IFDnewVol2 2.6102cdcdddad1

Ideas
Volume 2
for Design
Vol. 2, No. 1
Generate Realistic Models For LED

Current Versus Voltage
Designers of lighting drivers require a good approach to modeling LED
device performance in the first quadrant of the I/V plane. For engineering
applications, there are three approaches to consider: the linear, bias-voltage
plus resistor equivalent, a quadratic equation, and an exponential equation.
PWM-Based Analog Calculator Provides Four-

Quadrant Multiplication, Division
You can build an analog calculator using a pulse-width modulator (PWM)
to perform accurate four-quadrant multiplication and division. While this
approach won’t help you ace any math tests, it demonstrates some useful
sub-circuits that extend the functionality of the LTC6992 TimerBloxvoltage-
controlled PWM.
Driven Shield Enables Large-Area

Capacitive Sensor
This low-cost, large-area capacitive sensor is the wide-scale complement to
the ubiquitous, finger-operated touchscreen sensor and can detect presence
through plywood or wallboard.
Do you have Electronic Design is always on the lookout for new ideas. Do
an Idea for you have one? Our Ideas for Design articles are short and to
Design for the point, often with a single figure or program listing to help
Electronic explain the idea. If you would like to submit one, you can
Design? check out the details at
https://www.electronicdesign.com/contribute.
We look forward to seeing your ideas.
for Design
Ideas
KENG C. WU | SWITCHING POWER INC., [email protected]
Generate Realistic Models

For LED Current Versus Voltage
LEDS ARE NONLINEAR devices with I/V curves that resemble the THE QUADRATIC MODEL
curves of rectifier diodes. Designers of lighting drivers require The curve of Figure 1 has a concave portion that resembles
a good approach to modeling the LED device performance in one arm of a parabolic curve. It can be expressed with a qua-
the first quadrant of the I/V plane. dratic equation of the form a2vF2 + a1vF + a0. The key is the
For engineering applications, there are three approaches to determination of three coefficients: a2, a1, and a0.
consider: the linear, bias-voltage plus resistor equivalent; a To determine these coefficients, you can use the well-known
quadratic equation; and an exponential equation. Let’s use the Least Square Polynomial Curve-fitting algorithm of linear
three techniques with Cree’s XLamp XM-L LED as the LED to algebra. To do this, select more than a dozen data points from
be modeled (Fig. 1). Figure 1 (15 in this case) and place them in two matrixes.
One is a 3x15 rectangular matrix C (Equation 2). The other
THE LINEAR MODEL is a 1x15 column vector iF (Equation 3) where vFj , for j = 0 to
This model is the simplest, but the most inaccurate. It con- 14, are the corresponding LED forward voltages at the selected
sists of a dc bias voltage plus a resistor (Fig. 2a). The equiva- forward currents given in vector iF. The three coefficients are
lent mathematical form is given by: then given by a column vector of:
a = (C · CT)–1C · iFT (4)
(1)
Using software tools such as Matlab from the Mathworks or
MathCAD from Mathsoft, you can easily compute Equation 4.
You select two data points (vF, iF) in coordinate form (i.e., For this example, we obtain:
3.28 V, 2.6 A and 2.71 V, 0.2 A). Then, R is evaluated and
yields 0.2375. The selection also implies a = 0.2 and b = 2.71. 24.849
With all three parameters (a, b, R) properly assigned, the linear a = –20.093 (5)
model yields Figure 2b. Evidently, the linear model offers the 4.058
advantage of simplicity, but suffers in accuracy.
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
C := vF0 vF1 vF2 vF3 vF4 vF5 vF6 vF7 vF8 vF9 vF10 vF11 vF12 vF13 vF14 (2)
(vF0)2 (vF1)2 (vF2)2 (vF3)2 (vF4)2 (vF5)2 (vF6)2 (vF7)2 (vF8)2 (vF9)2 (vF10)2 (vF11)2 (vF12)2 (vF13)2 (vF14)2
iF := 200 m 400 m 600 m 800 m 1000 m 1200 m 1400 m 1600 m 1800 m 2000 m 2200 m 2400 m 2600 m 2800 m 3000 m (3)
3000 3
2800 2.8
2600 iF iF 2.6
2400 2.4
2.2
Forward current (A)
2200 VF
Forward current (mA)
2000 2
1800 R 1.8
VF 1.6
1600 1.4
1400 1.2
1200 1
1000 0.8
800 0.6
600 VBias 0.4
400 0.2
200 0
0 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6
2.00 2.20 2.40 2.60 2.80 3.00 3.20 3.40 3.60 a. b. Forward voltage (V)
Forward voltage (V)
2. The linear model (terminal view) is simplest and easiest to understand,
1. Modeling the I/V curve of the Cree XLamp XM-L LED with increasing but also the least accurate (a). The linear model results in this “curve,”
accuracy is the objective of this analysis. which clearly has some shortcomings (b)
ELECTRONIC DESIGN
IdeasForDesign
3 3
2.8 2.8
Note that the first element of vector 2.6 2.6
2.4 2.4
a is coefficient a0, etc., which leads to 2.2 2.2
Forward current (A)
Forward current (A)

the quadratic model of Figure 3. This 2 2
1.8 1.8
approach yields significant improve- 1.6 1.6
ments in the model across the whole 1.4 1.4
1.2 1.2
operating range of the LED. 1 1
0.8 0.8
0.6 0.6
The exponenTial Model 0.4 0.4
Figure 3 appears to bend more near 0.2 0.2
0 0
the low current, which is where the expo- 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6
nential model may offer further improve- Forward voltage (V) Forward voltage (V)
ment. This model comes in the form of 3. The curve based on a quadratic model curve 4. The exponential model further refines the
aeb·vF + c with three unknown parameters is much closer to the reality of the I/V curve for quadratic model and provides a better approxi-
a, b, and c to be determined. Again, soft- this LED. mation in the low-current region of the LED.
ware tools such as MathCAD help you
find those parameters numerically. Overall, the best fit near the low-cur- prediction from any given model, since
Under the specialized regression sec- rent region may be found between the almost all such analytical efforts are an
tion, an exponential regression statement quadratic and the exponential models. It attempt to represent the complexities of
expfit ( vF, iF, vg) can find all three param- is unrealistic to expect a single, perfect nature.
eters, given a data set in vectors and initial
guess value vg, also a column vector. For
this example, a = 9.66 · 10–3, b = 1.818, KenG C. Wu has a BS from Chiaotung university, Taiwan, and an MS from northwestern
and c = –1.157 are obtained. This results university, evanston, ill. he has published four books and holds seven u.S. patents.
in the exponential model of Figure 4.
ElEctronic DEsign Go To elecTronicdesiGn.com

for Design
MARK THOREN | LINEAR TECHNOLOGY CORP.
PWM-Based Analog Calculator Provides

[email protected]
Ideas
Four-Quadrant Multiplication, Division
YOU CAN BUILD an analog calculator +5 V
4 6 VCC = 5 V
using a pulse-width modulator (PWM) In Out
VEE = ground
to perform accurate four-quadrant LT6654-1.25
0.1
multiplication and division. While this µF
G G 1 1/3 74HC4053
approach won’t help you ace any math 2
1 2
tests, it demonstrates some useful sub- 3
circuits that extend the functionality +5 V +5 V
GND
of the LTC6992 TimerBlox voltage- V : REF_LO to REF_HI = GND
5
IN 3 5
controlled PWM. 0:100% duty cycle +
1 1 V+
MOD
The LTC6992-1 translates a 0-1 V 4
– LTC2054 +5 V 4 6 PWM
2 DIV OUT
input at the MOD pin into an output 3 output
GND SET
with a 0% to 100% duty cycle at a fre- GND
quency of 3.81 Hz to 1 MHz. A resis- 2 LTC6992-1

50k
tor at the SET pin and a resistor divider 0.1
µF GND
at the DIV pin control this frequency. 1M
In some applications, the LTC6992
R1
will be in the feed-forward path of 100k GND GND
a closed-loop control system (as in
a motor-speed controller), so its 1%
typical linearity provides consistent
overall loop performance. 1. This basic linearized PWM generator, without an external feedback mechanism, can still provide
Figure 1 shows a basic linearized accuracy of 0.1%.
PWM generator for applications where
accurate PWM is required without an external feed-
back mechanism. This circuit easily achieves 0.1% DIVCODE PROGRAMMING
PWM accuracy. The output of the LTC6992 con- DIVCODE POL NDIV Recommended fOut R1 (kΩ) R2 (kΩ)
trols one section of a 74HC4053 triple single-pole 0 0 1 62.5 kHz to 1 MHz Open Short
double-throw (SPDT) analog switch whose output
1 0 4 15.63 to 250 kHz 976 102
is switched between ground and an LT6654-1.25
reference. An integrator compares this signal to the 2 0 16 3.906 to 62.5 kHz 976 182
control input. The output duty cycle will settle on 3 0 64 976.6 Hz to 15.63 kHz 1000 280
a value that equals the fraction of the 1.25-V refer- 4 0 256 244.1 Hz to 3.906 kHz 1000 392
ence that’s present at the input. The term “fraction” 5 0 1024 61.04 to 976.6 Hz 1000 523
implies that this circuit performs division, as the
6 0 4096 15.26 to 244.1 Hz 1000 681
output PWM duty cycle is VIn/VRef.
Figure 2 extends this concept, with X as the input 7 0 16384 3.815 to 61.04 Hz 1000 887
(numerator) and Y as the reference (denominator). 8 1 16384 3.815 to 61.04 Hz 887 1000
An LT1991 configured in a gain of –1 provides a 9 1 4096 15.26 to 244.1 Hz 681 1000
precise negative copy of Y, extending operation to
10 1 1024 61.04 to 976.6 Hz 523 1000
four quadrants (positive and negative X and Y),
with duty cycle = 50% × [1 + (X/|Y|)]. 11 1 256 244.1 Hz to 3.906 kHz 392 1000
As with any physical realization of division, a 12 1 64 976.6 Hz to 15.63 kHz 280 1000
zero value for the denominator (Y) will produce an 13 1 16 3.906 to 62.5 kHz 182 976
undefined output. A negative voltage applied to the 14 1 4 15.63 to 250 kHz 102 976
Y input inverts the polarity of the feedback signal
15 1 1 62.5 kHz to 1 MHz Short Open
to the integrator, which requires another inversion
ELECTRONIC DESIGN GO TO ELECTRONICDESIGN.COM

IdeasForDesign
10 somewhere in the loop to ensure feed-

Z back is negative.
4
– 1 6 1 The DIV pin also can invert the
LT1991 2 10k
3 + PWM polarity (a 0- to 1-V input =
3
Output = X/Y × Z 100% to 0% duty-cycle output), in
1, 2, 3 0.1
Y GND µF
(reference)
addition to selecting one of eight NDiv
values to set the frequency. The NDiv
GND
magnitudes are mirrored around VCC/2,
10 where swapping values in the resistor
Buffered
4 +5 V PWM
divider inverts the transfer function
– 1 6 1 1
LT1991 2 2
output while maintaining the same divider
3 +
3 3 value (see the table).
1, 2, 3 An LT1671 comparator detects the
GND GND
polarity of the Y input and sets the
3 4
polarity by switching the divider poten-
tiometer’s excitation accordingly, main-
LT1671 7404D
3 1M taining correct operation. Note that a
+
1 1 2 10-turn potentiometer works well for
2 experimentation. You could replace
–
7404D
“Flip-mode” circuit it with a fixed resistor once you have
GND
Duty cycle = 50% × (1 + X/|Y|) selected the desired NDiv.
+5 V
The “Z” input is multiplied by the
5
X
3
+
X/Y quotient by supplying the inputs
1 1 V+
MOD to another switch with Z and –Z. (Once
4 4 6 Supply connections
–
LTC2054
3
DIV OUT
LT1991: VEE = –5 V, VCC = +5 V
again, an LT1991 provides precision
SET inversion.) This is a “pulse width/pulse
GND LT1671: V– = –5 V, V+ = +5 V
50k 2 LTC6992-1 74HC04: VCC = +5 V height” multiplier, also with four-quad-
0.1 1M LTC2054HV: V+ = +5 V, V– = –5 V rant operation.
µF
GND 74HC4053: VEE = –5 V, VCC = +5 V Figure 3 shows the absolute error
1 of the circuit at a 1.5-kHz frequency,
GND 2
100k = 1/3 74HC4053
3 sweeping X from –Y to +Y for values
of Y from –3 V to +3 V, while holding
2. This enhancement to the circuit of Figure 1 extends the analog multiplication/division to all four Z constant at 5 V. Even with Y at 0.5
analog signal quadrants. V (where error sources are more sig-
nificant), the worst-case error is about
0.03 0.6% and rapidly improves with larger
values of Y. Error sources include the
0.04% error of the LT1991, mismatch
0.02
in switch resistance between its two
positions compared with the resistance
0.01
of the downstream filter’s resistance,
and the response of the LT1991 outputs
to switching transients, whose effect
will vary with PWM frequency.
Error (V)
–0.01 MARK THOREN is

the applications engi-
neering manager for
Y = –0.5 V Y = –2 V Y=3V Y = 1.5 V
–0.02 mixed-signal products
Y = –1 V Y = –2.5 V Y = 2.5 V Y=1V
at Linear Technology.
Y = –1.5 V Y=0V Y=2V Y = 0.5 V He has a BS in agri-
–0.03
cultural/mechanical
–100 –90 –80 –70 –60 –50 –40 –30 –20 –10 0 10 20 30 40 50 60 70 80 90 100
engineering and an
X input (% of Y)
MS in electrical engineering, both from
3. This plot of absolute error shows better than 0.1% for large values of Y. the University of Maine.
ELECTRONIC DESIGN
IdeasForDesign
DEV GUALTIERI | TIKALON LLC, LEDGEWOOD, N.J. [email protected]
Driven Shield Enables

Large-Area Capacitive Sensor
CAPACITIVE SENSORS ARE common in today’s con- R2 +5
2.2k
sumer electronics, since they are used in many
touchscreen applications. Most of the circuits 4 8
Frequency
R5
output
for such applications are designed for small-area 7
10k
capacitors and finger contact operation.
+5
This circuit is designed for large-area touch R1 6 IC1 3
220k 555 timer R6
plates that can be activated at a distance and hid- R3 – 47
2 Driven
den behind walls or inside other structures. It uses 10k IC2
shield
+
the “driven shield” concept that was popular in the
C1 Ground
early days of audio engineering, when high-imped- 220 pF 1 R4
1M IC2 = TLV2461
ance sound sources such as crystal microphones
Capacitor
were used. Shielded cables are needed to connect plate
these devices, but the cable capacitance limits the
high-frequency response. 1. In this large-area capacitive sensor, C1 and any parallel capacitance set the frequency of
The solution to this problem is to go back to the the 555 astable multivibrator IC1, while IC2 buffers the voltage at the capacitor to drive
physics of a capacitor to find a way to reduce the the shield plate. (The pin-outs for IC1 are for the “N” package.)
cable capacitance. Driving a conductor between the
outer shield and the inner, signal conductor with the same voltage should have a high gain-bandwidth product to allow a faithful
as the signal conductor largely eliminates the capacitance. NASA representation of the signal voltage without phase shift.
used this driven-shield idea in a large-area capacitive sensor two Figure 2 shows the construction of the sensing plate. These
decades ago (John M. Vranish, et al., US Patent No. 5,116,679). are the dimensions used in the author’s tests, but quite a lot of
The circuit in Figure 1 performs the main capacitance- variation is permissible. A practical implementation would be
sensing task, while leaving detection to a microcontroller. IC1 to build the circuitry onto an edge of the ground plane.
is a 555 timer wired as an astable multivibrator with as small The frequency response decreases in a nearly linear fash-
a timing capacitance as possible. With the component values ion with respect to inverse position (Fig. 3). In the author’s
shown, the frequency is slightly above 10 kHz. IC2 buffers the test, hand contact reduced the frequency to 10.47 kHz. Hand
voltage on this capacitor to drive a shield plate. contact through a 0.375-in. (10 mm) thick piece of plywood
For IC2, resistors R3 and R4 decrease the buffered voltage or gypsum wallboard resulted in a frequency change of 10%,
by about 1%. This prevents oscillation that may occur because which is easy to detect. Although frequency-detection ICs are
of the external capacitance. R6 is included since some opera- available, a simpler, less expensive solution is to use a micro-
tional amplifiers have difficulty driving a capacitive load. IC2 controller to detect a change in signal period.
Position (inch)
Double-sided ∞ 2 1 0.67 0.5
5 in. FR4 PCB Capacitor plate
7 in. 7 in. 12.3
Driven shield
Copper 0.1 in. Ground
ground
plane 9 in.
Frequency (kHz)
12.2
2. In the constructed sensor, a double-sided printed-circuit board (PCB) has
one conductor as the signal plate and the other as the driven shield. The
semi-infinite ground plane, which is slightly larger in area than the sensing 12.1
or shield plates, can be another PCB or an aluminum plate.
12
DEV GUALTIERI received his PhD in solid-state
0 0.5 1 1.5 2
science from Syracuse University in 1974. He
Inverse position (inch–1)
now does various computer, electronic, and
embedded systems projects at his consulting 3. The circuit response is a close-to-linear inverse-position relationship
company, Tikalon LLC. between hand position and the sensing plate.

Vol. 2, No. 2
Test Setup Checks Transistors’ hFEs When Tight

Control Is Important
This circuit uses pulsed current to check transistor hFE on an oscilloscope.
The circuit creates the biasing conditions specified by the transistor’s
manufacturer..
Multiple Power Supplies Fortify High-Side

Gate Control
A switching power supply was devised to ensure high-side FET gate control,
particularly in light of a wildly floating high-side rail. The design’s “VeeDrive”
delivers a sharp and muscular square-wave that’s robust enough to drive
several charge pumps.
High-Side Switch Provides Overvoltage Protection

With Only Four Components
Low-voltage microcontrollers often need to control higher-voltage loads via
high-side switching. This simple two-transistor circuit enables that control and
provides over-voltage protection.
Add Short-Circuit Protection, Diagnostics To

Automotive High-Side/Low-Side Driver
Smart switches for automotive applications are costly, may not be readily
available for 24-V operation, and often come in unneeded dual and quad
configurations. By adding a few standard, low-cost components to a basic
driver, you can make a smarter one which includes short-circuit protection
and built-in diagnostics.
you have one? Our Ideas for Design articles are short and to
an Idea for
the point, often with a single figure or program listing to help
Design for
explain the idea. If you would like to submit one, you can
Electronic
check out the details at
Design?
for Design
Ideas
CHEE HUA HOW | TDK MALAYSIA
Test Setup Checks Transistors’ hFEs

