design

Edited by Bill Travis

ideas The best of
design ideas
8 Check it out at:
www.edn.com

Combine two 8-bit outputs to make one 16-bit DAC

Steve Woodward, Chapel Hill, NC
nexpensive, 16-bit, monolithic DACs tion posed a challenge, because, at design

I can serve almost all applications.

However, some applications require
unconventional approaches. This Design
finalization, the NPSL included no ap-
propriate, flight-qualified, 16-bit DACs,
and the budget included no funds for cer-
Combine two 8-bit outputs
to make one 16-bit DAC ..................................85
LED driver provides
Idea design concerns circuitry I recently tification of new devices. I escaped from software-controlled intensity ..........................86
designed for a tunable-diode laser spec- this impasse by exploiting two fortuitous
trometer for a Mars-exploration applica- facts: The update rate of the two DACs Improve roll-off of Sallen-Key filter ................88
tion. The control circuitry included two was only tens of hertz, and the 69RH- AC-coupling instrumentation amplifier
16-bit DACs that interface to the radia- 051A had a number of uncommitted, 8- improves rejection range of differential
tion-hardened, 8051-variant 69RH051A bit, 14.5-kHz PWM outputs. These out- dc input voltage..................................................88
microcontroller. Because of the intend- puts made one 16-bit DAC; a second pair
Simplify computer-aided engineering
ed space-flight-qualified specification, of PWM bits and an identical circuit
with scientific-to-engineering conversion ....94
everything in the design had to consist made the other (Figure 1).
solely of components from the NPSL Hex inverter IC1’s VCC rail connects to 1.5V battery powers white-LED driver ..........96
(NASA parts-selection list). This restric- a precision 5V reference. The inverter’s Simple VCOM adjustment uses
any logic-supply voltage ..................................96
IC2 5V
C4 69RH051 A Publish your Design Idea in EDN. See the
27 pF 40 What’s Up section at www.edn.com.
18 VDD2
XTAL1
0.1 ␮F
14.756 MHz Y1 20
VSS
19
XTAL2 EA
31 outputs are accurate analog square waves.
C5 9
RST The low-order PWM-signal output,
27 pF 28
VDAC
A15
27
PWM0, of the 8051 controls the V3
A14
0 TO 5V R3 R2 R1
A13 26 square wave, and the high-order PWM
(REFERENCE) 200k V4 3.92k V1 1M 25
11
TX0
A12
A11 24 output, PWM1, controls the V1 square
10 23
RX0 A10
22
wave. R2 and R6 passively sum the two
4 3
IC1 A9
A8 21 square waves in the ratio R2/R6⫽3290/1
HC04
16
million⫽1/255 to produce V4, duplicat-
WR
10
IC1
11 17 ing the 28 ratio of the 16-bit sum. This ac-
R0
HC04
ALE 30
tion makes the dc component of V4 equal
12 13 32
to 5V(REF)(PWM0⫹255PWM1)/256.
IC1 AD7
8
CEX4 AD6 33 Thus, if you write the 0 to 255, high-or-
C8 HC04 34
R5 0.0056 R4
7
CEX3 AD5
AD4 35 der byte of a 0 to 65,535, 16-bit DAC set-
␮F V 6
200k 2 3.92k CEX2 36 ting to the CEX1 register of the 8051 and
2 AD3
IC1 1 PWM1 5
CEX1 AD2 37
HC04
PWM0 4
CEX0 AD1 38 write the 0 to 255, low-order byte to
C9 39
0.01 ␮F
8 IC1 9
AD0 CEX0, a corresponding 16-bit analog
R 6 V3 HC04 12 29 representation appears in the dc compo-
INTO PSEN
1M
R7 nent of V4. The accuracy of the R2-to-R6
1M 6
IC1 5 ratio is the only limit on the monotonic-
HC04
ity and accuracy of this circuit. For ex-
5V (REFERENCE) ample, one part in 25,500⫽14.5 bits for
14
1%-tolerance R2 and R6 and a full 16 bits
Figure 1 VCC for 0.3% tolerance or better. But the sto-
IC1
C1B HCO4 ry doesn’t end there. Two problems re-
0.1 ␮F GND
Two PWM outputs from a 7 main.
microcontroller combine to The first problem is the extraction of
form a monotonic 16-bit DAC. V4’s desired dc component from all—or
www.edn.com September 30, 2004 | edn 85
design
ideas
at least 15 or 16 bits⫽99.995%—of the
undesired square-wave ac ripple. The R3-
C9 lowpass filter does some of this work.
If you make C9 large enough, in princi-
ple, the filter could do the whole job. The
reason this simple approach wouldn’t
work is that, to get such a large ripple at-
tenuation of approximately 90 dB with a
single-stage RC filter would require an
approximately 300-msec time constant
and a resultant 3-sec, 16-bit settling time.
This glacial response time would be too
slow even for this undemanding applica-
tion. To speed things, the R4, R5, R7, C8
network synthesizes and then sums V2: an
inverse-polarity duplicate of V4’s 14.5-
kHz ac component. This summation ac-
tively nulls out approximately 99% of the
ripple. This nullifying action leaves such
a small residue that an approximately 2-
msec and, therefore, approximately 25-
msec-settling-time R3C9 product easily
erases it.
The other problem is compensation for
the low, but still nonzero, on-resistance of
the HC14 internal CMOS switches, so
that the resistance doesn’t perturb the
critical R2-to-R6 ratio. This issue is of no
particular concern for R6, because the R6-
to-on-resistance ratio is greater than
10,000-to-1, making any associated error
negligible. This situation is not the case
for R2, however, in which, despite the
triple-parallel gates, the R2-to-on-resist-
ance ratio is approximately 300-to-1,
which is small enough to merit attention.
Load-cancellation resistor R1 provides
such attention. R1 sums a current into the
R2 driving node that, because it is equal in
magnitude but opposite in phase to the
current through R6, effectively cancels the
load on the R2 drivers. This process makes
the combined on-resistance approxi-
mately 100 times less important than it
otherwise would be. The result is a sim-
ple, highly linear and accurate voltage-
output DAC with a respectable, if not
blazingly fast, settling time of approxi-
mately 25 msec. And the most important
result, in this case, was a parts list with an
impeccable NPSL-compliant pedigree.왏

