November 30, 2018
This blog post was first published by United Silicon Carbide (UnitedSiC) which joined the Qorvo family in November 2021. UnitedSiC is a leading manufacturer of silicon carbide (SiC) power semiconductors and expands Qorvo's reach into the fast-growing markets for electric vehicles (EVs), industrial power, circuit protection, renewables and data center power.
By Dr. Anup Bhalla, VP Engineering at UnitedSiC
Physics both gives and takes away. Devices built with materials that exhibit wide band-gaps, such as silicon carbide (SiC), offer designers the gift of transistors that can sustain high power densities thanks to their combination of low conduction and switching losses, high operating junction temperatures and fast switching speeds. These high power densities are attractive for enabling smaller power control and conversion circuits, but can't be achieved by choice of semiconducting material alone. Such devices need to be offered in packages with low thermal resistances, such as the TO-247, so that the heat they dissipate can be carried away easily. Unfortunately, and this is the bit where physics takes away again, connections to the TO-247 package often have high inductances, which can limit switching speeds.
Physics gives and physics takes away, and in this case, offers a solution to the inductance issue using a technique called a Kelvin connection. However, such connections must be implemented carefully if physics isn't going to take away their advantages once more.
The Kelvin connection
The Scots/Irish experimentalist Lord Kelvin cared about the precision with which physical phenomena, such as electrical currents, could be measured. He understood that to measure a low resistance using Ohm's Law, by checking the voltage drop it caused given a defined current, he had to take the voltage measurement exactly at the resistor with connections separate from those carrying the current. This method became known as making a Kelvin connection.
The Kelvin connection was originally intended to measure a static voltage at the right point in a circuit, but it can also be used to inject a voltage at the right point. For example, when driving the gate of a MOSFET switching at high frequency, the source connection to the device is a common point for the gate-drive voltage and drain-source current. If there is a common source inductance L (as in Figure 1), then changes in the current will affect the gate voltage in a way that is proportional to the inductance L and the rate of change of the current. When the gate is driven off, the voltage developed across the inductance L acts to hold the gate on longer, slowing the current fall. Conversely, during turn-on, the voltage across the inductor L acts to slow down the rate of rise of current.
Managing the impact of lead inductance
The inductance L can be attributed to the internal bond wires of the MOSFETs and is usually around 1nH per mm. If the device has leads, as with the TO-247 package, these external connections also add to L.
Figure 1: Common-source inductance can cause gate-voltage transients that slow down switching currents
When switching times were measured in microseconds, amps of switched current only produced transients of millivolts, which left the gate-drive voltage almost unchanged. However, wide band-gap (WBG) devices can switch tens of amps in a few nanoseconds, producing transients of around 2-5V per nH of connection inductance. If this transient adds to the gate drive, it stops the MOSFET from being switched off, risking ringing and even device failure.
The effect can be reduced with Si MOSFETs if the gate is driven to a negative voltage when off, perhaps as much as –10 V, to defeat the bias reduction caused by the voltage spike. This causes greater gate-drive power dissipation, which also scales with the total gate-drive voltage swing. It's more of a problem with WBG devices made using SiC or gallium nitride, which can only support negative drive voltages of about –3 V. The solution is to make a Kelvin connection to ensure that the gate-drive return is as close as possible to the source connection of the MOSFET die. Although this is easy with chip-scale packaging, if manufacturers want to use the TO-247 package for its superior heat-dissipation characteristics, they must add a fourth lead to make the Kelvin connection (Figure 2).
Figure 2: The fourth lead of this TO-247 package offers a Kelvin connection to the source
Faster switching enables greater efficiency
Controlling lead inductances, and their potential impact on gate biases, by using Kelvin connections means wide bandgap devices can be run at their true switching speeds without having to implement negative gate voltages. This simplifies drive circuits. The effect is dramatic: when the SiC JFET cascodes from UnitedSiC are put in three-terminal packages, the device has to be slowed down to maintain its reliability. Implemented in a four-terminal package with Kelvin connections, the current slew rates can exceed 5000 A/µs, enabling greater efficiencies without affecting the gate-drive signals.
As physics both gives and takes away, even in the TO-247 package, there is still the device lead inductance to contend with, usually with a small snubber across the drain-source to stop voltage overshoots in the power path. The gate-drive loop must also be carefully laid out to minimize its inductance and prevent pick-up from external magnetic fields caused by the main commutation loop.
Which Kelvin connection?
There are other practical issues with implementing Kelvin connections. If the gate-drive return is the main system 0 V, tied to the power ground, it will probably be inconvenient to also make this common point the Kelvin connection to the switch. If the power circuit is a full-bridge, there will be at least two low-side devices, each with a Kelvin connection, so which one should be connected to system 0 V? This issue gets more complex if resistive current sensing is used in the device source lead: if the Kelvin connection is the system 0 V, the voltage generated across the resistor is negative.
One way to fix this is to isolate the gate drive through an optocoupler or transformer, as is necessary for any high-side drives anyway. If such isolation is used for the low side, the Kelvin connection can float, isolated from system 0 V (Figure 3). Using a transformer also means designers can generate a negative off-state gate drive if desired and a positive drive scaled to its optimum value by adjusting the turns ratio.
Figure 3: Isolated transformer gate drive
More give than take
Making Kelvin connections to leaded wide band-gap devices means they can be delivered in TO-247 packages, which support high power dissipation. This brings us closer to a switch that is electrically ideal and yet can be practically used at high power levels.
Have another topic that you would like Qorvo experts to cover? Email your suggestions to the Qorvo Blog team and it could be featured in an upcoming post. Please include your contact information in the body of the email.