[email protected]
When Tight Control Is Important

THE SPREAD OF h FE values in a batch of bench evaluation of hFE may be required codes. Sometimes these measurements
transistors may be wide enough to cause during the system’s development stage. are made using pulsed collector current.
unreliable performance during mass Designers may have to screen samples This idea describes how to make such
production of a system. Consequently, over a spread of several years of date measurements. For example, the QSZ2
Q1 Vtest Vtest R1
transistor (2SB1695) specification states
+10 V R2a 27
2N2907A 12, 0.5 W three biasing requirements on hFE: IC =
0.1 A, VCE = 2 V, and ∆tw = 1 ms. The
+ R9
C8 C7 10k C1 R2b 22 circuit in Figure 1 meets these require-
U1
0.1 µF 3.3 µF R10 0.22 µF LM385DZ Q2 ments for hFE measurement for a QSZ2
1k R13 MPSA63
Q5 100k
pnp transistor.
On 2N2222A R4 +VRC1 Zener U1 and its associated components
On 10k
R12 R11 RC1 form a current regulator that sources 0.1 A
10k 2k 7.521 to the succeeding stage. During the test
–VRC1
V D1 BAT46
period (∆tW in Figure 2), U2’s feedback
–VRB1 +VRB1 CE(DUT) Q4
RB1 2N7000
mechanism trims the base resistance of
C6 270 pF R3 the device under test (DUT) while main-
3k
8.25 X1
DUT C2
R7 taining its VCE at the required 2 V. This is
100k Vtest C5 QSZ2 1 µF R14 Dummy
0.1 µF 100 expressed in Equation 1, in which RDSQ3
LT1077 Q3
VNIV 3 +V 7 is automatically tuned against IB(DUT) and,
+ 499-P44-20
V U2 socket
therefore, hFE and VBE(DUT) variations:
ref 2 6 R5
–– VN2222LL Wells-CTI
R8 4 –V 1k
160k R6 C3
C4 U5
0.1 µF 62k LMV431ACZ 3 µF
D3
D2
+10 V 1N4148 +10 V 1N4148 C9
+10 V C11
10 nF Equation 2 then allows you to deter-
2.2 µF
R16 +
100k QB mine the value of the DUT’s hFE by mon-
S1
C10
0.1 µF
R15 itoring VRB1:
R18 100k T2A T1A VSS
VCC T2B T1B
10k BB QA QA
QB BA
BB U3b Falling edge
U3a AA triggers
AB
QB CD4538 QA scope
CD4538 CDA
+10 V CDB
RST RST
R19
C13 R17 C12
10k
10k 0.1 µF
0.1 µF
Basically, resistor RB1’s value is set as
C16 large as possible to minimize the oscil-
D5 1 nF +10 V C14 lation of IB(DUT) upon transition and to
1N4148 0.1 µF
QB U4b
1. This test circuit provides the biasing condi-
U4a On QB Dummy
R20 100k tions needed to evaluate a transistorÕs hFE
R21 20k HEF4071B
HEF4071B according to the manufacturerÕs criteria. In
QA this example, the device under test is a QSZ2
C15
D4 1 nF (2SB1695), and the values chosen are specific to
1N4148 that device.
ELECTRONIC DESIGN
IdeasForDesign
BB
[Press S1]
t
0
2. The actual test period
QB is during the interval ∆tW,
t during which a 980-µs
0 pulse is generated at QA.
∆td (about 13 µs) ensures
QA there is enough time to
[Trigger]
t make the measurement.
0
On
t
0 ∆tB ∆tW Complete test cycle
∆td
Pre-bias phase: Q1, Q2, and Q4 activated
Testing phase: Q1, Q2, Q3, and DUT activated
Measurement phase: Q1, Q2, Q3, and DUT activated
maximize the signal level to ease VRB1 monitoring. The hFE of

a QSZ2 ranges from 270 to 680, so IB(DUT) falls between 147 5 %PD[ =
(9
&('87 ² 9%(PLQ ) = ( ² ) =
µA and 370 µA. ,%PLQ
Next, assume the worst-case VBE(DUT) variation of 0.5 V to
0.8 V. Then:
Choose RB1 so that it is slightly below RB(min).
The circuitry around U3 forms two cascading one shots that,

when triggered by S1, generate a 220-ms pulse (∆tB ≈ R16 ×
C11) at the QB output. This is followed by a 980-µs pulse, ∆tW
(determined by R15 and C9), at QA. The first pulse pre-biases
and stabilizes the current source while the second one provides
the needed test pulse for the DUT.
During ∆tB, Q4 is activated to sink the 0.1 A of U1, and R3
is selected so the effective voltage drop across R3, D1, and Q4
(and thus VCE(DUT)) is slightly less than 2 V (about 1.7 V in this
example). This forces U2’s output and the DUT to stay “low”
and “off,” respectively.
At the end of ∆tB, ∆tW starts and Q4 is turned off, transfer-
ring the test current to the DUT. U2 then raises VCE(DUT) to 2
V and regulates it.
The high-to-low transition at the QA output after ∆tW trig-
gers the measuring device (in our example an oscilloscope)
(Fig. 3). VRC1 and VRB1, which correspond to IC and IB of the
DUT, are recorded upon QA’s trigger. R21 and C15 create ∆td
(about 13 µs), which ensures enough time to make the mea-
surement before Q1 is gated “off.”
For more, see “Anoop’s Analysis” with the online version of

this article at electronicdesign.com.
Chee hua how, design engineer, specializes in ana-

log and power electronics design. he has an MSee from
3. The falling edge of QA triggers an oscilloscope, which uses VRB1 (trace Multimedia university, Malaysia, and a BSee from Coventry,
A) and VRC1 (trace B) to calculate hFE (trace D). university, the u.K.

IdeasForDesign
Steven P. Hendrix, Pe | Hx EnginEEring LLC [email protected]
Multiple Power Supplies

Fortify High-Side Gate Control
For a recent design project, a robust power supply was needed Logic GND turns off Q612 and disables the low-side drive,
to control the gates of high-side FETs. Optocouplers readily allowing VeeDrive to settle to a steady state just below V12.
translate the logic signals up to the correct control voltages This will save power during those times that don’t require the
to drive the FETs. However, to ensure that the FETs turn on GateDrive supply.
and off cleanly, the design required a solid –15 V relative to a Figure 2 shows one of the four charge pumps using VeeDrive
wildly floating high-side rail. to provide a floating supply that follows a specific floating rail
A simple Zener diode and resistor, possibly with a FET buf- VFLOAT. Because the gate drive must be at least 10 V below
fer to ensure adequate current, might do the job under some V FLOAT, and some drops occur in the diodes, a two-stage
conditions. In certain circumstances, though, the required gate
drive would dip slightly below the lowest other supply rail in V12
the system, which is in this case GND, since there are no nega-
tive supplies. At other times, it could be almost 100 V above Q611
SI2334DS
all other supply voltages. Since the design in question was to VeeDrive
be used in a production scheme, the solution had to stay away Q610
from proprietary, expensive, single-source parts. DMP2160
The design already included a switching supply to effi- Q612
SI2334DS
ciently convert from an incoming supply at approximately 12 VeeEnable
V, down to an internal logic supply at 4 V (Fig. 1). The switch Logic GND
node of that supply, at the junction of U610 pin 3, L610, and Logic GND
D611, swings between the incoming 12-V rail and a diode
D611
drop below ground—approximately –0.5 V. This could pro- CMS03 V4
TP610
vide the feed for a charge pump. However, the 4-V supply U610 L610
can’t be lightly loaded, which would cause insufficient drive at MIC4680 68 µH 1.2 A
2 IN SW 3 V4
this point to support a charge pump. V12
C612
To guarantee a solid drive for the charge pump, power is 220 μF R613 C613 C615
22.6k 15 pF 1 μF
GND
derived directly from the 12-V supply and ground using tran- C610 1 4
SHDN FB + C614
1 μF 1.23 V
sistors Q610 and Q611. Note that these aren’t configured in R612
220 μF
5
a traditional inverter configuration, which could suffer from 10k
shoot-through when both are turned on simultaneously during
the gate’s voltage transition. Logic Logic Logic Logic Logic Logic Logic
GND GND GND GND GND GND GND
Instead, transistors Q610 and Q611 are simple source-fol-
lowers. As the gate voltage rises more than about a volt above 1. The 12-V downswitcher, which is already in the design, provides a
the present value of net VeeDrive shown in Figure 1 at the square-wave to generate VeeDrive from V12 with about a 10-V swing.
source terminals of Q610 and Q611, Q611 turns on, pulling up
VeeDrive so it remains about 1 V below the gate voltage all the VFLOAT
way up. On the falling edge at the switch node, the gate voltage
falls, Q611 is turned solidly off, and Q610 begins to turn on
as the gate voltage falls about 1 V below VeeDrive. Therefore,
VeeDrive follows the gate voltage down to within about 1 V of C123 C124
100 µF 100 µF
Logic GND. C125
D120 1 μF
As a result, VeeDrive provides a sharp and muscular square- D122 D123 18 V 150 V
wave that swings between about 1 V and 11 V (relative to BAS70 BAS70
GateDrive
Logic GND), for a total of 10 V p-p. It’s robust enough to
drive several charge pumps, if the need arises for multiple sup- C121 C122
1 μF 1 μF
plies. In fact, this one square-wave provided all four supplies VeeDrive
required by the design at hand.
Q612 is an optional component. Raising net VeeEnable, the 2. The charge pump generates a robust supply that always tracks 15 V below
gate drive to Q612, to a logic high voltage turns on Q612 and VFLOAT, even if that voltage is below ground. The voltage also can be placed
enables the charge pump action. Lowering VeeEnable back to above VFLOAT by simply reversing the diodes and the polarized capacitors.
ElEctronic DEsign
IdeasForDesign
charge pump is employed. My actual design includes four cop- C121 and C122 ensures that GateDrive maintains a fairly con-
ies of the charge pump shown in Figure 2 to produce GateDrive stant offset from VFLOAT. This will hold true while the flying
relative to each of four different floating supply rails. A single capacitors rapidly charge or discharge to the appropriate levels
VeeDrive drives all four rails. to continue the charge pump action.
Starting with accumulators C123 and C124 discharged, Because the flying capacitors are ceramic (and thus
when VeeDrive goes low (1 V), C121 charges up through C123 unconcerned about polarity), GateDrive will remain 15 V
and the right half of dual diode D122 to approximately VFLOAT below VFLOAT, even if VFLOAT should fall all the way to Logic
– 1.5 V (after accounting for an approximate 0.5-V drop of GND or below. The flying capacitors must, however, have a
the diode). When VeeDrive goes high (11 V), C121 discharges sufficient voltage rating so that the highest value of VFLOAT
through the left half of D122 to VFLOAT – 10.5 V (11 V – the can occur in practice.
0.5-V drop in the diode). As cycling continues, charge contin- The output voltage at GateDrive is very consistent in prac-
ues to be drawn through C123, gradually charging up C123. In tice and nicely matches the theoretical values after accounting
steady state, with no current being drawn off, C123 will charge for the various drops in the feed. There are several ways to
up to 9 V, leaving its negative end at VFLOAT – 9 V. adjust this voltage, though. Stages can be easily added to or
The second-stage charge pump, consisting of C122, D123, removed from the charge pump to produce different multiples
and C124, operates similarly to produce an additional 9-V of the input voltage. If the input voltage is adjustable, finer
offset. As a result, GateDrive is 18 V below VFLOAT. In actual granularity is possible.
practice, of course, some current is being drawn off and there It’s easy to configure GateDrive when it must be above rath-
are more resistive drops, so the actual voltage settles close to er than below VFLOAT. Simply flip the polarized accumulator
15 V. Should the voltage somehow exceed 18 V, Zener D120 capacitors to allow for the new state, and reverse the diodes.
clamps the voltage to no more than 18 V, which guarantees that With those simple changes, the design will produce GateDrive
the GateDrive’s 20-V maximum limitation isn’t surpassed. at approximately VFLOAT + 15 V.
C125 helps to suppress switching noise caused by sudden
loads on the supply during switching.
This arrangement can be readily adapted to many situa- Steve Hendrix is the principal engineer at Hx engineering
tions. If VFLOAT can change, even fairly rapidly, the large size LLC. He has a BS in computer science and mathematics from
of accumulators C123 and C124 relative to flying capacitors the U.S. Air Force Academy.
WilliAm sWAnson | [email protected]
High-Side Switch Provides Overvoltage

Protection With Only Four Components
Although microcontrollers generAlly run at low voltages, ing a current of (VOH – VBE)/R1, or about 1.2 mA for a 3.3-V
such as 3.3 V, they often need to control loads running at high- microcontroller. Most of this current flows through Q2’s base,
er voltages like 12 V. If switching the low side isn’t an option,
the microcontroller needs a way to control a high-side switch Q2
2N4403
from its low-voltage output. 4 to 40 V 0 to 15 V
The circuit in the figure provides a simple solution. Besides VIn VOut
using few components, it provides overvoltage protection as a

bonus. As long as the input voltage stays below approximately
D1
15 V, the output voltage will equal the input voltage (minus the 12 V
Q1
tiny VCE drop across transistor Q2). Microcontroller
2N4401
Once the input voltage exceeds that limit, however, the cir-
cuit begins acting as a low-dropout regulator (LDO), limiting
the output voltage to 15 V. This feature is useful if the input
R1
voltage occasionally exceeds the rated load voltage, but the 2.2k
load switch needs to remain efficient otherwise.
Transistor Q1 and resistor R1 form a current sink. The
microcontroller’s output voltage (VOH) and Q1’s base-emitter The microcontroller turns the load off by controlling a high-side switch formed
drop (VBE) create a stable voltage across the resistor, produc- by Q1 and Q2. The current-sink action of Q1 controls the load voltage.

IdeasForDesign
which controls the load. This current drops to zero when the voltages and pick the one that works. A good starting point is
microcontroller output goes low, switching the load off. to pick Zener voltage VZ to equal to the desired output voltage
Zener diode D1 provides an alternative path for R1’s cur- minus the microcontroller voltage. This is just an overvoltage
rent. If the diode’s breakdown current is VZ, it will begin to protector, so great accuracy should not be necessary.
conduct when the output voltage exceeds VZ + VOH – VBE, or If the overvoltage protection isn’t important, simply remove
about 14.6 V for a 12-V diode and a 3.3-V microcontroller. the Zener diode. The resulting load switch uses only three
Since the voltage across R1 is constant, the diode effectively components and retains the advantage of driving Q2’s base
“steals” current from the base of Q2, reducing the amount of with an actual current source. Even if the input voltage chang-
current flowing into the load. This negative feedback causes es, Q2’s base current remains constant.
the circuit to act like a voltage regulator.
To apply this circuit to your application, adjust R1 so Q2’s
base current equals the maximum load current (IMAX) divided
by Q2’s gain (β), or R1 = β ×(VOH – VBE)/ IMAX. Be sure to WILLIAM SWANSON graduated in 2005 with
select a transistor for Q2 that can dissipate the heat generated a BSEE from Cal Poly San Luis Obispo. He cur-
during an overvoltage event. rently designs boards and writes firmware for
There is no straightforward way to calculate the exact output an early stage startup company.
voltage, so the quickest approach is to try a few Zener diode
VISHWAS VAIDYA | TATA MOTORS [email protected]
Add Short-Circuit Protection, Diagnostics

To Automotive High-Side/Low-Side Driver
AUTOMOTIVE APPLICATIONS NEED “smarter” drivers for their Adding a few inex- +12 V +12 V
switches, which handle loads ranging from a fraction of an pensive components
ampere to several amperes. Not only are these smart chips cost- to a basic switch driv-
R2 R4
ly, but many times, they also aren’t available for the 24-V appli- er enhances it with 1k 1k
cations used in this application area. Further, they usually come additional operational 1
Load
in quads and pairs, so designers are paying for unused switches. features. 5 U1
The simple circuit in the figure adds a few low-cost com- U1 4N37
optocoupler
ponents to a “dumb driver” to make it a smart one, with added 4N37
optocoupler 2
short-circuit protection and built-in diagnostics (see the fig- R1
4
ure). Transistor Q1 is an inexpensive driver that can withstand 100
Q1
5V MPSA42
a few hundred volts and drive a few hundred milliamps. The A B
Microcontroller R3
microcontroller output goes high to latch the transistor on via pin D1 2.2k
the phototransistor of optocoupler U1. It next configures the 4.7 V
output pin into an input, while U1 keeps the transistor on by
supplying its base current through R2.
U1 has its LED in the load circuit of Q1 for detecting the
short circuit. Under short-circuit conditions, it is “starved” of short is persisting. Depending on the application, a suitable
current, turning off the phototransistor driving power-switch retry regime can be worked out based on the end application.
Q1’s base and protecting it. The CPU input pin can sample the During normal operation, the CPU can easily turn off the
occurrence of the short circuit at its convenience as needed, switch by pulling the output low. The CPU can issue a logic
such as every few seconds. low by turning the pin into output again whenever it needs to
The voltage at point A changes from a few hundred mil- switch Q1 off during normal operation.
livolts during short circuit to 4.7 V during normal operation,
providing a clear indication to the CPU’s input port. For retry-
based recovery from short circuit, the CPU may turn itself into VISHWAS VAIDYA is an assistant general manager of the
an output again at its leisure for restoring the “on” status and electronics division at Tata Motors Ltd. He is based in Pune,
recovering the switch from its “trip-off” condition. The CPU India. He holds a master’s degree in control engineering from
pin again now goes into input mode to sense if the external the Indian Institute of Technology, New Delhi, India.
ELECTRONIC DESIGN
Vol. 2, No. 3
SPI Eases Interfacing For 5-By-5 Dot-Matrix Display

The circuit interfaces a display that features four 5-by-5 dot-matrix characters
by using the display’s serial peripheral interface and a programmable system
on a chip (PSoC). The SPI reduces firmware overhead and enhances clock
stability.
Use An LED Dot-Graph Display To Complement

Your DVM Readout
The digital voltmeter is a great tool, but it can show confusing results when
the input voltage is unstable or erratic; by adding an auxiliary circuit driving
a string of LEDs, you can easily also get a sense of the variations in input
voltage.
Lamp Eliminates Need For Limit Switches In Flap-

Motor Control
A flap controller needs current limiting to protect the motor when the flap
reaches full closure. A small incandescent lamp can perform this task without
the need for more costly limit switches or a microcontroller.
for Design
Ideas
GIRISH CHOUDANKAR | EMPHATEC INC., MARKHAM, ONT., CANADA
SPI Eases Interfacing

[email protected]
For 5-By-5 Dot-Matrix Display

WHEN DISCUSSING DISPLAYS, seven-segment displays and LCDs internal RAM. Asynchronously, the RAM is read by the char-
are among the first that come to mind. Each type has its pros acter multiplexer at the rate defined by the strobe.
and cons. But generally, seven-segment displays cannot dis- The microcontroller supplies a string of bit-mapped char-
play letters, and LCDs tend to be bulky. Recently, I came acters (Fig. 2a, b, and c). Each character consists of six 8-bit
across a four-character 5-by-5 dot-matrix, serial-input display words. The first word is the character address. It’s followed by
(the SCDV5542 from Osram) that has several advantages. the row data, which represents the on/off status of each LED
The display can display alphanumeric characters and mea- in the column. Each bit of the 8-bit word consists of a 3-bit
sures 10.16 by 19.91 mm (0.400 by 0.784 in.) with 3.12-mm opcode (D7-D5) and 5 bits (D4-D0) of column data or charac-
(0.123-in.) characters. It has a serial peripheral interface (SPI) ter address or control word data.
and allows high-speed data input. It features eight levels of To load data, bring the Load line low, which puts data in the
dimming, an internal/external clock capability, and decoders, serial register (Fig. 2d). The shift action occurs on the low-
multiplexers, and an LED driver. to-high transition of the serial clock (SDCLK). The least sig-
The device consists of a CMOS IC with control logic and nificant bit (D0) is loaded first. After one word is transferred,
drivers for the four 5-by-5 characters (Fig. 1). Each individual the Load line is set to high for opcode decoding. The decoded
LED dot is addressable, so the user can create special char- opcode latches D4-D0 in the character address register. The time
acters. The device requires only four lines from a microcon- between loads must be at least 600 ns. The character address
troller. The IC accepts decoded serial data that is stored in register bits (D0-D4) and row address register bits (D7-D5)
direct the column data bits (D4-D0) to a spe-
cific RAM location.
Four 5-by-5 characters The load character address tells the display
Row decoder 10 the location (Table 1). The load column data
and driver
controls the individual LED dot (Table 2).
0 1 2 3
This feature allows you to create your own
10 special symbols.
Column logic and driver The display has seven brightness levels,
Mux CLK_I/O defined by the percentage of full bright-
Counter chain
Mux ClockSEL
_____ and timing logic
ness that can be achieved: 100% (0xF0),
Brightness and
Reset
lamp test
53% (0xF1), 40% (0xF2), 27% (0xF3), 20%
(0xF4), 13% (0xF5), and 6.6% (0xF7). To
Oscillator Power down
control the brightness level, change the duty
factor of the strobe pulse. Table 3 shows an
Display multiplexer
Serial example line.
data D7-D6-D5 D4-D3-D2-D1-D0 5
opcode data
In the power-down mode, the display
Serial brightness is set to 0% and the clock to the
And
Column output
data clock 3 5
row address decode
RAM internal multiplexer is stopped, reducing the

Write character/
____ Opcode (100 bits,

Character
Load decode address
5 write 20 by 5, quiescent current, ICC, to 50 µA (Table 4).
register read 10 by 10) The display is reactivated by loading a new
brightness level control word into the display.
Control During power-down mode, data can be writ-
word 4
address
ten to RAM.
register Table 5 illustrates the operation of the lamp
Software test mode, which sets all the LEDs to 53%
clear
brightness. This operation can be cleared by
loading a brightness level control word.
1. The SCDV5542 four-character 5-by-5 LED dot-matrix display stores decoded serial data in a 100- Finally, the software-clear mode clears the
bit internal RAM. Users can address each dot individually so they can generate custom characters. address register and the RAM (Table 6). The
ELECTRONIC DESIGN
IdeasForDesign
Table 1: load CharaCTer address
Opcode Character address

Operation
Hex
load
D7 D6 D5 D4 D3 D2 D1 D0
1 0 1 0 0 0 0 0 A0 Character 0
1 0 1 0 0 0 0 1 A1 Character 1
1 0 1 0 0 0 1 0 A2 Character 2
1 0 1 0 0 0 1 1 A3 Character 3
Table 2: load Column daTa
Opcode Column data

Operation load
D7 D6 D5 D4 D3 D2 D1 D0
0 0 0 C0 C1 C2 C3 C4 Character 0
Table 3: display brighTness
Opcode Control word

Operation
Hex
load
D7 D6 D5 D4 D3 D2 D1 D0
1 1 1 1 0 0 0 1 F1 53%
Table 4: power down
Opcode Control word

Operation
Hex
Level
D7 D6 D5 D4 D3 D2 D1 D0
1 1 1 1 1 1 1 1 FF 0% brightness
Table 5: lamp TesT
Opcode Control word

