
7
FN8087.3
April 21, 2006
Board Mounting Considerations
For applications requiring the highest accuracy, board
mounting location should be reviewed. Placing the device in
areas subject to slight twisting can cause degradation of the
accuracy of the reference voltage due to die stresses. It is
normally best to place the device near the edge of a board,
or the shortest side, as the axis of bending is most limited at
that location. Obviously mounting the device on flexprint or
extremely thin PC material will likewise cause loss of
reference accuracy.
Noise Performance and Reduction
The output noise voltage in a 0.1Hz to 10Hz bandwidth is
typically 30μV
P-P
. This is shown in the plot in the Typical
Performance Curves. The noise measurement is made with
a bandpass filter made of a 1 pole high-pass filter with a
corner frequency at 0.1Hz and a 2-pole low-pass filter with a
corner frequency at 12.6Hz to create a filter with a 9.9Hz
bandwidth. Noise in the 10kHz to 1MHz bandwidth is
approximately 400μV
P-P
with no capacitance on the output,
as shown in Figure 16. These noise measurements are
made with a 2 decade bandpass filter made of a 1 pole high-
pass filter with a corner frequency at 1/10 of the center
frequency and 1-pole low-pass filter with a corner frequency
at 10 times the center frequency. Figure 16 also shows the
noise in the 10kHz to 1MHz band can be reduced to about
50μV
P-P
using a 0.001μF capacitor on the output. Noise in
the 1kHz to 100kHz band can be further reduced using a
0.1μF capacitor on the output, but noise in the 1Hz to 100Hz
band increases due to instability of the very low power
amplifier with a 0.1μF capacitance load. For load
capacitances above 0.001μF the noise reduction network
shown in Figure 17 is recommended. This network reduces
noise significantly over the full bandwidth. As shown in
Figure 16, noise is reduced to less than 40μV
P-P
from 1Hz
to 1MHz using this network with a 0.01μF capacitor and a
2k
Ω
resistor in series with a 10μF capacitor.
Turn-On Time
The ISL60007 devices operate with ultra-low supply current
and thus the time to bias up internal circuitry to final values
will be longer than with references that require higher
current. Normal turn-on time is typically 4ms. This is shown
in Figure 18. Since devices can vary in supply current down
to 250nA, turn-on time can last up to about 6ms. Care
should be taken in system design to include this delay before
measurements or conversions are started.
Temperature Coefficient
The limits stated for temperature coefficient (tempco) are
governed by the method of measurement. The
overwhelming standard for specifying the temperature drift of
a reference is to measure the reference voltage at two
temperatures, take the total variation, (V
HIGH
- V
LOW
), and
divide by the temperature extremes of measurement
(T
HIGH
- T
LOW
). The result is divided by the nominal
reference voltage (at T = 25°C) and multiplied by 10
6
to yield
ppm/°C. This is the “Box” method for specifying temperature
coefficient.
C
L
= 0
400
350
300
250
200
150
100
50
0
1
10
100
1000
10000
100000
N
P
)
FIGURE 16. NOISE REDUCTION
C
L
= 0.001μF
C
L
= 0.01μF
& 10μF + 2k
Ω
C
L
= 0.1μF
V
IN
= 3.0V
V
IN
V
O
GND
ISL60007
0.01μF
10μF
2k
Ω
0.1μF
10μF
FIGURE 17.
0
0.5
1
1.5
2
2.5
3
3.5
0
2
4
6
8
10
12
TIME (ms)
V
I
O
400nA
V
IN
570nA
250nA
FIGURE 18. TURN-ON TIME (+25°C)
ISL60007