98
7 Designing Application-Specific Architectures
sensitive experiments. Furthermore, the application-specific architectures never
require a payload re-injection, which simplifies the MPU. Overall, the resulting
architectures constitute a significant improvement.
7.6 Conclusion
This chapter proposed the design of application-specific architectures for a set
of experiments. These application-specific architectures address drawbacks of
currently considered ring architectures such as large execution times or the need
of complex payload re-injections. In order to derive the respective architectures, a
method was proposed which tackles the underlying complexity using SMT-solvers.
Evaluations showed that the proposed method enables designers to efficiently
generate architectures satisfying their needs as well as quality criteria. Moreover, the
evaluations confirmed the advantages of application-specific architectures compared
to the ring architecture.
The obtained application-specific architectures do not yet specify how the
modules and connections are implemented. For example, an architecture does
not consider how the connections are implemented in the form of channels and
bifurcations or how it is ensured that the droplets flow along the desired directions.
However, this constitutes a dimensioning task and, hence, the methods proposed in
Chap. 4 and [49] can be applied. How these methods can be integrated is discussed
in Chap. 9.
After dimensioning the architecture, a microfluidic network results where the
bifurcations realize multiple paths through this network. Along these paths, the
different operations on the droplet are executed and, hence, different experiments
can be triggered. That means, in order to execute an experiment, a payload droplet
has to be routed along the respective path. For this purpose, a droplet sequence
consisting of headers and the payload needs to be injected in a coordinated manner.
Determining such a droplet sequence is described in the next chapter.
7 Designing Application-Specific Architectures
sensitive experiments. Furthermore, the application-specific architectures never
require a payload re-injection, which simplifies the MPU. Overall, the resulting
architectures constitute a significant improvement.
7.6 Conclusion
This chapter proposed the design of application-specific architectures for a set
of experiments. These application-specific architectures address drawbacks of
currently considered ring architectures such as large execution times or the need
of complex payload re-injections. In order to derive the respective architectures, a
method was proposed which tackles the underlying complexity using SMT-solvers.
Evaluations showed that the proposed method enables designers to efficiently
generate architectures satisfying their needs as well as quality criteria. Moreover, the
evaluations confirmed the advantages of application-specific architectures compared
to the ring architecture.
The obtained application-specific architectures do not yet specify how the
modules and connections are implemented. For example, an architecture does
not consider how the connections are implemented in the form of channels and
bifurcations or how it is ensured that the droplets flow along the desired directions.
However, this constitutes a dimensioning task and, hence, the methods proposed in
Chap. 4 and [49] can be applied. How these methods can be integrated is discussed
in Chap. 9.
After dimensioning the architecture, a microfluidic network results where the
bifurcations realize multiple paths through this network. Along these paths, the
different operations on the droplet are executed and, hence, different experiments
can be triggered. That means, in order to execute an experiment, a payload droplet
has to be routed along the respective path. For this purpose, a droplet sequence
consisting of headers and the payload needs to be injected in a coordinated manner.
Determining such a droplet sequence is described in the next chapter.