The integration model falls short for gate drivers used in switching circuits for power converters. Discrete semiconductors keep noise in check and place the driver close to the switch.
BY RUDYE MCGLOTHLIN
Director of Marketing, Power Products
Component integration has been the driving force of the semiconductor industry for more than 60 years. It’s right there in the industry term integrated circuit, or IC. And year after year, diligent circuit designers, engineers, and product marketers look for opportunities to take chips to the next level of integration to reduce cost, shrink device and board size, and minimize bill-of-materials (BOM).
Why not? There are many good reasons and advantages for system designers to integrate more functionality into an IC. First is convenience. Soldering down one chip is always better than having to solder down two. Next is interoperability. Integrated components are, of course, designed to work together. There is no need to worry about matching digital interfaces, impedances or messy glue logic. Finally, the cost is a big incentive for component integration. Cost reduction has been the promise of integration realized now in economical computing systems and low-cost microcontrollers (MCUs) with an ever-increasing slew of functions.
When functions are complementary in achieving a system goal, then integration makes a lot of sense. The integration of high-performance op amps with analog-to-digital converters (ADCs) is a good example. The next step is integrating these analog components with an MCU. Together, they accomplish a system requirement with all the advantages of integration.
However, not all integration incurs advantages without significant disadvantages or tradeoffs. In some cases, the better choice for a system design may be to continue with discrete components. Often, the deciding factor in whether to integrate or not is the effect of noise on the various components. Sensitive analog measurement integrated with noisy switching components rarely results in an improved system.
Another instance when integration comes into question is when there are parts of the system that are space-critical. It’s generally related to the parasitic capacitors, loops, and inductors in the system. When one parameter must be minimized, it often takes precedence over any advantages that may be gained by integration.
Finally, the cost benefit of integration can sometimes reverse. This situation is seen with power MOSFETs in which discrete components end up being cheaper than equivalent integrated devices because of the specialized fab process and packaging associated with them.
Case study: isolated gate driver
A common component that exemplifies the advantages of discrete over integrated components is the isolated gate driver. Isolated gate drivers are used when switching high-voltage rails in power conversion systems.
Besides the requirements associated with effective driving of switch gates — fast current sourcing, low propagation delay, and high transient immunity — there are also distinct requirements associated with the isolation, such as package spacing.
There are clear reasons why an isolated gate driver is not a good candidate for integration into its paired system controller. For example, the fast, high-voltage switching of a field-effect transistor (FET) gate is inherently noisy. The gate voltage on the high-side switch travels through the entire range between the lower rail and the upper rail during the typical switching cycle. In some areas of the switching cycle, it can change by hundreds of volts or more in tens of nanoseconds or less.
This fluctuation produces huge transients on the gate driver output. Dedicated gate drivers are designed to reject these transients, but introducing this noise into the package can affect all circuits present on the die. If those circuits were sensitive analog circuits or time-critical digital circuits, they would be overwhelmed and their functions would be fruitless.
Fig. 1: A discrete isolated gate driver optimizes the system in several ways: A) Controller package can be reduced and no space along the package is wasted; B) gate drivers are optimized to reject transient noise and prevent the controller from experiencing it; and C) parasitic induction and capacitance is minimized by placing the driver close to the FET.
Another reason that integration is not an option for these components is that the gate driver needs to be close to the switch that it is responsible for (Fig. 1). The switch used and its associated requirements for heatsink mass and airflow often set the size for the switching subsystem. For switching half bridges, and especially for full bridges, integrated components make it impossible to locate the gate driver close to all of the FETs being used — at least two, but often four or more, devices.
When designing a half-bridge or full-bridge circuit, component placement and PCB layout are critical to performance. To get the best performance, current return paths and the effects of parasitic elements — stray capacitance and inductance — must be minimized. Parasitic capacitance and inductance are unavoidable, but keeping the driver close to the FET minimizes adverse effects.
Finally, the unique creepage requirements associated with the galvanic isolation deter integration of this component (Fig. 2). Creepage is defined as the spacing along the package between exposed metal on the outside of the IC. Generally, as the bus voltage increases, the creepage gets larger. Typical creepage for isolated gate drivers runs from about 4 to 8 mm and even larger.
Fig. 2: Problems with integrated isolated drivers include A) wasted package space due to creepage requirements; B) high-voltage transient noise coupling in package to sensitive digital signals; and C) layout requirements increasing distance to FET and worsening parasitic effects.
In the theoretical case of integrating an isolated gate driver, this creepage requirement places a large burden on the rest of the components. Integration with a system controller would require the package to grow in size, and a large area left free of pins or exposed metal might reduce creepage.
That might significantly reduce the peripherals available to controllers that usually have pins around four sides of a device with functions assigned to each. Increasing the package size and accommodating the requirements of the isolation barrier will surely increase the system cost.
Silicon Labs offers several families of high-performance discrete isolated gate drivers. Some include options for single gate drivers that can be placed very close to the power switch. Other families have high-side/low-side pairs, which provide the same benefits of a discrete driver in noise immunity and cost optimization. However, care must be taken in layout of these devices to maintain symmetric parasitic environments.
The Si827x driver family from Silicon Labs, for example, provides a very high level of transient noise immunity. The device operates as expected even in the presence of 200-kV/μs common-mode transients. Other gate driver families, such as the Si8239x, offer up to 5-kV isolation ratings in packages with 8-mm creepage.
Integrated vs. discrete
Integration of components into a more capable single device makes sense in many cases. Integration of analog and mixed-signal functions, memory, and high-performance digital logic has been a boon for the semiconductor industry for decades.
Fig. 3: Discrete isolated gate drivers provide basic functionality for efficient switching, including under-voltage lock out (UVLO) and enable pin (EN) for asynchronous control. Dual isolated gate drivers add more functions, such as overlap protection, but need more care in layout.
The integration model falls short in some application cases, though. Gate drivers used in switching circuits for power converters must remain discrete components to keep noise from interfering with system controller functioning and to allow drivers to be placed close to switches to reduce parasitic effects.
Using isolated gate drivers as discrete components in a system design can reduce overall system costs due to the unique package size requirements. Attempting to integrate these components creates a distinct burden that can only be addressed with expensive, non-standard packaging.