Power Supply Design Seminar

Current Sensing Solutions for

Power Supply Designers
Topic Categories:
Basic Switching Technology
Design Reviews – Functional Circuit Blocks

Reproduced from
1997 Unitrode Power Supply Design Seminar
SEM1200, Topic 1
TI Literature Number: SLUP114

© 1997 Unitrode Corporation

© 2011 Texas Instruments Incorporated

Power Seminar topics and online power-

training modules are available at:
[Link]/seminars
 Current Sensing Solutions for Power Supply Designers
By Bob Mammano

sensing applications, there are many issues facing the

ABSTRACT designer. Among these include:
While often considered a minor overhead func-
tion, measuring and controlling the currents in a • The need for AC or DC current information
power supply can easily become a major contributor • Average, peak, RMS, or total waveform of the
to the success or failure of a design. To aid in current
achieving optimum solutions for this task, this topic • Isolation requirements
will review the many strategies for current sensing, • Acceptable power losses in the measurement
describe the options for appropriate sensing devices, process
and illustrate their application with practical design • Accuracy, stability, and robustness
techniques. The distinction between current control • Bandwidth and transient response
and fault protection will be explored and applied to • Mechanical considerations
various power supply topologies with emphasis on • Implementation cost
performance-defining characteristics.
Since there has been no published report thus far
CURRENT SENSING IN POWER
of any single current monitoring approach which is
SUPPLIES
clearly superior in all the aspects given above, an im-
While it would be easy to classify the design of
portant task in any new project is to prioritize the re-
the typical power supply as a voltage regulation
quirements as each options will have its strengths and
problem, experienced designers recognize that very
weaknesses in satisfying this list of issues. With
high up on any priority list of tasks are issues related
these characteristics in mind, let us first look at some
to the measurement and control of current. And, as it
of the possibilities in measuring current.
turns out, these are usually where more of the diffi-
culties lie. At the risk of being too philosophical, let SENSING AND MEASURING
me suggest that the measurement of voltage is a CURRENT
"passive" activity in that it can readily be done at al-
RESISTIVE SENSING
most any place in the system without affecting per-
formance. Current measurements, however, are Discrete Resistors for Current Sensing
"intrusive" in that we have to "insert" some type of In almost all cases, current sensing means devel-
sensor and, in so doing, run a greater risk of affect- oping a voltage signal which is representative of the
ing the system we are trying to monitor or control. current flowing at the particular place of interest in
Which is, of course, justification for this topic in the circuit. Thus, a current sensing resistor should be
which we will attempt to present a compendium of considered as a current-ta-voltage converter which, in
current sensing devices, techniques and applications order to perform its function, must be selected on the
as an aid to designers of a broad range of power sys- basis of a set of attributes such as the following:
tems. • Low value in order to minimize power losses
There are many reasons for monitoring current in (Rcs<1 n for this discussion)
power supplies, and as many or more uses to make of • Low inductance because of high dildt
the information once obtained. At least three of these • Tight tolerance on initial value and low tem-
which will be discussed herein, pertain to the use of perature coefficient for accuracy
current sensing for fault protection, for current-mode • High peak power rating to handle short duration
high current pulses
controlled voltage regulation, and for commanding
• High temperature rating for reliability
current as a goal in itself. In all of these current

Texas Instruments 1 SLUP114

 Low value resistors in today's market consist of ~ - The first criteria in determining a sense re-
wirewound, thick film, thin film, metal element, and sistor's value is often the voltage threshold of the
perhaps even other technologies. They come in a following circuitry which is going to operate on the
wide variety of form factors including axial and ver- sensed current information. The simpler Ie control-
tical through-hole mounted, power and chip surface lers, for example, will have a defined voltage range
mounted devices, two terminal and four terminal for current sensing, or a specific threshold for initi-
(Kelvin), on metallic or alumina substrates, and some ating over-current protection, and the value of the
with integral heat sinks. Representative samples of sensing resistor is then uniquely defined by the value
this variety are pictured in Figure 1. The following of the current to be measured. In more sophisticated
discussion is offered as an aid to understanding some applications where some signal amplification is
of the considerations in selecting and applying these available, the choice would be made on the basis of
devices to perform the function of current sensing. minimizing the voltage drop across the sense resistor.

Caddock Electronics ' Type MV lsotek 's PMA, PMD Surface Mount Caddock Electronics ' Type SR10 Cur-
Power Film Resistor. Design uses four terminal package, 10mm x 6mm rent Sense Resistor. Kelvin connection
interdigitated non-inductive termina- x 1.5mm. Resistance down to 3mI2, vertical mount 1W non-inductive de-
tions. Power ratings from 1.5W to with <30ppmi'C TCR sign. O.008W to 1.0W, 1% standard
lOW tolerance.

Isotek's RTO Heat Sink Base Mount lRC's OARS series Surface Mount
two and four terminal configurations. sense resistor. As low as 5mD stan-
Uses TO-238 package and screw ter- dard, rated at 1W with <30ppmi'C
minal connections. Rated to 50W with TCR. Mounting configuration mini-
<50ppm TCR and <2nH inductance mizes TCR of copper leads

Figure I A sampling of commercially available current sensing resistor products.

Texas Instruments 2 SLUP114

This process may be driven by the voltage itself - as connect to it. Note that these derating curves are
in low headroom regulators, for example - or by unique to the style of resistor, its value, its material,
power loss considerations stemming from either and its mounting configuration.
thermal requirements or efficiency goals. In either
case, minimizing the voltage drop requires a trade-off
150
study as there is almost always an accuracy or cost Cl
<C
penalty incurred. For example, amplifier or com- 0
...J
parator input offset voltage variation represents a Cl 100
w
fixed range of ambiguity which becomes an ever
more significant variable as the sense voltage is re-
I-
<C
a::
I- 120·C
"""~1 17S·C -
SOLDER

~F\W-
duced. Other errors which might not be noticed with z 50
w CONTINUOUS
higher voltage drops include losses in contacting the 0
a:: RATING OF ;
FR · 4CB _ ,
~
sensing resistor and, potentially, even the effects of w
a.
o
MA TERIALS I !
thermal EMF generation at the connections.
o 50 100 150 200
Another important consideration relative to re-
AMBIENT TEMPERATURE (·C)
sistor value is the variability of resistor materials
chosen by the various manufacturers of these devices. Figure 2 - A typical power derating curve for a current
These low values of resistance demand specialized sense resistor
materials and the manufacturers have responded with
a wide range of proprietary formulations, all with Peak or pulsed power capability is a function of
variations in their characteristics. Even within a par- energy (watt-seconds) because it is energy which cre-
ticular part type from a particular manufacturer, dif- ates heat, not just power alone. Peak power can
ferent resistor values may be made with different therefore be hundreds or thousands of times higher
materials with resultant differences in performance. than steady state power as long as the pulse duration
This is an important point and will be illustrated fur- is limited. An example of this is shown in the curves
ther in the discussion which follows. of Figure 3 which describe a one-watt surface mount
device from IRe which can have a pulse power rating
Power Ratin~ - This parameter is often the driving of up to 40 watt-seconds. (Note that the discontinuity
indicator for selection of the proper technology for a in the curve is due to a material change at 0.0150.)
sense resistor. An important part of the power con-
siderations is an understanding of the nature of the
current waveform to be experienced. The device may 50

be intended to sense a DC current but may also need

(.) 40
to survive - or perhaps even measure - transient peak w
CI)
\ ......
currents caused from PWM switching, output ca- ..:. 30
I-
\ ......
pacitor charging, inrush current surges, load tran- «
s:
sients, short circuit peaks, and random externally
caused surge currents.
20
\

"'- '" .....

'--....
~
10
Data on steady state power dissipation capability 0.01 0.02 0.03 0.04 0.05
is most often supplied as a curve giving allowable
RESISTANCE (0)
power rating versus ambient temperature, where the
curve is designed to manage the temperature rise in
Figure 3 - Pulse-power, or energy rating, for a one-watt
the resistor. Figure 2 gives an example of a typical
IRe resistor series.
derating curve for a sense resistor making note of the
temperature limitations of the PC board under the
resistor and the melting point of the solder used to

Texas Instruments 3 SLUP114

 Some technologies and package options may vice has a finite and non-linear characteristic, the
limit the surge current rather than the power, due to industry practice is to specify TCR as a linear rela-
limiting constraints on leads, wire bonds, or in the tionship between specified temperatures, such as
case of film devices, the rating of the film itself. This 20°C to 60"C, where 20°C is the reference tempera-
is an area where it is important not to make assump- ture. Since current sense resistors often operate at
tions but to gather specific data from the supplier of temperatures substantially above ambient tempera-
the particular device proposed for use, and then to ture, designers should be wary of TCR specifications
refine this information with measurements made in which do not provide curves over their entire operat-
the unique environment and mounting arrangement of ing range.
the proposed application.
TCR is normally specified in units of parts per
Temperature Coefficient of Resistance - The TCR million per degree centigrade (ppmfC) and can be
of a sense resistor is an 'important parameter when translated into accuracy by using the formula:
accuracy is required. All materials undergo a resis-
. ppm
tivity change with temperature but this change is dif- M (%) ::: (Toperatlflg - Treference ) x- 6- x 100%
10
ferent and unique to each. Moreover, it is typically
In addition to the TCR of the resistor's element,
non-linear and - particularly at the higher tempera-
low-valued sense resistors can experience an addi-
tures - can be either positive or negative as shown in
tional error caused by the much higher TCR of the
the representative curves of Figure 4. It should be
copper connections. Copper has a TCR of
noted that manufacturers have come a long way with
+3900ppmfC and to illustrate the impact of this, con-
this technology and by combining various materials,
sider the resistor of Figure 5 which itself has a re-
and using unique processes, can now produce sensing
spectable TCR of IOppmfC but with one m Q of
resistors with TCR's as close to zero as most any de-
total contact resistance (2Rcu), exhibits a much
signer would want. However, when a particular de-
poorer performance according to the curve shown.

f.g( % )

0.1
· 20 20 40 60 80 100
700
.A"f, ~ ~ llT7 ~
T('C)

av
· 0.1
........
., · 0.2 0/ ~ 600

~
-0.3 Kt 500

U
MANGANIN ~co. 400

a
.J.~ +3ppmlr 1 II: 300 \.
f.g(%) ... ~ ,
0.01
....../1 r1~ '" 200
-20 20 ... .('~ 1,-7/
-7" ~:;t..-?"" V
/,
)00
1ppmk"

,.- 100
~~
~
, -0.01
" ,
-
/ 1/ /

~,
---
~ T/'C
~ 1ppmk'1

0 ' \
-0.02 " 0 0,05 0.1 0,15 02
-', _3Ppmk-1
" -0 .03

ZERANIN
SPECIFIED RESISTANCE OF DEVICE [WI

Figure 5 - The impact of contact resistance on total

sensed TCR.
Figure 4 - Variations of TCR with temperature for
two different materials produced by ISOTEK.

