Raging trench warfare

Increasing cell densities, reducing RDS(on) and reducing die size continue to be the main challenges facing the designers of trench MOSFETs. But many fear that the limits on what can be achieved with existing structures may soon be coming to an end and the question will soon have to be faced that, given the expense of investing in designing new structures, will it be worth it for what could be very small benefits in performance.

There have already been some dramatic improvements over the years. Take cell density, for example, where for devices of 30V and below a density of a billion cells per square inch is not unknown. Compare that with 15 years ago when a density of eight million cells per square inch was leading edge. Even with higher voltage parts in the 200-250V range, densities of 50-100 million cells per square inch are not uncommon.

True, at the higher voltages, increasing cell density is not as important because other factors are more prevalent, such as the epitaxial layer structure and the substrate.

“As you increase the voltage, the cell density becomes less important,” said David Gray, head of the power MOSFET group at Vishay Siliconix. “The epitaxial influence becomes more dominant as you increase the voltage. So, if you increase the density at higher voltages, there is very little effect on the on-resistance.” Also, increasing the cell density still does increase the capacitance, which can have an overall detrimental effect on the device’s performance.

“At each voltage, there is an optimum cell density,” said Gray. “At lower voltages, though, you are still getting benefits in overall performance as you increase the density. It is certainly possible to double the current density to two billion per square inch, but you have to look at the cost versus performance benefits.”

The other aspect is switching frequency, and Gray believes there is more work in R&D departments in trying to improve switching frequency than there is trying to reduce the on-resistance.

“This is where there is a lot more IP on the design,” he said. “There are more elaborate structures now. Not many of the devices look like a traditional trench FET; they are more elaborately designed.”

The main contributor to switching losses is parasitic capacitance, usually between the gate and the source: “All the new developments are in trench MOSFETs because the idea is to have better performance in terms of parasitic losses,” said Marcello Sgroi, Product Marketing Manager at STMicroelectronics. He said that for 30V products, all the main players were working towards an on-resistance of around 1mΩ; the range is just over that to about 3.4mΩ.

“We are working at getting below 1mΩ,” he said. “With silicon it might not be possible. We may have to change the structure. There are physical limitations.” This may, he said, lead to a complete change of technology, which was not really needed in the market today.

Chris Bull, Marketing Manager at Texas Instruments, added: “We are not a million miles from the limit you can go to in reducing RDS(on). But how much investment do you need to get to the next big jump and is it really worth it? Will it be too expensive? There is room to improve, but we are getting to the end of the learning curve. We are still improving but the pace will slow.” He said that there would now be more emphasis on improving the switching performance. “RDS(on) is always the first thing the customer asks,” he said. “The next is cost. And then, depending on the customer, the switching performance. Just choosing the lowest resistance is not always the best choice for the application.”

At high current, conduction losses due to the on-resistance are much more important than switching losses, and so switching losses take a low priority. But at low current, the switching losses become more important. And the higher the switching frequency, the greater the loss because it occurs on each cycle. This is why this type of MOSFET is preferred in applications that need a lower frequency.

“Look at a classic 30V trench FET for a synchronous buck converter,” said Bull. “The high-side and low-side switches need very different specifications. The low side needs low resistance. For the high-side FET, you have a die that is smaller with a two to three times higher resistance. But here you have to worry about switching losses. The planar FETs we have are probably the best in the world for the high side but not for the low side.”

At lower power, lateral transistors tend to work better but at higher power trench transistors are the better option. However, there is a grey area in the middle where both can work depending on the application.

“Practical designs balance switching loss and conduction loss and may trade off switching frequency to optimise the balance point,” said Andrew Smith, Senior Product Marketing Manager at Power Integrations. “When you look at the power loss of the transistor, you have the trench transistor with low conduction loss and the lateral transistor with low switching loss. It depends on the application where which will be most beneficial.”

However, at say 25-65W applications, both conduction and switching losses are important, so there is no clear benefit for either technology. The switching frequency, though, does play a role in this region. For the same power loss, a lateral transistor can switch at higher frequencies because the switching loss per cycle is lower. Higher frequencies also allow smaller transistors, so the cost is lower. With trench MOSFETs you need a larger transformer because they switch at lower frequencies.

However, using techniques such as resonant switching reduces losses, increasing the applications for trench MOSFETs: “If you use these kinds of techniques,” said Smith, “you can move the trench MOSFET up to the higher frequencies.”

Techniques such as charge balancing are used in most modern MOSFETs to reduce the contribution of the epitaxial region. Above 400V, the super-junction technology pioneered by Infineon has produced a large reduction in on-resistance by using charge balancing. This can actually be applied at all voltages, but the effect is less when the voltage is lower.

MOSFETs also have an intrinsic diode, called the body diode, between the drain and source. In a lot of DC/DC converters, that body diode conducts on every cycle and that causes losses. That is why many now put a Schottky diode in there but there is still a lot of work going on to improve the performance of that diode. Basically, they are trying to reduce the forward voltage and the charge when conducting.

“If they can improve both of these, they can improve the performance of the MOSFET,” said Gray. “Every time we release a new family, we spend a lot of time trying to optimise the performance of the body diode.”

Another factor in the design of these products is the packaging, and more particularly how that is linked to the silicon. According to Sgroi, just switching from wires to ribbons in some types of packaging can improve the on-resistance by as much as 22%. This can be reduced even further by using clips.

On trying to drive down cost, some manufacturers are trying to look at reducing the number of mesh levels: “Say you have ten mesh levels on device,” said Bull, “if you can reduce that to seven or eight, that takes away cost.”

Designers of trench MOSFETs are realising that their quest to bring down the on-resistance could soon come to an end without large investments in deigning new structures. However, there are other techniques that are being explored to improve the performance of these devices and it is into these that most current research is concentrating. Without a pressing market design, this situation is likely to continue.