EL5236, EL5237
14
FN7833.0
March 31, 2011
This approximate analysis is intended to show the importance of
working with relatively low resistor values if the low noise of the
EL5236 is to be retained. It also shows why, with an inverting
configuration, it is important to either keep a low impedance on
the non-inverting input and/or add a noise shunting capacitor
across it.
DC Precision
The EL5236 offers extremely low input offset voltage and input
offset current. To take full advantage of the very low offset
current (<±500nA), the source resistance looking out of the two
inputs must be matched. Figure
40 shows the output DC offset
analysis circuit.
If Rs = Rf||Rg is imposed on the design, the total output offset
will be given by Equation
4:Putting in the specified worst case limits of ±3mV for the offset
voltage and ±500nA for the offset current into a NG = 2 and
Rf = 402 condition would give an output DC error envelope of
±6.2mV. This is assuming the Rs is set to 201. To change
Figure
37 to a 201 Rs, add a 175 in series with the V+ node
from the 50 termination. This will reduce the output offset
induced from the Ib terms to the ±0.2mV part of the ±6.0mV
computed above – at the cost of a bit higher input noise.
A second issue would be the tempco of the output offset voltage.
To the extent that the output is dominated by the offset voltage
term, its drift will dominate. The specified typical input Vos drift is
-300nV/°C. Continuing this example, that would give a typical
output drift of -0.6V/°C. Over a +50°C ambient range, this
would map to only a 30V shift in the output offset voltage.
Active Filter Designs
Being a low gain stable wideband VFA op amp, the EL5236 is
particularly suited to differential I/O active filters, as shown in
Figure 1. That relatively complex example gives a 5th order Butterworth filter as part of an output stage interface to a
complementary DAC output current. These DAC output stages
generate both a common mode voltage and a differential signal
at the termination resistors to ground. The design of Figure
1gets the real pole as part of that termination then implements
the two complex pole pairs as an SKF stage followed by an MFB
stage. This gives much better stop band rejection using the 2nd
MFB stage and allows an easy place to introduce a common
mode level shift to remove the DAC output common mode. This
was used to return the final output to be a ground centered
differential signal. The MFB design of the output stage uses a
feedback capacitor inside the filter that normally expects a unity
gain stable op amp for implementation. Adding the two
capacitors to ground on the inverting inputs of this stage shapes
the noise gain up at higher frequencies holding this stage stable.
Simpler designs are possible as shown in Figure
41. This is a single
stage gain of 2 Butterworth filter with a 20MHz cutoff. Even with the
300MHz gain bandwidth product of the EL5236, this is a fairly high
frequency filter to attempt with this relatively limited amplifier
bandwidth margin. In this case, the Rf = Rg = 649 is also being set
to get bias current cancellation for improved output DC precision
along with the necessary gain of 2 setting for the design.
This design was produced using the Intersil online active filter
designer which includes an amplifier bandwidth adjustment in
the R1 and R2 values. It is a general active filter tool tailored to
the available precision and high speed op amps from Intersil. It is
available at the following link:
As shown in the simulated vs. ideal curves of Figure
42, the
design is doing a very good job of matching the ideal response
through 80MHz. All SKF filters deviate from the ideal roll-off at
higher frequencies due to the increase in output impedance as
the amplifier bandwidth is approached.
-
+
Rg
Vos
Rs
Rf
ib
Vio
ib
ISL5236
FIGURE 40. OUTPUT DC OFFSET ANALYSIS CIRCUIT
f
os
io
os
R
I
NG
V
*
±
±
=
(EQ. 4)
-
+
97.6
649
Rg
226
+VS
ISL5236
Vo
Vi
-VS
Rf
47pF
FIGURE 41. GAIN = +2V/V, 20MHZ 2nd ORDER BUTTERWORTH
LOW PASS ACTIVE FILTER
10
5
0
-5
-10
-15
-20
GAI
N
(
d
B)
1E7
1E8
FREQUENCY (Hz)
FIGURE 42. SIMULATED vs IDEAL FILTER RESPONSE
COMPARISON