AD8345
Rev. B | Page 13 of 20
BASEBAND I AND Q CHANNEL DRIVE
The I channel and Q channel baseband inputs should be driven
differentially. This is convenient as most modern high-speed
DACs have differential outputs. For optimal performance at
VS = 5 V, the drive signal should be a 1.2 V p-p differential
signal with a bias level of 0.7 V; that is, each input should swing
from 0.4 V to 1 V. If the AD8345 is being run on a lower supply
voltage, then the peak-to-peak voltage on the I and Q channel
inputs must be reduced to avoid input clipping. For example, at
a supply voltage of 2.7 V, a 200 mV p-p differential drive is
recommended. This results in a corresponding reduction in
output power (see Figure 3). The I and Q inputs have a large
input bandwidth of approximately 80 MHz. At lower baseband
input levels, the input bandwidth increases (see Figure 4).
If the baseband signal has a high peak-to-average ratio (such as
CDMA or WCDMA), then the rms signal strength must be
backed off from this peak level in order to prevent clipping of
the signal peaks.
Clipping of signal peaks tends to increase signal leakage into
adjacent channels. Backing off the I and Q signal strength, in
the manner recommended, reduces the output power by a
corresponding amount. This also applies to multicarrier
applications where the per-carrier output power is lower by
3 dB for each doubling of the number of output carriers.
The I and Q inputs have high input impedances because they
connect directly to the bases of PNP transistors. If a dc-coupled
filter is being used between a DAC and the modulator inputs,
then the filter must be terminated with the appropriate
resistance. If the filter is differential, then the termination
resistor should be connected across the I and Q differential
inputs.
REDUCTION OF LO LEAKAGE
Because the I and Q signals are being effectively multiplied with
the LO, any internal offset voltages on these inputs result in
leakage of the LO. The nominal LO leakage of −42 dBm, which
results from these internal offset voltages, can be reduced further
by applying offset compensation voltages on the I and Q inputs.
(Note that LO feedthrough is reduced by varying the differential
offset voltages on the I and Q inputs, not by varying the nominal
bias level of 0.7 V.) The reduction is easily accomplished by
programming (and then storing) the appropriate DAC offset
code. This does, however, require dc coupling the path from the
DAC to the I and Q inputs. (DC coupling is also advantageous
from the perspective of I and Q input biasing if the DAC is
capable of delivering a bias level of 0.7 V.)
The procedure for reducing the LO feedthrough is simple. In
order to isolate the LO in the output spectrum, a single
sideband configuration is recommended (set I and Q signals to
sine and cosine waves at, for example, 100 kHz; set LO to
FRF − 100 kHz). An offset voltage is applied from the I DAC
until the LO leakage reaches a trough. With this offset level
held, an offset voltage is applied to the Q DAC until a (lower)
trough is reached.
LO leakage compensation holds up well over temperature.
Figure 10 shows the effect of temperature on LO leakage after
compensation at ambient.
Compensated LO leakage degrades somewhat as the frequency
is moved away from the frequency at which the compensation
was performed. This is due to the effects of LO to RF output
leakage, which is not a result of offsets on the I and Q inputs.
SINGLE-ENDED I AND Q DRIVE
Where only single-ended I and Q signals are available, a
differential amplifier such as the AD8132 or AD8138 can be
used to generate the required differential drive signal for the
AD8345.
Although most DACs have differential outputs, using a single-
ended, low-pass filter between the dual DAC and the I and Q
inputs can be more desirable from the perspective of
component count and cost. As a result, the output signal from
the filter must be converted back to differential mode and
possibly be rebiased to 0.7 V common mode.
Figure 30 shows a circuit that converts a ground-referenced,
single-ended signal to a differential signal and adds the required
0.7 V bias voltage. Two AD8132 differential op amps configured
for unity gain are used. With a 50 Ω input impedance, this
circuit is configured to accept a signal from a 50 Ω source (for
example, a low-pass filter). The input impedance can be easily
changed by replacing the 49.9 Ω shunt resistor (and the
corresponding 24.9 Ω resistor on the inverting input) with the
appropriate value. The required dc-bias level is conveniently
added to the signal by applying 0.7 V to the VOCM pins of the
differential amplifiers.
Differential amplifiers, such as the AD8132 and AD8138, can
also be used to implement active filters. For more information
on this topic, refer to the data sheets of these devices.