Texas Instruments 4 SLUP114

Four-Wire (Kelvin) vs. Two-Wire Resistors A SMD two-terminal resistors, and in either case, the
four-wire Kelvin resistor solves this problem (as well technique is to define a pair of sense leads placed
as the basic inaccuracy of contact resistance) by away from the high current path to exclude as much
eliminating the error of the board trace, the solder of the contact voltage drop as possible. While this
connection, and the leads of the resistor itself, by approach is clearly not as good as using a true four-
bringing out two Kelvin sense leads as shown in Fig- contact resistor, the error - particularly with surface
ure 6. As long as the measuring system presents a mount metal strip devices - can be extremely small.
reasonably high impedance to the sense leads C and
D, then the measured voltage drop is only that of the Paralleling Four-Wire Resistors Occasionally,
resistive element, and all the effects of contacting the when very high power is required, paralleling Kelvin
resistor are bypassed. Clearly, this technique be- resistors may be an appropriate solution to distribute
comes almost mandatory at the higher current levels the heat. When this is done, it is worth noting that
and many unique package solutions are offered with ballast resistors must be used in the sense lines to
four contacts (see Figure 1). insure that these lines do not inadvertently conduct
significantly higher currents than expected. Refer-
ring the schematic of Figure 8, and remembering that
CU RRENT the whole point of the Kelvin contacts is because of
• the variability's of the contact resistance, without the
RCU R ELEMENT RCU ballast resistors, unequal currents through the contact

~ ='---1>---'l/'.A~---= ~
resistances could equalize in the sensing elements by
flowing through the interconnected sense lines which
typically are not designed for high currents. The
Figure 6 - Afour-wire resistor with Kelvin sense leads. lower the Kelvin resistor value, the higher the error
introduced by mismatched connections. The ballast
The added accuracy of a four-wire connection
resistor should be on the order of 1000 times the ele-
comes at an a higher cost, however, as four-contact
ment resistance and can have a low power rating
resistors are more costly, both to manufacture and to
since its presence insures that any currents in these
install. One economical compromise is to simulate a
leads will be quite small.
four-wire connection - as much as possible - with the
printed circuit board layout as shown in Figure 7.
The patterns are shown for use with either axial or ACU RElem.,,1 ACU

AbaUast Rba ll asl

Lea de d D ev ice T ra ce Pa tl ern

~ ~
Curr Oul
I--- + SENSE - + -

c 0
Rballasl Rballasl
SMO Devlct! (Sha de d) Palt er n AC U AEla men l ACU
SMO De vice must contact all 4 pads as shown

Figure 8 Using ballast resistors when paralleling Kelvin

resistors.
Accuracy is imp roved" distan ce
"d O Is la rge enough to prev ent C o
solder bridging between pa ds. Self-Inductance - A low inductance type of resistor
becomes extremely important in any application
Figure 7 - Simulating a four-wire resistor using the PC where the current waveform can exhibit a high dildt -
board trace. an issue inherent in most switching power supply
circuits. Any inductance in the resistor, when ex-

Texas Instruments 5 SLUP114

posed to a high current slew rate, will result in an inches. Another difficulty is that copper has a tem-
inductive "step" voltage superimposed upon the sense perature coefficient of 0.39 % / DC, meaning that the
voltage as shown in Figure 9. This may merely con- voltage drop for a given current will increase some
tribute an error or, more commonly, prematurely trip 20% for a 50°C temperature rise. However, for those
an overload-detecting comparator, or ring below applications where this technique is appropriate, the
ground and latch up the sense circuitry. Most sense following design data is provided [1]:
resistors on today's market are manufactured with The resistance of a metal path at a given tem-
"non-inductive" techniques but it should be remem- perature is governed by the equation:
bered that this is a subjective term and, the higher the
current, the less inductance it takes to create prob- R(T) = S(T) Length
lems. Width x Thickness

where S(n is the basic resistivity of the material.

Current For the normal copper used in PCB technology, its
value is
+
A 0----1 V'v---Q B
S = 1.7241 X 10--6 n -cm at 20°C
and the more complete description is
Current
S(T)=1.7241X1O·6 [1 + 0.0039(T-20)] ncm

Since PC boards use sheet copper of a known

Voltage thickness, it is helpful to divide the resistivity by the
thickness (in centimeters) to obtain a Sheet Resis-
Figure 9 - Resistor inductance combined with high dildt tance, Rs, which now has the units of Ohms per
can cause voltage spikes. square. The advantage is that now the resistor's de-
Current Sensing With a Copper Resistor sign can be done in only two dimensions - length and
Rather than add a separate component for current width - which are the dimensions that the designer
sensing, there is certainly an incentive for using the can control. For one-ounce copper, the thickness is
copper trace in a printed circuit board as a low-value typically 0.0036 cm, and the 20°C value of a given
resistor for this purpose. If we can accomplish this, resistor then becomes:
not only will we - of necessity - suffer a smaller volt-
age drop and associated power loss, but now sensing R(in milliohms) =0.4789 X length
current comes for free, saving all the costs associated width
with buying and installing a discrete resistor. But of
where length and width can be in any units as
course, there are difficulties, most notably that the
long as they are the same. And, of course, two-ounce
resistance of copper is VERY low, meaning that the
copper is twice as thick, so Rs is half the above value.
signal voltage will either be so small as to require
Note that, as mentioned above, dimensional toler-
relatively challenging amplification, or that the
ances directly affect the resistor's value. Edge con-
length of this resistor may add cost due to the PC
trol of the PC etch process will usually define the
board area required. This area usage is also impacted
minimum acceptable width, and one must always
by the tolerances required of the resistor's value.
insure that a cost-cutting buyer doesn't allow a thin-
While a narrow resistor pattern would decrease the
ner copper sheet midway through a production cycle.
length and thus also the area, PCB etch tolerances
The use of a low value sense resistor implies that
usually dictate a minimum width of 0.015 to 0.025
the current in the resistor may be quite high. A cop-

Texas Instruments 6 SLUP114

per etch on a PCB will self heat due to the power dis-
sipated by the resistor. This temperature rise above r -- - -- -- - - -- -- -- ---- -- l
I UCC3926 CURREN T SE NSOR I
ambient can be determined from the current and the
total resistor area by using guidelines from MIL-
Q-- CHIP SUPPLY :
I
I
STD-275E [2]. This information was used to gener- I
ate Table 1 which provides the required dimensions SIGN
for a loz PCB resistor given a maximum current and
desired voltage drop. This data assumes a tempera- POS OU T
ture rise of 30°C above an ambient temperature of 60
NEG OUT
degrees.
AUTO ZERO
TEMP COMPENSAT ED
Desired Vo ltago Drop GA IN TRIMMED
DIFF AMP
10mV 25mV 50mV
AMPS PCB Etch LenQth (in) Width (in)
1 O. t62 00405 0.810 0.0 10
2 0.243 0.608 1.215 0.Q15
3 0.405 1.0t3 2.025 0.025
4 0.648 1.620 3.240 0.040
5 0.891 2. 228 4.456 0.055
6 1.053 2.633 5.266 0.065 I
I
7 1.377 3.443 6.886 0.085 IL _ __ ___ __ _ __ ___ __ ____ _ J
8 1.70 1 4.253 8.506 0.105
9 2.025 5.063 10.126 0.125
10 2.430 6.076 12.152 0.150
Figure 10 - The UCC3926 includes the currefll sensing
11 2.754 6.886 13.772 0.170
12 3.078 7.696 15.392 0. 190
resistor and the signal-conditioning circuitry together in a
single IC surface-mount package.
Table 1 . Dimension solver f or given ClIrrelll and voltage
based on aI/owing a 30°C tempera ture rise above a 60 °C
ambiefll.

Using an IC Leadframe as a Current Sensing Re-

sistor
There is another approach to using copper as a
current sensing resistor, exemplified by Unitrode's
proposed UCC3926 Integrated Current Sensor, illus-
trated with the simplified block diagram of Figure 10.
This product takes the tack of integrating the
sensing resistor with the amplification and level
shifting circuitry normally required to develop a us-
able current signal. Since low-value resistance is not
readily compatible with IC processing, a specially
designed copper leadframe shown in Figure II is
used to form the resistor. While the current to be
measured must now pass through the IC package, the
design of the sensor and it's signal-conditioning cir-
cuitry as an integrated product allows several benefits
as listed below:
Figure 11 - A specially-designed lead/rame for the
The custom leadframe design allows the safe UCC3926 forms the current sensing resistor.
conduction of as much as 20 Amps with a maxi-

Texas Instruments 7 SLUP114

 mum total voltage drop of 30 millivolts, signifi-
cantly minimizing both power loss and internal ---- ----------
UCC3582
-- - ----,I
heating.
The leadframe resistor voltage drop is measured
with Kelvin contacts provided by internal bond-
ing wires.
A chopper-stabilized, high-gain, auto-zero ampli-
fier is included to deliver a 0.5V signal from a
15A input current. The common-mode range of
the sensor can be either at ground or at a positive
voltage rail of up to 14V. Sensed current can be
bi-directional.
Fig ure 12 . Th e UCC3582 eliminates th e need for a
The amplifier is trimmed after assembly to cancel
separate currefll sensillg resistor by "sing the Rds(all ) of
the effects of both resistor tolerances and gain, as
the FET.
well as packaging induced shifts. This product is
specified to measure current with an accuracy of
±2%. Using the Inductor as a Sense Resistor
Another interesting circuit idea for measuring
The amplifier has built-in temperature compen-
sation for the TC of the copper resistor and will current without a discrete sensing resistor has been
tend to track any self-heating as the chip is proposed by Pacholok [3] and is particularly appro-
mounted in the same package. priate for low voltage power converters where a volt-
age drop of several hundred millivolts for current
Using FET Drain-Source Resistance as a Sensor sensing would represent a significant efficiency loss.
With the successful utilization of power MOS-
This approach, illustrated with a boost converter cir-
FETs as the switching devices for the current we
cuit in Figure 13, recognizes that the voltage drop
wish to monitor, and the recognition that a FET is a
across a power inductor consists of two parts:
majority-carrier device with a fairly linear on-
resistance, the prospect of using the Rds(on) of the
1. VI = L dI
FET as the current sensing resistor has gained recent . dt
interest. An illustration of this approach is shown in
Figure 12 which shows a portion of the Unitrode
UCC3582, a simple synchronous buck regulator, in
which short circuit protection is provided without the where the "R" is the copper resistance of the in-
need of a separate current sensing resistor. With this ductor's winding. These two voltages add to form
device, the drain-source voltage of the external high- the total voltage across the inductor, VT. By adding
side switch is monitored during the on-time of the a closely coupled second winding with an equal
switch, and a voltage greater than 450 mV will initi- number of turns, but with minimal current loading,
ate pulse-by-pulse current limiting. While this is a this winding will have only VL across it. By then
very low-cost solution, one must recognize that the summing the voltages from the two windings such
actual current threshold will vary with unit-ta-unit that the VL terms cancel out, one is left with only
variation of FET parameters, gate drive amplitude, VR, a voltage determined by the inductor current, and
and temperature. While these tolerances are probably the appropriate signal for a current-mode control
too great to allow this technique to be used for accu- loop. A simple current mirror circuit will convert
rate control, it may be perfectly acceptable if the only this differential VR term to a single-ended, ground-
issue is survivability in a fault condition. referenced signal compatible with the current sensing
pin of a UC3842 or similar device.

