The benefits of developing all boards of a system concurrently on a single CAD canvas.
A multi-board system comprises two or more interconnected PCBs in a single enclosure. Typically, the boards will have very different roles. For example, if you consider e-mobility (i.e., the industry trend of switching over to electric drive trains and actuation in the automotive, aerospace and other transportation sectors), many modules are multi-board systems. One board will be a controller. Another will be for switching in and out potentially high current loads.
While they share many common design and manufacturing considerations, the PCBs will warrant special attention when it comes to their specific roles. In this respect, the controller board might be very high-density and feature BGA devices (with hundreds of balls each), flip-chip devices, wire-bonded die and embedded components (i.e., the PCB substrate contains structures with resistive and/or capacitive properties).
The controller board may also feature high-speed digital and possibly even RF signals, and ensuring signal integrity through impedance matching will be of paramount importance. As for the power board, it may need to handle hundreds of amps, so thermal management may be the biggest challenge.
All in one. Modules containing PCBs with dedicated functions have been around for decades. All e-mobility sectors are now looking at how the boards are physically connected. A current trend is to mount them on top of each other, with the controller board placed as if it were a large BGA device on the power board.
The advantages of board-stacking are many. Most benefits derive from no longer needing mechanical connectors and cables. For example, there are cost savings because there are fewer parts; assembly is simplified (also a cost-saver), and reliability is improved through virtue of not having “mechanical” links formed using connectors and cables.
These benefits come at a price, however. The design process is now more complex, and we are, for all intents and purposes, designing two PCBs at the same time – at least the first-time around, after which one or both boards might be used in subsequent projects.
Clearly, the physical connections – i.e., where the module’s bumps connect with solder-paste pads on the main board – must align precisely and be electrically correct, with no signal mismatches between boards.
Designing the boards in a single environment – again, there might be two or more – has considerable advantages. Some ECAD tools support such concurrent design in a single canvas. At the schematic level, connector symbols are used to represent interfaces between boards. These connectors initially exist as a temporary component on each board and build up the connectivity between both.
Typical flow. For layout purposes, start with the controller board. Define its shape and size, and place all standard components. Most would be placed on the top side but, as mentioned, embedding (if only partially) is an option. Some design tools automatically create a temporary component for placement on the underside of the controller board. This temporary component exists initially as connection point crosses, an indication of pad (or ball, if you prefer to think BGA) positions and padstacks.
Regarding placement of each pad, the signal’s “intent” must be a factor. In other words, for a signal that must connect with the base board, routing from the pad of a topside component to the nearest pad on the underside might be practical for the controller board. However, it may not be as convenient for the base board to receive the signal from that physical location.
Next, the base board would be assigned a shape and dimensions. Again, all standard components would be placed. The controller board would be assigned to the temporary component of the base board schematic and subsequently placed at its intended position.
As with the assigning of signals to pads on the controller board, the same considerations apply for the base board. Signal integrity must be maintained through, for example, impedance matching for high-speed signals. Here, the beauty of working in a single canvas is impedance matching can be managed for signals that transition between boards.
Indeed, design rule-checks (DRCs) should flag an infringement, such as failing to maintain a suitable galvanic isolation distance when routing high current signals. Accordingly, if the pad of one board is reassigned, then executing an “update” will keep the two design parts in sync. For example, to reduce crosstalk, it is sometimes necessary to switch the +ve and -ve halves of an LVDS signal over. If the signal transitions boards, the inversion needs to be made on both. Electrical updates are not the only ones that can be easily implemented. Sometimes, it is necessary to move one or more pads relative to the others, and we have this flexibility if the underside of the controller board and its position on the base board exist as temporary components. The traditional BGA grid is a default starting pattern.
Stacked in your favor. The benefits of designing the PCBs for a multi-board module – at the schematic entry and board layout levels – are considerable, and the practice is proving increasingly popular. Whether the top board is regarded as a “component” or as part of the single board (that happens to have different design requirements and therefore rules) is academic, and electrical and physical continuity (pads) is assured throughout.
Also, it does not matter if the board technologies vary considerably. For example, if the base board of a module will be close to a heat source – such as a combustion engine or power inverter – it can be prototyped in FR-4, be functionally verified on the bench, then the design re-spun for ceramic. The controller board would then be a forced-air-cooled component on that board.