LED driver provides software-controlled intensity

Neda Shahi and Bjorn Starmark, Maxim Integrated Products, Sunnyvale, CA
ecent advances in VIN

R
D1 SMB R1
operating efficiency 1
L1
2
CMSH1-100 51
VCC
have expanded the use C1 22 ␮H C2 C3
of LEDs from one of mere 3.3 ␮F ELL6RH 0.47 ␮F R2 0.1 ␮F LED1
1206 Q1 1210 16k 1206
indicators to becoming 10V IRLL110 100V 100V
GND
driving forces in electronic SOT223

lighting. Increased reliabili- LED2

ty and ruggedness (versus VIN
C4 11 10
other lighting technologies) 0.1 ␮F VIN GATE
R3
gives the LED a bright fu- 4 56k
CA⫹
ture indeed. Vendors in re- 5 Q2 MMBT5551
CS⫺ SOT23
cent years have introduced IC1
12
many ICs for driving LEDs, TO CS DACOUT
6
Q3 MMBT3906 10k
R4 LED(N-1)
1
but the problem of driving CONTROLLER SCLK FB 7 SOT23
2 DIN R5
serial chains of LEDs has re- COMP
8 510k
3 9
ceived less attention. One CL GND R6 LED(N)
12k
approach to that problem MAX1932
C5 R7
adapts a bias-supply IC 12QFN
0.33 ␮F 10k
Figure 1
for APDs (avalanche
photodiodes) to provide
adjustable-current, soft- The APD driver, IC1, provides high-voltage LED modules with software-adjustable intensity control.
ware shutdown, and logic
indication of open-circuit faults (Figure sheet help you select components for the these conditions correspond to a full-
1). This design reconfigures the APD-bias step-up dc/dc converter. The current-ad- scale output of 39 mA and a resolution of
IC, IC1, to allow its low-voltage DAC out- justment transfer function is: 150 ␮A. The three-wire serial interface
put to modulate the high-voltage, cur- that controls IC1 allows you to shut down
rent-sense feedback via a high-voltage- IC1 by writing code 00hex to the DAC.
output transconductance stage com- The circuit also provides an output-volt-
prising Q2 and Q3. These two comple- age limit. If an LED fails open, the R5-R7
mentary transistors provide first-order where VCL is the current-limit threshold divider limits the output voltage, in this
temperature-compensation sufficient for (2V), CODE is the digital code to the case, to 50V. Simultaneously, the CL pin
the application. DAC in decimal format, and IOUT is the goes high to indicate the open-fault con-
Equations from the MAX1932 data desired output current. For this circuit, dition.왏
86 edn | September 30, 2004 www.edn.com
design
ideas
Improve roll-off of Sallen-Key filter
Doug Glenn, Teledyne, Lewisburg, TN
he well-documented Sallen- mately one-third to one-fourth