Texas Instruments 8 SLUP114

 I+-VL--+I
NL Np _ NL
VIN R ~ Np RESIST ANCE Your

E : 4nl0·7 NASf :K,Sf

Voul : tcSEdl : 1<,,1

Figure 13 - By adding an additional winding. the IR drop
in the inductor can be used Jor current sensing. Figure 14 - The Rogowski Coil delivers a terminal voltage
proportional to the rate-oj-change oj the current passing
While this technique has saved the power loss through its loop.
and cost of an additional current sense resistor, rec-
ognize that the resistance which is used is the copper E= NA dl
JW dt
wire in the inductor and the temperature coefficient
of copper's resistivity applies. There may also be
where J1 0 is the permeability of free space, N is
some inaccuracies as the current signal is a small
the number of turns/meter of the coil, A is their
voltage derived as the difference between two much
cross-sectional area in square meters, and I is the
larger voltage signals.
current passing through the loop. This expression for
the voltage can then be written as
MAGNETIC SENSING DEVICES
Ever since it was discovered that a current flow- E=K dI
ing in a conductor generated a magnetic field, there , dt
were techniques developed to use that field as a way
of measuring the current which created it. It should with KI now defined as the coil sensitivity with
be obvious that one of the more significant benefits units of Volt-sec / Amp.
of using magnetic coupling for sensing current is the To reproduce the current waveform as a voltage,
electrical isolation which this achieves. Clearly, the an integrator such as shown in the above figure will
most popular and practical application of this ap- generate
proach is embodied in the current sensing trans-
former, but before discussing this component, it may
Vow
be useful to review some other techniques which
have been used to determine a current's value from
its associated magnetic field. if we define K2 as Kl / RC.
The Rogowski Coil Practical implementations of this technique
One of the earliest current transducers was the typically also incorporate a low-frequency roll-off to
Rogowski Coil, first described in 1912 and illustrated eliminate thermal noise and drift. Since the trans-
in Figure 14. [4][5] ducer is not able to reproduce the DC component of
This device consists of a single-layer coil, uni- the current, it operates in a manner similar to the AC
formly wound on a non-magnetic core which is either input of an oscilloscope.
flexible or formed into a circle which surrounds the The major benefit of a Rogowski Coil is that
conductor of the current to be measured. The termi- since the core is effectively air, there is no magnetic
nal voltage of this coil will be material to saturate and the coil's output remains lin-

Texas Instruments 9 SLUP114

ear for extremely high peak currents. This device is repetitively interrupted, allowing time for transformer
still used today for measuring high-energy current reset.
pulses or transients with high-frequency harmonic The current sense transformer operates the same
content since the upper bandwidth can extend into the as any transformer, with the exception that the pri-
megahertz range. mary signal is a current. In a typical switching con-
Current Transformers verter application, the primary winding will consist
With the availability of magnetic core materials, of a single tum, with the secondary turns providing
it could have been a simple but very significant ex- the current attenuation. The primary-to-secondary
tension of the Rogowski Coil technology to replace current relationship is given by
the air core with a material which concentrates the
Is Np
magnetic flux inside the coil. With the flux con- -=-
tained within the coil instead of passing through it, a Ip Ns
direct relationship between the coil current and the
current in the conductor generating the field was and the resultant output voltage is defined by the load
achieved and the current transformer was born. The resistor, Rcs, as
similarities in the structures can be seen in the sketch
of a current transformer shown in Figure 15, but with
Vo = RCS(~)
the coil now terminated with a low impedance, the
induced voltage becomes irrelevant and the output The equi valent model for a current sense trans-
voltage is instead determined by the coil current and former is usuall y as shown in Figure 16 where, in
the load resistance. [6] addition LO the primary and secondary tu rn , the
magneti zing inductance and interwinding capac itance
Intogration Path
have been added to the secondary.
Magnetic
Core
Secondary Winding

Vout

Np
IDEAL TRANSFORMER TURNS RATIO
Ns -
LM = MAGNETIZING INDUC T ANCE
Cw = INTERWINDING CAPACITANCE

Figure 15· A terminated current sense transformer Figure 16 . The typical electrical model for a current
will provide a voltage signal directly related to primary sense transformer.
current.
Core Issues - The current in the primary is trans-
Current sense transformers, of course, provide formed to the secondary, but the transformer must be
two important benefits over simple resistive sensing - excited through the action of the magnetizing current.
isolation between control and power and lower power The magnetizing current is supplied by the primary
losses. The lower power dissipation of a current and is therefore subtracted from the current available
sense transformer allows a much higher signal level, to drive the secondary. This leads to one of the de-
significantly improving the signal-to-noise environ- sign criteria for the transformer - low magnetizing
ment of the control system. The fact that a current current - achieved by providing a large magnetizing
sense transformer cannot work with DC currents is inductance. Additional requirements are typically
usually not particularly troubling when used in high saturating flux density, low core loss, and, of
switching power converters where the currents are course, small size and low cost.

Texas Instruments 10 SLUP114

 In most high frequency applications (f > 50kHz), winding needs a resistor, Rcs, to convert the secon-
ferrite will be a good choice for the core material, but dary current to a voltage adaptable to the needs of the
at least one alternative is a Ni-Fe tape wound core. following current control circuitry. The purpose of
Materials such as Supermalloy have very high per- the diode is to allow the transformer's magnetizing
meability and saturating flux density compared to current to induce a high reverse voltage on the secon-
ferrites. Their permeability rolls off with frequency, dary when the primary current goes to zero in order
however, which gives the advantage to ferrite at fre- to provide a fast reset for the transformer's core, and
quencies above 50kHz. Tape wound cores also tend also to isolate this negative reset voltage from the
to be more expensive than ferrites for the same core control circuit.
size. On the other hand, tape wound cores can still be
cost effective at low frequencies where their higher
Vd
permeability and saturation level may allow for a
......---l .:>1-- .......- - 0 Ve
much smaller core than would be required if ferrite
were used. Common choices for ferrite include type Res
J or W materials from Magnetics, Inc. or H5B2 from
TDK.
Current sense transformer cores are almost al-
ways toroids. This shape minimizes leakage induc- Figure 18 Trallsformillg primary current to a control
tance and there is no loss of permeability due to an voltage.
air gap which would be unavoidable with any type of
cut core configuration. Toroids are normally more Step one - Determine the Turns Ratio - Usually the
expensive to wind but this disadvantage is minimized design begins after the rest of the converter circuitry
with a single turn primary and the availability of has been defined. Initial parameters such as the
many "off-the-shelf' standard designs. An example maximum secondary voltage are usually determined
of a simple high-frequency current sense transformer by the control circuit. For example, if using the
design is shown in Figure 17. [7][8] UC3842 Current-Mode PWM Controller, the maxi-
mum current input signal is 0.9 Volts. If we allow a
20% margin for current limiting, the maximum op-
erational signal, Vc, becomes O.7V. Similarly, the
turns ratio is usually constrained by the desired sec-
ondary voltage, the primary current, and the need to
keep the control circuit power to a minimum. A rea-
sonable goal might be to keep the power dissipation
in the sense resistor to less than 62mW (so a 1I8W
resistor could be used). (Purests will note that we are
using peak values where accuracy would dictate the
use of rms, however the simplification is easily justi-
Figure 17 - A typical current sense transformer applica- fied and could be corrected if needed for a critical
ble to switching power supplies - Courtesy of Pulse Engi- design.) If the maximum primary current was to be 5
neering, Inc. Amps, and the full-scale secondary voltage is O.7V,
then two equations in two unkl10wns can be written
Current Sense Transformer Design Procedure as
To illustrate a typical current sense transformer
design process, the schematic of Figure 18 will be Is2 Res == O.062Watt
used. Since the transformer is transforming current -
with zero voltage on the primary - the secondary IsRes == O.7Volt

Texas Instruments 11 SLUP114

 Solving these equations yields Rcs = 7.9 ohms. which can be met with 33 AWG wire. The total cop-
Once the sense resistor is determined, the secondary per area is then
current and hence the turns ratio can be calculated by
Area (total) = 50T x 0.2 x 10.3 + IT x 10 X 10.3
0.7V = 20 X 10-3 cm 2
Is '" 89 rnA
7.90.
and if we use a fill factor of 0.4, a core with a win-
Ip X Np 5A x lT dow area of 50 X 10-3 cm 2 would be appropriate,
Ns = = 56 turns
Is 89mA such as a Magnetics, Iric. Core Number 40601. This
core has a cross-sectional area, Ae, of 0.21 cm' and
A comment here about turns ratio. As the secon- an effective path length, 1!. m, of 1.3 cm.
dary turns go up, the magnetizing inductance will
increase, thereby reducing the error, and the power Step Three - Calculate Flux Density - The concern
dissipation will go down, but the interwinding ca- here is to insure that the core will not be driven into
pacitance will tend to increase. However, it is often saturation. Using Faraday's law and assuming
desired to use a standard, off-the-shelf transformer. switching frequency of 200kHz (5 usec total period)
Common turns ratios are 50: I and 100: I and, al- and a maximum duty cycle of 40%, we can calculate
though it might make for a slightly higher resistor the maximum flux density as
dissipation, to keep this example simple we will
choose a turns ratio of 50. This gives a secondary
current of 100mA (5A150) and to maintain the same (Vc + Vdiode) X ton 8
B = xl0
control voltage of 0.7V, the resistor value will be re- Ns x Ae
duced to 70. .
1.4V x 40% X 5usec 8
Step Two - Calculate the Wire Size - As a first cut, a = 50T X 0.21cm2
X 10 = 27 Gauss
current density of 500 Alcm' can be used. Remem-
ber, the real parameter of interest is the power dissi-
pation, or how much power loss can be tolerated in This is certainly well below saturation and quite
the windings. Once you've determined the wire di- typical of current sense transformer design. Since the
ameter, pick a core that can handle the copper area. secondary voltage is low and the turns ratio relatively
high, the flux density should be low.
Returning to our example, if the primary current
is SA, the wire size required is Step 4 - Calculate Magnetizing Inductance
With initial penneability, ~, of 10,000 for
Wire cross-section area the type W ferrite core, magnetizing inductance
5A is given by the following COS equation:
=----- 2
500 A/cm
2
Ns Ae x 10-9
Lm= 4 Jrf.l 1!.
which corresponds to 17 AWG wire size. Typically, m
this is a single tum and therefore the total resistance
502 x .021 -9
will be quite small. A smaller wire size could be = 4 Jr X 10,
000
X x 10
used. The secondary wire size is 1.3
=5mH

Area =
51sar = 2(X) x l 0-6 cm
2
I turn
The peak magnetizing current is then:
SOOA/cm2

Texas Instruments 12 SLUP114

 to limit the maXImum voltage, with these existing
paths, it is usually unnecessary and serves only to
further lengthen the reset time. Another possible so-
This translates into a transfonner error: lution, if needed, is the use of a zener as the rectify-
ing diode
I This issue of reset time and/or voltage can be-
% error =...!!!... x 100 = 0.6%
Is come important in applications where the duty cycle
can approach 100%. In boost converters at startup,
Since the magnetizing current is assumed to and power factor controllers where the duty cycle
start at zero and increase with the duration of the attempts to go to 100% with every zero-crossing of
primary current pulse, this error represents the the power line, reset time can become critical and
droop in the secondary current wavefonn, definitive steps may need to be taken to limit the
reached at the end of the pulse.
maximum duty cycle to something less than 100% in
order to keep the current sense transformer from satu-
Step Five - Reset Voltage - The last step is to de- rating. Saturation, if it does occur, can be a serious
termine the reset voltage required after each current problem in a current control application as it will
pulse. The cardinal rule of magnetics is that the volt- cause the sensed current signal on the secondary to
seconds must be equal. Therefore, diminish which, in tum, signals the control circuit to
force more current, which can lead to catastrophic
Vreset X toff = (Ve + Vdiode) X ton results.
Rearranging terms, we can calculate the reset DC Current Transformers
voltage which will appear as Wouldn't it be nice if we could combine the low
power loss and isolation of the current transformer
Vreset = (Ve + Vdiode) ton with the ability to accurately measure DC currents as
toff accomplished with the shunt resistor? Actually, sev-
= (0.7 + 0.7) 40% /60% = 0.93V eral solutions for a "DC transformer" have been pro-
posed in the past, one of which is shown in Figure
For this example, the reset voltage of less than 19. [9]
1V illustrates why, in most DCIDC converter appli-
cations, core reset is not a problem. If we were to
extend the maximum duty cycle to 95%, this example
would still only require a reset voltage of 27V. In the
typical application with a diode in series with the
transformer's secondary, one might assume that the
theoretical reset voltage would be determined by the
breakdown voltage of the diode as the magnetizing
inductance will force the flow of magnetizing current
somewhere in the secondary circuit when the primary
current is interrupted. This high voltage would, of Figure 19 - Using two cores with an AC excitation voltage
course, make the reset time extremely short but there applied so that the magnetic fields cancel, implements
could be some reliability concerns. In actuality, a transformer with isolated, accurate response to DC
current.
however, instead of breaking down the diode, the
magnetizing current will flow into the interwinding
This transformer uses two cores with an AC ex-
capacitance and through the reverse recovery of the
citation voltage applied to generate opposite-polarity
diode to the sense resistor. While an additional re-
flux excursions in each core. Using square-loop
sistor could be placed across the secondary winding
cores such as permalloy 80 or a high-frequency fer-

