Digital Signal Processing Reference
In-Depth Information
TABLE 14-1. System Input Variables for the Signaling Example of Chapter 14
Model Term
Minimum
Typical
Maximum
Transmitter termination,
R
Tx
()
40
50
60
Receiver termination,
R
TT
()
40
50
60
Differential impedance,
Z
diff
()
84.0
100.3
117.9
Trace length,
L
(m)
0.254
0.381
0.508
Equalization coefficient, EQ
−
0
.
1
−
0
.
2
−
0
.
3
TABLE 14-2. Valid Equalization Settings
Setting
bit
3
bit
2
bit
1
bit
0
EQ
0
0
0
0
0
−
0
.
3000
1
0
0
0
1
−
0
.
2867
2
0
0
1
0
−
0.2733
3
0
0
1
1
−
0
.
2600
4
0
1
0
0
−
0.2467
5
0
1
0
1
−
0
.
2333
6
0
1
1
0
−
0
.
2200
7
0
1
1
1
−
0
.
2067
8
1
0
0
0
−
0
.
1933
9
1
0
0
1
−
0
.
1800
10
1
0
1
0
−
0
.
1667
11
1
0
1
1
−
0
.
1533
12
1
1
0
0
−
0
.
1400
13
1
1
0
1
−
0
.
1267
14
1
1
1
0
−
0
.
1133
15
1
1
1
1
−
0
.
1000
The number of terms in the second-order model,
k
,is
n
var
(n
var
−
1
)
k
=
1
+
2
n
var
+
(14-3)
2
In creating the model, we use response data from multiple observations that are
generated from different combinations of model inputs. The number of obser-
vations and choice of input combinations is typically determined via the
design
of experiments
(DOE) approach. As its name implies, DOE is a discipline that
allows us to design experiments that will result in the creation of accurate mod-
els that require a minimal number of observations to obtain a good fit. The
latter characteristic is particularly important for analyzing and designing signal-
ing links. Simulations for fully coupled interconnects with nonlinear transceiver
models can take several minutes to complete. With that in mind, a technique that
minimizes the number of simulations needed to create a reliable response surface
model will reduce development time and CAD infrastructure costs. As was the
case with response surface modeling, DOE is a topic for which entire textbooks
Search WWH ::
Custom Search