IEEE Solid-States Circuits Magazine - Spring 2019 - 20

response, while the green curve corresponds to the composite channel/
de-emphasis response, and the blue
curve presents the composite channel/DFE response. As shown in Figure 8(a), all three equalizers are capable
of flattening the composite response
to a frequency beyond 1 GHz (further
extension is possible). In this case,
any of the solutions might suffice,
with the final design choice depending on other factors, such as the circuit power or silicon footprint. On the
other hand, Figure 8(b) demonstrates
that neither CTLE nor de-emphasis
is capable of compensating for the
insertion loss null resulting from the
excessive signal reflections in the
multidrop memory configuration.

The DFE, however, is able to span the
null, providing a significant increase
in received SI.
Figure 9 provides the more familiar time-domain presentation of the
unequalized channel (red) and composite channel/equalizer responses
(cyan = CTLE, green = de-emphasis,
and blue = four-tap DFE) for the
multidrop server channel response
reported in Figure 8(a) and represented in Figure 6. As shown, both
CTLE and de-emphasis suppress the
intersymbol interference (ISI), as
measured by the reduced variation
of the high and low signal levels (e.g.,
less voltage distribution) relative to
the unequalized channel, yet neither opens the eye. In contrast, the

-5

-5
Channel

Channel
Four-Tap DFE

-10

Four-Tap DFE
Spans the Null

-10
-15
(dB)

-15
(dB)

four-tap DFE demonstrates a clear
ability to open the data eye, despite
the complex channel behavior.
The tap-count choice, while ultimately imposed by the standard,
was specified intelligently as a function of both the anticipated channel
behaviors and circuit design tradeoffs
(including area, power, and routing
congestion to each DFE block). Figure 10 presents the channel-based
argument, while Figure 11 highlights a
portion of the circuit considerations.
Figure 10 compares the target response
of an ideal zero-forcing equalizer (corresponding to the multidrop channel
previously introduced), with the simulated DFE response for one tap, four
taps, and 10 taps. As the DFE tap count

De-Emphasis
-20

De-Emphasis

-20

CTLE

CTLE

-25

-25

-30

-30
0

0.5

1

1.5
2
Frequency (GHz)
(a)

2.5

3

0

0.5

1

1.5
2
Frequency (GHz)
(b)

2.5

3

FIGURE 8: The effectiveness of CTLE, de-emphasis, and four-tap DFE applied to (a) a well-behaved low-pass channel and (b) a typical multidrop server channel.

1
0.9
0.8
0.7
0.6
-100

0

100 200 300 400 500 600
Time (ps)
(a)

100 200 300 400 500 600
Time (ps)
(b)

100 200 300 400 500 600 700
Time (ps)
(c)

FIGURE 9: A time-domain comparison of the relative equalizer effectiveness over the multidrop server channel presented in Figure 8(b):
(a) CTLE, (b) de-emphasis, and (c) four-tap DFE.

20

S P R I N G 2 0 19

IEEE SOLID-STATE CIRCUITS MAGAZINE
IEEE Solid-States Circuits Magazine - Spring 2019

Table of Contents for the Digital Edition of IEEE Solid-States Circuits Magazine - Spring 2019