Texas Instruments 13 SLUP114

rite, the excitation voltage will drive the cores into The waveforms for this example are shown with
and out of saturation with opposite polarities at each squarewave excitation although the technique was
half-cycle. A simplified illustration of the operation initially described by Storm [10] for use with sine
on the B-H curves for each core is shown with the waves. And, of course, it should be obvious that this
two curves superimposed in Figure 20. The set and transformer will work with primary currents in either
reset points are determined by the excitation voltage direction. The secondary signal is rectified and the
on the two secondary windings connected in series, result is a DC current directly related to the absolute
but offset by the operating point defined by the pri- value of the primary current by the turns ratio_ Im-
mary current common to both cores. portantly, this transformer can be made quite accurate
as, to a first order, excitation voltage levels, loading,
8
temperature, and core characteristics do not enter into
the equations. Of course, in today's technology, the
T 1 Reset Point limited high frequency capability of magnetic cores
T 1 Operating do represent a serious restriction for applicability to
Point set by Ip switching power supplies.
H
However, Severns [11] has taken this approach
one step further, with potential applicability to high
T 2 Operating
frequency switching power supplies, with his design
T2 Reset
Point
1
Point set by Ip
of a unipolar version of this function using a single
core as shown in Figure 22.
-~.:::;...~
TS

Figure 20 The 8-H Loop operation of the DC current

transformer.

As each core alternately comes out of saturation,

".1f"'"
the changing flux induces a current in the secondary +
proportional to that flowing in the primary. This op- Dt
eration is best visualized with the aid of the time-
based waveforms of Figure 21.
PUL SE
GENERATOR 02
Ns

"
J Np.t

SAMPLE
AND
HOLD OR Yo

+Es Rs l" }y, AYERAGE

FIL TER

Excitation -
Voltage 0
TJi T5 3TJi 2T5
2 2
Figure 22 - A unipolar DC current sensor with high-
frequency capability can be built with a single core.
I
+8 ...:L The operation of this circuit is as follows:
T, flux

T2 flux The pulse generator is run with a very short duty-

L>8
-8 , ,I , ,IT cycle - typically less than 10%. While it is in the

~~
: ~~
:t
+I5 off state the switch, Ql , is open and the primary
Secondary 0 current, at any value above some minimum, will
Current -Is
hold the core in saturation and there will be no
Figure 21 - The offset operating points of the two cores secondary current.
allow the excitation voltage to pull each core alternately
out of saturation to induce a secondary current.

Texas Instruments 14 SLUP114

2. When the pulse generator is on, Ql closes and the The basic relationship describing the output volt·
reversal in voltage across the secondary allows age from a Hall sensor is
the core to come out of saturation with a current
Is = Ip I Ns. At the same time, Vo is developed
by sampling the voltage on Rs caused by Is.
where I is the bias current through the sensor and B is
the magnetic flux passing through it. ~ is a coeffi-
This circuit could potentially operate with sam-
cient defining the sensitivity of the sensor and it in-
pling frequencies up to the megahertz region, allow-
cludes both the electron mobility and resistivity of
ing reasonable resolution of current waveshape with a
the semiconductor material as well as the geometry
bandwidth in excess of 100 kilohertz. An additional
benefit is that the low duty-cycle of the excitation
Vee
voltage acts as a multiplier to the turns ratio and al-
Saturation
lows very high primary current without a corre-
/
sponding number of secondary turns. [12]

HALL SENSORS IL"

/Quiescent
Outpvt
Another element with potential application for /' Voltage

sensing DC current with electrical isolation is the

Hall-effect sensor, a device which is currently pro-
./
V
duced in silicon as a semiconductor component r---
Saturation
where a certain amount of interfacing circuitry can be
a
integrated along with the sensing device into a single ·6 a +6
Magnotic Field (Gauss)
silicon chip. The Hall-effect principle is that a volt-
age will be produced across a current-carrying semi- Figure 24 - The characteristic curve for a Hall sensor
conductor material under the influence of a magnetic designed for linear applications.
field . This principle is illustrated in Figure 23 by
observing that current I flowing from bottom to top in dimensions of the particular sensor design. Most
the Hall device is deflected at right angles to an im- sensor devices used for linear applications today in-
posed magnetic flux B, generating an induced voltage clude added integrated circuitry to develop the bias
V across the device. [13] current and provide a differential amplifier for the
output voltage. It should be noted that the discussion
herein will be limited to linear current sensing appli-
cations and excludes switching and position sensing
uses - by far the biggest markets for these devices.
Linear devices are typically bi-directional with a
characteristic transfer function as shown in Figure 24.
Depending upon the particular design, this curve can
show a linear output voltage from zero to five volts
with a flux excursion of + to - 1000 Gauss.

On first glance, a Hall device would look like the

perfect answer to a current sensing problem. It is a
Figure 23 - A Hall-effect sensor develops an output volt-
age by the modulating effect of a magnetic field on a bias small-sized, low cost device with a linear response to
current. a magnetic field, capable of measuring DC currents,
and with inherent isolation capability. The problems

Texas Instruments 15 SLUP114

all show up in the fine print, and include the follow- This expression is for a round conductor where R
ing: is the distance in centimeters from the center of the
• The sensitivity is low, unusually requiring conductor to the point of sensing, I is the current in
some form of "concentrator", which adds Amps, and B is the flux density in Gauss. The prac-
costs. tical limitations of this equation are easily seen in
• Mechanical positioning can be tricky and that the flux density from a current of 100A would
may also add manufacturing costs. only be 100Gauss at 0.2 cm from the center of the
• The range of linearity is limited, and there is conductor. A high-current solution proposed by Rip-
often an offset at zero current. pel [14] is illustrated in Figure 25 where he necked
• The devices are sensitive to mechanical down the cross-section of a flat conductor so he could
stresses and ambient temperature variations, place the sensor close to the center, while still keep-
as well as unit-to-unit tolerances. ing the resulting hot-spot within acceptable bounds.
• The sensing element has a large junction ca- For lower currents, some form of field concen-
pacitance which typically limits the maxi- trator will likely be required. Using a high-
mum frequency response to 20 - 40 kilohertz. permeability core with a gap into which the Hall sen-
• While isolated from the current they would sor can be placed allows usable flux densities to be
measure, these sensors do require some form achieved at much greater distances from the conduc-
of power supply for their own consumption. tor's center. Figure 26 illustrates this technique. [15]
The next step for greater sensitivity at lower cur-
However, within these confincs, practical solu- rents would be to reduce the size of the conductor and
tions have been found. Some of these are outlined in wrap multiple turns around the field-concentrating
the following examples. core as shown in Figure 27. While this may appear
The simplest approach would be to just place the as an unacceptable complication, it should be noted
Hall sensor against the conductor containing the cur- that with a winding on a gapped core, we have cre-
rent to be measured. However the value of the flux ated an inductor which, of course, could be the cireuit
level at this point is defined, from Ampere's Law, by inductor whose current we are trying to measure in
the first place. The only issue is whether the thick-
I
B=O.2x-. ness of the gap - which must be large enough to allow
R the Hall sensor to fit - is aeceptable for the design of
the inductor.

Current
Carrying
Conductor

Figure 25 By reducing the area of the conductor, the

Hall sensor can be placed close enough to the center to Figure 26 - Practical separations between the Hall
be usable as a current sensor in the 100 - 500 Amp sensor and the conductor can be achieved with a field
range. concentrator made from a magnetic core.

Texas Instruments 16 SLUP114

 this is turned into an output voltage by passing
/ Current Carrying Conductor through a sense resistor as shown in the figure. By
keeping the net flux at zero, most of the sensor toler-
ances are canceled and, in addition, the frequency
response is also enhanced.
Before leaving the subject of Hall-effect current
sensors, it might be interesting to mention an innova-
tive application proposed by Carrasco, et al [16] in
order to get an accurate current sensor with a band-
width from DC to beyond several hundred kilohertz.
This solution, which is illustrated in Figure 29, uses a
core with a Hall sensor and three windings. The first
is the primary and conducts the current to be meas-
' - - - Magnetic Core ured. This, of course, could be a single tum. The
second is the compensating coil used to linearize the
Hall-effect device as described above. While this
Figure 27 - The flux from the current to be measured
improves the response of the Hall sensor, it will still
can be further multiplied by adding turns to the core.
fall off at something less than 100kHz. The third
In the above applications, the Hall sensor is es- winding forms a conventional current sensing trans-
sentially operating as an open-loop input device and, former which, while unable to pass DC, has a good
as was stated earlier, the variability's and non- high frequency capability. This signal is then scaled
linearity's of Hall-effect devices contribute to a sub- and added to that of the Hall output for a resultant
stantial inaccuracy in the resultant measurements. A very wide bandwidth as shown.
technique for minimizing or eliminating much of this
SENSE-FET TECHNOLOGY
inaccuracy is to enclose the sensor in an overall feed-
With the advent of power MOSFET technology,
back loop as shown in Figure 28. In this circuit, a
another alternative for sensing current in power sys-
low-current secondary winding is added to the core
whose function is to develop a magnetic flux in op-
position to that developed by the primary conductor.
The feedback loop forces the total magnetic field Vo
within the core to zero and the current it takes to do

Hall Output Current Tmnslormor

Figure 29 - This current sensing scheme uses a Hall

Figure 28 - With a secondary winding to cancel the sensor to add DC response to the high frequency capa-
primary field, accuracy and response of the Hall sen- bility of the current transformer, for a resultant broad
sor is improved. bandwidth.

Texas Instruments 17 SLUP114

terns has become available. Since a power MOSFET
is constructed with multiple source contacts which r-------'J DR AIN
inherently have good current balancing between I~I
them, the current through one will match, according GA TEo--J
to their relative numbers, the current in all the others. SENSE K SOURCE

So by bringing one source out of the package on a

separate pin, the current in that pin will track that in
the power source pin with a ratio that can be made
accurate to within 3%. This current ratio is typically
made in the range of 1000 to 1500 so that the power (a)
loss in the sensing circuitry is negligible. This device
can best be visualized as two units in parallel as
shown in Figure 30 where the sensing device has
I-------'J DR AIN
source area A while the power device has an area of
I I~
nA. Note that the power source also includes a Kel- GATE O--J
SENSE K SOURCE
vin contact to keep impedances in the power line Rs
from contributing to a mismatch error.

