Analog Devices AN-753 Application Notes

AN-753

ADC

INPUT

(VIA MUX)

TOUCH

SCREEN

V

REF

V

CC

X+

REF+

SINGLE-ENDED METHOD

RATIOMETRIC METHOD

REF–

Y+

Y–

GND

APPLICATION NOTE

One Technology Way • P.O. Box 9106 • Norwood, MA 02062-9106 • Tel: [phone] • Fax: [phone] • www.analog.com

Conguring the AD7877

by Susan Pratt

INTRODUCTION

The AD7877 touch screen controller is a 12- bit suc­cessive approximation ADC with a synchronous serial
interface and low on resis tance switches for driving
touch screens. The AD7877 features direc t bat ter y
measurement on two inputs, temperature and touch­pressure measurement.

The AD7877 has many user-programmable conversion
controls, including variable acquisition time, rst conver­sion delay, and averaging. It is ideal for battery-powered
systems, such as personal digital assistants, smar t
phones, and other portable equipment with resistive
touch screens.

The AD7877 requires conguration via the on- board
registers to fully utilize its features. This application
note goes through the conguration process step by
step, and explains how to congure the device for your
application.

SINGLE-ENDED OR RATIOMETRIC MEASUREMENT

The main task of any touch screen controller is to take
accurate measurements from the touch screen. On the
AD7877, there are two measurement methods available:
single- ended or ratiometric. The difference between the
two methods is in the ADC reference voltage. In single ­ended mode, the ADC is referenced to ground and to
V

. (V

REF

an external reference). For the differential measurement,
the ADC is referenced to the screen excitation voltage.

REV. 0

can be either the internal reference voltage or

REF

Figure 1. Measurement Methods

The advantage of using the single -ended method is that
the touch screen does not need to be powered except
during signal acquisition. Signal acquisition takes only
a portion of the total conversion time. This method can
result in signicant power savings in the system, as the
screen itself can draw more than 1 mA while powered.
This method also has some disadvantages. Voltage
drops across the internal switches cause errors in the
measurement, as they reduce the apparent excitation
voltage across the screen. In addition, this method
can only be used when VCC is close to V
the ADC is saturated (when VCC > V

REF

the full range of the ADC is not used (VCC < V

, otherwise

REF

), or conversely,

). The

REF

single - ended method is particularly suited for battery­operated systems where power is at a premium. It is only
recommended for applications that use the nger as an
input device, and therefore do not need high accuracy
measurements.

Using the touch screen excitation voltage as the ADC
referen ce gives a ratiometric measurement of the
input signal. This method is more accurate than the
single - ended method because voltage drops across the
internal switches do not have any effect on the measured
results. However, the screen needs to be powered at all
times in order to provide the ADC with its reference. For
applications such as handwriting recognition, where the
measured touch screen position must be accurate, the
ratiometric method is recommended.

AVERAGING

The signals measured from the touch screen can be
noisy. To minimize low level noise, the AD7877 can
automatically perform averaging on the measured input
signals. The AD7877 can be programmed to perform one,
four, eight, or sixteen consecutive measurements on the
same channel, and write the averaged result to the result
register. The more averages the part performs, the longer
each measurement takes. When choosing the amount of
averaging in the system, the time available for conver­sions needs to be taken into account. The amount of
averaging required will depend on the noise factors in
the specic system, but in general, at least four averages
are recommended for each measurement.

AN-753

–3–

AN-753

FIRST CONVERSION DELAY = 8.19ms

FIRST CONVERSION DELAY = 1.024ms

FIRST CONVERSION DELAY = 500ns

REV. 0

STOPACQ

In most applications, the touch screen is placed directly
on top of the LCD. Periodic noise from the display can
interfere with the touch screen measurements. As the
LCD horizontal lines are written, noise is generated that
can be coupled onto the touch screen. This specic
type of noise is generally framed by an LCD control
signal related to the horizontal refresh phase. HSYNC
and Vcom are the two typical control signals found on
different types of LCDs. The noise can be framed by the
high or low period of these signals, or it may be worse
at the signal transitions.

If the user sees periodic noise in the system, an analysis
should be performed to determine if the noise is coming
from the LCD, and whether there is a control signal
available that frames the noise. This control signal
should be tied directly to the AD7877 StopAc q pin,
provided the signal level is suitable. Even if the control
signal does not frame all the noise but frames at least
some of it, greater accuracy will be obtained by using
the StopAcq feature.

The StopAcq feature can be used with any signal to
control the acquisition period of the AD7877. Its recom­mended use is in prevention of noise pickup from the
LCD as described earlier, but that does not preclude its
use for other purposes.

Figure 2 graphically illustrates the ef fect of the rst
conversion delay on the accuracy of the touch screen
results. The correc t first conversion delay value for
the particular touch screen used in this experiment is

1.024 ms. The samples taken from the touch screen using
this value are in green and clearly show that all points
on the touch screen can be measured correctly. Plenty
of samples are gathered all over the screen.

If the rst conversion delay is set too small, then the
range of values read from the screen is severely limited.
In this case, the signal has not had time to settle, so the
input value is being measured while the signal is still
ramping up to its true value. As illustrated by the red
samples, vast areas of the screen will never register as
being touched.

If the rst conversion delay is too large, then not enough
samples are measured to gain suf cient information
from the screen. The black samples illustrate this. While
positions from the full screen area can be measured, the
time between samples is too great. If the touched posi­tion was to move quickly, not enough samples would be
taken to track these changes. Applications such as writ­ing on the screen would be impossible with too large a
rst conversion delay.

FIRST CONVERSION DELAY

The nature of the touch screen measurement process
means that voltages are being switched on and off the
touch screen pins in rapid sequence. To get both X and
Y positions, voltage needs to be switched across rst the
X then the Y layer, while the Y+ or X+ input is switched
to the ADC. Once the voltages are switched, the input
signal will require a settling time before it reaches its
true value. Taking a measurement before the signal is
settled gives an erroneous result.

The AD7877 can be programmed to add an automatic
delay before each touch screen measurement com­mences. The delay should be long enough to allow the
input signal to settle. The AD7877 rst conversion delay
can be 500 ns, 128 µs, 1.024 ms, or 8.19 ms. The user
should calculate the settling time of the input signal and
then choose a rst conversion delay that is at least as
long. Note that the choice of rst conversion delay will
add to the total conversion time. The rst conversion
delay happens automatically before the rst conversion
is performed on the AD7877, before each touch screen
positional measurement, and after the nal conversion
in a sequence.

Figure 2. Effect of First Conversion Delay on
Touch Screen Results

–2–

REV. 0