T
10
Key active filter is a staple of 0 the value of R1, and then adjust
BREADBOARD
analog design. This Design Idea –10 R4 as needed to allow use of
shows a way to obtain better roll-off –20 SIMULATED standard capacitor values. The
THREE POLE
by adding just a few common pas- –30 graph in Figure 2 shows the im-
SIMULATED
sive components. Figure 1a shows –40 provement in the cutoff rate of
a typical implementation of a –50 the filter; the result is a quasi-el-
three-pole, lowpass version. In op- –60 liptic response. A breadboard of
eration, you adjust the ratio of ca- 100 1000 10,000 100,000 the circuit in Figure 1b uses 5%
pacitors C1 and C2 to give a parts. The measured results
Figure 2 The improved cutoff rate of the filter results
peaked response for the two show good agreement with the
poles within the feedback loop. The in a quasi-elliptical response. Spice simulation. To take ad-
peaked response compensates for vantage of the faster roll-off,
the initial roll-off in the third pole formed quency, F⫽1/(2␲R4C4), is equal to ap- just scale the frequency and impedance
by the R3-C3 section at the input. In Fig- proximately twice the desired cutoff fre- to your application. The highpass dual of
ure 1b, a twin-tee notch filter replaces the quency. this circuit works as well as the lowpass
R3-C3 section at the input. The notch fre- Select a value for R4 that’s approxi- version.왏

C1 R4 R5 C1
0.1 ␮F _ 3600 3600 0.1 ␮F _
R3 R1 R2 R1 R2
10k 10k 10k C6 10k 10k
+ 0.02 ␮F +

C3 C2 C2
0.015 ␮F 470 pF C4 R6 C5 470 pF
0.01 ␮F 1800 0.01 ␮F
(a)
(b)

Figure 1 The addition of a twin-tee network (b) considerably improves the roll-off rate of the circuit (a).

AC-coupling instrumentation amplifier improves

rejection range of differential dc input voltage
Francis Rodes, Olivier Chevalieras, and Eliane Garnier, ENSEIRB, Talence, France
he need for conditioning low- proach to removing the differential dc strumentation amplifier is to maintain

T level ac signals in the presence of

both common-mode noise and dif-
ferential dc voltage prevails in many ap-
plications. In such situations, ac-coupling
to instrumentation and difference am-
plifiers is mandatory to extract the ac sig-
nal and reject common-mode noise and
differential dc voltage. This situation typ-
ically occurs in bioelectric-signal acqui-
sition, in which metallic-electrode polar-
ization produces a large random dif-
ferential dc voltage, ranging from
⫾0.15V, which adds to low-level biolog-
ical signals. Input ac-coupling is one ap-
content. But this technique requires
adding a pair of capacitors and resistors
to ac-couple the inputs of the difference
the mean output voltage at 0V. To do so,
you insert an autozero feedback loop,
comprising IC4, RFB, and CFB, in a classic
amplifier. The manufacturing tolerances three-op-amp instrumentation amplifi-
of these components severely degrade the er. This feedback loop produces a fre-
CMRR (common-mode-rejection ratio) quency-dependent transfer function:
of the amplifier. If cost is not an issue, you
could perform an initial trim, but this op-
eration is useless for biological applica-
tions plagued by wide variations in elec-
trodes and tissue impedances. The Consequently, the ac-coupled instru-
differential topology in Figure 1 ad- mentation amplifier behaves as a high-
dresses these problems (Reference 1). pass filter with a ⫺3-dB cutoff frequen-
The principle of this ac-coupled in- (continued on pg 92)