DRAIN

7 (b)

Figure 31 Two circuit approaches for utilizing

sense-FETs.
A nA
An alternative approach is to use a forcing ampli-
fier to hold the sensing source pin at a virtual ground
potential as shown with circuit (b). While much
SOURCE more accurate, this circuit yields a negative voltage
SENSE KEL VIN
for the current signal, possibly creating further com-
Figure 30 - An equivalent representation of a sense-FET plications. While appearing to be a useful current
as a small device in parallel with a well matched but much sensing solution, the widespread usage of sense-FETs
larger unit. has been impeded by their limited availability and
higher cost. For further details, reference should be
Two of the most common approaches to using
made to the manufacturer's application information.
sense-FETs for current sensing are illustrated in Fig-
[17]
ure 31. While circuit (a) appears very simple, it
points out an important limitation in the use of sense-
FETs in that for the current ratio to hold accurately, CIRCUIT TOPOLOGY ISSUES
the voltage on both sources must not differ by much LOCATIONS FOR SENSING CURRENT
more than 2-300 mY. This puts a low maximum
Non-Isolated Switching Regulator Circuits
value on the voltage across Rs which, in tum, im-
In picking the location for a current sensing ele-
poses some restrictions on the allowable input offset
ment an obvious first consideration should be to
voltage for the comparator. It also illustrates the im-
make sure that the circuit branch selected contains
portance of the Kelvin contact to isolate this low
the information desired. For example, in a PWM
threshold from voltage drops in the power ground
regulator, the current in the power switch is not the
line.
same as the average current in the load, and while it

Texas Instruments 18 SLUP114

 sensing easier, however, short-circuit protection is
not possible. The issues here relate to grounding. If
V IN 9 I 9 .--- the system ground is as shown, the voltage drop
(' 1 (bl
across the sensing element will subtract from the load
voltage and regulation will be affected. Also the
ability to parallel power supplies to a common load
will be degraded. This is a popular location where
Figure 32 - A typical power switching stage with pos- the load might be a battery as both charging and dis-
sible current sense locations.
charge currents can be monitored with dual-polarity
might be easier to sense the current in the lower sensing.
switch of a half-bridge topology, this location may
not provide output short circuit protection. Beyond Location (Q - Here the problems relate to separating
this, several possible locations for current sensing in the load ground from the input. Now paralleling
a power switching stage are shown in Figure 32 , and loads is practical but multiple power supplies cannot
the considerations, pro and con, are discussed below: be driven from the same source voltage, and the po-
larity of the current sense signal is negative with re-
Location (a) - The current here, and in the return spect to load ground.
line, is total input current - both DC and ripple. It is
fine for input power calculations and power factor Transformer-Coupled Power Stages
control, but the DC requires a resistor sense. The The schematics shown in Figure 33 offer several
altemati ves for current sensing strategies in higher
high side location is good for short circuit protection
power applications. While full-bridge circuits are
but does require a high common-mode input to the
sensing circuitry . shown, it should be fairly easy to extrapolate to half-
bridge, push-pull, and two-switch forward and fly-
Location (b) - This is switch current and a trans- back topologies. In these cases, current sensing
former is acceptable due to the pulsing current wave- transformers are often preferable and justifiable over
form. While there is a high common-mode voltage resistive sensing, both to save the power loss in a re-
here, at least it is as steady as the input voltage. This sistor shunt and to use the transformer's isolation to
is a good location for both peak current-mode control allow level shifting from the high-voltage rail down
and overall short-circuit protection. to low-voltage controls. The distinctions between
these circuits are described below:
Location (c) - This branch will yield average load or
inductor current but it is probably the worst place for Circuit (a) - If resistive sensing is proposed, this is
sensing due to the widely-swinging common-mode probably the most practical solution as the voltage
voltage - from below ground to the input voltage. levels are referenced to ground. There is a cautionary
point to be made, however, in that if direct ground-
Location (d) - The voltage range at this point is sta- referenced gate drive circuitry is used to drive the
bilized by the output voltage but the common-mode lower FET switches, the transient gate currents
capability must also include ground for startup and through the gate-source capacitance will show up in
short-circuit conditions. This is the ideal location for the current sensing circuitry, perhaps requiring extra
average current-mode control as the current is both filtering or blanking. This problem often dictates the
instantaneous inductor current and average load cur- use of gate-drive transformers which can be con-
rent. The DC value precludes the use of transformer nected directly from gate to source, bypassing the
sensing. current sensor.

Location (e) - This branch, of course, is true load Circuit (b) - This circuit suffers from the same gate·
current and the low common-mode voltage makes drive issue as Circuit A but, in addition, the use of ~

Texas Instruments 19 SLUP114

 (b) (e)

0 0

LL JJ LL JJ LL
J ]1 II[
FL 0
FL 0
J9 FL
(d) (e) (f)

Figure 33 Various current sensing possibilities with higher power topologies.

transfonner in this location (and in Circuit C follow- output power transfonner). As a result, a senes
ing) there is added concern in keeping the trans- blocking capacitor is often added to this circuit.
fonner from saturating with the uni-directional
current pulses. While this location is in series with Circuit Ce) - This circuit uses two sensing primaries
the switches, there are two paths for current flow on a common core which alternate to allow the de-
which alternate and the only time allotted for reset- velopment of a uni-polarity signal on the secondary
ting the core is the deadband when all four switches with minimal expense. Since the core is driven in the
are off. same direction with each pulse, transfonner reset
time must still be assured.
Circuit Cc) - No gate drive problems but the poten·
tial for saturation. Circuit <D - While this approach bears the cost of
two current transfonners, it frees the circuit from all
Circuit Cd) - This would appear to be an ideal appli- the problems described above. The two secondaries
cation for a current transfonner as it will measure can be paralleled for a un i-polarity output.
true load current (reflected to the primary side) and
the current in the sensing circuitry alternates polarity Obtaining DC Average Current with Current
with each half cycle. Again, a cautionary note - if Sense Transformers
there are any unbalances in the two sides of the Two Alternating Current Sense Transfonners Rec-
bridge, a net DC current could flow in the center leg ognizing that a current sense transfonner requires
potentially saturating the current transfonner (and the reset time, its use in power supplies must be to sense

Texas Instruments 20 SLUP114

"switched" current which is typically not the same as ductor current for both DC and AC excursions. A
the "average" current delivered to the load. One way current transformer is still applicable for the charging
around this limitation is to use two sensing trans- 'portion of the waveform while the power switch is
formers as shown in Figure 34, where one is placed conducting. The output from the current transformer
to measure the current in the power switch and the is turned into a low-impedance voltage waveform by
other senses current in the free-wheeling diode which the action of Rcs in Figure 35, and applied to C2 to
flows when the switch is off. By summing their out- give the rising portion of the waveform. When the
puts as shown in the figure, a true inductor current power pulse is terminated by the action of the PWM
signal is constructed which is accurate both in wave- control, the current transformer's output goes to zero
form and in DC (or average) value. and the transformer is reset but, during this time, the
voltage on C2 discharges through R2, approximating
.--DI--_ _ _--T"-'VOUT the downslope of inductor current as it delivers cur-
rent to Co and the load. Note that this application
Il~
allows the use of a current transformer to sense DC
0----
load current on the secondary where the typical
higher current levels would make the use of a sense
resistor difficult. This could be particularly helpful
where sensing for load sharing is required.

Figure 34 - Two current sensing transformers can be used

to measure a DC inductor current by summing the switched
waveform in both the switch and diode.
AVERAGING INDUCTOR CURRENT
WITH A SINGLE CURRENT TRANSFORMER
Synthesizing Current Information - The measure- Simulating Inductor Downslope

ment of current has become more important as the

use of this information has gone past fault protection
to become a part of the regulating control loop. Even
with peak current-mode control, the current signal
could be obtained from a measurement of switch cur-
rent which could readily be done with either sensing
resistors or current transformers. However, as the
dynamic regulation demands have become more
stringent, average current-mode control has become a Figure 35 - Generating an average inductor current
waveform by simulating the inductor downslope.
more practical algorithm. With average current-
mode control, the complete current waveform in the
A more accurate representation of inductor cur-
power inductor must be known, and with continuous
rent is achieved with the circuit of Figure 36 where
inductor current, conventional current transformers
the downslope is determined directly from the in-
were no longer applicable. This becomes even more
ductor through the use of a second winding on the
of a problem in isolated supplies where the inductor
inductor which generates a constant current discharge
would be on the secondary while the controller re-
of C2. This signal will now be directly related to the
sides on the primary side.
regulator's output voltage, even during startup and
A variety of solutions have been developed over
short circuit conditions. By turning the inductor into
the years to combat this problem. Most of them take
a transformer, isolation is achieved with both charge
the general form shown in Figure 35 where the wave-
and discharge signals and the current-control cir-
form of the current in the output inductor is replicated
cuitry could reside on either side of the isolation
as a voltage waveform on a capacitor. Properly done,
boundary, The UC3848 primary-side controller
this voltage waveform can accurately follow the in-

Texas Instruments 21 SLUP114

contains an even more sophisticated version of this This added amplifier stage is very helpful as it
current synthesizer. [18] can perform three separate functions as outlined be-
low:
MEASURING AVERAGE INDUCTOR CURRENT
Downslope Simulation Based On Actual Vo Level Shifting - The resistor dividers of RIIR2 level-
shift the differential current sense signal down to a
y'"----,
voltage level much closer to ground, allowing the use

0:=0± Yo
of an amplifier with an input range less than that of
the high voltage rail. While these dividers are se-
lected to keep the maximum voltage within the am-
plifier's specified range, for startup and short circuit
conditions the amplifier's inputs must also be capable
of operation at zero voltage.

Differential Amplification - The voltage across the

sense resistor must be held low to minimize power
Figure 36 - Measuring average inductor current with the loss, but now this circuit will amplify that small dif-
downslope determined from the output voltage.
ferential voltage with a gain determined by the R3IRI
ratio in order to provide an adequate signal level for
OTHER CIRCUIT TOPOLOGY the sensing threshold of the control circuitry.
ISSUES
Output Signal Referencing - The control circuit
"High-Side" Current Sensing
processing the current information at Vc may not
The use of a current sensing resistor in series
necessarily accept a zero-voltage signal for a zero-
with either the input or output rail of a power system
current indication and the use of Vref allows the out-
can present additional problems due to the high
put of this amplifier to be offset to some higher volt-
common-mode voltage on the sense resistor with re-
age with respect to ground. Note that Vref could be
spect to ground-based control functions. The usual
ground, a DC reference, or it could also be a sinusoi-
solution to these problems is the incorporation of a
dal or other AC waveform reference.
differential level-shifting circuit similar to that shown
The determination of the nominal values for the
in Figure 37.
circuit components of Figure 37 is quite simple:

Vs The sense resistor value is determined with a

V IN + Res _ VO UT
knowledge of the maximum short circuit current,
R3B 10, and either the sense resistor's power dissipa-
tion limit, Pd, or the maximum allowable voltage
drop, Vs, as

Rcs = Pd(max) . or alternatively, Res = Vs(max)

lo(max) 2 lo(max)

~~vv---. VR EF 2. The divider Rl I R2 is normally selected by first

ignoring the effect of the gain-setting resistor,
Figure 37 - Level shifting a high-side current sense signal R3, and using the approximate expression below
to ground-referenced controllers. to insure that amplifier inputs remain below the
common-mode limit.

Texas Instruments 22 SLUP114

 R2 < Vem(amplifier max). matching, a common solution is to use ratio metric
Rl + R2 - Vo(max) + Vs(max) resistor packs for at least R 1 and R2 which, in addi-
tion to providing close ratio tolerances, track well
3. The gain and output offset are then set according over temperature.

