependent gain
has the asymptotes as shown in Figure 4 (b) and (c).
Equation (1) is generally valid. The switching period TS is
usually of the order of microseconds, given switching frequency of the order of a few hundred kHz. The ratio of the
inductance and the parasitic resistance is typically in the
order of a millisecond. For example, a 5µH, 15A inductor
has typically 5mΩ parasitic resistance, therefore, L/RL =
1ms.
If the gain is the case shown in Figure 4(b), OCP may be
tripped at a current level less than the desired value. An
example of this case is shown in Figure 5. The current rise
from 0A to 18A, but the sensor will output a signal indicating higher than 20A. If the current threshold is set at
20A, OCP will be tripped.
Equation (2) usually is very difficult to satisfy due to the
tolerances of all the variables. The tolerance of R can be
1%, 5%, or even higher. Cs has its initial tolerance and
dependency on the temperature. The inductance, L, has
initial tolerance as well as dependency on the dc biasing
current, which makes the inductance vary over a large
range.
Impact Of Parameter Uncertainty?
The impact depends on how the Cs voltage reacts to the
inductor current. Analysis shows that the response of the
Cs voltage to the inductor current is frequency dependent.
The current sensor gain is
Figure 5 - The response of the current sensor to an 18A step current in the inductor when L/RL > RCs
|T(jω)|
If the gain is similar to the case of Figure 4 (c), which has
a lower high-frequency gain than the low-frequency gain,
the response is in the opposite way. The response of the
sensor to a 25A step current is shown in Figure 6 (a). The
sensor exceeds the 20A threshold after 0.674ms delay.
RL
(a)
ω
|T(jω)|
L/RCs
RL
(b)
ω
RL/L
1/RCs
|T(jω)|
RL
(c)
L/RCs
ω
1/RCs
RL/L
Figure 4 - The gain of the new current sensor.
(a) L/RL = RCs; (b) L/RL > RCs; (c) L/RL < RCs
Copyright © 1998
Rev 1.1 07/98
Figure 6(a) - Response to a 25A step current when
L/RL < RCs
LinFinity Application Note
Page 3
AN-7
Loss-Less Current Sense Technique
This delay can, in fact, improve the noise immunity of the
sensor and, furthermore, avoid unnecessary OCP tripping
for a very short-duration over-current situation. An example is shown in Figure 6 (b), where the current exceeds the
20A threshold for a very short period. The output of the
sensor, because of the delay, does not. Therefore, dissatisfaction of equation (2) does not hurt OCP performance.
On the contrary, OCP behaves more desirably when the
sensor gain is in the case shown in Figure 4 (c).
as one might think. Because when the inductor temperature is low, so is the ambient, hence, the power MOSFETs
in the power supply are allowed to operate at higher current without damage.
This new current sense technique is more accurate than
1
1.0
0.95
R( T )
0.9
Rup ( T )
Rlow ( T ) 0.85
0.8
0.75 0.75
20
40
60
T
Figure 7(a) 20
- RL versus temperature
1.4
80
80
1.4
1.3
1
R( T )
1.2
1
Rup( T )
1.1
1
Figure 6(b) - Response to short duration transient
Rlow( T )
1
The important parameter in the new current sensor is the
accuracy of RL. When the inductor current changes from
one level to another one, the sensor output will change to
the new state, iLRL, within about 3 times of the time constant RCs. The OCP tripping current is
ITRIP = VTRIP/RL,
(5)
where VTRIP is the tripping voltage of the current comparator. Therefore, the tripping current is directly related
to RL.
When using large gauge magnetic wire, RL is can be controlled within reason. The diameter of the wire has a 1%
tolerance, resulting in only 2% tolerance in the resistance
per unit length. The length of the wire is in the order of
10cm so its error is negligible.
The copper wire resistance has a 0.39%/°C temperature
coefficient. The positive temperature coefficient results in a
higher resistance (lower OCP threshold) at higher temperature (which helps prevent thermal run away).
Figure 7 (a) shows the resistance variation from 20°C to
80°C, normalized at 80°C. The two dashed lines indicates
the 2% tolerance. Figure 7(b) shows the OCP threshold vs.
temperature, also normalized at 80°C. The temperature
variation is due to the ambient and the temperature rise of
the inductor in operation. The OCP threshold is 1.3 times
higher at 20°C than at 80°C. In practice, this is not as bad
Copyright © 1998
Rev 1.1 07/98
0.9 0.9
20
20
40
60
T
Figure 7(b) - Itrip versus temperature
80
80
using PCB trace. Using PCB trace has a poor tolerance due
to the copper foil thickness. For example, the 1-oz copper
PCB’s thickness is 34 ± 5µm (0.0014 inch ± 0.0002 inch).