88 edn | September 30, 2004 www.edn.com

design
ideas
VIN1 +
cy from the equation f⫽1/2␲RFBCFB. At R R
IC1
first glance, you might think that the out- _
put-autozeroing behavior of the ac-cou-
pled instrumentation amplifier is per- R2

fect. Unfortunately, the output _

R1
autozeroing capability of this circuit is IC3 VOUT
R2
strongly limited. You can determine this + CFB
limitation by expressing the output volt-
RFB
age as a function of the input signals and _ _
the integrator’s output voltage, VZ: IC2
R R
IC4
VOUT⫽(1⫹2R2/R1)(VIN1⫺VIN2)⫹VZ⫽AD VIN2 + +
VZ
(VIN1⫺VIN2)⫹VZ, where VOUT is the out-
put voltage. In this expression, AD⫽
1⫹2R2/R1 is the differential gain in Figure 1
the passband. At dc, the output voltage is This ac-coupled instrumentation amplifier accommodates only ⫾5-mV maximum input.
0V as long as the integrator’s output does
not reach its saturation voltage, VZ(MAX). as a function of the input signal and the
Therefore, setting the output voltage at integrator’s output voltage, VZ, becomes:
0V in the above expression yields the
maximum differential-input dc voltage
that this circuit can handle:

In the above equations, the additional

term, 1⫹R4/R3, is the gain of the active-
feedback stage.
Consider, for instance, the typical per- In this expression, AD⫽(1⫹2R2/R1) The new expressions for ⌬VIN(MAX) and
formance and constraints of a portable (1⫹R4/R3) is the new differential gain in AD(MAX) clearly show the advantages of
biotelemetry system: differential gain of the passband. Figure 2’s ac-coupled instrumentation
1000, ⫾5V split power supplies, and op At dc, the output voltage remains 0V as amplifier with active feedback: For an
amps with rail-to-rail output-voltage long as the integrator’s output does not identical differential gain, you can extend
swing. In this system, the application of reach its saturation voltage, VZ(MAX). the polarization-voltage range, ⌬VIN(MAX),
the formula for ⌬VIN yields a maximum Therefore, setting the output voltage at by a factor equal to the gain of the active-
differential-input dc voltage of only ⫾5 0V in the new expression for output volt- feedback stage. Conversely, for a given
mV. This limited performance is unac- age yields the new maximum differential- polarization-voltage range, ⌬VIN (MAX),
ceptable for biological applications, in input dc voltage and differential gain. you can increase the differential gain by
which you encounter differential-input They are, respectively: the gain of the active-feedback stage.
dc voltages of ⫾0.15V. The ac-coupled
instrumentation amplifier in Figure 2 1/2 LT1464
R4
R R 1/2 LMC662 100k
overcomes this limitation, thanks to the VIN1 +
100k 100k
+
addition of “active feedback,” which in- IC1 IC5
R3
cludes voltage divider R3-R4 and the as- _ _
1.5k
sociated buffer amplifier, IC5. With this R2
3.3k
arrangement, the following equations 1/2 LMC662
_
give the new transfer function and high- R1
R2 IC3 VOUT
pass cutoff frequency, respectively. 470
3.3k CFB
+
CMRR 1 ␮F
TRIM T RFB
4.7M
_ R RF _
100k 95k
IC2 IC4
VIN2 + 10k +
VZ
1/2 LT1464 1/2 LMC662