Amplifier Offset Volta~e - Any input offset voltage

Vc(max) =Vref + - R3(
1 ) will be multiplied by the closed-loop gain of the
Rl Res x Jo(max) opamp and result in a DC error at the output, Vc.
Note that the offset voltage specification is typically
As can be seen in this expression, when 10 0, = a fixed value so that its contribution to the output er-
=
Vc Vref , and then if Io(max) is, for example, the ror increases as the voltage drop on the sense resistor
short circuit current limit, then Vc(max) must be the decreases. For this example, the effect of offset volt-
voltage threshold of the shutdown (or other limiting age is shown in Figure 39(c).
function) of the following control circuitry.
The difficulties with this circuit come in under- Amplifier Offset Current - While matching of the
standing and controlling the sources of potential er- resistors will tend to cancel amplifier bias current
rors. These are at least five sources for serious error error, the offset current will be turned into an added
contributions which need to be considered. To illus- offset voltage error by the impedance level at the in-
trate their potential impact, a specific design was puts. Typically, this term is not very significant at
analyzed to determine the relationship of amplifier impedance levels in the lO kohm range.
output voltage to component tolerances. The as-
sumed circuit conditions and nominal values are Amplifier Common-Mode Rejection Ratio - As the
shown in Figure 38 and the errors investigated were RIIR2 divider lowers the common-mode voltage, it
the following: also lowers the range of input voltage applied to the
amplifier's inputs but, as with the offset voltage, its
VS - 0 .25V MAX
effect is multiplied by the closed-loop gain of this
VIN +
.010n
Vo _12V stage. Note that if a high common-mode instrumen-
10 - 25A tation amplifier were to be used - eliminating the
R1A R1B R3B need for the R2 resistors - a poor common-mode re-
lOOK lOOK 600K
jection could introduce significant error.

r I
" Resonant Switching
Ve( l o = o) _ 2V
+ Ve(lo -25)-3V In the ever-expanding search for new power to-
~

R2A R2B
pologies to improve the performance and reduce the
R3A
10K 10K 600K cost of power systems, one of the more popular de-
VAEF
~ velopments has been the use of resonant switching as
"* 2.0V

a means for reducing switching losses in PWM con-

Figure 38 - The impact of component tolerances can be verters. The implication of this technology to current
demonstrated with the values for this typical design. sensing is not a major issue but designers need be
aware that in order to get resonant transitions, there
Resistor Matchin~ - While the absolute values of the are circulating currents which continue to flow in the
resistors in this circuit are not critical, the ratios be- primary circuitry during the time a conventional
tween them are exceedingly so. Since we are looking power switching stage would be non-conductive. A
for millivolt signals and the resistors have substantial good illustration of this is the phase-shift controlled,
voltages across them, minor differences can create bridge topology shown 10 Figure 40. [19]
significant errors which are graphed in Figure 39(a)
and (b). Because of the importance of resistor

Texas Instruments 23 SLUP114

 ~
c
150 r-----~------r_----_.------r_----~------._----_.------._----_.----__.

'f
o
z vc Erro r vs Rl and R2 Resistor To lerance
E
g 1 00~----+-----~----~-----+----~~----+-----+---~~~~-+----~
~
e..-

50r-----r-----+-----r-~~ ..~~~~~~----~----~----1_----~

o 0.1 0.2 0.3 0.4 0.5 0 .6 0.7 0. 8 0.9

_ Vs = [Link], RI and R2 vary worst case
Rl and R2 Resister Tolera nce ('Yo)
• V S = 0.1 V, RI and A2 vary worst case
(a)

,
~
c
'f I I I I ••
o
0.8 - vc •• ~
.. ;,;,;:-
Z Error vs R3 Resistor Tolerance .~
E ~
g ••
~
#- 0.6
.~ ~
'"
0>
~ 0.4
Loll .......
~
'">!
~ 0.2
...,.,. ~
I"""

g ~
o~
w
o
> o 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
- Vs = [Link], R3 Varies
R3 Resister Tolerance ('Yo)
• Vs = O.W , R3 Varies
(b)

••
I I I I ••
VC Error vs 0p·Amp Offset Voltage

•••
• ••• • ••• •• ••
...
•• ••
•• ••
•••
• •••
2 4 6 8 10 12 14 16 18 20 22
- RiRl=6 Input Offset Voltage (mV)
• RiRl = 10 (e)

Figure 39 - These curves show the relationship a/the amplifier's output voltage accuracy to the tolerances a/the circuit
parameters.

Texas Instruments 24 SLUP114

 Q)
0

l ~
I
+ GATE A-t: 1-
GATE B I I ~
GATE C I I
Lac
GATEV~ i ___J
I I
VIN I I
T 1M _ L _

aB J Lao
IT -

01
L
I
02 I
+ I I I I I
c vo IL - , - - - - r - - T - - - - T - - ,-

10 11 12 13 14-10
CURRENT SENSE

Figure 40 - In a resonant-switched bridge circuit, current sensing must be moved away from the power switches.

An examination of the primary current waveform, IT, even when the power switch is off, and thus need to
shows current flowing in the power devices even be included in the measurement system. The sche-
when nothing is being delivered to the output induc- matic of Figure 41 is recommended for use with the
tor. As a result, primary-side load current must be UC3855 PFC Controller so that as the auxiliary switch
monitored in one of the power rails, as shown in the turns on - prior to activating the main power switch -
figure. current through both the resonating inductor and ca-
Another example where resonant switching re- pacitor, as well as that in the power switch, will flow
quires special current sensing considerations is the through the current sensing transformer. [20]
Zero-Voltage Switched, boost converter particularly
CURRENT LIMIT PROTECTION
applicable to high power PFC preregulators. In this
Having spent a considerable portion of this paper
case, the resonant currents flow in the boost inductor,
discussing the measurement of current, it would now
seem appropriate to address some of the uses to
VIN LeOOST
which this information would be put. And, of course,
CURRENT SENSE high on this list should be current limit or overload

Tf f?l l~
protection for power supplies. While most everyone
recognizes the need for short circuit protection, there
are many potential strategies to consider for its ac-
complishment and the following information should
-.J~ =r Cr help to place these in perspective.
QMAIN ~ Current Limiting
Constant Current Limiting - One of the most obvi-
Figure 41 - This circuit, recommended for the UC3855 ous approaches to overload protection is to invoke a
PFC Controller, insures that the current transformer will constant current limit at a definable threshold. This
measure the resonant current as well as that through the changes the power supply's output from a voltage to
main power switch. a current source and would provide the same protec-

Texas Instruments 25 SLUP114

tion whether the fault was a m:inor overload or a dead
short. The implementation of constant current limit- Con stant Current Limiting
ing is typically done with a two-loop control system
Isc
as shown in Figure 42. This diagram shows two
feedback amplifiers, one for voltage and one for cur- I - - - -_ _ ~
"""'\
rent. They each use a voltage source as a reference
but the action of the current amplifier is to regulate
the voltage across the current sense resistor, thereby
i Internal
Vo Power
regulating the current. The operation of the circuit is Dissipation
that when the current is less than the limit threshold, Pd ~ Vl n x Isc \

that amplifier is inactive and the voltage amplifier is o L -_ _ __ _--'--_ _ _--'

in control. When the current tries to exceed the limit 10 - -

value, the current amplifier takes over and the voltage
Figure 43 - Output characteristics of a constant current
amplifier is disabled. This approach can be used in
limited power supply.
either linear or switching topologies with the only
difference being whether there is a pulse-width whatever it takes to keep the current constant. With a
modulator in the driver and power stage following the shorted output, the output voltage is of course, zero
anlplifiers. Note that it is load current that is being and a linear regulator in this mode will experience its
monitored in the circuit shown, not internal switch maximum internal dissipation as it is supporting the
current, so the next few waveforms do not apply to entire input voltage while simultaneously conducting
current-mode controlled applications. the maximum output current. This can often make
the thermal management problem in the fault mode
Res several times more difficult than when the regulator
is operating with normal loads.

Foldback Current Limiting - In an effort to ease the

fault power dissipation problem, power supply de-
V,N /-}-'-'=---=-='--'-+ VouT signers found ways to reduce the level of the maxi-
'--P!----<:. j mum current allowable as a function of the amount of
j COMP overload. This characteristic is called "foldback" for
reasons which are obviously seen in the characteristic

Figure 42 - Constant current limiting uses a two-loop Foldback Current Limiting

analog control.

This form of protection is typically favored by

the power supply user because he always knows what
to expect and it is easy to interface to a variety of
load characteristics. The supply, on the other hand, is Vo
i
often not very comfortable with this strategy as an
output short can result in higher than normal internal
power dissipation. The characteristics of a constant
current limited power supply are shown in Figure 43 o L---~------------~
10 _ _
where it can be seen that at all currents below the
threshold, the supply provides a constant output volt- Figure 44 A power supply with foldback current
age. When the load - for whatever reason - exceeds limiting.
the current limit, Isc, then the output voltage falls to

Texas Instruments 26 SLUP114

shown in Figure 44. With foldback current limiting,
the power supply behaves normally at current levels Rlldbaclc Latctl·Off At Start
up to where current limiting starts to kick in. At this
point, the output voltage begins to fall and, as it does,
the current limit value is also reduced until, when the
output is completely shorted, the available current
has been reduced to 20-30% of the normal maximum
load value. So as the output voltage falls - increasing
the internal voltage drop - the current is reduced and
the power dissipated within the unit is, if not con-
stant, at least much more manageable.
The implementation of foldback current limiting
10 _
is usually based upon modifying the current limit
threshold as a function of a voltage measurement
which changes as the supply goes into current limit- Figure 46 - A problem with foldback is the possibility
ing. There are at least two common approaches to for latch up when starting.
this, which are shown in Figure 45 . In schematic (a),
rent. Approximate equations for each case are given
the current limit is reduced with increasing (Vin -
which define the output current limit value as a func-
Vout), while in (b), a falling Vout reduces the cur- tion of the changing voltage levels.
While foldback current limiting has made the life
Rcs
I POWER I of the power supply designer much easier, the user
V'N
I STAGE I Vo
R1 now has to be much more concerned as to the nature
R2 of his load. The problem is that most "real world"
loads are not just resistive, particularly at startup or

J
~
VTH
1'1-
when recovering from an overload. Figure 46 illus-
trates a not uncommon situation where, as voltage is
applied to the load, the current increases in a non-
IF R2»R1. IPe'K.~(VTH-<V'N-Vo)B.11 linear trajectory. If the static load demands more cur-
cs R2
rent than the supply will provide - at any voltage less
<aJ
than the regulated value - the supply can latch up and
fail to start. This can be particularly troublesome and
Rc s
I POWER I
V'N
I S TAGE I Vo
CONSTANT CURRENT
R1 ~~LOAD
LIMITED SUPPLY

J
"
VTH
1'1-
CURRENT
LIMIT
ACTIVATION L.....,
97%
TIMER
? R2

-$- 1
1.
IF R2»R1. IPEAK'

(b)
~
cs
(VTH +Vo

Figure 45 - The current limit value can be made to vary

~1 )
2

+ -==.?-'
..
Figure 47 - A method for provldmg constant current limit-
ing with low average power dissipation.
with either Vout or (Vin - Vout).