The percentage error is 14%! The errors due to width and
length of the trace are negligible compared to the thickness error. Moreover, the current density in the copper is
usually higher than that in the inductor, hence the temperature rise is higher.
USING THE NEW CURRENT SENSE
TECHNIQUE
When using the new current sense technique, one should
consider the possible range of the parameters and make
sure the sensor gain is in the case shown in Figure 4 (a) or
(c). For example, assume R has 5% accuracy, L has a
value with possible maximum 2.5µH at low current and a
minimum1.1µH at high current, Cs has 10% tolerance, and
RL is 3mΩ at 20°C. The minimum dc sensor gain is 3mΩ,
and then the maximum ac gain should be selected to be
below the dc gain, i.e.
LinFinity Application Note
Page 4
AN-7
Loss-Less Current Sense Technique
Lmax/RminCsmin = 3 mΩ.
(6)
Vtrip
OCP
Therefore,
CMP
RminCsmin = 3mΩ/2.5µH = 1.2 ms.
(7)
Considering the tolerances of R and Cs, select RCs 15%
larger than 1.2ms, i.e. 1.4ms. One can select R = 3kΩ and
Cs = 0.47µF.
R=R1//R2
Q
D
L
Cs
RL*R2/(R1+R2)
In most case, RL is larger than the desired value. The desired value can be found from
Rs = VTRIP/ITRIP,
(8)
where VTRIP is the tripping voltage in the current comparator, as shown in Figure 8, and ITRIP is the tripping current.
The typical VTRIP of the LX166xA and LX1668 devices is
60mV. If the parasitic resistance is larger than the desired
value, one can put a resistor in parallel with Cs, as shown
in Figure 8 (a) that is equivalent to Figure 8 (b). Notice
that the equivalent RL changes to RL*R2/(R1+R2) and the
equivalent R is R1//R2, compared to those parameters in
Figure 2.
C3
Figure 8(b) - Equivalent Circuit
The first test was on a 2.5µH inductor. The inductor used a
Micrometals’ material 52 iron powder core and 8 turns of
AWG 18 wire. The measured results are shown in Figure 9.
The slope of the curve indicates the resistance. At low
current, the resistance is estimated to be 3mΩ. Notice that
the slope increases slightly at higher currents, due to the
higher temperatures.
TEST VERIFICATION
Two tests were carried out to verify the new current sense
technique. The first one was to measure the dc Cs voltage
when a dc current is flowing in the inductor. The second
was a short-circuit load test of a Pentium® II processor
power supply using LX1664 control IC.
Cs VOLTAGE (mV)
60
50
40
30
20
10
0
0
2
4
6
8
10
12
14
16
CURRENT (A)
Vtrip
OCP
CMP
Figure 9 - The measured Cs voltage versus inductor current
R2
R1
Q
D
L
Cs
RL
C3
Figure 8(a) - Complete Sense Circuit
Figure 10 - OCP Action. The top trace is output voltage (1V/div)
and the bottom is inductor current (10A/div)
Copyright © 1998
Rev 1.1 07/98
LinFinity Application Note
Page 5
AN-7
Loss-Less Current Sense Technique
The second test replaced the sense resistor in the LX1664
evaluation board with the new current sense. Since VTRIP =
100 mV in LX1664, we used a 5µH inductor that has an RL
of 5mΩ. The R and Cs were 10kΩ and 0.1µF ceramic capacitor respectively. The output was short-circuited when
the circuit was running and the transient waveforms are
shown in Figure 10. Notice that the current limit was at a
higher level at the beginning of the transient but drops to
20A after about 0.5ms because the sensor gain was in the
case of Figure 4 (c).
The new sensor does not affect the voltage-positioning
feature of LX166x family ICs. Figure 11 shows the output
voltage and a 15A step-current load. The voltage positioning is clearly shown.
DESIGN PROCEDURES FOR USING THE
NEW SENSOR
1.
2.
3.
4.
5.
Calculate the desired sense resistance RS with equation (8).
Select an inductor with RL ≥ RS.
Find the maximum value of Lmax/RL,min possible. Usually the maximum L is at zero current and minimum RL
is at room temperature. Take the tolerance of L and RL
into account as well.
Decide R and Cs so that the time constant Rmin*Cs,min
> Lmax/RL,min.
If RL > Rs, Select R1 and R2 so that R2/(R1+R2) =
Rs/RL and R1//R2 = R.
Figure 11 - Step load response. Top trace is output voltage
(100mV/div) and bottom is load current (5A/div)
Copyright © 1998
Rev 1.1 07/98
LinFinity Application Note
Page 6