Access Official Technical PDF

Share Summary:LinkedInEmail

Automotive power delivery for a Zynq UltraScale+ MPSoC is a system-design challenge, not simply a matter of selecting regulators with adequate output current. The processor system (PS) and programmable logic (PL) have separate power domains, multiple supply rails and defined sequencing requirements. The right architecture depends on the device, application load, functional-safety requirements and design priorities such as PCB area, bill of materials (BOM) cost, efficiency and scalability.

For automotive advanced driver assistance systems (ADAS) and autonomous-driving applications, these decisions also need to fit within a safety-engineering framework. The XA Zynq UltraScale+ family described in the application note is AEC-Q100 qualified and certified to ISO 26262 ASIL C; system-level safety measures still depend on the particular design and its requirements.

Start with the MPSoC power domains and sequence

The Zynq UltraScale+ MPSoC has two principal blocks: the PS and the PL. The PS can boot and operate without powering the PL. Within the PS, the low-power domain (LPD) must be operational before the full-power domain (FPD) can function, although those domains may be powered up simultaneously. The PL has its own power domain, known as PLPD.

Depending on device complexity and application, the MPSoC may need ten or more independent rails. Their required voltages and currents vary, as do the constraints on the order in which rails turn on and off. Following the device’s recommended sequencing is important for proper power-up and minimizing current draw. The power-off sequence should reverse the power-on order.

Designers should use the applicable Zynq UltraScale+ DC and AC switching-characteristics documentation to establish rail requirements and timing. Where sequencing is controlled externally, the control circuitry or microcontroller can generate regulator enables and monitor rails. Individual enable control for each rail supports more precise sequencing; a PMIC with dedicated enables for its regulators is one way to implement this approach.

Match pre-regulation to the application

The example architectures assume a 12 V automotive battery input. A pre-regulation stage first creates an intermediate supply, such as 5 V or 3.3 V. Downstream converters, controllers and PMICs then generate the MPSoC’s individual rails. The choice of intermediate voltage and pre-regulator depends in part on the minimum operating battery voltage and the total power required.

It is useful to plan PS and PL rails separately. The PS requirements in the two example application classes are relatively similar, while PL rail count and current demand can differ substantially. This makes a common PS solution practical in some designs, but often calls for a distinct, more scalable PL power architecture.

Lower-power designs

For the lower-power example, representative of ZU2EG and ZU3EG devices, total power is usually below 15 W. One or two buck regulators may be enough for pre-regulation. A two-output automotive PMU can provide a 3.3 V output for PS requirements and includes features such as output-voltage monitoring and a window watchdog. Its primary switcher can deliver up to 2.5 A at 3.3 V, which may suit PS rails but is not sufficient for many complete Zynq UltraScale+ applications.

Where the PMU’s additional functions are not needed, a higher-current buck regulator can provide an alternative. The application note describes a dual-mode regulator capable of 3.0 A at 5 V, potentially enough to supply both PS and PL rails in simpler designs. A single pre-regulator can reduce PCB area and BOM cost compared with separate pre-regulation for the domains. If the PL load requires more power, splitting PS and PL across separate intermediate supplies is an option.

Higher-power designs

The second example covers ZU5EG, ZU6EG and ZU7EG devices, intended for more complex workloads such as sensor fusion. In this case, the greater total power and wider variation in PL requirements favor separate intermediate supplies for PS and PL. The PS can use a PMU or a higher-output buck regulator, depending on system needs. PL rails can be generated with buck controllers selected for the required power and switching frequency.

For the controller choices discussed in the application note, the NCV891930 operates at 2 MHz and is presented for total power up to 30 W; the NCV881930 operates at a nominal 410 kHz and is presented for higher total power. These are architecture examples, not substitutes for validating regulator, switch and thermal choices against the actual rail loads and operating conditions.

Scale current while managing efficiency

Some PL rails can have especially high or variable current demand. For a rail such as VCCINT, a multiphase buck controller can expand current capability by adding phases as demand increases. Dynamic phase management can adjust active phases as load current changes, supporting efficiency across a range of operating conditions.

Power MOSFET selection also affects the design’s cost and efficiency. The switches driven by pre-regulation controllers, as well as devices used in downstream controllers, should be selected and dimensioned for the application’s actual current requirements. A schematic-level reference architecture is therefore a starting point; component choices need to follow the load and system constraints.

Make functional safety part of the power architecture

In a safety-critical system, power faults that could affect MPSoC or peripheral operation must be addressed through defined safety requirements. These should derive from system-level safety goals and specify, for example, acceptable rail voltages, correct start-up sequencing and required shutdown behavior. The application note gives example voltage tolerances, but the applicable limits must be established and validated for each design.

For an ASIL C system, continuous monitoring may be needed across most or all power rails. The example safety approach uses a monitor device to supervise voltages and perform sequential power-up, with monitoring and regulation kept sufficiently independent to reduce common failure modes. This can involve separate devices or a single device with independent functional blocks. The monitor’s response time must also allow the system to reach its safe state within the applicable Fault Tolerant Time Interval (FTTI).

Automotive-qualified components without an ASIL rating may be used in an ASIL decomposition approach when the safety requirements are allocated appropriately and independent monitoring is provided. The safety case must be developed for the complete system; component qualification alone does not establish system-level compliance.

Conclusion

A robust Zynq UltraScale+ power design begins with the device’s rail and sequencing requirements, then scales pre-regulation and downstream conversion to the expected PS and PL loads. Lower-power applications may benefit from a compact shared supply, while higher-power designs often call for separate domains and scalable controllers. In safety-critical systems, voltage monitoring, sequencing control and independence of safety mechanisms must be planned alongside power conversion. Validate the selected architecture against the application’s electrical and system-level safety requirements before implementation.

Access Official Technical PDF

About The Author