NOTE: R AND RF ARE ⫾1%; OTHERS ARE ⫾5%

Figure 2
This instrumentation amplifier can accommodate a differential-input
The expression for the output voltage range of ⫾0.34V.
92 edn | September 30, 2004 www.edn.com
design
ideas
The only drawback of this topology is
apparent in the expression for fC, the
highpass cutoff frequency.You multiply
this frequency by the gain of the active-
feedback stage. Therefore, to maintain a
given cutoff frequency, you must multi-
ply the time constant by a factor equal to
the active-feedback stage gain. This fac-
tor can be an issue in processing signals
whose spectrum includes low-frequency
components. In such applications, RFB
and CFB can reach prohibitive values.
Consequently, you must make a trade-off
between the time constant and the active-
feedback stage gain. The component val-
ues in Figure 2 are a typical example of
such a trade-off: The values are for an
EEG (electroencephalogram) amplifier
with ⫾5V split power supplies. The am-
plifier has a differential gain of 1000 and
a highpass cutoff frequency of 2.3 Hz,
and it can handle a differential-input dc-
voltage range of ⫾0.34V.
To obtain this performance, you set the
active-feedback stage gain and the differ-
ential-amplifier gain, respectively, to 67.6
and 15. With these gain values, the noise
performance of the ac-coupled instru-
mentation amplifier of Figure 2 is simi-
lar to that of a classic instrumentation
amplifier. This situation occurs because
the autozeroing and active-feedback
stages, IC4 and IC5, are after the input dif-
ferential stage, IC1 and IC2. Consequent-
ly, the gain of the differential stage rough-
ly divides their respective noise con-
tributions, which are therefore negligible.
You can use several low-noise op-amps
for IC1 and IC2. For portable bioteleme-
try applications, the LT1464 is a good
compromise for input-noise density,
noise-corner frequency, input-bias cur-
rent and current drain. (Respectively:
VNOISE⫽26 nV/公Hz, fC⫽9 Hz, IBIAS⫽0.4
pA, and ICC⫽230 ␮A.)
A theoretical analysis using the LT
1464’s noise parameters shows that un-
der worst-case conditions, the input-
noise voltage should not exceed 11 ␮V
rms. Tests on prototypes confirm this
prediction; the tests effectively measure
input-noise voltages of 3 to 6 ␮V rms. To
sum up, an ac-coupled instrumentation
amplifier with active feedback is well-
suited for applications requiring high dif-
ferential gain, a capability for handling
large differential-input dc voltages, and
low-noise performance.왏
Reference
1. Stitt, Mark, “AC-Coupled Instru-
mentation and Difference Amplifier,”
Burr-Brown, AB-008, May 1990.

Simplify computer-aided engineering

with scientific-to-engineering conversion
Alexander Bell, Infosoft International, Rego Park, NY
he simple yet useful The formula is in VB; you

T formula in this Design

Idea enables conver-
sion from scientific format
could use it in any VB/VBA-
backed software applications.
In this example, the function
(for example, 2.2⫺9), which is in a code module of an MS
is typical for CAE (com- Excel file (Figure 1). You
puter-aided-engineering), could also use it as an Excel
double-precision output Add-In (.xla) or a “pure-VB,”
values, into “human- compiled-DLL component.
friendly” engineering for- You can download Listing 1
mat (for example, 2.2 nF). and the Excel file from the
The engineering format is Web version of this Design
more suitable for bills of Idea at www.edn.com. The
material and other electri- input numerical value in this
cal and electronic-engi- formula has double-preci-
neering documents The formula appears in the formula bar, taking the first sion accuracy, and its range
and specifications. Figure 1 spans from approximately
numeric parameter from the column to the left. The unit of
The formula is rather electrical resistance, “ohm,” is a second parameter. ⫺1.79308 to ⫹1.79308, which is
straightforward. It takes sufficient for any practical
two parameters. The first is the numeri- cal-engineering practice. Listing 1, avail- engineering calculations. Note that the
cal value, and the second one specifies the able at www.edn.com, shows details of the maximum value is even bigger than the
unit of measurement—ohms, farads, or formula. The tricky part of the formula famous “googol,” which is represented by
henries, for example. Alternatively, it is the conversion to a decimal type after 100 digits.왏
could be of any random text, including an the formula calculates the ratio of two log
empty string, “.” The formula calculates values (Reference 1). This step ensures the Reference
the mantissa/order of magnitude and re- correct order-of-magnitude calculations 1. Bell, Alexander, “What’s wrong with
turns the text string, formatted in com- in cases in which the mantissa of the in- INT(LOG) in VBA?” Access-VB-SQL Ad-
pliance with commonly accepted electri- put value is close or equal to one. visor, October 2002, pg 65.
94 edn | September 30, 2004 www.edn.com
design
ideas
1.5V battery powers white-LED driver
Steve Caldwell, Maxim Integrated Products, Chandler, AZ
lthough white LEDs are common L1

