Understanding Tree and Crosspoint Matrix Architectures
Matrix implementation typically involves two common methods, often combined in various ways. It’s not always apparent whether a matrix is built using tree multiplexers (MUXs) or a crosspoint configuration, as functional diagrams almost always depict a crosspoint layout to simplify the representation of its functionality. Additionally, the driver provided with the product abstracts the hardware's complexity, making it easier for users to program and interact with the system.
Crosspoint Matrix Switch
In a crosspoint solution, a network of tracks connects the Y axis to the X axis through relays positioned at each intersection. This design is straightforward, easy to implement, and offers significant flexibility in enabling various connections. However, it may not be ideal for high-frequency applications due to the bandwidth limitations it imposes on the matrix.
In the example above, if Y1 is connected to X1 via the relay, the red track becomes a stub—essentially an unterminated transmission line. At low frequencies, this stub introduces additional capacitance, which can limit the maximum frequency that can be used without significant signal loss. Furthermore, the track itself acts as a transmission line. When its length equals a quarter of the signal’s wavelength, the open circuit at the far end is effectively transformed into a short circuit at the relay connection. This results in substantial path loss within the matrix, significantly impacting performance.
The length of the added stubs varies depending on the selected path. For instance, as shown, the path from Y1 to X4 has no excess track on the functional diagram and is therefore unaffected by this issue. However, even this path is not entirely free from defects. The relays at each intersection introduce parasitic capacitance, effectively creating short stubs that lead to signal loss. Additionally, the extent of this loss may depend on what is connected to the other relay terminals.
Some paths may involve multiple stub connections. For instance, the connection from Y2 to X2 clearly shows track stubs on both the Y and X tracks. On larger matrices, these stubs can become significantly longer.
In crosspoint matrices, PCB tracking is typically not designed as transmission lines with specific impedance. Instead, the design focuses on accommodating the current carrying capacity and the voltage rating of the switching system. Due to the dense tracking required in these matrices, the resulting transmission line impedance is generally quite low. Additionally, the high layer count of PCBs in these systems often results in thinner individual layers.
The layout may not always adhere to the straightforward arrangement depicted in the functional diagram. For instance, the path from Y1 to X4 might not represent the longest route in the tracking process.
This type of crosspoint switch is ideal for low-frequency applications where bandwidth is not a critical factor. It offers a highly cost-effective solution with minimal user limitations. Notably, it provides complete flexibility in establishing connections, whether X to X, Y to Y, or X to Y. This makes it particularly suitable for scenarios where an X-to-X connection is used to link instrumentation to the Unit Under Test (UUT).
Tree MUX Matrix Switch (Blocking Matrix)
When bandwidth (BW) is a critical factor, a different construction approach is necessary, often utilizing a Tree MUX architecture. For more details, refer to our guide: Comparing Tree and Conventional MUX Architectures.
In this design, the matrix is created by connecting a bank of Y-input MUXs to a corresponding bank of X-input MUXs. For instance, a 4x4 matrix requires 16 connection points, many of which intersect. As the matrix size increases, the number of required connections grows exponentially.
The Tree MUX is optimized for bandwidth, eliminating open circuit track lengths and minimizing the number of relays connected to each path. However, this design comes with increased costs due to its higher relay count and the complexity of its interconnection system. The point-to-point connections required can be expensive and challenging to implement, often resulting in lower-density designs.
Using a Tree MUX adds not only cost but also introduces a key limitation: the inability to establish X-to-X or Y-to-Y connections. Since the Tree MUX selects between one path or another, it cannot connect its common node to two tracks simultaneously. This restriction prevents the use of X-to-X or Y-to-Y configurations and also prohibits applying two signals to the same connection on the matrix. For this reason, it is sometimes referred to as a "Blocking Matrix."
The matrix is specifically designed for Y-to-X (or X-to-Y) operation, and the driver enforces this by locking out any attempts to create X-to-X or Y-to-Y connections, or to route two signals to the same port.
Non-Blocking Matrix
A variation of the Tree or Blocking matrix incorporates a sophisticated arrangement of power dividers and combiners. This design enables signals from two input ports to be combined into a single output port or allows one input port to connect to multiple output ports simultaneously. While such configurations address specific application needs, they are highly complex and come with significant challenges, such as elevated insertion loss due to the use of power dividers and combiners. To counteract this loss, amplifiers can be introduced, but this alters the system, making the matrices uni-directional.
Sparse Matrices
A sparse matrix typically refers to a matrix with a limited number of allowable paths at any given moment. For instance, consider a 6x6 matrix that, following the Blocking route mentioned earlier, supports six simultaneous paths and requires 12 SP6T switches. In contrast, a sparse matrix may permit only a single path at a time. In such a case, it can be constructed using two 6-to-1 switches with their common connections linked together.
Hybrid Solutions
Matrix designers sometimes use a combination of methods to implement their systems. For instance, they might construct an array of smaller crosspoint switches and route signals to them using a Tree MUX or a standard MUX with sufficient bandwidth. While this approach adds complexity, the driver software is designed to hide these intricacies, ensuring a seamless and user-friendly experience.
One example of a 12x8 matrix is shown below.
The simplified functional diagram on the left illustrates the matrix with isolation switches on both the X and Y axes. On the right, the matrix is depicted as being constructed from four subsidiary matrices, also known as block matrices.
There are numerous alternative approaches to designing a hybrid matrix, each aiming to balance cost and performance to suit specific applications.