Biomedical Engineering Reference
In-Depth Information
satisfies design specifications. The solution thus obtained yields a biochip
design with a 10 × 10 microfluidic array and an assay completion time of
390 s. Next, we compare it with a design synthesized using the routing-aware
defect-oblivious method described in Section 2.2. The metric for comparison
is the amount of defect tolerance exhibited by each design. For the defect-
oblivious case, we use the design shown in Figure 2.6b, which is a 10 × 10
microfluidic array with an assay completion time of 377 s.
We evaluate the defect tolerance of the two synthesized designs by inject-
ing random defects. A design is deemed to be robust if the injected defects
can be bypassed by partial reconfiguration. Defects can be classified based
on their impact on bioassay functionality.
The first category includes defects that affect only the unused cells in
the array. As the biochip functionality is not compromised, these defects
are referred to as
benign
. The second category refers to defects that cause
significant “fragmentation” of the array, whereby it is no longer possible to
relocate a microfluidic module to another part of the array due to lack of
availability of defect-free cells. These defects are referred to as
catastrophic
.
The third category includes defects that are neither benign nor catastrophic.
The microfluidic array can be reconfigured for such defects; hence, we refer
to these defects as
repairable
.
A biochip that contains only benign defects is placed in Group I. A biochip
that contains catastrophic defects is placed in Group II. Finally, a biochip that
contains only repairable and benign defects is placed in Group III. Let
N
t
be the total number of biochips in a representative sample, and let
N
i
be the
number of biochips in Group
i
, 1 <
i
< 3. Clearly,
N
1
+
N
2
+
N
3
=
N
t
. We next
define two ratios related to the defect tolerance capability of the synthesized
biochip: (1) robustness index
r
= (
N
1
+
N
3
)/
N
t
; (2) failure index
f
=
N
2
/
N
t
. The
goal of defect-aware synthesis is to maximize
r
and minimize
f
.
Resynthesis must be carried out for biochips in Group II, that is, for biochips
that suffer from catastrophic defects. Let the bioassay completion time before
(after) resynthesis be
T
1
(
T
2
). We define the time degradation
td
as follows:
td
= (
T
2
−
T
1
)/
T
1
. Another goal of defect-aware synthesis is to minimize
td
.
We take 100 simulated samples of a microfluidic biochip synthesized for
the protein assay with defect tolerance as a criterion and without defect
tolerance, that is, in a defect-oblivious manner. In each case, we randomly
inject defects by assuming that each unit cell is defective with probability
p
(
p
= 0.01, 0.05, 0.1) in our experiments. We then determine the ratios
r
,
f
, and
td
for both methods. The results are shown in Table 2.2.
As shown in Table 2.2, although the defect-tolerant design leads to slightly
higher assay times, this design leads to a DTI value of 0.8918, which implies
that almost 90% of the modules can be reconfigured if they are affected by
defects. This is a considerable improvement over the DTI value of 0.0144
obtained using the defect-oblivious method. This improvement is also
apparent from the comparison of the failure ratio (
f
), the robustness index (
r
),
and the time degradation (
td
) from Table 2.2.