A in a variety of lighting applications,

their 3 to 4V forward-voltage drop
makes low-voltage applications challeng- 1.5V
+
C1
10 ␮H

ing. Charge pumps and other ICs are 10 ␮F 5

R1
62 mA
1
available for driving white LEDs, but they LX 4
OUT
generally don’t work with the low supply 1
BATT C2 4 5
voltage of 1.5V in single-cell-battery ap- 10 ␮F RS+ RSⳮ D2
IC1 3 V
plications. The low-voltage circuit of Fig- D1 CC
MAX1722
3.9V IC2
ure 1 provides a current-regulated output 2 GND MAX4073T
suitable for driving white LEDs. 3 1
The boost converter, IC1, can sup- Figure 1 FB OUT
GND
ply load currents to 62 mA with input 2
voltages as low as 1.2V, making it suitable
for use with a 1.5V, single-cell battery. Be- Powered from a single-cell battery, this circuit provides a regulated output current suitable for
cause a white LED draws negligible load driving a white LED.
current until the output voltage rises
above 3V, the boost converter can start across R1 with a gain of 20. This high gain provides overvoltage protection at the
with input voltages as low as 0.8V. boosts efficiency by enabling use of a output. When the output voltage rises
By deriving feedback from a high-side small-valued current-sensing resistor. above the sum of the zener voltage (VZ)
current-sense amplifier, IC2, the circuit You can calculate the value of R1 from the and IC1’s 1.235V feedback voltage (VFB),
allows current regulation without sacri- desired output current: R1⫽1.235V/ the feedback voltage (Pin 3) rises and
ficing efficiency. IC2’s 1.8-MHz band- (20⫻IOUT). For 1.5V input and 62-mA causes IC1 to stop switching. Thus, for an
width also eliminates instability in the output, the circuit efficiency of Figure 1 open-circuit output, the output voltage is
feedback loop. IC2 amplifies the voltage is approximately 80%. Zener diode D1 regulated at VZ⫹VFB.왏

Simple VCOM adjustment uses

any logic-supply voltage
Peter Khairolomour and Alan Li, Analog Devices, San Jose, CA
ll TFT (thin-film-transistor) LCD

A panels require at least one appropri-

ately tuned VCOM signal to provide a
reference point for the panel’s backplane.
TABLE 1—OUTPUT-VOLTAGE RANGE
R2 tolerance (%), scale
ⳮ30, zero
ⳮ30 mid
R2 (k⍀
0
3.5
⍀) VCOM (V)
3.5
4.0
Step size (mV)
3.9

