88
7 Designing Application-Specific Architectures
7.3 Design of Application-Specific Architectures
This section first discusses the task of designing application-specific architectures
and the applied quality criteria. In order to handle the complexity of the design
task, the utilization of powerful Satisfiability Modulo Theories solvers (SMTsolvers [5]) is proposed. The second part of this section therefore briefly introduces
the background on SMT-solvers and outlines the general idea how to solve the
considered problem using SMT-solvers.
7.3.1 Design Task
Recall that the main objective is the realization of an application-specific architecture realizing a set of experiments := {φ 1 , φ 2 , . . .} as discussed in Definition 7.2.
Additionally, the architecture has to satisfy all physical constraints (i.e., maximally
allowed successor edges of a module are limited to two and no cycles without
including the MPU are allowed) and, at the same time, has to satisfy the designer’s
quality requirements.
In order to define the quality of an architecture, several design objectives can be
used. Here, the following quality criteria can be employed 1 :
• The connection depth, i.e. the maximum number of connections a droplet has to
traverse from the MPU to the modules and, eventually, back to the MPU. Hence,
the maximal connection depth is determined by the experiment that requires
the longest path in the application-specific architecture. The connection depth
indicates the number of channels a droplet has to traverse in order to execute the
experiment.
• The number of connections, i.e. |E|. This metric represents the complexity of the
structure to be realized in the microfluidic network.
• The number of modules used in the architecture, i.e. |V | except the MPU.
This metric is important since more modules in an architecture cause higher
fabrication costs.
Determining an architecture, which optimizes one or multiple objectives, is a
computationally hard problem. To handle this complexity, the deductive power of
solving engines is exploited, which are introduced next.
1 Note that the proposed method can easily be extended to support further objectives. Furthermore,
note that in [48] additionally a contamination quality criterion was introduced, which counts the
number of times a droplet passes a connection. However, this criterion is not well-suited for droplet
microfluidics as droplets confine the samples and, hence, prevent contamination.
7 Designing Application-Specific Architectures
7.3 Design of Application-Specific Architectures
This section first discusses the task of designing application-specific architectures
and the applied quality criteria. In order to handle the complexity of the design
task, the utilization of powerful Satisfiability Modulo Theories solvers (SMTsolvers [5]) is proposed. The second part of this section therefore briefly introduces
the background on SMT-solvers and outlines the general idea how to solve the
considered problem using SMT-solvers.
7.3.1 Design Task
Recall that the main objective is the realization of an application-specific architecture realizing a set of experiments := {φ 1 , φ 2 , . . .} as discussed in Definition 7.2.
Additionally, the architecture has to satisfy all physical constraints (i.e., maximally
allowed successor edges of a module are limited to two and no cycles without
including the MPU are allowed) and, at the same time, has to satisfy the designer’s
quality requirements.
In order to define the quality of an architecture, several design objectives can be
used. Here, the following quality criteria can be employed 1 :
• The connection depth, i.e. the maximum number of connections a droplet has to
traverse from the MPU to the modules and, eventually, back to the MPU. Hence,
the maximal connection depth is determined by the experiment that requires
the longest path in the application-specific architecture. The connection depth
indicates the number of channels a droplet has to traverse in order to execute the
experiment.
• The number of connections, i.e. |E|. This metric represents the complexity of the
structure to be realized in the microfluidic network.
• The number of modules used in the architecture, i.e. |V | except the MPU.
This metric is important since more modules in an architecture cause higher
fabrication costs.
Determining an architecture, which optimizes one or multiple objectives, is a
computationally hard problem. To handle this complexity, the deductive power of
solving engines is exploited, which are introduced next.
1 Note that the proposed method can easily be extended to support further objectives. Furthermore,
note that in [48] additionally a contamination quality criterion was introduced, which counts the
number of times a droplet passes a connection. However, this criterion is not well-suited for droplet
microfluidics as droplets confine the samples and, hence, prevent contamination.