Operation
Hex
Level
D7 D6 D5 D4 D3 D2 D1 D0
1 1 1 1 1 0 0 0 F8 Lamp test
Table 6: sofTware Clear
Opcode Control word

Operation
Hex
Level
D7 D6 D5 D4 D3 D2 D1 D0
1 1 0 0 0 0 0 0 C0 Clear

IdeasForDesign
Example: serial clock = 5 MHz, clock period = 200 ns
264 clock cycles, 52.8 μs
(a) Character 0 Character 1 Character 2 Character 3
66 clock cycles, 13.2 μs

Character 0 Row 0 Row 1 Row 2 Row 3 Row 4
(b)
address column data column data column data column data column data
11 clock cycles, 2.2 µs 11 clock cycles, 2.2 µs
Character address Opcode Time Column data Opcode Time

(c) between D0 D1 D2 D3 D4 D5 D6 D7 between
D0 D1 D2 D3 D4 D5 D6 D7 loads loads
0 0 0 0 0 1 0 0 600 ns (min) C4 C3 C2 C1 C0 600 ns (min)
Load
Serial Clock
clock period
Data
(d)
D0 D1 D2 D3 D4 D5 D6 D7 Time between loads
IO
2. The bit-mapped characters supplied by the microcontroller occur within certain timing specifica-
tions. The times in this example are based on the use of a 5-MHz clock (a period of 200 ns).
display is blanked and the character reg- project I opted for the SPI hardware
ister address is set to character 0. The instantiation to reduce firmware over-
internal counter and control word regis- head. This choice also ensures the clock
ter are unaffected. is stable and isn’t influenced by other
The example circuit uses a Cypress priority events.
CY8C27443-24PXI programmable Instantiating SPI simplified the dis-
system-on-chip (PSoC) and a program play’s communication block. Below is
written in C. I chose the PSoC because the flow chart for firmware implementa-
it enables designers to select individual tion. (The code for the demo project can
hardware blocks. The serial interface to be downloaded from the online version
the display is straightforward and easy. of this article at electronicdesign.com.)
There are two ways to interface the
display. The first is to connect a micro- • Power up the display.
controller via the serial port in mode 0. • Bring Reset low (for at least 600 ns)
The serial port control register will be a to clear the multiplex counter, address
simple shift register. Serial data enters register, control word register, user
and exits through the RXD pin, and the RAM, and data register. The display
TXD pin outputs the clock. will be blank. The display brightness
The second method interfaces the dis- will be set to 100%.
play via the SPI in mode 0. Besides MISO • If different brightness is desired, load
(master in slave out) and clock, the dis- the proper display brightness control
play needs Reset and Load control lines. word from Table 4.
I chose SPI for this project because of its • Load the character address into the dis-
inherent advantages over serial port (Fig. play (Table 1).
3). The circuit uses a MAX1232 as a reset • Load column data into the display
controller, but you can choose any other (Table 2).
method for properly resetting the micro- • Repeat steps 4 and 5 for rest of the digits.
controller.
As noted, the PSoC enables you to Finally, one display CLK_I/O line
choose your own hardware modules, can drive 15 slave CLK_I/O lines, so
which speeds development. For this you can cascade displays to increase the
ElEctronic DEsign
IdeasForDesign
C1 +5 V +5 V
1 μF 16 V
1 14
SDCLK
____ SDCLK
____ Gnd
+ C2 2 Data 13
1 P1A Load Load Data
100 nF 3 VCC 12
50 V X
4 DS1 11
X SCDV5542 VCC
5 VCC 10
28 X
____ _______
6 9 + C3
1 VCC 27 RST CLKSEL 22 μF
P0[7] P0[6] 7 8 C4
2 26 Gnd CLK_I/O 35 V
10 nF
P0[5] P0[4]
3 25 50 V
P0[3] P0[2]
4 24
___ P0[1] P0[0]
5 23
Reset P2[7] P2[6] SDCCLK
____ 6 22
Load P2[5] U1 P2[4] Data
7 21
P2[3] CY8C27443 P2[2]
8 20 C5
P2[1] P2[0]
9 19 2 P1B +5 V 100 nF
SMP XRES 50 V
10 18
P1[7] P1[6] R1
11 17
P1[5] P1[4] 100k 3
12 16 1% P1C ___
P1[3] P1[2] VCC 1
P2D 4 15 5 P1E PB
13
P1[1] P1[0] 6 ____
VSS RST 2
P1F U2 TD
3
14 6 5 RST MAX1232 TOL
7 P1G ___ 7
ST
Gnd P2D
4
3. The author opted to use a SPI instantiation for the display because it reduced the firmware over-
head required.
___
RST
VCC
___ ______ ___ ______

RST CLK_I/O CLKSEL RST CLK_I/O CLKSEL
Up to 14 displays
SCDV5542 in between SCDV5542
____ ____
Data SDCLK Load Data SDCLK Load
Data
SDCLK
0
A0
A1 Address Address decode 1-14
decoder
A2
15
A3
__ __
LD CE
4. The display’s CLK_I/O line can drive up to 15 slave CLK_I/O lines, so designers can easily cascade
multiple units.
project’s display capabilities. Figure 4 2. Cypress Semiconductor

shows the block diagram for interfacing CY8C27443-24PXI, www.cypress.
multiple displays. com/?mpn=CY8C27443-24PVXI
3. Cypress Semiconductor PSoC Designer,
SOURCES www.cypress.com/?id=2522
1. Osram SCDV5542 display, cata- 4. Maxim Integrated MAX1232, www.maxi-
log.osram-os.com/media/_en/ mintegrated.com/datasheet/index.mvp/
Graphics/00034132_0.pdf id/1286
GIRISH CHOUDANKAR works at Emphatec Inc., a Toronto-based

design house for industrial control interfaces and switch-mode
power supplies. He holds a bachelor’s degree in electronics engi-
neering from Mumbai University, India.

for Design
Ideas
NEDJELJKO LEKIC AND ZORAN MIJANOVIC | UNIVERSITY OF MONTENEGRO, PODGORICA, MONTENEGRO
Use An LED Dot-Graph Display

To Complement Your DVM Readout
+12 V
LEDS OFTEN ARE used to form a dot/bar graph dis-
R1
play.1-4 You also can use an LED dot-graph display VIn
1M
D1
as a complement to a digital voltmeter (DVM) with Digital voltmeter

D2
the LM3914 dot/bar display driver and nine LEDs.5 C1
+12 V
1 μF U1
This is useful where the input voltage is unstable, so D3
RH1 V+
the voltmeter indications are variable and the result REF OUT
LED1
is hard to read and understand. LED2
D4
R
By adding an associated circuit, the DVM measures 1.2k LM3914 LED3
LED4 D5
and displays the approximate mean value of the input R2 REF ADJ LED5
C2 LED6
voltage, while the LED dot-graph provides informa- 1 μF 1M P LED7 D6
tion about the voltage’s stability. This approach is 10k LED8
LED9
suitable for rapidly testing power supplies. RLO LED10 D7
Mode
The input signal passes through low-pass filter SIGIN V– D8
R1C1 to a digital voltmeter (Fig. 1). The filter’s out-
R3
put is the mean value of the input signal, so the volt- 1M
D9
meter indicates it is stable, and it is easy to read. (If
the digital voltmeter has a built-in averaging option,
the low-pass filter is not required.) Capacitor C2 1. The LED dot-graph display uses a standard display driver preceded by a high-pass
passes the variable component of the input voltage to filter. It visually complements a digital voltmeter’s numeric output.
U1, and the LED dot graph that U1 drives indicates
the intensity of these variations.
Figure 2 (obtained by simulation) shows how the low-pass REFERENCES
and high-pass filters pass the mean and variable components of 1. “LED bar-graph display represents two digits,” Ajoy Raman,
the input signal, respectively. If the input voltage is stable and Bangalore, India, ('1, September 22, 2011.
doesn’t have any ripple, only central LED D5 lights up. 2. “PIC microprocessor drives 20-LED dot- or bar-graph display,”
As voltage ripples increase, the adjacent diodes begin to Noureddine Benabadji, University of Sciences and Technology,
flicker. For larger ripples, more of the adjacent diodes light. Oran, Algeria, ('1, September 1, 2006.
Potentiometer P controls the LED dot-graph’s sensitivity, with 3. “Drive 16 LEDs with one I/O line,” Zoran Mijanovic and Nedjeljko
maximum sensitivity of 0.125 V when P is set to 0 Ω. Nine of Lekic, University of Montenegro, ('1, June 9, 2011.
the 10 available LED outputs are used. 4. “Drive 12 LEDs with one I/O line,” Charaf Laissoub, Maisons Alfort,
France, ('1, February 4, 2010.
5. “LM3914 Dot/Bar Display Driver,” National Semiconductor,
February 2003, http://bit.ly/naDCRG.
NEDJELJKO LEKIC works at the University of Montenegro,

Department of Electrical Engineering. He holds a PhD in electrical
and computer engineering from the University of Montenegro. He
has been working on designing identification systems based on
RFID and/or biometrical identifiers as well as microcontroller sys-
tems for industrial automation.
2. The simulation of the waveforms of VIn (V[n001]) as well as the volt-
age at the low-pass filter output (junction of R1/C1) (V[n002]) and at the ZORAN MIJANOVIC has been working as a professor with the elec-
high-pass filter output (junction of R2/R3/C2) (V[n003]) show how the tro-technical faculty at the University of Montenegro. He has designed
low-pass and high-pass filters pass the mean and variable components of many microcontroller devices for automation and control. He has a
the input signal, respectively. PhD in electronics from the University of Belgrade.
05.02.13 ELECTRONIC DESIGN

IdeasForDesign
GUNNAR ENGLUND | GKE ELEKTRONIK AB, GRANBERGSDAL, SWEDEN [email protected]
Lamp Eliminates Need For Limit

Switches In Flap-Motor Control
A FRIEND OF a friend asked me how to best open and close a The friend said he had tried that already, but he ran into
flap at the end of a tube: pneumatics, hydraulics, electrical, problems with limit switches. The need for a constant closing
or what? It was an automotive application and no hydraulic force at the end of travel was another complication. Adding a
system or compressor was available. So, the answer seemed microcontroller was too complicated, he said.
obvious: a 12-V dc, small-gear motor. There seemed to be no simple answer. But eventually a
solution came to me. And all it requires is a low-wattage, low-
1. A low-wattage, low- voltage incandescent lamp added to the motor circuit (Fig. 1).
VPlus/minus
voltage incandescent The relationship between voltage and current in a small incan-
lamp in the flap’s Low-voltage descent lamp is very nonlinear (Fig. 2). The tungsten wire in the
lamp
motor-control circuit DC lamp has a low resistance when cold. That’s in the region up to
motor
eliminates the need around half a volt. Cursors CI and CII indicate that the slope in
for limit switches or a that region is 0.439/1.267 = 0.35, while the slope between the
VMinus/plus
microcontroller. origin and CIII (nominal voltage and current) is 2.12.
So, the lamp acts as an automatically changing resis-
tor with a low resistance when the load is low and a
resistance six times higher when fully loaded. That’s
exactly what we need to run an actuator without limit
switches, microcontroller, transistors, or whatever.
Figure 3 shows flap position, motor voltage, and
motor current in a close/rest/open/rest/power-off
cycle. The motor current’s variations (red trace) reflect
a badly aligned coupling and noise in the commuta-
tor. That is always present in small dc motors. But the
motor used in this application is worse than most other
motors in that respect. Despite that, the movement
(grey) is smooth and well-defined.
When the flap hits the mechanical stop (flap closed),
the motor stops and the back electromagnetic field
2. The lamp’s current-voltage curve (vertical versus horizontal) illustrates the non- (EMF) drops to a low value, which means that current
linearity that makes this solution possible. would have risen to perhaps five or 10 times the run-
ning current, depending on which motor and which
supply is chosen. That, of course, would destroy the
motor and is why limit switches and controllers are
needed to switch off, or reduce, current.
By adding a simple low-wattage and low-voltage
incandescent lamp to the circuit, you can avoid this
problem and keep the motor current at a low level that
the motor can tolerate indefinitely. You can even use
the lamp as a visual indicator that the flap has reached
its closed (or open) position. And, the reduced current
needed to keep the flap tightly closed comes without
even asking for it.
GUNNAR ENGLUND, electrical engineer, worked

with ABB and Siemens for many years and is now
3. A plot of the circuit’s operation shows the smooth, well-defined movement of the an independent consulting engineer. He is mostly
flap despite the variations in the motor’s current. active in research and problem solving regarding
EDM in bearings (bearing currents).
ELECTRONIC DESIGN GO TO ELECTRONICDESIGN.COM 73

Vol. 2, No. 4
Single-Cell Regulated Charge Pump Draws Low

Quiescent Current
A charge pump (or Q pump) uses a flying-capacitor architecture to boost a
single cell’s voltage to a regulated 2.5V, yet operates at low quiescent current.
This approach is a good fit for microcontroller applications where the load
is usually in sleep mode and quiescent battery drain is more important than
efficiency at higher loads. The two-stage design has efficiency in the 40 to
70% range—comparable to an LDO—yet consumes just microamps when the
load is not active.
Hex Buffer, MOSFETs Build A High-Power, Lossless,

Virtual Ground
This circuit uses a standard hex inverter to create an oscillator and Class-D
amplifier, which in turn drives MOSFETs that can efficiently supply the bipolar
rails and sink the ground current.
Graphical, Numerical Techniques Size LED

Array’s Drive
The nonlinear relationship between forward voltage and current in an LED
makes it difficult to properly size the drive current and current-limiting resistor,
but useful analysis can be derived from the exponential diode model.
Electronic
Design?
ideas
Single-Cell Regulated Charge Pump
Draws Low Quiescent Current
for design
BRUCE D. MOORE | ALERT SOLUTIONS CO., DAVIS, CALIF. [email protected]
CAPACITOR-BASED CHARGE PUMPS (OR Q-pumps) generally efficiency. This might be the case when powering a microcon-
aren’t useful for sourcing large amounts of current, but they troller that spends most of its time in sleep mode.
work well in niche micropower applications where space is at a Low-voltage microcontrollers such as those in the PIC24
premium. They work best in applications where the output volt- or MSP430 families are generally powered from a regulated
ages are integer multiples of the input voltage. The integer mul- supply voltage such as 2.5 V. If clocked slowly, they might
tiples, then, are operating points that result in peak efficiency. draw as little as 25 μA or 50 μA. In standby mode with only the
However, Q-pumps can also work well when they are pow- real-time clock running, the current can be vanishingly small,
ered from a variable input such as a battery, particularly when often less than a microamp. This is a good application for the
quiescent battery drain is more important than heavy-load regulated two-stage Q-pump described here, which boosts a
V DD
single alkaline or nickel-metal-hydride (NiMH)
VOUT cell to 2.5 V.
0.95 to 2.5 V V1 2.5 V at 50 μA The “wings” of a Q-pump, called “flying capac-
C3 itors,” connect first to the input and then to the
C5
10 μF 22 μF output. If the capacitor is stacked on the input
voltage, it forms a voltage doubler. In the case of a
C1 regulated charge pump with a fixed output volt-
10 μF C4
C2 age, the voltage across the flying capacitor may
2.2 μF 2.2 μF
differ significantly from the voltage across the
SN74AUP1G14 output-filter capacitor.
When you connect two capacitors that are
R5 initially charged to different voltages, you get
1M
a spark or power dissipation in the switches as
the capacitors equalize in voltage. This is why a
NXP3008NBKT simple voltage doubler typically exhibits better
R6 efficiency than a regulated Q-pump.
1.65M This regulated Q-pump has an on-demand
V DD 1%
oscillator, a feedback regulation loop made from
10 2 an op amp and reference, and a two-stage pump
9 –
R2 R4 /LHDET 1 circuit, plus two flying capacitors, C2 and C4
100k 1.5M 5 3 (Fig. 1). The first pump stage is driven directly by
4 + R7
+ –
499k the Touchstone Semiconductor TS12011 com-
6 8
1% parator, which forms the oscillator, while the
R1 7
2M VREF second stage is driven by an inverter powered
R3 from the output voltage of the first stage. The
C6
2M TS12011
100 pF full-load efficiency varies from 70% to 40% over
All diodes = BAS52-02V a 1- to 2.5-V input range, which is comparable to
a linear regulator.
1. The low-voltage regulated charge pump uses the “flying capacitor” topology The TS12011 analog building block requires
(C2 and C4) to “stack” and thus increase the output voltage. very low supply currents (3.2 μA typical) and
68 06.06.13 ELECTRONIC DESIGN

operates well down at the sub-1-V levels needed for single-
cell operation. The comparator output stage has good drive-
current capability down below 0.8 V VDD, which is an uncom-
mon feature that allows us to drive the first stage directly from
the oscillator.
The SN74AUP family of logic gates similarly uses super-low
power. The MOSFETs were also carefully chosen for low-volt-
age operation, with low gate-threshold voltage specifications
and low gate-charge characteristics for low switching losses.
Using these high-performance components, the no-load qui-
escent current is a mere 8 μA.
The RC oscillator is stopped when the output is in regula-
tion via its /LHDET latch input (Fig. 2). When the output is
above the regulation threshold set by the reference and feed-
back divider, the op-amp output drives the latch input low, 2. The oscillator controls how much charge is pumped and thus the
latching the comparator output in the high state and stopping output voltage: VOUT (top waveform), comparator output (middle),
the oscillator. and op-amp output (bottom).
A large hysteresis band was chosen for the oscillator, result-
ing in a large signal swing on the timing capacitor C6. This ing loss in headroom, a problem when trying to multiply up to
achieves the most efficient size-versus-current operating point 2.5-V output levels.
for the oscillator. The maximum frequency is nominally set at The rectifier choice is further complicated in this low-qui-
1 kHz with the component values shown, but can be adjusted escent application by leakage currents, which effectively load
up to about 3 kHz, where it becomes limited by the propaga- down the pump and increase idling current. The BAS52-02V
tion delay through the comparator. Schottky diode has a good combination of low reverse leakage,
The amount of charge transferred with each cycle and the low forward drop, small packaging, and wide availability.
switching frequency determine the output current. Accurate Reverse leakage current at high temperature is the weakness
calculations for the output impedance or output current from of Schottky diodes, but lab measurements show that the typi-
a regulated two-stage pump are complex and need a large cal BAS52 is very good on this parameter, at less than 1 μA at
spreadsheet, but you can make some oversimplifications to 50°C. For even lower reverse leakage, there is the BAS40-02V,
get into the right range. Assuming the capacitor is completely but the tradeoff is about 75-mV higher forward-voltage drop.
charged and discharged with each cycle (which obviously isn’t The capacitor values and switching frequency are chosen to
true), then: be somewhat excessive for charge transfer, so they don’t get in
the way of the voltage-drop/headroom issues. As a result, the
Q=C×V frequency is low and the capacitors are relatively large, which
I=Q×f also keeps switching losses in check. More importantly, a long
on-time allows the voltage across the diode to reach a mini-
so: mum, with the tail end of the forward voltage/current curve
being a little less than 0.2 V for the BAS52.
I=C×V×F From the simplified charge-transfer equations above, an
overly optimistic estimate for first-stage current is:
where C is the flying capacitor value, V is the applied voltage
during the charging phase, F is the oscillator frequency, and (VDD – VFWD – VSAT1) × C3 × F =
I is the output current you’d get if you could remove all the (0.95 V – 0.2 V – 0.05 V) × 2.2 μF × 1 kHz = 1.5 mA
stored charge during the discharge phase (roughly equivalent
to a short-circuit load current). which is more than sufficient to meet a 50-μA load-current
The modified Dickson Multiplier two-stage pump topology requirement. The best-possible peak-output voltage that you
used here is a quadrupler, and both stages must be designed to can achieve for the intermediate V1 tap is:
transfer the needed amount of charge at the worst-case low-
battery voltage. The first stage is the most important, because 2VDD – 2VFWD – VSAT1(high) – VSAT1(low)
the voltage applied to the capacitor is low. Any drop across the
first rectifier subtracts from the applied voltage with a result- Similarly, the second-stage peak voltage for VOUT is:
GO
GO TO ELECTRONICDESIGN.COM 69
IdeasforDesign
V DD
quadruple the lower end of the voltage range
VOUT
C5 of a single cell. We can improve this, and get to
C1 C3
10 μF 10 μF 22 μF sub-0.9 V operation, by several means: adding
another pump stage; changing the output volt-
age to 2.2 V (but the microcontroller may not be
C4
able operate this low); or adding a synchronous
BSH205 2.2 μF rectifier across the first diode.
From The third option is chosen here, due to its
feedback divider good efficiency at the operating “sweet spots”
U1 From
inverter
(Fig. 3). A low-threshold P-channel MOSFET
R1
with low gate charge shunts the first pump
C2
1M 2.2 μF diode, reducing the forward drop from 0.2 V to
approximately 0.01 V. This adds a few hundred
NXP3008 extra millivolts right at the front end, where it’s
needed most.
A NOR-based logic circuit provides non-
R2 R3 overlapping gate-drive signals to the synchro-
U2 1M 1M U3 nous switch. The goal is to be certain that the
MOSFET gate is never driven high when the
U1 = SN74AUP1G32
flying capacitor is high. The amount of dead
C6 C7
47 pF 47 pF U4 U2 = ½ SN74AUP1G02 time is set by the 1 MΩ × 47 pF RC time con-
U3 = ½ SN74AUP1G02 stants. Here, the dead time is very long since the
U4 = SN74AUP1G14
All diodes = BAS52 1-kHz oscillator is so slow.
Charge pumps offer an alternative to induc-
tor-based boost regulators and can fit nicely
From TS12011 comparator
with the dual nature of many microcontroller
loads, with their low-current RTC mode versus
3. To achieve operation below 0.9 V, the first stage is modified by adding a synchro- heavier but infrequent run mode. The challenge
nous rectifier across the first diode. of designing them for low input voltages can
be met by using the right low-voltage components, choosing
2V1 – 2VFWD –VSAT2(high) – VSAT2(low) sensible switching frequencies, and taking full advantage of
the V-I characteristics of the rectifiers.
Thus, the best-possible peak-load voltage at the minimum
input voltage is:
VOUT = 4VDD – 4VFWD – 4VSAT =

(4 × 0.95 V) – (4 × 0.2 V) – (2 × 0.05 V) – (2 × 0.1 V) = 2.7 V
IDEAS FOR DESIGN WANTED
Send us your Ideas For Design. We’ll pay you $150 for every Idea For
Design that we publish. In addition, this year’s top design as selected
where VSAT1 and VSAT2 are the voltage drops across the com-
by our readers will earn an additional $500, with two runners up each
parator and the AUP inverter output stages in the ON state, receiving $250. You can submit your Ideas For Design via:
respectively. Again, this analysis is oversimplified but points
out the limits of operation. • E-mail: [email protected]
This math shows that there is not a lot of room for tem-
perature variation and component tolerances when trying to OR BY
• Postal mail to:
Ideas For Design
BRUCE D. MOORE is a consulting analog engineer for Alert Electronic Design
1166 Avenue of the Americas, 10th Floor
Solutions Co., doing new product definition and application
New York, NY 10036
circuit development for amplifier and power-supply ICs. He gradu-
Go to www.electronicdesign.com for our submission guidelines.
ated with a BSEET from Heald Engineering College, San Francisco.

