Data Sheet AD8139
Rev. C | Page 23 of 26
Driving a Capacitive Load
A purely capacitive load reacts with the bondwire and pin
inductance of the AD8139, resulting in high frequency ringing
in the transient response and loss of phase margin. One way to
minimize this effect is to place a small resistor in series with
each output to buffer the load capacitance (see Figure 58 and
Figure 63). The resistor and load capacitance form a first-order,
low-pass filter; therefore, the resistor value should be as small as
possible. In some cases, the ADCs require small series resistors
to be added on their inputs.
–13
–12
–11
–10
–9
–8
–7
–6
–5
–4
–3
–2
–1
0
1
2
3
4
10M 100M 1G
5
FREQUENCY (Hz)
CLOSED LOOP GAIN (dB)
R
S
= 30.1Ω
C
L
= 15pF
R
S
= 60.4Ω
C
L
= 15pF
R
S
= 60.4Ω
C
L
= 5pF
R
S
= 0Ω
C
L, dm
= 0pF
R
S
= 30.1Ω
C
L
= 5pF
V
S
= ±5V
V
O, dm
= 0.1V p-p
G = 1 (R
F
= R
G
= 200Ω)
R
L, dm
= 1kΩ
4679-076
Figure 63. Frequency Response for
Various Capacitive Loads and Series Resistances
The Typical Performance Characteristics that illustrate transient
response vs. the capacitive load were generated using series
resistors in each output and a differential capacitive load.
Layout Considerations
Standard high speed PCB layout practices should be adhered
to when designing with the AD8139. A solid ground plane is
recommended, and good wideband power supply decoupling
networks should be placed as close as possible to the supply pins.
To minimize stray capacitance at the summing nodes, the copper
in all layers under all traces and pads that connect to the summing
nodes should be removed. Small amounts of stray summing-node
capacitance cause peaking in the frequency response, and large
amounts can cause instability. If some stray summing-node
capacitance is unavoidable, its effects can be compensated for
by placing small capacitors across the feedback resistors.
Terminating a Single-Ended Input
Controlled impedance interconnections are used in most
high speed signal applications, and they require at least one
line termination. In analog applications, a matched resistive
termination is generally placed at the load end of the line. This
section deals with how to properly terminate a single-ended
input to the AD8139.
The input resistance presented by the AD8139 input circuitry is
seen in parallel with the termination resistor, and its loading
effect must be taken into account. The Thevenin equivalent
circuit of the driver, its source resistance, and the termination
resistance must all be included in the calculation as well. An
exact solution to the problem requires the solution of several
simultaneous algebraic equations and is beyond the scope of
this data sheet. An iterative solution is also possible and simpler,
especially considering the fact that standard 1% resistor values
are generally used.
Figure 64 shows the AD8139 in a unity-gain configuration
driving the AD6645, which is a 14-bit, high speed ADC, and
with the following discussion, provides a good example of how
to provide a proper termination in a 50 Ω environment.
The termination resistor, RT, in parallel with the 268 Ω input
resistance of the AD8139 circuit (calculated using Equation 19),
yields an overall input resistance of 50 Ω that is seen by the signal
source. To have matched feedback loops, each loop must have
the same RG if they have the same RF. In the input (upper) loop,
RG is equal to the 200 Ω resistor in series with the (+) input plus
the parallel combination of RT and the source resistance of 50 Ω.
In the upper loop, RG is therefore equal to 228 Ω. The closest
standard 1% value to 228 is 226 Ω and is used for RG in the
lower loop. Greater accuracy could be achieved by using two
resistors in series to obtain a resistance closer to 228 Ω.
Things get more complicated when it comes to determining the
feedback resistor values. The amplitude of the signal source
generator VS is two times the amplitude of its output signal when
terminated in 50 Ω. Therefore, a 2 V p-p terminated amplitude
is produced by a 4 V p-p amplitude from VS. The Thevenin
equivalent circuit of the signal source and RT must be used when
calculating the closed-loop gain, because in the upper loop, RG is
split between the 200 Ω resistor and the Thevenin resistance
looking back toward the source. The Thevenin voltage of the
signal source is greater than the signal source output voltage
when terminated in 50 Ω because RT must always be greater than
50 Ω. In this case, RT is 61.9 Ω and the Thevenin voltage and
resistance are 2.2 V p-p and 28 Ω, respectively. Now the upper
input branch can be viewed as a 2.2 V p-p source in series
with 228 Ω. Because this is a unity-gain application, a 2 V p-p
differential output is required, and RF must therefore be 228 ×
(2/2.2) = 206 Ω. The closest standard value to this is 205 Ω.
When generating the Typical Performance Characteristics data,
the measurements were calibrated to take the effects of the
terminations on the closed-loop gain into account.