Power Delivery Architecture
Different voltage is required for different components on a PCB. Different voltage rails are required to supply those components. So, proper power delivery architecture is required. Power delivery architecture can be adopted depending upon three different criteria: Efficiency, power density and cost. Different architectures have been discussed here.
- A typical power delivery architecture consists of a single front end regulator supplying all the different voltage rails (i.e. multichannel buck regulator).
Instead of using a single multichannel buck regulator, multiple ICs can also be used. But if too many voltage rails are required, using separate IC for each voltage rail increases the BOM cost and reduces power density. The main disadvantage with this architecture is that if the distance between the component supply node is far from the regulated voltage node, there will be IR drop when the component draws current. Now, i.e. if the component requires 0.9 V input and due to IR drop the voltage drop of 0.3 V happens, then the component might not be functional. Over which there is always chance of ground bounce and cross coupling.
- Nowadays people are using point of load distributed bus architecture.
This architecture uses two step conversion process. First the input voltage is converted to a known loosely regulated voltage (i.e. 12 V for 24 V). This voltage reaches near to the components through distributed bus. There near the component, second step of conversion happens (i.e. 12 V to 3.3 V). The output voltage of the second stage is tightly regulated. This architecture will increase the BOM cost. This architecture suffers very less from ground bounce and cross coupling problems.
Grounding and Layout guidelines
PCB for embedded application contains analog, digital and RF components. So, grounding should be proper. Otherwise there can be interference between them which may leads to system response far from intended. Components need to be classified according to their functionality and same type of components should be connected to the same ground. Ferrite bead needs to be placed in between different grounds (i.e. analog and digital ground). Ferrite beads need to be designed such that it can provide very high attenuation for the interfering signals. So, the current will flow through the intended path. It reduces the loop area for go and return of any signal and simultaneously reduces the radiated EMI.
There are two different philosophies regarding signal grounding schemes
- Single point grounding and
- Multipoint grounding
- A single point ground system is one in which subsystem ground returns are tied to a single point within the subsystem.
There are two possible way to do that. The first one is daisy chain approach. This technique has a problem of permitting common impedance coupling between two subsystems. The second approach is the star connection approach. But in this case also,
individual ground conductors will have some impedance depend on the length of the connections. In a distributed system, these connection wires may need to be long so that it can be connected to a single point. This increases the ground impedance as well as the loop area. So, it will produce coupling between subsystems.
- In multipoint ground system, a large conducting plate serves as the return of ground current.
Individual grounds of different subsystems are connected at different points to this plate. The main assumption for multipoint ground system is that impedance between points of ground return plate to which individual subsystem grounds are connected has very low impedance at the frequency of interest.
When two subsystems are connected at ground points which are far away, due to high frequency current flowing through the ground, there will be voltage drop across two grounds. So, it is required to put high frequency decoupling capacitor near to every subsystem such that high frequency is supplied by these decoupling capacitor locally.