The exact value of VCOM varies from pan-
el to panel, so the manufacturer must pro-
gram the voltage at the factory to match
the characteristics of each screen. An ap-
propriately tuned VCOM reduces flicker
and other undesirable effects. Tradition-
ally, the VCOM adjustment used mechani-
cal potentiometers or trimmers in the
voltage-divider mode. In recent years,
however, panel makers have begun look-
ing at alternative approaches because me-
chanical trimmers can’t provide the nec-
essary resolution for optimal image
fidelity on large panels. They also require
a physical adjustment that technicians on
96 edn | September 30, 2004
ⳮ30, full 7 4.5
30, zero 0 3.3 6.8
30, mid 6.5 4.2
30, full 13 5.1
the assembly line usually perform. This VCOM-adjustment process, resulting in
adjustment is not only time-consuming, lower manufacturing cost and higher
but also prone to field failures arising from product quality. Unfortunately, many
human error or mechanical vibration. panels operate at higher voltages, and the
A simple alternative to achieving the choice of available supply voltages is lim-
increasing adjustment resolution for op- ited. The system implementation for a 5V
timal panel-image fidelity is to replace supply is straightforward (Figure 1).
the mechanical potentiometer with a dig- Without a 5V supply, the circuit can be-
ital potentiometer. Using digital poten- come more complex.
tiometers, panel makers can automate the This Design Idea shows a simple way
www.edn.com
design
ideas
that you can use any available logic sup- stores the desired potentiometer setting plications. For systems that have no 5V
ply to power the potentiometer provid- into the EEPROM. The AD5259 uses a supply, many designers would be tempt-
ing the VCOM adjustment. The 6- or 8-bit 5V, submicron CMOS process for low ed to simply tap off the potentiometer’s
AD5258/59 nonvolatile digital poten- power dissipation. It comes in a space- series-resistor string at the 5V location.
tiometer demonstrates this approach. An saving 10-pin MSOP, an important fea- This approach is not viable, because, dur-
I2C serial interface provides control and ture in low-cost, space-constrained ap- ing programming (writing to the
VCC SUPPLIES POWER 14.4V
⬃3.3V 14.4V TO BOTH THE
⬃3.3V MICROCONTROLLER
5V AND THE LOGIC R1
R1 SUPPLIES OF THE
DIGITAL POTENTIOMETER 70k
C1 70k
C1 ⬃5V
1 ␮F AD5259 1 ␮F AD5259
R6 R5 VDD R6 R5
10k 10k VDD
_ 10k 10k _
VLOGIC IC1 VLOGIC IC1
AD8565 AD8565
SCL R2 R2
CONTROLLER + 3.5V<VCOM<4.5V SCL 3.5V<VCOM<4.5V
10k CONTROLLER 10k +
SDA SDA
GND GND

R3 R3
25k 25k

Figure 1 Figure 2
A digital potentiometer makes it easy to adjust VCOM to the A separate VLOGIC pin makes it possible to derive the VDD supply from
desired value. the potentiometer’s resistor string.

98 edn | September 30, 2004 www.edn.com

design
ideas
VDD
EEPROM), the AD5259’s VLOGIC pin typ- RDAC1 RDAC1
A1
ically draws 35 mA. It cannot draw this VLOGIC RDAC EEPROM REGISTER
current level through R1 because the volt- DGND
W1

age drop would be too large. For this rea- B1

son, the AD5259 has a separate VLOGIC pin

that can connect to any available logic SCL
8 DATA
supply. In Figure 2, VLOGIC uses the sup- SDA
I2 C
SERIAL 8
ply voltage from the microcontroller that INTERFACE CONTROL
is controlling the digital potentiometer. AD0
Now, VLOGIC draws the 35-mA program-
AD1
ming current, and VDD draws only mi- COMMAND-DECODE LOGIC
croamps of supply current to bias the in-
ternal switches in the digital potentio- POWER- ADDRESS-DECODE LOGIC
ON
meter’s internal resistor string. If the pan- RESET
el requires a higher VCOM voltage, you can CONTROL LOGIC
add two resistors to place the op amp in
a noninverting gain configuration. Figure 3 This block diagram shows the digital potentiometer’s inner workings.
The digital potentiometer has ⫾30%
end-to-end resistance tolerance. Assum- 4V⫾0.5V, with a maximum step size of target specification. Also, because the
ing that the tolerances of R1, R3, and VDD 10 mV. As Table 1 shows, the circuit in digital potentiometer’s logic supply
are negligible compared with those of the Figure 2 guarantees an output range of matches the microcontroller’s logic lev-
potentiometer, you can achieve the range 3.5 to 5.4V with a step size within ⫾10 els, the microcontroller can read data
of output values that Table 1 shows. As- mV. And, despite the ⫾30% tolerance of back if desired. Figure 3 shows a block di-
sume that the desired value of VCOM is R2, the midscale VCOM output meets the agram of the digital potentiometer.왏

100 edn | September 30, 2004 www.edn.com