![background image](http://html.mh-extra.com/html/national-semiconductor/lmh6624/lmh6624_manual_710362014.webp)
Application Section
(Continued)
R
f
||R
g
= R
seq
for bias current cancellation.
Figure 4
illus-
trates the equivalent noise model using this assumption.
Figure 5
is a plot of e
ni
against equivalent source resistance
(R
seq
) with all of the contributing voltage noise source of
Equation 2. This plot gives the expected e
ni
for a given (R
seq
)
which assumes R
f
||R
g
= R
seq
for bias current cancellation.
The total equivalent output voltage noise (e
no
) is e
ni
*
A
V
.
(2)
As seen in
Figure 5
, e
ni
is dominated by the intrinsic voltage
noise (e
n
) of the amplifier for equivalent source resistances
below 33.5
Ω
. Between 33.5
Ω
and 6.43k
Ω
, e
ni
is dominated
by the thermal noise (e
t
=
√
(4kT(2R
seq
)) of the external
resistor. Above 6.43k
Ω
, e
ni
is dominated by the amplifier’s
current noise (i
n
=
√
(2) i
n
R
seq
). The point at which the
LMH6624’s voltage noise and current noise contribute
equally occurs for R
seq
= 464
Ω
(ie., e
n
/
√
(2)i
n
). For example,
configured with a gain of +20V/V giving a −3dB of 90MHz
and driven from R
seq
= 25
Ω
, the LMH6624 produces a total
equivalent input noise voltage (e
ni
x
1.57
*
90MHz) of
16.5µV
rms
.
If bias current cancellation is not a requirement, then R
f
||R
g
need not equal R
seq
. In this case, according to Equation 1,
R
f
||R
g
should be as low as possible to minimize noise.
Results similar to Equation 1 are obtained for the inverting
configuration of
Figure 2
if R
seq
is replaced by R
b
and R
g
is
replaced by R
g
+ R
s
. With these substitutions, Equation 1 will
yield an e
ni
referred to the non-inverting input. Referring e
ni
to the inverting input is easily accomplished by multiplying
e
ni
by the ratio of non-inverting to inverting gains.
NOISE FIGURE
Noise Figure (NF) is a measure of the noise degradation
caused by an amplifier.
(3)
The Noise Figure formula is shown in Equation 3. The addi-
tion of a terminating resistor R
T
, reduces the external ther-
mal noise but increases the resulting NF. The NF is in-
creased because R
T
reduces the input signal amplitude thus
reducing the input SNR.
(4)
The noise figure is related to the equivalent source resis-
tance (R
seq
) and the parallel combination of R
f
and R
g
. To
minimize noise figure.
•
Minimize R
f
||R
g
•
Choose the Optimum R
S
(R
OPT
)
R
OPT
is the point at which the NF curve reaches a minimum
and is approximated by:
R
OPT
≈
e
n
/i
n
NON-INVERTING GAINS LESS THAN 10V/V
Using the LMH6624 at lower non-inverting gains requires
external compensation such as the shunt compensation as
shown in
Figure 6
. The compensation capacitors are chosen
to reduce frequency response peaking to less than 1dB.
INVERTING GAINS LESS THAN 10V/V
The lag compensation of
Figure 7
will achieve stability for
lower gains. Placing the network between the two input
terminals does not affect the closed-loop nor noise gain, but
is best used for the inverting configuration because of its
affect on the non-inverting input impedance.
20058921
FIGURE 4. Noise Model with R
f
||R
g
= R
seq
20058922
FIGURE 5. Voltage Noise Density vs. Source
Resistance
20058924
FIGURE 6. External Shunt Compensation
LMH6624
www.national.com
14