IdeasforDesign
Hex Buffer, MOSFETs Build A High-

Power, Lossless, Virtual Ground
LOUIS VLEMINCQ | BELGACOM, EVERE, BELGIUM [email protected]
A VIRTUAL GROUND IS useful whenever you need to create a MOSFETs. Inverter U1 provides the non-inverting input of the
bipolar supply but the dc source is unipolar, as is often the amplifier, as the inverting input is implicit and is represented
case with battery-operated equipment. Sometimes, the solu- by the CD4069’s mid-supply potential. Capacitor C3 provides
tion is as easy as using a high-ohm resistive divider to provide positive feedback and allows the circuit to self-oscillate.
a mid-supply potential. But if the ground impedance needs The entire control circuit has a floating supply referenced
to be lower, the simple approach can be enhanced by using a to the output. Diodes D3 to D6 and resistor pair R6/R7 gener-
buffer amplifier. ate this supply from the dc input. R2 and R3 create the virtual
When the ground created this way sees high imbalance cur- ground potential. In this example, they are equal, as will gener-
rents, though, things become more complicated. The buffer ally be the case, but asymmetrical supplies also can be created.
not only needs to be a power buffer, it will also have to dis- For greater accuracy, a trim potentiometer can be added
sipate the result of any imbalance. This is wasteful, and it may to the midpoint of R2/R3. With the values shown, the pulse-
even require bulky, inconvenient heatsinks. width modulation (PWM) frequency is approximately 45
This circuit addresses these issues simply, cheaply, and kHz, which can be modified by changing C3.
effectively (see the figure). It retains the resistive divider plus The power MOSFETs and inductor must be sized accord-
buffer configuration, but with a big difference. The buffer is ing to the expected output current. If a very clean ground is
now a self-oscillating Class D amplifier, capable of provid- required, an additional LC filter can be used after L1 and C6/
ing many amperes of ground current and of “recycling” any C7 to clean up the remaining switching residues.
imbalance back to the supply. The circuit is compact and effi-
cient, while providing a stiff, accurate, and powerful ground. LOUIS VLEMINCQ works as a design engineer in Physcial
The oscillator is built around a simple CD4069 CMOS hex Layer Engineering at Belgacom, Evere, Belgium.
buffer. Most of its inverters act as gate drivers for the power
+21 V
V
D5
R2 R6 1N4148
V DD
47k C3
D1 680/
1 nF Red R4 1W
12k
U2 M1
IRF530 +
C6
U3 C1 470 μF/
100 nF
C4 + D3 L1 60 V
Unipolar supply
R1
Bipolar output
BZX84C6V2L 100 μH/

+ 4.7k U1 U4 3A
22 μF/10 V
GND
CD4069B 22 μF/10 V + D4
V1 U5
42 V BZX84C6V2L
C5
C2 +
U6 100 nF
C7
M2 470 μF/
IRF9530 60 V
R5
VSS
D2 12k R7
R3 Red D6
680/
47k 1W 1N4148
– 21 V
V
V
Although it requires more components than the conventional resistor-divider approach, this Class D oscillator/amplifier (based on a CD4069
hex inverter) drives power MOSFETs that make the virtual ground much more efficient.

ideas
Graphical, Numerical Techniques
Size LED Array’s Drive
KENG WU | SWITCHING POWER INC. [email protected]
for design
L E D - BA S E D L I G H T I N G I S R E P L AC I N G separating the two key I-V relationship E

9V
incandescent and gas-discharge lamps equations, v + RxIF = Vs and IF = aebv + c. L$ Q DH P

F
in many situations. As practical lighting For example, using a Lumileds Luxe-
sources, and considering their direc- on Rebel ES LED with a 3.2-V dc source With the array model, the switch-
tional nature, LED fixtures usually use and a desired drive current of 0.5 A, the mode buck-regulating current driver
multiple LEDs arranged in an array, horizontal current line intercepts the in a closed-loop configuration (Fig. 4)
with “m” LEDs in a serial string and “n” LED curve at 2.92 V (Fig. 3). Therefore, can be described by a set of equations
such strings in parallel (Fig. 1). For the the series resistor will be R x = (3.2 – starting at the feedback point, circling
array configuration and the exponen- 2.92)/0.5 = 0.56 Ω. around the control loop, and ending at
tial I-V characteristics model of a single However, the graphical approach the driver output:
LED,1 it’s additionally possible to have a would not handle an LED array effec-
similar model representing the loading tively, and the numerical method works • Current sensing:
of such an array. better. To represent the array current iA,
Since the single LED is a diode with the exponential model for the (m × n) Vf = iA • RSense (4)
a nonlinear forward-voltage/current, array is modified, since voltage drops
driving it with a fixed voltage source is across current-limiting resistor Rn and • Error amplifier:
generally not recommended. Rather, sense-resistor R Sense are usually small
a serial (current-limiting) resistor is compared with driving-source Vs: Ver = A(VRef – Vf ) (5)
needed to equalize the power source and
the load (Fig. 2). However, sizing this • Effective error:
resistor requires a significant analyti- R1 Rn 9HU ² 9)
cal manipulation. Using the single-LED 9HI % &75 % 5 H
5G
exponential model, a Kirchoff ’s voltage D1
law (KVL) equation around the circuit • PWM ramp:
loop gives: 90D[ ² 90LQ
95DPS W 90LQ
W
% 7V
5 [ DH E9)
F
9) 9V
which can be rearranged to: Dm

Rx
IF
9 ²9
5 [ VE9) )
DH
F VS VF
R Sense
where a, b, and c are model parameters
for a selected LED, as defined in the
reference.
This is a challenge to solve symboli- 1. Establishing the correct resistor size for 2. Even for a single LED driven by a voltage
cally, but it can be solved numerically each string of an array of n strings, where source, the exponential current-voltage
with mathematical software tools. Alter- each strings has m LEDs, is challenging relationship complicates selection of the
natively, it can be solved graphically by due to the nonlinear nature of the diodes. series-resistor value.

• Duty cycle, D: *'+
*'*
*')
VRamp(D • Ts) = Vef (8) )'2
)'1
<nkk^gm!:"
• Power stage: b? ! o " )'0
)'/
9,Q
% 1 V % ' 9V Bk ! o " )'.
1S )'-
)',
With effort, all the equations can be consolidated into a )'+
single, closed-loop equation. Equation 10 is a transcendental )'*
equation with an exponential term that prevents us from solv- )
+'/ +'0 +'1 +'2 ,') ,'* ,'+
ing it analytically, but it can be solved using computational
OhemZ`^!O"
techniques and software tools.
3. The intersection of the desired current line and the forward cur-
Reference rent/voltage curve (red) leads to the unique solution to the non-
1. “Generate Realistic Models for LED Current Versus Voltage,” Keng C. Wu;
linear equations of the LED model. (The blue line is the reverse-bias
Electronic Design, Vol. 61, No. 1, p. 52, Jan. 10, 2013, http://electronicde-
sign.com/power/generate-realistic-models-led-current-versus-voltage current/voltage relationship.)
9V KENG C. WU has a BS from

E
$ 95H I ² Q DH P
& % 5 6HQVH ² 9) Chiaotung University, Taiwan,
% &755H²90LQ and an MS from Northwestern
5G University, Evanston, Ill. He has
9,Q published four books and holds
% 1V % 9V
1S 90D[ ² 90LQ seven U.S. patents.
VIn VS
D1
Np NS D2
Figure 1
array
Vf
Ver –
Driver VMax A
+
VMin TS VRef
VCC Rd
PWM
–
RSense
+ VF
CTR
Vef
Re
4. Analysis of the switch-mode buck-regulating current driver in a closed-loop configuration for driving an array yields a set of equations
that are better solved using numerical computation rather than an analytical, closed-form approach.
GO
GO TO ELECTRONICDESIGN.COM
Vol. 2, No. 5
Simple Addition Permits Voltage Control Of DC-DC

Converter’s Output
This article presents apples-to-apples test results for noise densities for
several low-noise discrete, SMT n-JFETs. The tests used the same operating
conditions so that designers can make a fair comparison for their applications.
Simple NiCd Battery Charger Includes Charge

Indication
This circuit uses a microcontroller and simple algorithm managing voltage
regulator ICs and discrete components to control charging of nickel-cadmium
batteries. It provides a steady charging current matched to the battery
capacity while avoiding overcharging; it also includes LEDs that indicate the
status of the charging process to the user.

ideas
Simple Addition Permits Voltage
Control Of DC-DC Converter’s Output
EUGENE PALATNIK | ITEC ENGINEERING [email protected]
for design
IN A STANDARD DC-DC converter, a resis- 5 5 VOut = 3 Vfb – 2 Vr (4)

tor divider typically defines a fixed out- 92XW ² 9U 9IE ² 9U
5
put voltage. However, applications like R3 and R4 have the same value, 10
programmable output voltage power Since R1 = 20 kΩ and R2 = 10 kΩ, kΩ, so amplifier U2’s output voltage is:
supplies and motor control circuits Equation 2 can be simplified to:
require dynamic control of the dc-dc Vr = 2 VRef – VC (5)
converter’s output voltage. The circuit VOut – Vr = 3(Vfb – Vr) (3)
described here allows control of the con- where VRef is the reference voltage gen-
verter’s output voltage, VOut, with a con- or: erated by U3 after resistor divider R7/R8.
trol voltage, VC.
In a conventional dc-dc buck con- C5 0.1 μF
2 1 L1
verter, VOut is: VIn VIN BOOT 2.2 μH
8
C2 U1 PH VOut
5 5 10 μF 3 TPS54332
92XW 9IE 6
EN
D1 C6
5 5 22 μF R1
COMP VFB B230-13
4 9 20k
SS PGND
so VOut is fixed by the values of R1 and Vfb
R9
R2 (Fig. 1). 20.5k
The added circuitr y in Figure 2 C4

0.01 μF R2
C3
enables users to control the same dc- 820 pF
10k
dc converter’s output voltage using VC.

In this case, R2 is not connected to the
ground but, rather, to Vr. Equation 1 1. The output voltage in a conventional dc-dc buck converter is fixed and depends on the
then becomes: resistor divider, R1/R2.
2 1
VIn VIN BOOT L1
C2 C5 0.1 μF 2.2 μH
10 μF 8
3 PH VOut R3
EN U1 D1 10k
2. The added cir- TPS54332 B230-13
6 5 C6
cuitry in this ver- COMP VFB
22 μF R1 +5 V R4
sion of the dc-dc 20k 10k
4 9 R5
SS PGND 7 2
51 VC
converter permits 6 –
R9 Vfb +5 V
control of V Out by + 3
20.5k U2 4
varying a control TLC271
R2
–5 V R7 R6
voltage, VC. 10k 2k
C4 10k
C3 Vr VRef
0.01 μF
820 pF
C1 R8
1 μF 3.16k U3
LM4040D25

Combining Equation 4 and Equation 5: Finally, C1 lowers U2’s output impedance at high frequencies,
maintaining the stability of U1’s feedback loop.
VOut = 3 Vfb – 4 VRef + 2 V (6) The added circuitry allows users to control the buck con-
verter’s output voltage, VOut, in the range of 0 to 5 V with a con-
To simplify Equation 6, choose components that make: trol voltage, VC, in the range of 0 to 2.5 V. Similar circuitry can
be designed
3 Vfb = 4 VRef (7) EUGENE PALATNIK is for use with a
an electrical engineer and boost convert-
Then Equation 6 becomes: president of ITEC Engineer- er, or any other
ing LLC. He is active in the dc-dc con-
VOut = 2 VC (8) design of electronics for verter, as long
medical, scientific, and as its feedback
The internal voltage reference of U1 is 0.8 V. By choosing R7 industrial applications. voltage pin is
= 10 kΩ and R8 = 3.16 kΩ, VRef = 0.6 V, satisfying Equation 7. accessible.
Simple NiCd Battery Charger

Includes Charge Indication
ABEL RAYNUS | ARMATRON INTERNATIONAL INC. [email protected]
RECHARGEABLE NICKEL-CADMIUM (NICD) the battery during charging. Also, more full, to avoid overcharging. This simple
BATTERIES are widely used in consumer energy must be supplied to the battery and inexpensive charger overcomes
electronics because of their high energy than its actual capacity to compensate both problems. The cheapest and saf-
density, long life, and low self-discharge for energy loss during charging. est way to charge a NiCd battery is to
rate. Standard NiCd cells can be charged However, two problems must be charge at 10% of its rated capacity per
at different rates: a fast charge with high addressed when designing a charger hour for 16 hours. The battery pack
current, or overnight with low current. for them: how to set the proper charg- used contained two AA-size 1200-mAh
Regardless of the charge speed, a ing-current value, and how to stop the NiCd cells, so the battery should be
steady current should be provided to charging process when the battery is charged with 120-mA current.
IC2
In the charging circuit of Figure 1, a
LM78L05 constant charge current is generated by
+5 V 1 3
a current regulator comprising IC3 (an
+9 V
C1 3 LM317 LDO) and resistor R3, where R3
0.1 μF 2
1 IC3 Buzzer is 1.25 V/120 mA, about 10 Ω. Switch-
LM317
ing MOSFET Q1 (IRF520) was chosen
2 because of its very low open-state (con-
1
LED4 2
R3 ductive) impedance of 0.3 Ω.
100% pA5 10
The best charging practice is to use a
LED3
3 120 mA timer to prevent overcharging to con-
75% pA4 IC1 3
Battery + R2 tinue past 16 hours. This approach does
MC68HC908QT1 7
LED2 4 under 200k
50% pA0 charge – 2 Q2 not require an end-of-charge sensor,
pA3 2N3904
LED1 6 2 and it ensures a full charge. The tim-
25% pA1 R1
100k Q1 1 ing function is performed by microcon-
1 IRF520 troller IC1, which also reports the state
8
3 of charge via the LEDs.
Any microcontroller could be used in
1. The constant charging current is produced by an LDO and resistor and gated by Q1, which this project. Here, the inexpensive eight-
in turn is managed by an output of the microcontroller. A quartet of LEDs, also microcon- pin Motorola (Freescale) MC68HC-
troller-driven, indicates charge status to the user. 908QT1 microcontroller was used.
GO
Ideas for Design
Each charging step is indicated by lighting a correspond- Start
ing LED. The number of steps is determined by the number

of available outputs of the microcontroller, without adding Initialization
any extra components. Since the microcontroller has five Start charge
outputs, one of them is used for charge triggering, and so Main
No
the four can be used for charge indication. To minimize the TOF = 1?
number of components, LEDs with built-in resistors are used Yes
(WP710A10YD5V, www.kingbrightusa.com). Increment counter
To make the process more visual, these LEDs should be Yes

CNT >1/4 MAX_CNT?
located in line with the outline of the battery drawn around No
them, so lighting the LEDs one by one will clearly indicate the Toggle LED1
progress of charging. It is reasonable to choose the time inter-

vals to be equal, with the LEDs indicating 25%, 50%, 75%, and Set LED1 on
100% of battery-charge time. Yes

CNT >1/2 MAX_CNT?
The program starts blinking the corresponding LED at No
the beginning of each time interval and up to the end of each Toggle LED2
interval. After that, it lights the LEDs steady on. When charg-
ing is over, all four LEDs are on, so the user knows the charge Set LED2 on
status at any time. (As an optional feature, a buzzer could be Yes

CNT >3/4MAX_CNT?
added to provide an audio signal when charging is over.) No
The microcontroller program of Figure 2 is straightforward. Toggle LED3
The assembler code listing can be found with the online ver-
sion of this article at http://electronicdesign.com/power/simple- Set LED3 on
nicd-battery-charger-includes-charge-indication. Yes
CNT >MAX_CNT?
The LED blinking period is set at one second. The built-in No
oscillator of the microcontroller generates a frequency of 12.8 Toggle LED4
MHz and provides a one-cycle duration of 312.5 ns. By setting
the timer prescaler to 64 and timer modulo register to 50,000
Set LED4 on
(C350H), the timer overflow (TOF) period is equal to one sec-
Stop charge
ond (0.3125 μs × 64 × 50,000). The program toggles the LED
at each TOF period. End
The overnight “long” charge lasts 16 hours, with counter
constant MAX_CNT calculated as 16 × 60 × 60 = 57,600 2. The flowchart shows the straightforward level-check/step-
(E100H). Any maximum charge time can be set in the same through iteration sequence of the code for driving the charge-indi-
way. Obviously, it’s not convenient to wait for 16 hours to test cation LEDs.
the program, and a period such as 20 minutes, for example,
would be more practical.
For that shorter period, constant MAX_CNT should be set IDEAS FOR DESIGN WANTED
to 20 × 60 = 1200 (04B0H). The duration of each of four time
intervals will then be automatically set by firmware once the Send us your Ideas For Design. We’ll pay you $150 for every Idea For
maximum charge time is entered.
This approach is very flexible and can be applied to charge receiving $250. You can submit your Ideas For Design via:
any NiCd battery by choosing resistor R3 accordingly. In addi- t&NBJMSJDIBSEHBXFM!QFOUPODPN
tion, nearly any type of microcontroller can be used, because OR BY
the program is simple and uses only standard instructions. t1PTUBMNBJMUP
Ideas For Design
&MFDUSPOJD%FTJHO
1166 Avenue of the Americas, 10th Floor
ABEL RAYNUS is an engineer with Armatron International New York, NY 10036
Inc., Malden, Mass. Go to www.electronicdesign.com for our submission guidelines.

