84
7 Designing Application-Specific Architectures
7.1 Overview of Existing Architectures
In order to realize an experiment, the payload droplet (i.e., the droplet containing
the biological sample) has to traverse a sequence of modules [23], which realize
elementary operations such as mixing m (mixes the biological sample within
a droplet), splitting s (breakup of droplets), fusion f (the fusion of two or
multiple droplets), detecting d, or heating h. For implementing the experiments,
the following architectures have been considered thus far:
• In [23], first theoretical architectures are proposed, which are inspired by computer networks. More precisely, this work proposes a ring, star, and grid architecture but no detailed design guidelines. Moreover, this work proposes information
encodings using droplets, i.e. encodings based on the presence/absence of
droplets, the distance between droplets, the size of the droplets, or the substance
composing the droplets.
• In [6] and [28], microfluidic bus architectures are proposed. Bus architectures
cascade bifurcations and, by this, allow a payload to be routed to a single
module executing an operation. More precisely, in [6] a bus architecture is
proposed, which uses a droplet by size addressing scheme. Furthermore, this
work provides dedicated design guidelines and a performance evaluation. In [28],
a bus architecture is proposed, where a droplet by distance addressing scheme is
used to address different modules. In order to avoid the coalescences of droplets
in the bus network, additionally, a medium access control is proposed in [28].
Finally, this work validates the resulting network and addressing schemes by
CFD simulations. A comprehensive review of microfluidic bus networks is given
in [58].
• In [60], a ring network is proposed which allows to execute multiple modules on
the droplet by a single traversal of the ring. More precisely, the ring architecture
connects the modules in series and the passive droplet routing mechanism allows
to decide whether a module is executed or skipped. Furthermore, a droplet reinjection mechanism closes the ring, which allows payloads to traverse the ring
multiple times.
Overall, the ring architecture is the only architecture which allows to route a payload through multiple modules with one injection of the payload. Therefore, the ring
architecture represents thus far the best choice to implement experiments consisting
of multiple operations. In the following, the ring architecture is considered in more
detail and, eventually, is taken as reference for application-specific architectures.
Consider the ring architecture as sketched in Fig. 7.1a. Here, the microfluidic
processing unit (MPU [23]) abstracts the pump producing the continuous flow,
the droplet-on-demand components, and a mechanism for re-injecting droplets.
Then, a droplet cyclically traverses the ring in order to execute operations realized
by modules. The network node presented in [60] allows to either execute or
skip a module and, by this, addresses a major drawback of before proposed ring
architectures, which limited a payload droplet to be sent to a single module only.
7 Designing Application-Specific Architectures
7.1 Overview of Existing Architectures
In order to realize an experiment, the payload droplet (i.e., the droplet containing
the biological sample) has to traverse a sequence of modules [23], which realize
elementary operations such as mixing m (mixes the biological sample within
a droplet), splitting s (breakup of droplets), fusion f (the fusion of two or
multiple droplets), detecting d, or heating h. For implementing the experiments,
the following architectures have been considered thus far:
• In [23], first theoretical architectures are proposed, which are inspired by computer networks. More precisely, this work proposes a ring, star, and grid architecture but no detailed design guidelines. Moreover, this work proposes information
encodings using droplets, i.e. encodings based on the presence/absence of
droplets, the distance between droplets, the size of the droplets, or the substance
composing the droplets.
• In [6] and [28], microfluidic bus architectures are proposed. Bus architectures
cascade bifurcations and, by this, allow a payload to be routed to a single
module executing an operation. More precisely, in [6] a bus architecture is
proposed, which uses a droplet by size addressing scheme. Furthermore, this
work provides dedicated design guidelines and a performance evaluation. In [28],
a bus architecture is proposed, where a droplet by distance addressing scheme is
used to address different modules. In order to avoid the coalescences of droplets
in the bus network, additionally, a medium access control is proposed in [28].
Finally, this work validates the resulting network and addressing schemes by
CFD simulations. A comprehensive review of microfluidic bus networks is given
in [58].
• In [60], a ring network is proposed which allows to execute multiple modules on
the droplet by a single traversal of the ring. More precisely, the ring architecture
connects the modules in series and the passive droplet routing mechanism allows
to decide whether a module is executed or skipped. Furthermore, a droplet reinjection mechanism closes the ring, which allows payloads to traverse the ring
multiple times.
Overall, the ring architecture is the only architecture which allows to route a payload through multiple modules with one injection of the payload. Therefore, the ring
architecture represents thus far the best choice to implement experiments consisting
of multiple operations. In the following, the ring architecture is considered in more
detail and, eventually, is taken as reference for application-specific architectures.
Consider the ring architecture as sketched in Fig. 7.1a. Here, the microfluidic
processing unit (MPU [23]) abstracts the pump producing the continuous flow,
the droplet-on-demand components, and a mechanism for re-injecting droplets.
Then, a droplet cyclically traverses the ring in order to execute operations realized
by modules. The network node presented in [60] allows to either execute or
skip a module and, by this, addresses a major drawback of before proposed ring
architectures, which limited a payload droplet to be sent to a single module only.