Texas Instruments 27 SLUP114

 RSENSE PNP PASS DEVICE
VIN o---T"''IM..-,-----------------, r-------------------------~~V O U T

V IN CS·
6 ---- -- --- -- - - ,
UC1833 :
I
I FIB
Co
lOI1 F

R2

1.-. ._ _ _ _ _ _ _ _ _ _

Figure 48 - A duty-cycle current limited linear voltage regulator.

difficult to predict when multiple supplies are used this type of current limiting has also been success-
with the potential of interactions between them. One fully applied to self-resetting circuit breaker func-
technique for solving this problem is the addition of a tions. There is one concern to duty-cycle current
time delay to the foldback portion of the circuit in limiting, however, in that the circuit has to be de-
order to let the supply start with constant current signed with a knowledge of the maximum output ca-
limiting and then apply foldback if a fault lasts longer pacitance expected. Too much capacity could
than the expected startup transient prevent startup under load unless the minimum on-
time is set appropriately.
Duty-Cycle Current Limiting - Another method for
providing current limiting with low internal power Pulse-by-Pulse Current Limiting - Pulse-width
dissipation is to combine constant current limiting modulating controllers have long used a technique to
with a low duty-cycle interruption circuit as shown in limit overload by limiting time instead of current
Figure 47. [21] The idea here is that under a fault The block diagram of Figure 49 provides an illustra-
condition the controller will allow conduction for tion of this technique. The oscillator initiates each
only a short interval and then shut down for substan- output pulse by resetting the PWM latch. Normally,
tially longer time (typically 3% on and 97% off). the pulse width is then determined by the ramp wave-
Under these conditions, even though the peak power form rising above the voltage amplifier's output and
may be high, the average is a fraction of that value. thus setting the PWM latch and terminating the out-
A diagram showing a typical implementation of put pulse. However, through a parallel path to the set
this type of fault management is shown in Figure 48 terminal of the PWM latch, the output pulse can be
where it is applied to a linear voltage regulator. Note terminated prematurely by an overcurrent signal from
that the current sense signal is used in both an analog the current sense comparator. This technique can be
loop to regulate the fault current, and in a digital form very effective and, of course, very easy to implement
to activate the timer. In addition to power supplies, with switching power supplies, however two consid-

Texas Instruments 28 SLUP114

 only Vout. Therefore, a falling output voltage from
an overload can diminish the ability to completely
"reset" the inductor with each pulse and the net result
is an increasing current with increasing overload.
Preventing this usually means insuring that the in-
ductor's on-time volt-seconds stays less than the volt-
seconds for reset during the off-time. While this
would appear difficult with a direct short circuit on
the output, the fact that the output rectifier voltage is
still present may provide adequate reset voltage to
prevent saturation - at least for power supplies with
Figure 49 - Pulse-by-Pulse current limiting with a low regulated output voltages. A more positive solu-
PWM controlled switching regulator. tion is to increase the off-time as the supply's output
voltage falls (due to an overload) by lowering the
erations must be kept in mind: First, since there is no switching frequency in a way that insures both ade-
analog current limiting, only residual circuit induc- quate reset time for the transformer and a low aver-
tance limits the rise in current from a hard short dur- age current to the fault. This technique is called
ing the delay it takes for the current loop to react, and "Frequency Foldback" and is implemented in the
secondly, in a continuous fault condition, the power UCC3884 Advanced Current-Mode Controller.
stage will continue pulsing although, hopefully, with
very narrow pulses.
V,N

I'
Current-Mode Peak Current Limiting - Pulse-by- Your

pulse limiting is even easier with peak current-mode

control, as shown in Figure 50, because a clamp on
1t~1 ¥)
MIN PULSE WIDTH
the output of the voltage error amplifier is all that is LIMITED BY
TURN·OFF DELAY
needed to limit the maximum current in the sensing
TON VIN STOFF N Vo
circuit. The basic operation of this topology is to
terminate each power pulse when the current builds
Res
to a level established by the error amplifier and
clamping the amplifier's swing therefore clamps the
peak current. There is a distinction which needs to be Figure 50 - Current-mode control makes peak current
made here in that with peak current-mode control, it limiting easily accomplished.
is the peak inductor current which is controlled,
while the load sees the average value, and there can Over-Current Shutdown
be a substantial difference between these levels - An alternate protection strategy to continuous
particularly as the pulse width is reduced to a small operation in a fault mode with the current (hopefully)
value. Since each pulse starts afresh by turning on under control, is just to shut down the power supply
the power switch, and that switch can't begin to tum when a fault is sensed. Within this strategy, there are
off until the current loop has sensed its peak thresh- still at least two options: Shutdown followed by an
old, the delay time of the tum-off function will result automatic restart, or latch off with the requirement
in a minimum controllable pulse width. This can im- for a manual restart from an operator by, for example,
pact the protection function by allowing continuous recycling input power. Shutting down the supply -
narrow pulses to "walk" the inductor up the B-H regardless of whether it is temporary or permanent -
curve toward saturation. This occurs because induc- still has many possibilities for its implementation.
tor current in the typical step-down power supply
ramps up according to (Vin - Vout), and down with

Texas Instruments 29 SLUP114

Soft-Start Hiccup Protection - Most switching sup- Figure 51 where the action is defined by the control
plies - and some linears as well - have in their control circuit power source. R(start) is made so large that,
circuitry a startup time constant (relatively long in while current through it will start the circuit, it is in-
comparison to the switching frequency) which can be adequate to maintain continuous operation, and that
exercised to recycle in the event of a fault. The pri- energy is derived from a secondary winding on the
mary purpose of this time constant is to allow a slow output inductor. The polarity is such that this auxil-
turn-on (soft-start) to minimize startup transients, iary winding delivers Vout (times the turns ratio) to
however, it obviously will also be useful in startup supply Vcc to the control circuit so, when Vo is
after a fault shutdown. The usual approach is to pro- shorted, Vcc will fall, eventually forcing the control-
vide two levels of current sensing by monitoring the ler to stop, at which time a restart is initiated. The
current waveform with two comparators with differ- hiccup time constant is then established by the energy
ent thresholds. The lower threshold invokes pulse- storage capacitor, Cc.
by-pulse current limiting allowing the circuit to con-
tinue to operate through a minor overload. The sec- Delayed Shutdown - A problem with using two dif-
ond comparator, with a higher threshold, sets a latch ferent current thresholds to make the decision as to
which discharges the soft-start capacitor and thereby whether the circuit goes into current limiting or a
reduces the output pulse width to zero. Resetting the shutdown mode, is that a fast fault transient may go
latch then allows the circuit to initiate another slow through both thresholds before the controller can re-
startup. If the system is set up to automatically reset act. One technique used to minimize what might be
the latch when the soft-start capacitor is fully dis- nuisance shutdowns is to provide a time delay before
charged, the operation is called "hiccup" protection allowing the higher threshold to initiate shutdown,
because, with a continuous fault, the circuit will al- however that can cause concerns as to how fast the
low a short burst of high-frequency pulses as the circuit will react to a major short. The control cir-
system tries to start, separated by relatively long pe- cuitry developed for the UCC3570 attempts to solve
riods while the soft-start capacitor is being dis- this issue by providing three levels of protection.
charged and then slowly recharged. Although this (See Figure 52) The first is pulse-by-pulse current
will continue as long as the fault is present, the aver- limiting which occurs at a threshold of 200mV and
age power should be much less than with pulse-by- merely reduces the normal pulse width under over-
pulse current limiting. load. The second is a shutdown signal but at a sig-
A simple implementation of hiccup fault protec- nificantly higher threshold - 600mV. This will react
tion is possible with any PWM control circuit which quickly in the event of a serious fault. The third level
has a low-current startup mode. This is shown in

CUAAENT L1Iot IT
INPUT D:;:C'-1-_ _ _ _ _-,

R START
SOH START
OISCliARGE

Vee

Res
UCC3 57 0
~-------------- ----- ---------------

Figure 51 - Using energy from the output inductor to Figure 52 - The UCC3570 provides three levels of over·
supply the control circuit can automatically provide hic- current protection.
cup fault protection.

Texas Instruments 30 SLUP114

 ~ Parasitic Capacitance
Leading-Edge Filter
Diode Recovery
c Sign~icant Distortion for
~ I ~ } Magnetising Current Narrow Pulses
o
U"
<l>
VJ
-
~ :::. : ~ Output Inductor Current
Load Current
C
<l>
(f) '- - - - - ~ Eddy Cu rrent
Leading-Edge Btanking
Power Pulse Width
Current Sense is "Blind"
during blanking
Figure 53 - A typical current waveform on the pri·
mary side of an isolated power supply.

is to count the current limiting signals to allow a fi-

Figure 54 - Two methods for combating the leading·
nite time in current limiting, and then shut down. edge current waveform spike, and the problems associ·
This will allow the supply to ride through some pe- ated with each.
riod of overload, but not permit continuous operation
54 shows two commonly used techniques. However,
in a fault condition.
as noted in the illustrations, both of these must be
CURRENT CONTROL applied with some care as too much low-pass filter-
Sensed Current Waveform ing will distort the waveform while leading-edge
Increasingly, current information in a power sup- blanking allows for an uncontrollable minimum pulse
ply is being used for more than merely fault protec- width.
tion. The most obvious example is in the widespread Current Control Applications
use of current-mode control as the overall control Current-Mode Control - Figures 55 and 56 are in-
algorithm for the normal operation of the power sup- cluded merely to illustrate the simplified architectural
ply. In this mode, more information than just the differences between peak and average current-mode
peak value of the current is required - the complete control. In peak mode, the waveforms are very im-
waveform is important and waveform fidelity can portant for reasons described above. With average
become an issue. When sensing current, it is impor- current-mode control, the added gain of the current
tant to understand all that is contained within the amplifier and the integration, or averaging, of the
waveform observed. Figure 53 attempts to illustrate
the components of a primary-side switched current as PEAK CURRENT-MODE CONTROL
the sum of the output inductor current, and the power YO

transformer's magnetizing current and eddy current

losses. (The actual load current is the average of the
inductor ripple current.) When implementing peak
current-mode control, it is helpful to know that the
transformer's magnetizing current increases the slope
of the current signal, and thus contributes to the nor-
mally-required slope compensation.
Another important but not so beneficial compo-
nent of the current waveform is the leading-edge
spike caused by parasitic capacitance in the power
switch, interwinding capacitance in the power trans-
former, and recovery current from the output rectifi- Figure 55 - Peak current-mode control architecture nor-
ers. It is easy to see that this spike needs to be mally senses switch current.
isolated from any fault sensing circuitry and Figure

Texas Instruments 31 SLUP114

current signal helps to make this topology substan-
tially more forgiving. Note that peak CIM senses
AVERAGE CURRENT-MODE CONTROL
switch current, allowing the easy application of cur-
rent sensing transformers, while average CIM looks
at inductor current and thus needs a DC current
sensing scheme. [22]

Power Factor Correction - In a typical high power

factor boost preregulator, as shown in Figure 57, cur-
rent sensing performs several important functions.
The high frequency ripple waveform provides for
stable operation of the PWM controller, the average
VCA -.A-.A...A.
low frequency signal insures that the input waveform y, V V V v
follow that of the AC line voltage, and, although not
shown in the figure, peak current control also pro-
vides protection for the power switch. In addition to Figure 56 - Average current-mode control adds a
current amplifier and needs DC current sensing.
sensing steady-state current, the circuit must also

C=2POIlF

Rcs
C5
RS

VAC

R1 R4
WAVE
M
FORM U
L...-----+--------l B L ICP GATE
T RAMP DRIVE
C osc- VV1./
I

C1 1_ C2 _ R3 I
: ICP=~B
1_ IL ____________________ _ UC3854 .1
_ _________

Figure 57 - Average current-mode gives the best performance in a PFC controller.