Vol. 2, No.6
Circuit Generates High-Frequency Sine/Cosine

Waves From Square-Wave Input
Direct digital synthesis ICs can convert square waves to sine/cosine waves
but their complexity is a disadvantage. This circuit uses simple CMOS logic
and two switched-capacitor filters to perform the conversion.
Determine Equivalent ESR, Ripple Voltage, And

Currents For Unequal Capacitors In Parallel
Capacitors are used in parallel to provide special performance attributes;
however, it’s difficult to determine critical performance specifics of ESR,
high-frequency ripple voltage, and individual RMS currents when they have
unequal values. This numerical approach provides a straightforward solution
to calculating the values based on published specifications.
Graphically Determine The Output Signal Level Of

An RC Filter
The need to know what the output filter’s value will be is especially important
for variable switching power supplies operating at different duty cycles. This
discussion provides a method for calculating that output.

ideas
Circuit Generates High-Frequency
Sine/Cosine Waves From Square-Wave Input
JOHN R. AMBROSE | MIXED SIGNAL INTEGRATION CORP. [email protected]
for design
ALTHOUGH QUITE A FEW direct digital synthesis (DDS) ICs can Two separate divider circuits are used. The 74HC393A
generate high-frequency sine waves, their complexity excludes divides the 50-MHz clock to 12.5 MHz. The 74HC390A is a
them from many designs. However, designers can use simple dual divide-by-2 and divide-by-5 device. By combining the
high-frequency CMOS logic and two switched-capacitor filters 74HC390 with the 74HC74A dual flip-flop, the 50-MHz clock
to create a sine/cosine generator. With newer filters, a 1-MHz can be divided to 500 kHz.
output at 1.7 V p-p is possible. The 74HC74A provides a Q and /Q output at half the fre-
The example circuit uses an MSHFS6 5-V, low-power 12.5:1 quency of the divide-by-25 output of the 74HC390A. Divid-
switched-capacitor filter with selectable Butterworth, Bes- ing the 74HC74A output by 2 with the divide-by-2 blocks in
sel, or elliptic filters in the lowpass mode and full-, 1/3-, or the 74HC390A creates two square waves –90° apart. Figure
1/6-octave filters in the bandpass mode. Since the lowpass 2 shows a 100-MHz square-wave input, a 12.5-MHz output
mode would cause a 3-dB loss of the signal output, the circuit for the filter clock, and 1-MHz sine and –cosine square-wave
uses the 1/6-octave bandpass filter, which is selected by tying output before the dividers. Resistor-divider circuits reduce
pins 1 and 3 high on the MSHFS6 (Fig. 1). the amplitude from rail to rail to prevent generation of distor-
+5 V
+5 V U3 C7
0.1 μF
U4 14 1 Clk AA VDDD 16
VCC 2 Rst A Clk AB 15
2 5 3 14
D Q QAA Rst B R1
3 74HC74A 6 4 13 10k
Clk /Q Clk BA QAB
+5 V 1 2 12
/Clr 5 R2
1 QBA Clk BB Cosine out
4 10k
/Pre 6 11
Gnd J1 QCA QBB 2
7 10
7 QDA QCB 1 C1
R11 8 9 0.1 μF
VSSD QDB J3 +5 V
10k
+5 V
R12 74HC390A
U2 R3
10k 1M
1 8 R4
FSEL In
2 7 100k
Out AGnd
+5 V C4 3 6
+5 V Type VSS C2 R5
0.1 μF +5 V C8 0.1 μF 0.1 μF
U1 4 5 100k
Sine out Clk VDD +5 V
R9 1M R8
1 8 14
2 FSEL In 100k
2 6 MSHFS6 C3
1 Out AGnd VCC 0.1 μF
3 7 1 3
J2 Type V SS C5 R10 /Clk QA
4 5 0.1 μF
Clk VDD 100k 2 4
Clr QB R6
MSHFS6 5 1k
+5 V QC
6
QD
Gnd R7
C6
0.1 μF 7 74HC393A 1k
1. Instead of a DDS IC, the sine/cosine generator uses simple CMOS logic and two switched-capacitor filters to provide a 1-MHz output at 3.0 V dc.

2. The 1-MHz sine and –cosine outputs of the generator (channels 3 3. Channels 1 and 2 show the outputs of the two switched-filter
and 4) result from the 100-MHz square-wave input (channel 1). capacitors with an input clock of nearly 50 MHz.
tion in the filters. The use of ac coupling at the MSHFS6 filter distortion on the sine output was only 0.1%. Although the
inputs ensures smoothed square waves centered around the 74HC390A and 74HC393A have a guaranteed maximum
filters’ analog ground. operating frequency of 50 MHz at 6 V, Mixed Signal Integra-
Figure 3 shows the output of the two filters with an input tion Corp. and other companies have found that spec to be
clock of nearly 50 MHz. If the inverted cosine is not accept- very conservative.
able, an op amp at the cosine filter output or the inverter at pin In this application, a 100-MHz input clock achieved the
13 of the 74HC390A can correct it. desired divide-by-4 and divide-by-100 needed to operate the
The Lissajous curve for the two outputs (Fig. 4) indicates newer MSVHFS6 switched-capacitor filter at 3.3 V. The only
that the phase circle matches the 89.1° reading in Figure 3. change needed was to reduce VDD to 3.3 V and replace the 5-V
Using a Krohn-Hite 6900B distor- MSHFS6 filters with the 3.3-V MSVHFS6.
tion analyzer and a 1-MHz Krohn The input clock was increased to 100 MHz.
Hite lowpass filter (to remove the Figures 5 and 6 show the filter outputs’
clock), the circuit’s total harmonic phase relationship in time and as a Lissa-
jous curve.
5 . T h e o r i g i n a l c i rc u i t u s e d t h e
MSHFS6 switched capacitor, but it
also works with the newer MSVHFS6 JOHN R. AMRBOSE is the vice president
version, which runs on 3.3 V rather of applications and system engineering at
than 5.0 V. This screen shows the two Mixed Signal Integration Corp.
outputs’ phase relationship in time.
4. The Lissajous curve for the circuit’s two outputs shows that the 6. The Lissajous curve for the circuit using the MSVHFS6 3.3-V filters
phase circle matches the 89.1° found in Figure 3. shows the outputs’ phase relationship.
Ideas for Design
Determine Equivalent ESR, Ripple Voltage,

And Currents For Unequal Capacitors In Parallel
ALEXANDER ASINOVSKI | MURATA POWER SOLUTIONS, MANSFIELD, MASS. [email protected]
CAPACITORS OFTEN ARE CONNECTED in parallel in power elec-

tronics to decrease high-frequency ripples, current stress,
power dissipation, and operating temperature, as well as to
shape frequency response and boost reliability. Yet designers Cs1 Cs2 Cs3 Cs4 Cs1 = Cs2 = Cs3 = Cs4 = C
have three critical questions about this technique: R s1 = R s2 = R s3 = R s4 = R s
t8IBUBSFUIFFRVJWBMFOUWBMVFTPGDBQBDJUBODF$se and equiv-

alent series resistance (ESR) Rse? 1. For an array of N identical capacitors in parallel, determining the
t8IBUJTUIFIJHIGSFRVFODZSJQQMFWPMUBHF total equivalent capacitance and ESR values is straightforward. For
t8IBUBSFUIFJOEJWJEVBM3.4DVSSFOUT unequal capacitors, the calculation can be difficult.
If all N capacitors in the parallel connection are identical 5 VN

;VN (3)
(Fig. 1) XJUIFRVBMDBQBDJUBODFWBMVFT$sk$BOEFRVBM&43
5 SN
5 VN
values Rsk = Rs, then for k = 1, 2, . . . N the answers are clear:
where:
t$se is directly proportional to the number of capacitors N:
$se/$ BOE3se is inversely proportional to N: Rse = Rs/N.
;VN (4)
/I&VN
t3JQQMFWPMUBHF7 3.4WBMVF
JT
is the reactance of the individual capacitor.
9 , 5 VH
;VH (1) "GUFSJOEJWJEVBMQBSBMMFMDBQBDJUBODFT$pk and resistances
Rpk are calculated according to Equations 2 and 3, equivalent
for a sinusoidal current excitation i(t) = I √2 sin (2πft) with QBSBMMFMDBQBDJUBODF$pe DBOCFFBTJMZGPVOEBTUIFTVNPG$pk:
frequency f, where Xse G$se) is the reactance of the 1
FRVJWBMFOUDBQBDJUPS$se BOE3.4WBMVF* BOEJOEJWJEVBM3.4 &SH - &SN (5)

N
currents in the capacitors are identical: Ik = I/N.
8IFOUIFDBQBDJUPSTJOUIFQBSBMMFMDPOOFDUJPOBSFOUJEFO- 5IFSFBMQBSUPGFRVJWBMFOUBENJUUBODFDBOCFGPVOEBTUIF
UJDBM XJUIEJGGFSFOUDBQBDJUBODF$ sk and ESR Rsk values, the sum of admittances 1/Rpk. Rpe can be obtained as a reverse
TPMVUJPOUPUIFQSPCMFNJTOUUSJWJBM5IFEJSFDUBQQSPBDIJTUP value of that sum:
obtain an analytical expression for the input impedance of the
5 SH 1
parallel connection in the algebraic form Z = Re Z – j ImZ = (6)
Zse Z and use the formulas Rse = Re Z, Xse*N; BOE$se = -5
N SN
1/(2πf Xse).
A less complicated approach is based on the conversion of 5IFTZTUFNTFRVJWBMFOUTFSJFTDBQBDJUBODF$se and ESR Rse
TFSJFT$sk, RskDPOOFDUJPOTUPFRVJWBMFOUQBSBMMFM$pk, Rpk con- DBOCFGPVOECZDPOWFSTJPOPGUIFQBSBMMFM$pe, Rpe connection
OFDUJPOT5PPCUBJOSFMBUJPOTIJQTCFUXFFO3pk and Rsk, and UPUIFFRVJWBMFOUTFSJFTDPOOFDUJPO$se, Rse5PPCUBJOSFMBUJPO-
BMTPCFUXFFO$pkBOE$sk, set the admittance Ypk of the parallel TIJQTCFUXFFO$seBOE$pe and also between Rse and Rpe, set
$pk, Rpk pair and admittance YskPGUIFTFSJFT$sk, Rsk pair con- impedance Zpe PGUIFQBSBMMFM$pe, Rpe and impedance Zse of
nections equal to each other: Ypk = Ysk, Re (Ypk) = Re (Ysk), and UIFTFSJFT$se, Rse connections equal to each other: Zpe = Zse,
Im (Ypk) = Im (Ysk
5IFO Re Zpe = Re Zse, Im Zpe = Im Zse5IFO
Csk
Cpk = 2

R (2) ; SH
1 + sk &VH &SH
(7)
Xsk 5 SH

5 SH 1A
5 VH
5 SH

(8) Cs1 = Cs2 = Cs3 = 22 ѥF, 4 mΩ
; SH Cs1 Cs2 Cs3 Cs4
Cs4 = 100 ѥF, 8 mΩ
where:

; SH (9)
/I&SH 2. This example of four capacitors, with three identical and one differ-
ent, illustrates how the computation scheme works in practice.
is the reactance of the equivalent parallel capacitor Cpe (Equa-
tion 5). • For equivalent parallel capacitance Cpe, its reactance Xpe and
Based on this analysis, the calculation procedure for equiva- equivalent parallel resistance Rpe of the structure according
lent series capacitance Cse, ESR Rse, voltage ripples V, and RMS to Equations 5, 9, and 6, we calculate Cpe = 115 μF, Xpe = 6.9
currents Ik in the capacitors is: mΩ, Rpe = 13.9 mΩ.
• According to Equations 7 and 8, the equivalent series capaci-
• Calculate reactances of individual capacitances according to tance Cse and ESR Rse are Cse = 143.4 μF, Rse = 2.76 mΩ.
Equation 4. • For RMS ripple voltage V based on Equation 1, we obtain V
• Determine equivalent parallel parameters Cpk, Rpk of the = 12.4 mV.
capacitors based on Equations 2 and 3. • RMS currents according to Equation 10 in ceramic and
• Calculate equivalent parallel capacitance Cpe of the struc- polymer capacitors are respectively: I1 = I2 = I3 = 341 mA,
ture, its reactance Xpe, and equivalent parallel resistance Rpe I4 = 1.1 A.
according to Equations 5, 9, and 6.
• Calculate equivalent series capacitance Cse and ESR Rse of This shows the technique can easily determine the param-
the structure according to Equations 7 and 8. eter values in each of the capacitors.
• Obtain RMS ripple voltage V using Equation 1.
• Calculate RMS currents Ik in the capacitors based on:
ALEXANDER ASINOVSKI is principal engineer at Mura-
9
,N (10) ta Power Solutions Inc., Mansfield, Mass. He holds BSEE and
5 VN
;VN MSEE degrees from State Technical University, St. Petersburg,
Russia, and a PhD from the University of Telecommunications,
Note that ESR values Rsk are strong functions of frequency. St. Petersburg.
A designer should use ESR data specified by capacitor manu-
facturers at a given frequency of operation, such as the data for
ceramic and polymer aluminum electrolytic capacitors from
Murata Manufacturing Co. Ltd. (MMC) (http://ds.murata. IDEAS FOR DESIGN WANTED
co.jp/software/simsurfing/en-us/index.html).
To illustrate the calculation procedure, let’s determine
equivalent parameters, voltage ripple, and current distribution
for a parallel connection of three ceramic capacitors (GRM-
receiving $250. You can submit your Ideas For Design via:
21BR60J226ME39L) and one polymer capacitor (ESASD-
40J107M015K00) from MMC (Fig. 2). Using the data f = 200 • E-mail: [email protected]
kHz, Cs1 = Cs2 = Cs3 = 22 μF, Rs1 = Rs2 = Rs3 = 4 mΩ, Cs4 = 100 OR BY
μF, Rs4 = 8 mΩ, I = 2 A, then: • Postal mail to:
Ideas For Design
• For reactance of each individual capacitance according to Electronic Design
Equation 4, we have Xsi = Xs2 = Xs3 = 3.6 mΩ, Xs4 = 0.8 mΩ. 1166 Avenue of the Americas, 10th Floor
• Equivalent parallel parameters Cpk, Rpk of the capacitors New York, NY 10036
based on Equations 2 and 3 are Cp1 = Cp2 = Cp3 = 21.7 μF, Rp1 Go to www.electronicdesign.com for our submission guidelines.
= Rp2 = Rp3 = 331 mΩ, Cp4 = 49.7 μF, Rp4 = 16 mΩ.
ideas
Graphically Determine The Output
Signal Level Of An RC Filter
GREGORY MIRSKY | ATLAS-MATERIALS TESTING COMPANY [email protected]
for design
WHEN DESIGNING FILTERS, engineers very often need to know ²7²
what the filtered value at the output will be. This is especially 9ORZ Q 9K Q² H
important for variable switching power supplies operating at
different duty cycles. The analysis presented here discusses where τ is the filter’s time constant and τ1 is the pulse duration
the calculation of the filtered output for RC filters, which are (Fig. 1). For simplicity, define:
common in feedback circuits, current transformers, output
²7²
devices, and other circuits. However, the analysis easily can be
H
extended to other kinds of filters.
²
It’s well known that for rectangular-pulse inputs the rms
² H
values of the unfiltered output parameters, like voltage and
²
current, are proportional to the square root of a duty cycle. But

H
the designer of a feedback filter for a switching power supply
must consider that the filter output is not an rms value but τ1 = DT (6)
rather proportional to the pulse’s duty cycle. You can adjust the
filter’s parameters to obtain an rms value, but it would be valid where D is the duty cycle of the input pulse train;
only for one duty-cycle value.
The analysis employs several assumptions: τ = kT (7)
• The filter works in a continuous-conduction mode, which where k is the number of periods for the filter time constant.
means the filter’s time constant is much greater than the So, Equations 1 and 2 become:
pulse’s repetition rate (period).
• The filter capacitor’s charge and discharge time constants are Vhn = VAβ +Vlownγ (8)
the same. That means the impedance sourcing signal to the
filter is much lower than the filter resistance and the filter’s
load impedance is much higher than the filter’s resistance,
which you can easily obtain by using an operational ampli- VA
Vh0
fier as a decoupling component. Vh1
• The filter’s voltage grows in exponential steps (charging

and discharging) that become smaller as the output voltage
Voltage
approaches its limit value.

• The filter is loaded with an impedance that is so high that it
can be ignored.
We did the analysis with MathCAD 15, and anyone with the Vlow1
appropriate license can reproduce the results. Vlow0
The expressions for the nth values of the filtered voltage’s 0 τ1 T
Time
maximum, Vh, and minimum, Vlow, are:
² 1. The plot of the input pulses (dashed lines) and filter output (solid lines)
9K Q 9ORZ Q
9$ ² 9ORZ Q ² H

indicates the definition of the parameters used in the calculations.
Ideas for Design
9K Q ² 9ORZ Q
Vlown = Vh(n–1)α (9) 9OLP LW Q 9ORZ Q

To derive the equation for the limit value of the filter output,
you must use the recurrent equations for the low and high You can calculate the limit for the normalized output volt-
values and calculate the average, which is based on the initial age using MathCAD tools:
parameters only. The equations for six high/low pairs are: '² '
² ²
9$
9$ H N
² 9$ H N
² 9$ H N
Vh0 = VAβ
/LPLW
²
N
Vlow0 = 0 9$ H ²
Vh1 = VAβ(γα + 1) or: ²

'
'² ²

Vlow1 = VAβα H N
² HN
H N
²
/LPLW

²
N
H ²
Vh2 = VAβ(γ2α2 + γα + 1)
Vlow2 = VAβα(γα + 1) (See the derivation of Equations 13 and 14 in “Normalized
Output Voltage Limit Derivation,” p. 91.)
Vh3 = VAβ(γ3α3 + γ2α2 + γα + 1) Equation 14 is a transcendent equation with two variables,
Vlow3 = VAβα(γ2α2 + γα + 1) which is hard to solve symbolically. But you can find the limit
value graphically by fixing the value of k and plotting the limit
Vh4 = VAβ(γ4α4 + γ3α3 + γ2α2 + γα + 1) as a function of the duty cycle, D. If k = 100.0:
Vlow4 = VAβα(γ3α3 + γ2α2 + γα + 1)
²' ' ² ²
H N
² H NH N

H N
²
Vh5 = VAβ(γ5α5 + γ4α4 + γ3α3 + γ2α2 + γα + 1)
/LPLW'
²
N
Vlow5 = VAβα(γ4α4 + γ3α3 + γ2α2 + γα + 1) H ²
Note from the above equations that: A plot of this function shows that the limit of the normal-
ized RC filter output is the duty cycle, D (Fig. 2). To create an
Q²
9ORZ Q 9$ P P

example, assign values to the variables in Figure 1:
P
VA = 1 V
and:
Q D = 0.6
9K Q 9$ P P

P
T = 20 μs
The average for the output voltage is:
τ1 = DT
1.0 0.6
k = 50
Voltage as a function of D
0.8
0.4
τ = kT
0.6
Limit (D)
Vlimitn
0.4 Then:
0.2 0.2
α = e–(T – τ1)/τ = 0.992
0
0 0.2 0.4 0.6 0.8
Duty cycle (D) β = 1 – e–τ1/τ = 0.012
0 100 200 300 400
2. The normalized output voltage of the n
RC filter is the duty cycle value of the 3. The limit value of the filtered voltage is τ = e–τ1/τ = 0.988
rectangular input pulses. The result does equal to the input voltage’s amplitude times
not depend on the value of k. the duty cycle. Then:

Q Q²
Q²
9$ P P
² 9$ P P

OLP 9$
Q
P P

P

P
9
P
To plot the graph for the limit, assign n =

1…400. Then:
NORMALIZED OUTPUT VOLTAGE LIMIT DERIVATION
Q²
9ORZ Q 9$ P P

Equation 12 can be interpreted as:
P
Q Q²
Q 9$ P P
² 9$ P P

9K Q 9$ P P

OLP 9$
Q²
P P

P P

P Q
P
9K Q ² 9ORZ Q
9OLPLW Q 9ORZ Q

Its solution is:
)LJXUHVKRZVWKHUHVXOWLQJSORW

VLJQXP9$ LI

GREGORY MIRSKY is a principal electrical
engineer at Atlas Materials Testing Company. OLP
9$ Q Q
²
9 $
Q Q
² LI

Q ² ²
He holds an MS from the St. Petersburg Baltic
Technical University, Russia, and a PhD from the
Moscow State Pedagogical University. Only the lower expression is of interest:
OLP
9$ Q Q
²
9 $
Q Q
²
IDEAS FOR Q ² ²
DESIGN WANTED Its solution is:

Send us your Ideas For Design. We’ll pay you

$150 for every Idea For Design that we publish. LI
In addition, this year’s top design as selected by
our readers will earn an additional $500, with

9$ ²

LI
two runners-up each receiving $250. You can
²
submit your Ideas For Design via: 9$

LI
²
• E-mail: [email protected]
OR BY
Again, only the lower expression is of interest:
• Postal mail to:
Ideas For Design ² '7 7²'7
Electronic Design 9$ ² H N7
H N7

1166 Avenue of the Americas, ² ² '7 ²7²'7

10th Floor H N7
H N7
²
New York, NY 10036
Go to www.electronicdesign.com
for our submission guidelines.
which can be converted into Equation 13.
Vol. 2, No.6
Circuit Generates High-Frequency Sine/Cosine

Waves From Square-Wave Input
Direct digital synthesis ICs can convert square waves to sine/cosine waves
but their complexity is a disadvantage. This circuit uses simple CMOS logic
and two switched-capacitor filters to perform the conversion.
Determine Equivalent ESR, Ripple Voltage, And

Currents For Unequal Capacitors In Parallel
Capacitors are used in parallel to provide special performance attributes;
however, it’s difficult to determine critical performance specifics of ESR,
high-frequency ripple voltage, and individual RMS currents when they have
unequal values. This numerical approach provides a straightforward solution
to calculating the values based on published specifications.
Graphically Determine The Output Signal Level Of

