Datasheet

LTC6957-1/LTC6957-2/
LTC6957-3/LTC6957-4
30
6957f
For more information www.linear.com/LTC6957-1
applicaTions inForMaTion
designed system would result in compromised spectral
performance. This often catches designers by surprise
because the mechanisms above are typically additive and
linear, which result in filtering and additional spectral com-
ponents, but don’t by themselves create phase modulation.
Unfortunately, any limiter, including the LTC6957, will,
through its nonlinear action, transform additive terms into
phase modulation. When a small tone is added to a large
pure tone, the larger tone will appear to have its amplitude
and phase modulated at a rate equal to the difference of
the two frequencies. Pass this through a limiter and only
the phase modulation remains.
In large complex systems, it may be impractical to eliminate
all potential corrupting of the clock signals. In such a case,
a narrow band filter placed at the inputs of the LTC6957
can remove the unwanted spectral components that are
far enough away from the fundamental.
Close-in spectral anomalies will likely be impervious to
such filtering. Therefore, it is doubly important to keep an
eye out for modulating mechanisms. If the clock is routed
through CMOS logic gates, the power supply used for that
gate will AM modulate the signal at the very
least. The
modulation could manifest itself as sideband tones if the
power supply has repetitive disturbances, common with
switching power supplies, or it could manifest itself as
random noise if the noise of a linear regulator is too high.
Another source of corruption in large systems or labora-
tory measurements is the use of flexible cabling, which
can have a low level piezoelectric effect that modulates the
electrical length in response to mechanical vibration.
Rigid or semi-rigid cabling and PCB routing can be used
to eliminate this source of signal corruption.
AM to PM Conversion at the LTC6957 Inputs
The LTC6957 input stage has some AM to PM conversion,
but as seen in the Typical Performance Characteristics
section, even at 300MHz this is less than 0.5°/dB. One
source of AM to PM conversion at the LTC6957 input is
the optional lowpass filtering, because the upper sideband
and the lower sideband will be attenuated by slightly
different amounts. This difference is quite small for low
offset frequencies, but the difference grows both as the
frequency of the modulation increases, and as the carrier
frequency approaches the filter cutoff frequency where the
filter has a steeper roll-off.
Therefore, if
small amounts of AM are known to be present
and an unacceptable level of PM is seen at the LTC6957
output, it may be helpful to change the input filter setting
to a higher cutoff frequency.
Cross Talk from Loading at the LTC6957 Outputs
Another mechanism to be aware of in the LTC6957 is
cross-modulation of the outputs. Except for the CMOS
LTC6957-3/LTC6957-4, there is minimal direct AM or
PM modulation of the outputs by the power supply. In
the CMOS case, the V
DD
power supply will directly AM
modulate the outputs, with a small amount of AM to PM
conversion.
The thing to be aware of here is that there can be load-
induced disturbances internal to the LTC6957 that can
modulate the other output. For instance, hooking up one
output to an ADC encode input and the second output to
the FPGA that performs the first DSP on the ADC outputs,
can result in considerable kickback of FPGA generated
signals into the LTC6957. If this cross-modulates over to
the other output, all kinds of deleterious effects may be
seen including tones, images, etc.
The CMOS LTC6957-3/LTC6957-4 are more susceptible
to
this than the LVPECL and LVDS (LTC6957-1/LTC6957-2). To
prevent this, a buffer can be placed between the LTC6957
and the FPGA, even one that compromises the full jitter
performance considerably. Because it is the ADC that is
doing the sampling—the FPGA clock input has enough
margin for error to qualify as a digital signal.