Datasheet

LTC6957-1/LTC6957-2/
LTC6957-3/LTC6957-4
24
6957f
For more information www.linear.com/LTC6957-1
Figure 8. LTC6957-2 Propagation Delay vs Overdrive
OVERDRIVE (mV)
0
500
t
PD
(ps)
600
700
800
900
1500
10
20
30 40
50
60 70
80 90
6957 F08
1400
1300
1200
1100
1000
100
t
PD
–
IN
+
OFFSETTED ±50mV
DC
IN
–
DRIVEN 100mV
P-P
FILTA = FILTB = L
t
PD
+
applicaTions inForMaTion
One important observation to take away from Figures 7a
to 7c is that while the worst filter settings for a given set
of conditions should certainly be avoided, it doesn't matter
nearly as much if the optimal or next to optimal filter set-
ting is used, because they are always fairly comparable in
terms of phase noise. So if a design will have an octave or
two range of amplitudes or frequencies, it is sufficient to
choose the filter setting whose range most closely matches
the application's range when using Tables 2 or 3 and the
noise penalty will not be severe anywhere in the range.
Evidently, the input filtering will not significantly help with
large and fast slewing input signals to the LTC6957. As
seen in Figure 1, the input has a differential pair before the
filters, so the limiting will already have happened before
the filter. Fortunately, with large input signals, performance
is typically better than with smaller input signals because
phase noise is a signal-to-noise phenomenon.
Input Drive and Output Skew
All versions of the LTC6957 have very good output skew;
the specification limits consist almost entirely of test
margins. Even laboratory verification of
the skew between
different
outputs is a challenging exercise, given the need
to measure within ±1ps. With electromagnetic propagation
velocity in FR-4 being well known as 6" per nanosecond,
the skew of the LTC6957 will be impacted by PCB trace
routing length differences of just 6mils.
The LTC6957 t
PD
and t
SKEW
are specified for a 100mV
step with 50mV of overdrive. This is common for high
speed comparators, though it may not reflect the typi-
cal application usage of parts such as the LTC6957. The
propagation delay of the LTC6957 will increase with less
overdrive and decrease with more overdrive, as would
that of a high speed comparator. To a lesser extent, hav-
ing the same overdrive but a larger signal (for instance a
differential input step of –200mV to 50mV) will increase
propagation delay, though this effect is smaller and can
usually be ignored.
A consequence of this behavior may be a perceived mis-
match between the propagation delay for rising versus
falling edges when driven with an AC-coupled input whose
duty cycle is not exactly 50%. The LTC6957 inputs are
internally DC-coupled, and as shown in Figure 1, biasing
is provided at ~64% of the supply
voltage. AC-coupled
input signals with a duty cycle of exactly 50% will see
symmetric levels of overdrive for the two signal directions.
If, for example, the input signal is a 100mV
P-P
square
wave with a duty-cycle of 48%, meaning it is high 48%
of the time and low 52% of the time, the DC average will
be 48mV above the low voltage level. This means the ris-
ing edge has 52mV of overdrive, and the falling edge has
48mV of overdrive.
As a result of this, the rising edge t
PD
will be faster than
the falling edge t
PD
. Fortunately, this will make the output
duty cycle closer to 50% than the input duty cycle. Figure 8
is from measurements on the LTC6957-2, with a 2V to
2.1V square wave on IN
–
, and with IN
+
set to various
DC voltages between those two levels. The X-axis is the
overdrive level for the t
PD
+ data, and is 100mV minus
the overdrive level for the t
PD
– data, to illustrate the level
of t
PD
changes that can unexpectedly occur with AC-
coupling. The lines are dashed where the measurement
uncertainty becomes large, when single digit millivolts
and picoseconds are being
measured
1
. As can be seen,
the t
PD
+/t
PD
– mismatch is very good at 50mV where the
two overdrive levels are the same.
1
Below 2mV to 3mV, the input offset and the small input hysteresis play a role too. Fortunately,
neither is large enough to be a concern in normal operation.