An RC Filter
The need to know what the output filter’s value will be is especially important
for variable switching power supplies operating at different duty cycles. This
discussion provides a method for calculating that output.

ideas
for design
+5 V
+5 V U3 C7
0.1 μF
2 5 3 14
D Q QAA Rst B R1
3 74HC74A 6 4 13 10k
Clk /Q Clk BA QAB
+5 V 1 2 12
/Clr 5 R2
4 10k
/Pre 6 11
Gnd J1 QCA QBB 2
7 10
7 QDA QCB 1 C1
R11 8 9 0.1 μF
VSSD QDB J3 +5 V
10k
+5 V
R12 74HC390A
U2 R3
10k 1M
1 8 R4
FSEL In
2 7 100k
Out AGnd
+5 V C4 3 6
+5 V Type VSS C2 R5
0.1 μF +5 V C8 0.1 μF 0.1 μF
U1 4 5 100k
R9 1M R8
1 8 14
2 FSEL In 100k
2 6 MSHFS6 C3
3 7 1 3
4 5 0.1 μF
Clk VDD 100k 2 4
Clr QB R6
MSHFS6 5 1k
+5 V QC
6
QD
Gnd R7
C6
0.1 μF 7 74HC393A 1k

jous curve.
Ideas for Design




;VN (3)
5 SN
5 VN
where:
;VN (4)
/I&VN
JT
9 , 5 VH

N
5 SH 1
N SN
1/(2πf Xse).
Im (Ypk) = Im (Ysk
Csk
Cpk = 2

R (2) ; SH
1 + sk &VH &SH
(7)
Xsk 5 SH

5 SH 1A
5 VH
5 SH

(8) Cs1 = Cs2 = Cs3 = 22 ѥF, 4 mΩ
Cs4 = 100 ѥF, 8 mΩ
where:

; SH (9)
9
5 VN
Ideas For Design
= Rp2 = Rp3 = 331 mΩ, Cp4 = 49.7 μF, Rp4 = 16 mΩ.
ideas
for design
²7²
H
²
² H
²

H
Vh0

Voltage

can be ignored.
Time
9K Q 9ORZ Q
9$ ² 9ORZ Q ² H

Ideas for Design
9K Q ² 9ORZ Q

² ²
9$
9$ H N
² 9$ H N
² 9$ H N
Vh0 = VAβ
/LPLW
²
N
Vlow0 = 0 9$ H ²
Vh1 = VAβ(γα + 1) or: ²

'
'² ²

Vlow1 = VAβα H N
² HN
H N
²
/LPLW

²
N
H ²
Vh2 = VAβ(γ2α2 + γα + 1)
²' ' ² ²
H N
² H NH N

H N
²
/LPLW'
²
N
Q²
9ORZ Q 9$ P P

P
VA = 1 V
and:
Q D = 0.6
9K Q 9$ P P

P
T = 20 μs
τ1 = DT
1.0 0.6
k = 50
0.8
0.4
τ = kT
0.6
Limit (D)
Vlimitn
0.4 Then:
0.2 0.2
α = e–(T – τ1)/τ = 0.992
0
0 0.2 0.4 0.6 0.8
Duty cycle (D) β = 1 – e–τ1/τ = 0.012
0 100 200 300 400

Q Q²
Q²
9$ P P
² 9$ P P

OLP 9$
Q
P P

P

P
9
P

1…400. Then:
Q²
9ORZ Q 9$ P P

P
Q Q²
Q 9$ P P
² 9$ P P

9K Q 9$ P P

OLP 9$
Q²
P P

P P

P Q
P
9K Q ² 9ORZ Q
9OLPLW Q 9ORZ Q

Its solution is:

VLJQXP9$ LI

9$ Q Q
²
9 $
Q Q
² LI

Q ² ²
OLP
9$ Q Q
²
9 $
Q Q
²
IDEAS FOR Q ² ²



9$ ²

LI
²

LI
²
OR BY
• Postal mail to:
H N7


10th Floor H N7
H N7
²
New York, NY 10036
ideas
for design
+5 V
+5 V U3 C7
0.1 μF
2 5 3 14
D Q QAA Rst B R1
3 74HC74A 6 4 13 10k
Clk /Q Clk BA QAB
+5 V 1 2 12
/Clr 5 R2
4 10k
/Pre 6 11
Gnd J1 QCA QBB 2
7 10
7 QDA QCB 1 C1
R11 8 9 0.1 μF
VSSD QDB J3 +5 V
10k
+5 V
R12 74HC390A
U2 R3
10k 1M
1 8 R4
FSEL In
2 7 100k
Out AGnd
+5 V C4 3 6
+5 V Type VSS C2 R5
0.1 μF +5 V C8 0.1 μF 0.1 μF
U1 4 5 100k
R9 1M R8
1 8 14
2 FSEL In 100k
2 6 MSHFS6 C3
3 7 1 3
4 5 0.1 μF
Clk VDD 100k 2 4
Clr QB R6
MSHFS6 5 1k
+5 V QC
6
QD
Gnd R7
C6
0.1 μF 7 74HC393A 1k

jous curve.
Ideas for Design




;VN (3)
5 SN
5 VN
where:
;VN (4)
/I&VN
JT
9 , 5 VH

N
5 SH 1
N SN
1/(2πf Xse).
Im (Ypk) = Im (Ysk
Csk
Cpk = 2

R (2) ; SH
1 + sk &VH &SH
(7)
Xsk 5 SH

5 SH 1A
5 VH
5 SH

(8) Cs1 = Cs2 = Cs3 = 22 ѥF, 4 mΩ
Cs4 = 100 ѥF, 8 mΩ
where:

; SH (9)
9
5 VN
Ideas For Design
= Rp2 = Rp3 = 331 mΩ, Cp4 = 49.7 μF, Rp4 = 16 mΩ.
ideas
for design
²7²
H
²
² H
²

H
Vh0

Voltage

can be ignored.
Time
9K Q 9ORZ Q
9$ ² 9ORZ Q ² H

Ideas for Design
9K Q ² 9ORZ Q

² ²
9$
9$ H N
² 9$ H N
² 9$ H N
Vh0 = VAβ
/LPLW
²
N
Vlow0 = 0 9$ H ²
Vh1 = VAβ(γα + 1) or: ²

'
'² ²

Vlow1 = VAβα H N
² HN
H N
²
/LPLW

²
N
H ²
Vh2 = VAβ(γ2α2 + γα + 1)
²' ' ² ²
H N
² H NH N

H N
²
/LPLW'
²
N
Q²
9ORZ Q 9$ P P

P
VA = 1 V
and:
Q D = 0.6
9K Q 9$ P P

P
T = 20 μs
τ1 = DT
1.0 0.6
k = 50
0.8
0.4
τ = kT
0.6
Limit (D)
Vlimitn
0.4 Then:
0.2 0.2
α = e–(T – τ1)/τ = 0.992
0
0 0.2 0.4 0.6 0.8
Duty cycle (D) β = 1 – e–τ1/τ = 0.012
0 100 200 300 400

Q Q²
Q²
9$ P P
² 9$ P P

OLP 9$
Q
P P

P

P
9
P

1…400. Then:
Q²
9ORZ Q 9$ P P

P
Q Q²
Q 9$ P P
² 9$ P P

9K Q 9$ P P

OLP 9$
Q²
P P

P P

P Q
P
9K Q ² 9ORZ Q
9OLPLW Q 9ORZ Q

Its solution is:

VLJQXP9$ LI

9$ Q Q
²
9 $
Q Q
² LI

Q ² ²
OLP
9$ Q Q
²
9 $
Q Q
²
IDEAS FOR Q ² ²



9$ ²

LI
²

LI
²
OR BY
• Postal mail to:
H N7


10th Floor H N7
H N7
²
New York, NY 10036
Vol. 2, No.7
Specialized Circuit Drives 150-V Piezoelectric

Motor Using Low-Voltage Op Amp
Use a basic op amp and discrete components to drive a piezoelectric motor
and provide circuit protection.
Model Diacs And Triacs For AC-Line Control

Diacs and triacs are often used for control of ac lines and are seeing
increasing use in Internet-enabled line switches. However, the traditional
approach of modeling them using bipolar transistors and diodes, and table-
based models, can be confusing and offers limited success. This article
explores functional models based on the LTspice simulator for improved
results.
Microcontroller Solves Complex Temperature

Polynomial Equations
This fast-executing program and simple circuit allows a PIC MCU to linearize
a thermistor and accurately determine and display the temperature from its
measured resistance.
the point, often with a single figure or program listing to help
Design for
Electronic
Design?
ideas
ALAN STUMMER | UNIVERSITY OF TORONTO [email protected]
for design
A PIEZOELECTRIC MOTOR is a linear motor with bidirectional The problem is the voltage. Fortunately, it is unipolar.
motion. It uses friction to grip the armature while a voltage is Unfortunately, it is +150 V (peak). The required current is
ramped to warp the piezoelectric material and move the arma- quite small. It has to be only enough to charge and discharge
ture. The voltage then is quickly removed. the 20-nF piezo element. A calculation using charge transfer
As the material springs back, it breaks away from the arma- (Q) shows that during the ramp phase:
ture and returns to its zero position, leaving the armature a few
micrometers further along its track. Repeat this at a kilohertz Q = It = CV
rate and for thousands of times.
While each of the motions is very small, after several sec- where:
onds you may see that the armature has moved, if you look
carefully. (Full disclosure: I had never heard of piezo motors t = 1 ms
before being asked to make a driver for one.) C = 20 nF
There are two drive waveforms, one for forward and the V = 150 V
other reverse: a sawtooth waveform with a slow linear rise
followed by a fast fall, and its complement with a fast rise and therefore:
slow linear fall. This was done using an op-amp triangle-wave
oscillator at 1 kHz, with diodes switched in to speed up either I = CV/t = 3 mA.
the rising or falling edges to about 5% of the cycle. The driver’s
required bandwidth is only 10 to 15 kHz. A 600-mW boost converter can be used to switch the +12 V
up to +200 V with a 3-mA load require-
ment. The most straightforward circuit
R2 R3 +200 V uses an op amp rated to at least 200 V. Op
499k 499k
amps are available with this voltage rat-
R4
16.5k ing, but they are meant for high-current
C1 +200 V
applications and are quite expensive.
C2 10 pF R5
10 pF 16.5k Q1 The circuit in the figure is much
ZVN4525G cheaper and based on a common op amp
R6 R8 Piezo
+12 V 16.5k D1 D2 used as a non-inverting amplifier. The
49.9 element
0 to
heart of the circuit is the current mirror
V+ R7 +12 V SMAZ16-13-F 150 V of R7, N-channel FET Q3, and series-
– 1.5k
connected R4, R5, and R6. (The rea-
Input + IC1 Q3 Q2
son for using three resistors in series is
0 to 10 V AD8065AR ZVN4525GTA ZVP4525G
V– explained below.)
R1
66.5k With Q3 in a common-gate configura-
tion and the gate at +12 V, its source volt-
age remains reasonably constant at +10
A low-voltage op amp can be used to drive a higher-voltage piezoelectric motor. D1 and D2 V (from +12 V less the Vgs(on)). Any out-
protect the circuit against load shorts, while resistors in series strings are used to reduce indi- put of op-amp IC1 less than that +10 V
vidual resistor dissipation and minimize their voltage coefficient of resistance. causes a voltage drop across R7, with the

current coming from R4 to R6. Since R7 has the same current 0805/2012 surface-mount (SMT) resistors can be used. The
as R4 to R6, there is a voltage gain of the ratio of (R4 to R6)/ feedback resistance composed of R2 and R3 is split into two
R7 or 33. The voltage at the bottom of R4 to R6 (the Q3 drain) devices to reduce the voltage coefficient of resistance. This is
now has the required voltage swing but a high impedance. a somewhat obscure effect, where a component’s resistance
A pair of complimentary FETs acts as followers to lower the actually changes slightly at higher voltages.
output impedance and boost current output. Negative feedback There is a small feedback capacitor directly on the op amp
is provided via R2 and R3, along with bandwidth limiting by C1. to provide stability. Without it, the parasitic capacitances
The overall closed-loop gain is set by (R2 + R3)/R1 + 1 = 16. (most notably the Miller capacitances C dg and the piezo’s
There are a few subtleties to the circuit. If something hap- capacitance) would cause enough phase shift for the op amp
pened to the piezo motor, such as the user shorting it, Zener to oscillate. Crossover distortion is not a problem because the
diodes D1 and D2 would protect the FET gates. The current- piezo element could not react to that bandwidth and does not
mirror high-side resistance provided by R4 through R6 is require a perfect waveform.
split into three distinct devices to handle the power levels so Another problem would be the infinite pole caused by a
purely capacitive piezo element. By taking the feedback direct-
ly from the sources of Q1 and Q2, resistor R8 would add a
ALAN STUMMER designs electronics for the University of Toron- more deterministic pole to the load and increase loop stability.
to’s quantum optics and pulse laser research labs, after years Once the somewhat finicky piezo motor was adjusted, the
working on analog controls for heavy industry and then telecom. final circuit worked well. The armature would move back and
forth with incredibly fine resolution.
Model Diacs And Triacs

For AC-Line Control
APPARAJAN GANESAN | CONSULTANT [email protected]
DIACS AND TRIACS often are used for line-voltage control. They The device continues in this state until the current through
also are getting additional interest as part of Internet- and it falls below It. At that point the flip flop is reset and the device
cell-phone-controlled power-line switches such as the Belkin switches once again to its off state, with a resistance Roff. (Note
WeMo Home Automation Switch. that the device is bi-directional.)
The traditional approach has been to model them using The triac is modeled in a similar manner (Figure 2 and
bipolar transistors and diodes.1 Table-based models have been Listing 2; Listing 2a is for the triac macromodel; Listing 2b is
used with varying success.2 The functional model approach the triac symbol). It has the same four parameters as the diac
shown here works well and has been used extensively with the
LTspice simulator (free, from Linear Technology).
In the diac schematic of Figure 1 and associated Listing
1, there are four parameters (Listing 1a is for the diac macro-
model; Listing 1b is the diac symbol; listings are available with
the online version of this article at electronicdesign.com):
Vt: voltage at which the diac triggers

It: current at which it turns off
Ron: resistance when it is on.
Roff: resistance when it is turned off
To turn on, the diac needs the voltage to exceed Vt. Once
it turns on, it needs the current to go below It to turn off. In
operation, the device starts as an open circuit. When the volt-
age across it exceeds Vt, the flip-flop is set, putting the device 1. In the diac circuit, the device starts as an open circuit and remains
in its on state with a low resistance of Ron. open until the voltage across it exceeds Vt
Ideas for Design
2. For the triac, the trigger voltage is on an independent port, in con- 3. The results (top) of the diac test circuit (bottom) show the sym-
trast to the diac. metrical operation of the circuit and allow for user modifications.
component, and it needs the voltage to exceed Vt to turn on. Figure 3 (bottom) and Listing 3b). As VIn goes lower, the device
Once it turns on, it needs the current to go below It to turn off. turns off once the current falls below It. It again reaches its on
The difference between the diac and the triac is that the triac state once the voltage exceeds Vt. The device is bidirectional
trigger voltage is on an independent port. and inherently symmetric. Also, the model can be modified to
It is difficult to design test circuits for these devices because, include current limits on voltages, as well as asymmetries.
as a result of their negative resistances, they usually oscillate or For the triac, the popular dimmer circuit is used for test,
provide limit cycles, which in turn makes it difficult for pro- with output file Figure 4 (top) and Listing 4a along with cor-
grams such as Spice to converge. The main diac characteristics responding schematic Figure 4 (bottom) and Listing 4b. The
of interest are breakover voltage, voltage symmetry, breakback load is a typical 100-W bulb, and the RC-time constant deter-
voltage, breakover current, and power dissipation.3 mines when the triac is triggered.
In the output file for the diac test circuit, VIn is initially at The diac in series with the gate is chosen to ensure the triac
its negative extreme, and VOut is low, as the device is in its on turns off completely. The model can be modified to include
state (Figure 3 (top) and Listing 3a, and corresponding schematic current limits, voltage asymmetries, dV/dt effects, and more.
By varying the RC-time constant, the duty cycle of the output
can be varied from 5° to about 170°, nearly spanning the full
0° to 180° theoretical limit. The device as modeled produces
identical results in the positive and negative half cycles.
REFERENCES
¸(:WPJL4VKLSMVY;YPHJZ¹(-7L[YPLHUK*OHYSLZ/`TV^P[a^^^YLHKIHN
JVTPU[\ZVM[HY[PJSLZ[YPHJ
¸4VKLSMVY+PHJPU6YJHK ¹^^^LKHIVHYKJVT[OYLHK O[TS
¸+PHJ;\[VYPHS¹(TLYPJHU4PJYVZLTPJVUK\J[VY^^^HTLYPJHUTPJYVZLTPJVT
PUMVYTH[PVU[\[VYPHSPUKL_WOW&[FPK$
P_aaL6I^V0N>
APPARAJAN GANESAN has an MS from the State Univer-

sity of New York (SUNY) at Stony Brook and has been designing
circuits for 40 years. He has worked at companies including
General Instrument, Bell Telephone Laboratories (Bell Labs),
4. The results (top) of the dimmer circuit (bottom) using a diac and RCA, and Analog Devices and has been an independent
triac show the operation possible with a standard resistive load.

ideas
Microcontroller Solves Complex
Temperature Polynomial Equations
for design
RICARDO JIMENEZ AND ROBERTO C. SOLORIO | INSTITUTO TECNOLOGICO DE MEXICALI [email protected]
WHEN YOU NEED TO measure temperature using thermistors, ,QSXW$1LVGHFODUHGDVDQDORJDQGWKHUHVWRIWKH,2OLQHV

you face the challenge of linearizing their response to get DUHGHFODUHGGLJLWDORXWSXWV7KHDQDORJWRGLJLWDOFRQYHUWHU
accurate readings. One of the best methods for linearizing a $'&LVFRQILJXUHGIRUELWVZLWKDVDPSOLQJWLPHRIV
thermistor is the polynomial “Steinhart-Hart” equation (S-H), 7KHWKHUPLVWRU¶VYROWDJHUHDGLQJLVVWRUHGLQELQDU\IRUPDW
which has an error of 0.1°C. The temperature range for this LQWKHYDULDEOH³YROW´ZLWKWKHLQVWUXFWLRQ$'&,1YROW
design is from 0°C to 100°C. 7KHQWKLVUHDGLQJLVPXOWLSOLHGE\WRFRQYHUWLWWRGHF
7KUHHWKHUPLVWRUFRHIILFLHQWV²DEDQGF²DUHUHTXLUHG LPDODQGLWLVVWRUHGLQDQHZYDULDEOHFDOOHG³Y´7RREWDLQ
WRLPSOHPHQWWKH6+HTXDWLRQ,IWKHPDQXIDFWXUHUGRHVQRW HDFKGHFLPDOGLJLWIURPYDULDEOH³Y´WKHFRPPDQGV',*
SURYLGHWKHFRHIILFLHQWV\RXFDQREWDLQWKHPE\VROYLQJWKH ',*DQG',*DUHXVHGWRVWRUHWKHVHUHDGLQJVLQWKUHHYDUL
6+HTXDWLRQ(TXDWLRQVDQGIRUWKUHHGLIIHUHQWWHP DEOHVFDOOHGGLJGLJDQGGLJUHVSHFWLYHO\
SHUDWXUHSRLQWV,QWKLVGHVLJQD3,&0&8ZLOOVROYHVXFK 7KHFRGHDYDLODEOHZLWKWKHRQOLQHYHUVLRQRIWKLVDUWLFOH
HTXDWLRQVWRSURYLGHDFFXUDWHUHDGLQJVLQPV)LJ DWHOHFWURQLFGHVLJQFRPVKRZVWKHVRIWZDUHSURJUDPZKHUH
7RLOOXVWUDWHZHZLOOXVHWKHQHJDWLYHWHPSHUDWXUHFRHI WKHIROORZLQJYDULDEOHVDUHGHFODUHGIRUSURFHVVLQJDQGVWRU
ILFLHQW17&/0WKHUPLVWRUZKLFKKDVWKHUHVLVWDQFH LQJWKHUHDGLQJV%$&/YROWYYSDWWHUQSDWWHUQ
DWWKUHHGLIIHUHQWWHPSHUDWXUHVVKRZQLQ7DEOHWith these SDWWHUQ,GLJLWGLJLWDQGGLJLW7ZRYDULDEOHVFRQY
three measurements, the temperature is converted to the Kel- DQGFRQYDUHUHTXLUHGIRUELQDU\WRGHFLPDOFRQYHUVLRQV
vin scale and substituted into three Steinhart-Hart equations: 7KH\LQFOXGHWKHELWUHVROXWLRQ$'&¶VOHDVWVLJQLILFDQWELW
/6%ZKLFKLVHTXDOWRP9

