The buffer gain is typically very close to 1.00 and is normally
neglected from signal gain considerations. It will, however, set
the CMRR for a single op amp differential amplifier configura-
tion. For the buffer gain α < 1.0, the CMRR = –20 • log(1 – α).
The closed-loop input stage buffer used in the OPA2683 gives
a buffer gain more closely approaching 1.00 and this shows up
in a slightly higher CMRR than previous current-feedback op
amps.
frequency response given by Equation 2 will start to roll off,
and is exactly analogous to the frequency at which the noise
gain equals the open-loop voltage gain for a voltage-feed-
back op amp. The difference here is that the total impedance
in the denominator of Equation 3 may be controlled some-
what separately from the desired signal gain (or NG).
The OPA2683 is internally compensated to give a maximally
flat frequency response for RF = 953Ω at NG = 2 on ±5V
supplies. That optimum value goes to 1.2kΩ on a single +5V
supply. Normally, with a current-feedback amplifier, it is
possible to adjust the feedback resistor to hold this band-
width up as the gain is increased. The CFBPLUS architecture
has reduced the contribution of the inverting input impedance
to provide exceptional bandwidth to higher gains without
adjusting the feedback resistor value. The Typical Character-
istics show the small-signal bandwidth over gain with a fixed
feedback resistor.
RI, the buffer output impedance, is a critical portion of the
bandwidth control equation. The OPA2683 reduces this
element to approximately 5.0Ω using the loop gain of the
closed-loop input buffer stage. This significant reduction in
output impedance, on very low power, contributes signifi-
cantly to extending the bandwidth at higher gains.
A current-feedback op amp senses an error current in the
inverting node (as opposed to a differential input error volt-
age for a voltage-feedback op amp) and passes this on to
the output through an internal frequency dependent
transimpedance gain. The Typical Characteristics show this
open-loop transimpedance response. This is analogous to
the open-loop voltage gain curve for a voltage-feedback op
amp. Developing the transfer function for the circuit of Figure
13 gives Equation 2:
Putting a closed-loop buffer between the noninverting and
inverting inputs does bring some added considerations. Since
the voltage at the inverting output node is now the output of
a locally closed-loop buffer, parasitic external capacitance on
this node can cause frequency response peaking for the
transfer function from the noninverting input voltage to the
inverting node voltage. While it is always important to keep
the inverting node capacitance low for any current-feedback
op amp, it is critically important for the OPA2683. External
layout capacitance in excess of 2pF will start to peak the
frequency response. This peaking can be easily reduced by
then increasing the feedback resistor value—but it is prefer-
able, from a noise and dynamic range standpoint, to keep
that capacitance low, allowing a close to nominal 953Ω
feedback resistor for flat frequency response. Very high
parasitic capacitance values on the inverting node (> 5pF)
can possibly cause input stage oscillation that cannot be
filtered by a feedback element adjustment.
RF
α 1+
RG
RF + RI 1+
Z(S)
VO
α NG
RF + RI NG
=
=
V
RF
I
1+
Z(S)
RG
1+
RF
(2)
NG = 1+
RG
This is written in a loop-gain analysis format where the errors
arising from a non-infinite open-loop gain are shown in the
denominator. If Z(S) were infinite over all frequencies, the
denominator of Equation 2 would reduce to 1 and the ideal
desired signal gain shown in the numerator would be achieved.
The fraction in the denominator of Equation 2 determines the
frequency response. Equation 3 shows this as the loop-gain
equation.
An added consideration is that at very high gains, 2nd-order
effects in the inverting output impedance cause the overall
response to peak up. If desired, it is possible to retain a flat
frequency response at higher gains by adjusting the feed-
back resistor to higher values as the gain is increased. Since
the exact value of feedback that will give a flat frequency
response at high gains depends strongly in inverting and
output node parasitic capacitance values, it is best to experi-
ment in the specific board with increasing values until the
desired flatness (or pulse response shape) is obtained. In
general, increasing RF (and then adjusting RG to the desired
gain) will move towards flattening the response, while de-
creasing it will extend the bandwidth at the cost of some
peaking. The OPA683 data sheet gives an example of this
optimization of RF versus gain.
Z(S)
= Loop Gain
(3)
RF + RI NG
If 20 • log(RF + NG • RI) were drawn on top of the open-loop
transimpedance plot, the difference between the two would
be the loop gain at a given frequency. Eventually, Z(S) rolls off
to equal the denominator of Equation 3, at which point the
loop gain has reduced to 1 (and the curves have intersected).
This point of equality is where the amplifier’s closed-loop
OPA2683
SBOS244H
19
www.ti.com
TI [ TEXAS INSTRUMENTS ]