Texas Instruments 32 SLUP114

 more thorough presentation. For all these contribu-
tions the author - and, one hopes, our readers - are
sincerely grateful.

BIBLIOGRAPHY
[1] Spaziani, Larry, "Using Copper PCB Etch for
Low Value Resistance", Unitrode Design Note
DN-71.

[2] MIL-STD-275E, "Printed Wiring for Electronic

Circuits", DOD Publication, Dec 31, 1984.

[3] Pacholok, D.R., "Novel Current-Sense Tech-

Figure 58 . Load sharing circuitry to allow multiple nique Boosts Converter Efficiency", PCIM
power modules to be paralleled. Magazine, July, 1990.

follow the zero crossings where the PWM command [4] Ray, W.F. and Davis, R.M., "Wideband AC Cur-
attempts to go to 100% duty cycle. [23] rent Transducer Measures High Current", PCIM
Magazine, September, 1994.
Load Sharin~ - Figure 58 illustrates a simplified
approach to adding load sharing to a power supply's [5] Bertsche, K.J., "A Modified Rogowski Coil for
control to allow reliable paralleling of multiple power Measurements of Hybrid Permanent Magnets",
modules. In this circuit, the DC output current is FERMILAB Conference Proceedings, 1996.
sensed and a voltage proportional to that value sent
[6] Waters, C., "Current Transformers Provide Ac·
out on a share bus to other modules intended to feed
curate, Isolated Measurements", PCIM Maga·
the same load. The voltage on the share bus is then zine, December, 1986.
used to adjust the reference voltage level in each
module that individual outputs such that they each [7] Billings, K.H., "Switchmode Power Supply
supply an equal contribution to the total load current. Handbook", McGraw-Hill, Inc., 1989.
[21]
[8] Williams, B.W., "Power Electronics", McGraw·
ACKNOWLEDGMENTS Hill, Inc., 1992.
Significant contributions to this topic were made
by many individuals and the editor would specifically [9] Gytri, A. and Undeland, T.M., "Low Cost DC
wish to recognize the participation of Bill Andreycak, and AC Current Transducer with Isolation and
Jim Noon, and Larry Spaziani of the Unitrode Appli- Good Linearity", PCI Conference Proceedings,
cation Engineering Department for the material March, 1982.
which they provided. Helpful references and a more
[10] Storm, H.F., "Magnetic Amplifiers", Wiley and
complete understanding of the magnetic structures Sons, 1955.
were the result of conversations with Lloyd Dixon of
Unitrode and Rudy Severns of Springtime Enter- [11] Severns, R., "Improving and Simplifying High
prises. In addition, Sam Davis, Editor of PCIM Frequency DC Current Sensors", IEEE APEC
Magazine, contributed several references from his Conference, 1986.
archives which were appropriate to this subject. Fi-
nally, the cooperation from the many venders who [12] Houldsworth, J.A., "Purpose-Designed Ferrite
Toroids for Isolated Current Measurements in
supply products applicable to current sensing, and
Power Electronic Equipment", Signetics Appli-
shared their information for this writing, has also cation Note.
helped to provide a much broader prospective and a

Texas Instruments 33 SLUP114

[13] Emerald, P. and Gilbert, J., "Introduction, The- [19] Andreycak, Bill, "Phase-Shifted, Zero-Voltage
ory, and Fundamentals of Hall Effect Sensors Transition Design Considerations and the
and Transducers", PCIM Conference, Septem- UC3875 PWM Controller", Unitrode Applica-
ber, 1995. tion Note U-136.

[14] Rippel, W.E., "A High Performance, Integrable [20] Noon, J.P., "Analysis and Design of a 250kHz,
Current Sensor", PCI Conference Proceedings, 500W Power Factor Correction Circuit Employ-
September, 1987. ing Zero Voltage Transitions", Unitrode Appli-
cation Note U-153.
[15] Drafts, Bill, "Current Sensing Solutions for
Power Conversion and Intelligent Motion Appli- [21] Mammano and Rodovski, "A New Linear
cations", PCIM Magazine, September, 1995. Regulator Features Switch-Mode Over Current
Protection", Unitrode Application Note U-116.
[16] Carrasco, J.A., et aI, "Improved Magnetic Cou-
pled Current Sensing Techniques", IEEE APEC [22] Dixon, Lloyd, "Average Current Mode Control
Conference, 1996. of Switching Power Supplies", Unitrode Appli-
cation Note U-140.
[17] Clemente, Ishii, and Young, "An Introduction to
the HEXSense Current Sensing Device", Inter- [23] Todd, Phil, "UC3854 Controlled Power Factor
national Rectifier Application Note AN-959. Correction Circuit Design", Unitrode Applica-
tion Note U-134.
[18] O'Connor, J.A., "The UC3848 Average Current
Mode Controller Squeezes Maximum Perform- [24] Jordan, Mark, "UC3907 Load Share IC Simpli-
ance from Single Switch Converters", Unitrode fies Parallel Power Supply Design", Unitrode
Application Note U-135. Application Note U-129.

Texas Instruments 34 SLUP114

TI Worldwide Technical Support

Internet
TI Semiconductor Product Information Center
Home Page
[Link]

TI E2E™ Community Home Page

[Link]

Product Information Centers

Americas Phone +1(972) 644-5580 Asia
Phone
Brazil Phone 0800-891-2616
International +91-80-41381665
Mexico Phone 0800-670-7544 Domestic Toll-Free Number
Note: Toll-free numbers do not support
Fax +1(972) 927-6377 mobile and IP phones.
Internet/Email [Link]/sc/pic/[Link] Australia 1-800-999-084
China 800-820-8682
Europe, Middle East, and Africa Hong Kong 800-96-5941
Phone India 1-800-425-7888
European Free Call 00800-ASK-TEXAS Indonesia 001-803-8861-1006
(00800 275 83927)
Korea 080-551-2804
International +49 (0) 8161 80 2121
Malaysia 1-800-80-3973
Russian Support +7 (4) 95 98 10 701
New Zealand 0800-446-934
Philippines 1-800-765-7404
Note: The European Free Call (Toll Free) number is not active
in all countries. If you have technical difficulty calling the free Singapore 800-886-1028
call number, please use the international number above. Taiwan 0800-006800
Thailand 001-800-886-0010
Fax +(49) (0) 8161 80 2045 Fax +8621-23073686
Internet [Link]/sc/pic/[Link] Email tiasia@[Link] or ti-china@[Link]
Direct Email asktexas@[Link] Internet [Link]/sc/pic/[Link]

Japan Important Notice: The products and services of Texas Instruments

Incorporated and its subsidiaries described herein are sold subject to TI’s
Phone Domestic 0120-92-3326 standard terms and conditions of sale. Customers are advised to obtain the
most current and complete information about TI products and services before
Fax International +81-3-3344-5317 placing orders. TI assumes no liability for applications assistance, customer’s
Domestic 0120-81-0036 applications or product designs, software performance, or infringement of
patents. The publication of information regarding any other company’s products
Internet/Email International [Link]/sc/pic/[Link] or services does not constitute TI’s approval, warranty or endorsement thereof.

Domestic [Link]/pic A122010

E2E and Unitrode are trademarks of Texas Instruments. All other trademarks are the
property of their respective owners.

SLUP114
 IMPORTANT NOTICE FOR TI DESIGN INFORMATION AND RESOURCES

Texas Instruments Incorporated (‘TI”) technical, application or other design advice, services or information, including, but not limited to,
reference designs and materials relating to evaluation modules, (collectively, “TI Resources”) are intended to assist designers who are
developing applications that incorporate TI products; by downloading, accessing or using any particular TI Resource in any way, you
(individually or, if you are acting on behalf of a company, your company) agree to use it solely for this purpose and subject to the terms of
this Notice.
TI’s provision of TI Resources does not expand or otherwise alter TI’s applicable published warranties or warranty disclaimers for TI
products, and no additional obligations or liabilities arise from TI providing such TI Resources. TI reserves the right to make corrections,
enhancements, improvements and other changes to its TI Resources.
You understand and agree that you remain responsible for using your independent analysis, evaluation and judgment in designing your
applications and that you have full and exclusive responsibility to assure the safety of your applications and compliance of your applications
(and of all TI products used in or for your applications) with all applicable regulations, laws and other applicable requirements. You
represent that, with respect to your applications, you have all the necessary expertise to create and implement safeguards that (1)
anticipate dangerous consequences of failures, (2) monitor failures and their consequences, and (3) lessen the likelihood of failures that
might cause harm and take appropriate actions. You agree that prior to using or distributing any applications that include TI products, you
will thoroughly test such applications and the functionality of such TI products as used in such applications. TI has not conducted any
testing other than that specifically described in the published documentation for a particular TI Resource.
You are authorized to use, copy and modify any individual TI Resource only in connection with the development of applications that include
the TI product(s) identified in such TI Resource. NO OTHER LICENSE, EXPRESS OR IMPLIED, BY ESTOPPEL OR OTHERWISE TO
ANY OTHER TI INTELLECTUAL PROPERTY RIGHT, AND NO LICENSE TO ANY TECHNOLOGY OR INTELLECTUAL PROPERTY
RIGHT OF TI OR ANY THIRD PARTY IS GRANTED HEREIN, including but not limited to any patent right, copyright, mask work right, or
other intellectual property right relating to any combination, machine, or process in which TI products or services are used. Information
regarding or referencing third-party products or services does not constitute a license to use such products or services, or a warranty or
endorsement thereof. Use of TI Resources may require a license from a third party under the patents or other intellectual property of the
third party, or a license from TI under the patents or other intellectual property of TI.
TI RESOURCES ARE PROVIDED “AS IS” AND WITH ALL FAULTS. TI DISCLAIMS ALL OTHER WARRANTIES OR
REPRESENTATIONS, EXPRESS OR IMPLIED, REGARDING TI RESOURCES OR USE THEREOF, INCLUDING BUT NOT LIMITED TO
ACCURACY OR COMPLETENESS, TITLE, ANY EPIDEMIC FAILURE WARRANTY AND ANY IMPLIED WARRANTIES OF
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, AND NON-INFRINGEMENT OF ANY THIRD PARTY INTELLECTUAL
PROPERTY RIGHTS.
TI SHALL NOT BE LIABLE FOR AND SHALL NOT DEFEND OR INDEMNIFY YOU AGAINST ANY CLAIM, INCLUDING BUT NOT
LIMITED TO ANY INFRINGEMENT CLAIM THAT RELATES TO OR IS BASED ON ANY COMBINATION OF PRODUCTS EVEN IF
DESCRIBED IN TI RESOURCES OR OTHERWISE. IN NO EVENT SHALL TI BE LIABLE FOR ANY ACTUAL, DIRECT, SPECIAL,
COLLATERAL, INDIRECT, PUNITIVE, INCIDENTAL, CONSEQUENTIAL OR EXEMPLARY DAMAGES IN CONNECTION WITH OR
ARISING OUT OF TI RESOURCES OR USE THEREOF, AND REGARDLESS OF WHETHER TI HAS BEEN ADVISED OF THE
POSSIBILITY OF SUCH DAMAGES.
You agree to fully indemnify TI and its representatives against any damages, costs, losses, and/or liabilities arising out of your non-
compliance with the terms and provisions of this Notice.
This Notice applies to TI Resources. Additional terms apply to the use and purchase of certain types of materials, TI products and services.
These include; without limitation, TI’s standard terms for semiconductor products [Link] evaluation
modules, and samples ([Link]

Mailing Address: Texas Instruments, Post Office Box 655303, Dallas, Texas 75265
Copyright © 2017, Texas Instruments Incorporated