7 7KHDOJRULWKPWKDWZDVGHYHORSHGWRJHWDWHPSHUDWXUHUHDG
D
E OQ5
F>OQ5 @ LQJXVLQJWKH6WHLQKDUW+DUWHTXDWLRQUHTXLUHVVHYHUDOVWHSV
)LUVWFompute the power supply voltage (v1), which is equal
7 to 48828 × volt2. Variable volt2 is read with the PIC’s 10-bit
D
E OQ5
F>OQ5 @
ADC. If the power supply is correct, volt2 = %1111111111,
and then v1 = 49,951,044. Now with the instruction DIV32,

7
D
E OQ5
F>OQ5 @
TABLE 1: THERMISTOR RESISTANCE
6ROYLQJWKHVHWKUHHHTXDWLRQV\LHOGVWKHFRHIILFLHQWVDE Resistance
Temperature (°C)
(kΩ)
DQGF7DEOH6XEVWLWXWLQJWKHFRHIILFLHQWVLQWKH6WHLQKDUW
2.7 23
+DUWHTXDWLRQZHJHWWKHWKHUPLVWRUHTXDWLRQ
27.5 8.8

7W 66 2.7
= ²
= ² /Q5W ² = ² >/Q5W@
RU TABLE 2: LM103 COEFFICIENTS
=
7W

OQ5W² /Q5W Coefficients Value
A 2.501 x 10±4
7KHFKDOOHQJHIRUWKH3,&)PLFURFRQWUROOHULVWR
B 3.505 x 10±4
XVH(TXDWLRQWRJHWWUXHWHPSHUDWXUHUHDGLQJV7KLVSURM
C ±1.415 x 10±7
HFWFRQVXPHVZRUGVRIWKHPLFURFRQWUROOHU¶VPHPRU\
Ideas for Design
the variable volt2 is divided by 10,000, giving a result equal to part). To compute it, we need the result of the successive divi-
4995, which represents the approximate value of 5 V from the sions “wu” by 2 (). Nonetheless this operation or process can-
power supply. not be applied directly, since dividing 47/2 results in 23.5, and
Second, the thermistor’s voltage drop is computed with the we lose the fractions part. (PIC Basic Pro does not work with
same process from the first step, but with another analog-to- fractions.) Therefore, to keep the fractions part, we create a
digital channel. Let’s say we have a reading in the ADC of 2.5 special subroutine that performs the following series:
V (adc = 1000000000). The voltage read is then 24,999,936,
which is divided by 10 with DIV32 for a result of 2499, which 47/2 = 23.5, 23.5/2 = 11.75,
approximates to 2.500 V. 11.75/2 = 5.875, 5.875/2 = 2.9375, 2.9375/2 = 1.46875
Third, we proceed to find the resistor’s value with:
5 = Y When dividing 47/2 = 23, with a remainder of 1, this “1”
5 appears in the first division of the five that will be performed.
( ² Y
Therefore, the fraction 1/25 = 0.03125. We cannot work with
To achieve that, we store in a variable called “dif ” the power’s fractions, though, so instead of dividing by 1 by 2 n, we take
supply voltage minus the thermistor’s voltage. Then we mul- the number 10,000 as a numerator, and we create a subroutine
tiply the thermistor’s voltage by the 10-kΩ fixed resistor, and that computes the denominator depending on the number of
we apply DIV32 to divide this last product by “dif,” resulting in division where appears a remainder of 1.
the thermistor’s resistance value. Then, we sum all the results of the divisions. To continue
Fourth, we now compute the base 2 logarithm from the with this example, we have the following five steps:
thermistor’s resistance with
Equations 7, 8, and 9:
x2 26 PortC.7 A1 35
5V 11
0 In Out VDD 27 PortD.4 B1 34
0 D RG1 32
17 PortC.2 C1 7
DIG 1
7805 Ref 18 PortC.3 D1 6
MCLR 1 19 PortD.0 E1 5
B1 + R3
0 25 PortC.6 F1 36
0 D 100k 24 PortC.5 G1 37

DP1
9V
0
0 VSS
12
30 PortD.6 A2 30
D 31
33 PortB.0 B2 29
10 PortE.2 C2 11
where a 0 , a 1 , and a 2 are the
DIG 2
13 PortA.7 D2 10
VDD 14 PortA.6 E2 9
next-digit mantissa in base 2.
PIC16F884/887
29 PortD.6 F2 31
Next, we find the character- 28 PortD.5 G2 32
R1
istic a0 or logarithm’s integer
DP2
COL
10k
with the function NCD, which
36 PortB.3 A3 25
delivers the most significant 37 PortB.4 B3 24
6 PortA.4 C3 15
DIG 3
bit (MSB) of a number. If we PortA.0 2

7 PortA.5 D3 14
decrement it by 1, we get the 8 PortE.0 E3 13
35 PortB.2 F3 26
base-2 logarithm character- 34 PortB.1 G3 27
istic. We know, for example, R2
DP3
10k at 25°C
that the base-2 logarithm for
number 47 is 5.554 using four VSS 40 PortB.7 A4 21
E
21 PortD.2 B4 20
DIG 4
significant digits. Thus, NCD 22 PortD.3 C4 19

D
23 PortC.4 D4 18
(47) = 6, where 47 in binary is 5 PortA.3 E4 17
C
00101111, and the MSB is in 39 PortB.6 F4 22

B
38 PortB.5 G4 23
the sixth position from left to
right. If we decrement it by 1, 9 PortE.1 DP3 16
20 PortD.1 COM 1
we get number 5, which is the 40
logarithm’s characteristic.
To get the logarithm, we 1. This simple circuit for a high-precision thermometer drives a glass LCD with high efficiency, using the
need the Mantissa (fractions PIC 16F887 microcontroller.

LINEARIZATION INSPIRES A SOLUTION
WE ORIGINALLY CONCEIVED of this idea as a solution

to the problem faced by engineers in trying to linearize a
thermistor’s response. We know there are analog methods
to achieve this, such as putting a parallel resistor across the
thermistor, constant current bias, or a combination of both.
Then you get an almost linear response defined by a straight
line. In this case, you have to design a signal conditioning
circuit based on op amps, which adds more hardware to the
project. Another option we had was using the thermistor’s
characteristic equation, but it is not accurate over the entire
temperature range. Therefore, we decided to use the vener-
able Steinhart-Hart equation.
With this in mind, we faced a new challenge: how to get
2. A signal at 40-Hz delivered by RC2 is used for driving the LCD’s the logarithm of a voltage reading. This had to be done with-
common pin. out getting overflow errors in the microcontroller. After trying
with different methods (Taylor series, Babylonian, etc.), we
• 47/2 = 23, with a remainder of 1; this is the first division of 5, found that the most appropriate algorithm for this project was
10,000/25 = 312.5 = 312 the one from Dr. G. Mayer in his paper “computing logarithms
• 23/2 = 11, with a remainder of 1; this is the second division digit-by-digit.” So the method was developed in base two,
of 4, 10,000/24 = 625 and at the end of the process, it was converted to base 10.
• 11/2 = 5, with a remainder of 1; 10,000/23 = 1250 Once this was achieved, the final step was to substitute the
• 5/2 = 2, with a remainder of 1; 10,000/22 = 2500 resulting Ln values into the S-H equation to get accurate
• 2/2=1, with a remainder of 0; there is no portion to add temperature readings.
The total algorithm’s processing time was 32 ms with the
,QWKLVFDVHWKHYDULDEOH³-D´LVHTXDOWRDQGLWVWLOO internal microcontroller’s oscillator set at 4 MHz. Some read-
QHHGVWREHLQFUHPHQWHGE\RQH7KHUHIRUHZHGLYLGH³-DÉ\ ers have been asking how to reduce the processing time.
DQGWKHQZHDGGXS7KHUHIRUH-D REWDLQ It can be reduced to less than 10 ms by selecting a higher
LQJDSSUR[LPDWHO\WKHRULJLQDOTXDQWLW\PHQWLRQHGDERYH oscillator’s frequency in the algorithm itself.
PXOWLSOLHGE\ZLWKIRXUVLJQLILFDQWGLJLWV We decided to drive a numerical LCD directly without
2QFHZHKDYHREWDLQHG0ZHQHHGWRJHWWKHPDQWLVVD external decoders for energy-saving applications. You can
XVLQJWKHSURFHVVGHVFULEHGLQ)LJXUH7KHGLYLVLRQE\DQ choose other types of displays to free the microcontroller’s
LVQRWFRPSOLFDWHGEHFDXVHZHDUHZRUNLQJLQEDVHDQGWKH ports for other control applications. We are currently working
GLYLVLRQLVSHUIRUPHGE\RU:KHQUDLVLQJWRWKHVTXDUH with tiny eight-pin based microcontrollers driving serial dis-
SRZHUZHPXVWDFFRXQWIRUZKHUHZHZLOOPXOWLSO\0DE\0D plays. The advantage of designing algorithms is that you can
DQGWKHQZHXVHWKDWGLYLVLRQE\ZLWK',9EHFDXVHWKH transfer them to any microcontroller, and you don’t need to
UHVXOWLVJUHDWHUWKDQELWV be tied to a specific structured language.
$IWHUWKLVSURFHVVLVSHUIRUPHGDUHTXLUHGQXPEHURIWLPHV
ZHQHHGWRFRQYHUWD QWRDGHFLPDOEDVH7RDFKLHYHWKLV
ZHPXOWLSO\E\WKHUHVSHFWLYHZHLJKWDQGVRRQ
VHTXHQWLDOO\$QGEHFDXVHZHFDQQRWZRUNZLWKIUDFWLRQVZH CORRECTION
PXOWLSO\E\HWF
)LQDOO\ZHDGGDOOWKHHOHPHQWVWRWKH0DQWLVVDDQLQGHFL “BUILD A HIGH-FREQUENCY Portable Spectrum Analyzer
PDOIRUPDWDQGGLYLGHWKHVXPRIDOORIWKHPE\:HVXPD Using Two Filter ICs” by in our Oct. 17 issue repeated Figure
îDQGZHVDYHLWLQYDULDEOH³O´VR³O´VWRUHVWKHYDOXH 2 as Figures 3 and 4. See http://electronicdesign.com/test-
RIWKHORJDULWKPLQEDVHRIWKHWKHUPLVWRU¶VLPSHGDQFH7KH amp-measurement/build-high-frequency-portable-spectrum-
ODVWVWHSWRREWDLQWKHQDWXUDOORJDULWKPLVWRDSSO\WKHHTXD analyzer-using-two-filter-ics for the correct figures.
WLRQVKRZQEHORZWRFKDQJHLWVEDVH
Ideas for Design
3. The yellow signal is the logarithm’s processing time of 33 ms. The

7-ms blue line is the time required to get a true reading by the MCU.
Thus, the total processing time is 40 ms.
ORJ[OQ OQ[
:HNQRZWKDWOQ VRZHMXVWQHHGWRPXOWLSO\
³OÉ\DQGWKHQGLYLGHLWE\XVLQJWKH',9
FRPPDQGWRVWRUHLWLQYDULDEOH³OP´ZKLFKLVWKHQDWXUDO
ORJDULWKPRIWKHUHVLVWRUPXOWLSOLHGE\
7RGULYHWKHFRPPRQSLQRIWKH/&'DIRUQH[WORRS
HQDEOHVWKHRXWSXW5&WREH³´)LJ7KHQWKHVXEURXWLQH
³'´LVLQYRNHGWRFRQYHUWWKHGHFLPDOGDWDWRDVHYHQVHJPHQW
FRGH7KH/22.83IXQFWLRQGHFRGHVWKHGHFLPDOYDOXHRI
GLJLWWRDVHYHQVHJPHQWIRUPDWWKDWLVVWRUHGLQYDULDEOHSDW
WHUQ1RZWKHYDOXHRI5&VWRUHGDVDE\WHLQYDULDEOH/LV
;25('ZLWKSDWWHUQWRGULYHWKH/&'LQSKDVH
7KLVSURFHVVLVUHSHDWHGZLWKWKHRWKHUWZRGLJLWVZKLOHD
PVSDXVHLVDGGHG7KHQ5&LVWXUQHGRIIDQGVWRUHGDV
DE\WHLQYDULDEOH/6XEURXWLQH'LVLQYRNHGDJDLQDQGWKH
SURFHVVLVUHSHDWHGZLWKWKHSKDVHRXWSXW5&FKDQJHGIRU
DQRWKHUPV$IWHUHYHU\/22.83IXQFWLRQHYHU\SLQLV
DVVLJQHGWRWKHFRUUHVSRQGLQJSRUW¶VSLQGHGLFDWHGWRGULYH
HDFK/&'VHJPHQW7RWDOWLPHUHTXLUHGLVPV)LJ
Reference
Computing Logarithms Digit-by-Digit, Mayer Goldberg, BRICS RS-04-17.
ISSN: 0909-0878
RICARDO JIMENEZ holds a master’s degree in elec-

tronics from ITM. He is an adjunct professor at Imperial
Valley College, Imperial, Calif. ROBERTO C. SOLORIO
holds a BS in electronics from ITM. He is a test engineer at
Skyworks in Mexicali, Mexico.
ideas
ALAN STUMMER | UNIVERSITY OF TORONTO [email protected]
for design
A PIEZOELECTRIC MOTOR is a linear motor with bidirectional The problem is the voltage. Fortunately, it is unipolar.
motion. It uses friction to grip the armature while a voltage is Unfortunately, it is +150 V (peak). The required current is
ramped to warp the piezoelectric material and move the arma- quite small. It has to be only enough to charge and discharge
ture. The voltage then is quickly removed. the 20-nF piezo element. A calculation using charge transfer
As the material springs back, it breaks away from the arma- (Q) shows that during the ramp phase:
ture and returns to its zero position, leaving the armature a few
micrometers further along its track. Repeat this at a kilohertz Q = It = CV
rate and for thousands of times.
While each of the motions is very small, after several sec- where:
onds you may see that the armature has moved, if you look
carefully. (Full disclosure: I had never heard of piezo motors t = 1 ms
before being asked to make a driver for one.) C = 20 nF
There are two drive waveforms, one for forward and the V = 150 V
other reverse: a sawtooth waveform with a slow linear rise
followed by a fast fall, and its complement with a fast rise and therefore:
slow linear fall. This was done using an op-amp triangle-wave
oscillator at 1 kHz, with diodes switched in to speed up either I = CV/t = 3 mA.
the rising or falling edges to about 5% of the cycle. The driver’s
required bandwidth is only 10 to 15 kHz. A 600-mW boost converter can be used to switch the +12 V
up to +200 V with a 3-mA load require-
ment. The most straightforward circuit
R2 R3 +200 V uses an op amp rated to at least 200 V. Op
499k 499k
amps are available with this voltage rat-
R4
16.5k ing, but they are meant for high-current
C1 +200 V
applications and are quite expensive.
C2 10 pF R5
10 pF 16.5k Q1 The circuit in the figure is much
ZVN4525G cheaper and based on a common op amp
R6 R8 Piezo
+12 V 16.5k D1 D2 used as a non-inverting amplifier. The
49.9 element
0 to
heart of the circuit is the current mirror
V+ R7 +12 V SMAZ16-13-F 150 V of R7, N-channel FET Q3, and series-
– 1.5k
connected R4, R5, and R6. (The rea-
Input + IC1 Q3 Q2
son for using three resistors in series is
0 to 10 V AD8065AR ZVN4525GTA ZVP4525G
V– explained below.)
R1
66.5k With Q3 in a common-gate configura-
tion and the gate at +12 V, its source volt-
age remains reasonably constant at +10
A low-voltage op amp can be used to drive a higher-voltage piezoelectric motor. D1 and D2 V (from +12 V less the Vgs(on)). Any out-
protect the circuit against load shorts, while resistors in series strings are used to reduce indi- put of op-amp IC1 less than that +10 V
vidual resistor dissipation and minimize their voltage coefficient of resistance. causes a voltage drop across R7, with the

current coming from R4 to R6. Since R7 has the same current 0805/2012 surface-mount (SMT) resistors can be used. The
as R4 to R6, there is a voltage gain of the ratio of (R4 to R6)/ feedback resistance composed of R2 and R3 is split into two
R7 or 33. The voltage at the bottom of R4 to R6 (the Q3 drain) devices to reduce the voltage coefficient of resistance. This is
now has the required voltage swing but a high impedance. a somewhat obscure effect, where a component’s resistance
A pair of complimentary FETs acts as followers to lower the actually changes slightly at higher voltages.
output impedance and boost current output. Negative feedback There is a small feedback capacitor directly on the op amp
is provided via R2 and R3, along with bandwidth limiting by C1. to provide stability. Without it, the parasitic capacitances
The overall closed-loop gain is set by (R2 + R3)/R1 + 1 = 16. (most notably the Miller capacitances C dg and the piezo’s
There are a few subtleties to the circuit. If something hap- capacitance) would cause enough phase shift for the op amp
pened to the piezo motor, such as the user shorting it, Zener to oscillate. Crossover distortion is not a problem because the
diodes D1 and D2 would protect the FET gates. The current- piezo element could not react to that bandwidth and does not
mirror high-side resistance provided by R4 through R6 is require a perfect waveform.
split into three distinct devices to handle the power levels so Another problem would be the infinite pole caused by a
purely capacitive piezo element. By taking the feedback direct-
ly from the sources of Q1 and Q2, resistor R8 would add a
ALAN STUMMER designs electronics for the University of Toron- more deterministic pole to the load and increase loop stability.
to’s quantum optics and pulse laser research labs, after years Once the somewhat finicky piezo motor was adjusted, the
working on analog controls for heavy industry and then telecom. final circuit worked well. The armature would move back and
forth with incredibly fine resolution.
Model Diacs And Triacs

For AC-Line Control
APPARAJAN GANESAN | CONSULTANT [email protected]
DIACS AND TRIACS often are used for line-voltage control. They The device continues in this state until the current through
also are getting additional interest as part of Internet- and it falls below It. At that point the flip flop is reset and the device
cell-phone-controlled power-line switches such as the Belkin switches once again to its off state, with a resistance Roff. (Note
WeMo Home Automation Switch. that the device is bi-directional.)
The traditional approach has been to model them using The triac is modeled in a similar manner (Figure 2 and
bipolar transistors and diodes.1 Table-based models have been Listing 2; Listing 2a is for the triac macromodel; Listing 2b is
used with varying success.2 The functional model approach the triac symbol). It has the same four parameters as the diac
shown here works well and has been used extensively with the
LTspice simulator (free, from Linear Technology).
In the diac schematic of Figure 1 and associated Listing
1, there are four parameters (Listing 1a is for the diac macro-
model; Listing 1b is the diac symbol; listings are available with
the online version of this article at electronicdesign.com):
Vt: voltage at which the diac triggers

It: current at which it turns off
Ron: resistance when it is on.
Roff: resistance when it is turned off
To turn on, the diac needs the voltage to exceed Vt. Once
it turns on, it needs the current to go below It to turn off. In
operation, the device starts as an open circuit. When the volt-
age across it exceeds Vt, the flip-flop is set, putting the device 1. In the diac circuit, the device starts as an open circuit and remains
in its on state with a low resistance of Ron. open until the voltage across it exceeds Vt
Ideas for Design
2. For the triac, the trigger voltage is on an independent port, in con- 3. The results (top) of the diac test circuit (bottom) show the sym-
trast to the diac. metrical operation of the circuit and allow for user modifications.
component, and it needs the voltage to exceed Vt to turn on. Figure 3 (bottom) and Listing 3b). As VIn goes lower, the device
Once it turns on, it needs the current to go below It to turn off. turns off once the current falls below It. It again reaches its on
The difference between the diac and the triac is that the triac state once the voltage exceeds Vt. The device is bidirectional
trigger voltage is on an independent port. and inherently symmetric. Also, the model can be modified to
It is difficult to design test circuits for these devices because, include current limits on voltages, as well as asymmetries.
as a result of their negative resistances, they usually oscillate or For the triac, the popular dimmer circuit is used for test,
provide limit cycles, which in turn makes it difficult for pro- with output file Figure 4 (top) and Listing 4a along with cor-
grams such as Spice to converge. The main diac characteristics responding schematic Figure 4 (bottom) and Listing 4b. The
of interest are breakover voltage, voltage symmetry, breakback load is a typical 100-W bulb, and the RC-time constant deter-
voltage, breakover current, and power dissipation.3 mines when the triac is triggered.
In the output file for the diac test circuit, VIn is initially at The diac in series with the gate is chosen to ensure the triac
its negative extreme, and VOut is low, as the device is in its on turns off completely. The model can be modified to include
state (Figure 3 (top) and Listing 3a, and corresponding schematic current limits, voltage asymmetries, dV/dt effects, and more.
By varying the RC-time constant, the duty cycle of the output
can be varied from 5° to about 170°, nearly spanning the full
0° to 180° theoretical limit. The device as modeled produces
identical results in the positive and negative half cycles.
REFERENCES
¸(:WPJL4VKLSMVY;YPHJZ¹(-7L[YPLHUK*OHYSLZ/`TV^P[a^^^YLHKIHN
JVTPU[\ZVM[HY[PJSLZ[YPHJ
¸4VKLSMVY+PHJPU6YJHK ¹^^^LKHIVHYKJVT[OYLHK O[TS
¸+PHJ;\[VYPHS¹(TLYPJHU4PJYVZLTPJVUK\J[VY^^^HTLYPJHUTPJYVZLTPJVT
PUMVYTH[PVU[\[VYPHSPUKL_WOW&[FPK$
P_aaL6I^V0N>
APPARAJAN GANESAN has an MS from the State Univer-

sity of New York (SUNY) at Stony Brook and has been designing
circuits for 40 years. He has worked at companies including
General Instrument, Bell Telephone Laboratories (Bell Labs),
4. The results (top) of the dimmer circuit (bottom) using a diac and RCA, and Analog Devices and has been an independent
triac show the operation possible with a standard resistive load.

ideas
Microcontroller Solves Complex
Temperature Polynomial Equations
for design
RICARDO JIMENEZ AND ROBERTO C. SOLORIO | INSTITUTO TECNOLOGICO DE MEXICALI [email protected]
WHEN YOU NEED TO measure temperature using thermistors, ,QSXW$1LVGHFODUHGDVDQDORJDQGWKHUHVWRIWKH,2OLQHV

you face the challenge of linearizing their response to get DUHGHFODUHGGLJLWDORXWSXWV7KHDQDORJWRGLJLWDOFRQYHUWHU
accurate readings. One of the best methods for linearizing a $'&LVFRQILJXUHGIRUELWVZLWKDVDPSOLQJWLPHRIV
thermistor is the polynomial “Steinhart-Hart” equation (S-H), 7KHWKHUPLVWRU¶VYROWDJHUHDGLQJLVVWRUHGLQELQDU\IRUPDW
which has an error of 0.1°C. The temperature range for this LQWKHYDULDEOH³YROW´ZLWKWKHLQVWUXFWLRQ$'&,1YROW
design is from 0°C to 100°C. 7KHQWKLVUHDGLQJLVPXOWLSOLHGE\WRFRQYHUWLWWRGHF
7KUHHWKHUPLVWRUFRHIILFLHQWV²DEDQGF²DUHUHTXLUHG LPDODQGLWLVVWRUHGLQDQHZYDULDEOHFDOOHG³Y´7RREWDLQ
WRLPSOHPHQWWKH6+HTXDWLRQ,IWKHPDQXIDFWXUHUGRHVQRW HDFKGHFLPDOGLJLWIURPYDULDEOH³Y´WKHFRPPDQGV',*
SURYLGHWKHFRHIILFLHQWV\RXFDQREWDLQWKHPE\VROYLQJWKH ',*DQG',*DUHXVHGWRVWRUHWKHVHUHDGLQJVLQWKUHHYDUL
6+HTXDWLRQ(TXDWLRQVDQGIRUWKUHHGLIIHUHQWWHP DEOHVFDOOHGGLJGLJDQGGLJUHVSHFWLYHO\
SHUDWXUHSRLQWV,QWKLVGHVLJQD3,&0&8ZLOOVROYHVXFK 7KHFRGHDYDLODEOHZLWKWKHRQOLQHYHUVLRQRIWKLVDUWLFOH
HTXDWLRQVWRSURYLGHDFFXUDWHUHDGLQJVLQPV)LJ DWHOHFWURQLFGHVLJQFRPVKRZVWKHVRIWZDUHSURJUDPZKHUH
7RLOOXVWUDWHZHZLOOXVHWKHQHJDWLYHWHPSHUDWXUHFRHI WKHIROORZLQJYDULDEOHVDUHGHFODUHGIRUSURFHVVLQJDQGVWRU
ILFLHQW17&/0WKHUPLVWRUZKLFKKDVWKHUHVLVWDQFH LQJWKHUHDGLQJV%$&/YROWYYSDWWHUQSDWWHUQ
DWWKUHHGLIIHUHQWWHPSHUDWXUHVVKRZQLQ7DEOHWith these SDWWHUQ,GLJLWGLJLWDQGGLJLW7ZRYDULDEOHVFRQY
three measurements, the temperature is converted to the Kel- DQGFRQYDUHUHTXLUHGIRUELQDU\WRGHFLPDOFRQYHUVLRQV
vin scale and substituted into three Steinhart-Hart equations: 7KH\LQFOXGHWKHELWUHVROXWLRQ$'&¶VOHDVWVLJQLILFDQWELW
/6%ZKLFKLVHTXDOWRP9

7 7KHDOJRULWKPWKDWZDVGHYHORSHGWRJHWDWHPSHUDWXUHUHDG
D
E OQ5
F>OQ5 @ LQJXVLQJWKH6WHLQKDUW+DUWHTXDWLRQUHTXLUHVVHYHUDOVWHSV
)LUVWFompute the power supply voltage (v1), which is equal
7 to 48828 × volt2. Variable volt2 is read with the PIC’s 10-bit
D
E OQ5
F>OQ5 @
ADC. If the power supply is correct, volt2 = %1111111111,
and then v1 = 49,951,044. Now with the instruction DIV32,

7
D
E OQ5
F>OQ5 @
TABLE 1: THERMISTOR RESISTANCE
6ROYLQJWKHVHWKUHHHTXDWLRQV\LHOGVWKHFRHIILFLHQWVDE Resistance
Temperature (°C)
(kΩ)
DQGF7DEOH6XEVWLWXWLQJWKHFRHIILFLHQWVLQWKH6WHLQKDUW
2.7 23
+DUWHTXDWLRQZHJHWWKHWKHUPLVWRUHTXDWLRQ
27.5 8.8

7W 66 2.7
= ²
= ² /Q5W ² = ² >/Q5W@
RU TABLE 2: LM103 COEFFICIENTS
=
7W

OQ5W² /Q5W Coefficients Value
A 2.501 x 10±4
7KHFKDOOHQJHIRUWKH3,&)PLFURFRQWUROOHULVWR
B 3.505 x 10±4
XVH(TXDWLRQWRJHWWUXHWHPSHUDWXUHUHDGLQJV7KLVSURM
C ±1.415 x 10±7
HFWFRQVXPHVZRUGVRIWKHPLFURFRQWUROOHU¶VPHPRU\
Ideas for Design
the variable volt2 is divided by 10,000, giving a result equal to part). To compute it, we need the result of the successive divi-
4995, which represents the approximate value of 5 V from the sions “wu” by 2 (). Nonetheless this operation or process can-
power supply. not be applied directly, since dividing 47/2 results in 23.5, and
Second, the thermistor’s voltage drop is computed with the we lose the fractions part. (PIC Basic Pro does not work with
same process from the first step, but with another analog-to- fractions.) Therefore, to keep the fractions part, we create a
digital channel. Let’s say we have a reading in the ADC of 2.5 special subroutine that performs the following series:
V (adc = 1000000000). The voltage read is then 24,999,936,
which is divided by 10 with DIV32 for a result of 2499, which 47/2 = 23.5, 23.5/2 = 11.75,
approximates to 2.500 V. 11.75/2 = 5.875, 5.875/2 = 2.9375, 2.9375/2 = 1.46875
Third, we proceed to find the resistor’s value with:
5 = Y When dividing 47/2 = 23, with a remainder of 1, this “1”
5 appears in the first division of the five that will be performed.
( ² Y
Therefore, the fraction 1/25 = 0.03125. We cannot work with
To achieve that, we store in a variable called “dif ” the power’s fractions, though, so instead of dividing by 1 by 2 n, we take
supply voltage minus the thermistor’s voltage. Then we mul- the number 10,000 as a numerator, and we create a subroutine
tiply the thermistor’s voltage by the 10-kΩ fixed resistor, and that computes the denominator depending on the number of
we apply DIV32 to divide this last product by “dif,” resulting in division where appears a remainder of 1.
the thermistor’s resistance value. Then, we sum all the results of the divisions. To continue
Fourth, we now compute the base 2 logarithm from the with this example, we have the following five steps:
thermistor’s resistance with
Equations 7, 8, and 9:
x2 26 PortC.7 A1 35
5V 11
0 In Out VDD 27 PortD.4 B1 34
0 D RG1 32
17 PortC.2 C1 7
DIG 1
7805 Ref 18 PortC.3 D1 6
MCLR 1 19 PortD.0 E1 5
B1 + R3
0 25 PortC.6 F1 36
0 D 100k 24 PortC.5 G1 37

DP1
9V
0
0 VSS
12
30 PortD.6 A2 30
D 31
33 PortB.0 B2 29
10 PortE.2 C2 11
where a 0 , a 1 , and a 2 are the
DIG 2
13 PortA.7 D2 10
VDD 14 PortA.6 E2 9
next-digit mantissa in base 2.
PIC16F884/887
29 PortD.6 F2 31
Next, we find the character- 28 PortD.5 G2 32
R1
istic a0 or logarithm’s integer
DP2
COL
10k
with the function NCD, which
36 PortB.3 A3 25
delivers the most significant 37 PortB.4 B3 24
6 PortA.4 C3 15
DIG 3
bit (MSB) of a number. If we PortA.0 2

7 PortA.5 D3 14
decrement it by 1, we get the 8 PortE.0 E3 13
35 PortB.2 F3 26
base-2 logarithm character- 34 PortB.1 G3 27
istic. We know, for example, R2
DP3
10k at 25°C
that the base-2 logarithm for
number 47 is 5.554 using four VSS 40 PortB.7 A4 21
E
21 PortD.2 B4 20
DIG 4
significant digits. Thus, NCD 22 PortD.3 C4 19

D
23 PortC.4 D4 18
(47) = 6, where 47 in binary is 5 PortA.3 E4 17
C
00101111, and the MSB is in 39 PortB.6 F4 22

B
38 PortB.5 G4 23
the sixth position from left to
right. If we decrement it by 1, 9 PortE.1 DP3 16
20 PortD.1 COM 1
we get number 5, which is the 40
logarithm’s characteristic.
To get the logarithm, we 1. This simple circuit for a high-precision thermometer drives a glass LCD with high efficiency, using the
need the Mantissa (fractions PIC 16F887 microcontroller.

LINEARIZATION INSPIRES A SOLUTION
WE ORIGINALLY CONCEIVED of this idea as a solution

to the problem faced by engineers in trying to linearize a
thermistor’s response. We know there are analog methods
to achieve this, such as putting a parallel resistor across the
thermistor, constant current bias, or a combination of both.
Then you get an almost linear response defined by a straight
line. In this case, you have to design a signal conditioning
circuit based on op amps, which adds more hardware to the
project. Another option we had was using the thermistor’s
characteristic equation, but it is not accurate over the entire
temperature range. Therefore, we decided to use the vener-
able Steinhart-Hart equation.
With this in mind, we faced a new challenge: how to get
2. A signal at 40-Hz delivered by RC2 is used for driving the LCD’s the logarithm of a voltage reading. This had to be done with-
common pin. out getting overflow errors in the microcontroller. After trying
with different methods (Taylor series, Babylonian, etc.), we
• 47/2 = 23, with a remainder of 1; this is the first division of 5, found that the most appropriate algorithm for this project was
10,000/25 = 312.5 = 312 the one from Dr. G. Mayer in his paper “computing logarithms
• 23/2 = 11, with a remainder of 1; this is the second division digit-by-digit.” So the method was developed in base two,
of 4, 10,000/24 = 625 and at the end of the process, it was converted to base 10.
• 11/2 = 5, with a remainder of 1; 10,000/23 = 1250 Once this was achieved, the final step was to substitute the
• 5/2 = 2, with a remainder of 1; 10,000/22 = 2500 resulting Ln values into the S-H equation to get accurate
• 2/2=1, with a remainder of 0; there is no portion to add temperature readings.
The total algorithm’s processing time was 32 ms with the
,QWKLVFDVHWKHYDULDEOH³-D´LVHTXDOWRDQGLWVWLOO internal microcontroller’s oscillator set at 4 MHz. Some read-
QHHGVWREHLQFUHPHQWHGE\RQH7KHUHIRUHZHGLYLGH³-DÉ\ ers have been asking how to reduce the processing time.
DQGWKHQZHDGGXS7KHUHIRUH-D REWDLQ It can be reduced to less than 10 ms by selecting a higher
LQJDSSUR[LPDWHO\WKHRULJLQDOTXDQWLW\PHQWLRQHGDERYH oscillator’s frequency in the algorithm itself.
PXOWLSOLHGE\ZLWKIRXUVLJQLILFDQWGLJLWV We decided to drive a numerical LCD directly without
2QFHZHKDYHREWDLQHG0ZHQHHGWRJHWWKHPDQWLVVD external decoders for energy-saving applications. You can
XVLQJWKHSURFHVVGHVFULEHGLQ)LJXUH7KHGLYLVLRQE\DQ choose other types of displays to free the microcontroller’s
LVQRWFRPSOLFDWHGEHFDXVHZHDUHZRUNLQJLQEDVHDQGWKH ports for other control applications. We are currently working
GLYLVLRQLVSHUIRUPHGE\RU:KHQUDLVLQJWRWKHVTXDUH with tiny eight-pin based microcontrollers driving serial dis-
SRZHUZHPXVWDFFRXQWIRUZKHUHZHZLOOPXOWLSO\0DE\0D plays. The advantage of designing algorithms is that you can
DQGWKHQZHXVHWKDWGLYLVLRQE\ZLWK',9EHFDXVHWKH transfer them to any microcontroller, and you don’t need to
UHVXOWLVJUHDWHUWKDQELWV be tied to a specific structured language.
$IWHUWKLVSURFHVVLVSHUIRUPHGDUHTXLUHGQXPEHURIWLPHV
ZHQHHGWRFRQYHUWD QWRDGHFLPDOEDVH7RDFKLHYHWKLV
ZHPXOWLSO\E\WKHUHVSHFWLYHZHLJKWDQGVRRQ
VHTXHQWLDOO\$QGEHFDXVHZHFDQQRWZRUNZLWKIUDFWLRQVZH CORRECTION
PXOWLSO\E\HWF
)LQDOO\ZHDGGDOOWKHHOHPHQWVWRWKH0DQWLVVDDQLQGHFL “BUILD A HIGH-FREQUENCY Portable Spectrum Analyzer
PDOIRUPDWDQGGLYLGHWKHVXPRIDOORIWKHPE\:HVXPD Using Two Filter ICs” by in our Oct. 17 issue repeated Figure
îDQGZHVDYHLWLQYDULDEOH³O´VR³O´VWRUHVWKHYDOXH 2 as Figures 3 and 4. See http://electronicdesign.com/test-
RIWKHORJDULWKPLQEDVHRIWKHWKHUPLVWRU¶VLPSHGDQFH7KH amp-measurement/build-high-frequency-portable-spectrum-
ODVWVWHSWRREWDLQWKHQDWXUDOORJDULWKPLVWRDSSO\WKHHTXD analyzer-using-two-filter-ics for the correct figures.
WLRQVKRZQEHORZWRFKDQJHLWVEDVH
Ideas for Design
3. The yellow signal is the logarithm’s processing time of 33 ms. The

7-ms blue line is the time required to get a true reading by the MCU.
Thus, the total processing time is 40 ms.
ORJ[OQ OQ[
:HNQRZWKDWOQ VRZHMXVWQHHGWRPXOWLSO\
³OÉ\DQGWKHQGLYLGHLWE\XVLQJWKH',9
FRPPDQGWRVWRUHLWLQYDULDEOH³OP´ZKLFKLVWKHQDWXUDO
ORJDULWKPRIWKHUHVLVWRUPXOWLSOLHGE\
7RGULYHWKHFRPPRQSLQRIWKH/&'DIRUQH[WORRS
HQDEOHVWKHRXWSXW5&WREH³´)LJ7KHQWKHVXEURXWLQH
³'´LVLQYRNHGWRFRQYHUWWKHGHFLPDOGDWDWRDVHYHQVHJPHQW
FRGH7KH/22.83IXQFWLRQGHFRGHVWKHGHFLPDOYDOXHRI
GLJLWWRDVHYHQVHJPHQWIRUPDWWKDWLVVWRUHGLQYDULDEOHSDW
WHUQ1RZWKHYDOXHRI5&VWRUHGDVDE\WHLQYDULDEOH/LV
;25('ZLWKSDWWHUQWRGULYHWKH/&'LQSKDVH
7KLVSURFHVVLVUHSHDWHGZLWKWKHRWKHUWZRGLJLWVZKLOHD
PVSDXVHLVDGGHG7KHQ5&LVWXUQHGRIIDQGVWRUHGDV
DE\WHLQYDULDEOH/6XEURXWLQH'LVLQYRNHGDJDLQDQGWKH
SURFHVVLVUHSHDWHGZLWKWKHSKDVHRXWSXW5&FKDQJHGIRU
DQRWKHUPV$IWHUHYHU\/22.83IXQFWLRQHYHU\SLQLV
DVVLJQHGWRWKHFRUUHVSRQGLQJSRUW¶VSLQGHGLFDWHGWRGULYH
HDFK/&'VHJPHQW7RWDOWLPHUHTXLUHGLVPV)LJ
Reference
Computing Logarithms Digit-by-Digit, Mayer Goldberg, BRICS RS-04-17.
ISSN: 0909-0878
RICARDO JIMENEZ holds a master’s degree in elec-

tronics from ITM. He is an adjunct professor at Imperial
Valley College, Imperial, Calif. ROBERTO C. SOLORIO
holds a BS in electronics from ITM. He is a test engineer at
Skyworks in Mexicali, Mexico.

IFDnewVol2 2.6102cdcdddad1

Uploaded by

Copyright:

Available Formats

IFDnewVol2 2.6102cdcdddad1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

IFDnewVol2 2.6102cdcdddad1

Uploaded by

Copyright:

Available Formats

Ideas

Generate Realistic Models For LED

PWM-Based Analog Calculator Provides Four-

Driven Shield Enables Large-Area

Generate Realistic Models

Forward current (A)

Forward current (A)

ElEctronic DEsign Go To elecTronicdesiGn.com

PWM-Based Analog Calculator Provides

quency of 3.81 Hz to 1 MHz. A resis- 2 LTC6992-1

ELECTRONIC DESIGN GO TO ELECTRONICDESIGN.COM

10 somewhere in the loop to ensure feed-

–0.01 MARK THOREN is

DEV GUALTIERI | TIKALON LLC, LEDGEWOOD, N.J. [email protected]

Driven Shield Enables

ELECTRONIC DESIGN GO TO ELECTRONICDESIGN.COM

Test Setup Checks Transistors’ hFEs When Tight

Multiple Power Supplies Fortify High-Side

High-Side Switch Provides Overvoltage Protection

Add Short-Circuit Protection, Diagnostics To

Test Setup Checks Transistors’ hFEs

When Tight Control Is Important

Testing phase: Q1, Q2, Q3, and DUT activated

Measurement phase: Q1, Q2, Q3, and DUT activated

maximize the signal level to ease VRB1 monitoring. The hFE of

For more, see “Anoop’s Analysis” with the online version of

Chee hua how, design engineer, specializes in ana-

ElEctronic DEsign Go To elecTronicdesiGn.com

Steven P. Hendrix, Pe | Hx EnginEEring LLC [email protected]

Multiple Power Supplies

WilliAm sWAnson | [email protected]

High-Side Switch Provides Overvoltage

using few components, it provides overvoltage protection as a

ElEctronic DEsign Go To elecTronicdesiGn.com

VISHWAS VAIDYA | TATA MOTORS [email protected]

Add Short-Circuit Protection, Diagnostics

SPI Eases Interfacing For 5-By-5 Dot-Matrix Display

Use An LED Dot-Graph Display To Complement

Lamp Eliminates Need For Limit Switches In Flap-

SPI Eases Interfacing

For 5-By-5 Dot-Matrix Display

RAM internal multiplexer is stopped, reducing the

____ Opcode (100 bits,

Table 1: load CharaCTer address

Opcode Character address

Table 2: load Column daTa

Opcode Column data

Table 3: display brighTness

Opcode Control word

Table 4: power down

Opcode Control word

Table 5: lamp TesT

Opcode Control word

Table 6: sofTware Clear

Opcode Control word

ElEctronic DEsign Go To elecTronicdesiGn.com

Example: serial clock = 5 MHz, clock period = 200 ns

264 clock cycles, 52.8 μs

(a) Character 0 Character 1 Character 2 Character 3

_ __ _ ______

IN A STANDARD DC-DC converter, a resis- 5 5 VOut = 3 Vfb – 2 Vr (4)

If all N capacitors in the parallel connection are identical 5 VN

FRVJWBMFOUDBQBDJUPS$se BOE3.4WBMVF* BOEJOEJWJEVBM3.4 &SH - &SN (5)

WHEN DESIGNING FILTERS, engineers very often need to know ²7²

indicates the definition of the parameters used in the calculations.