Forget Power Device Current Ratings, Calculate Application Losses

System designers are often charged with selecting the most suitable power device from a wide array of products, available with very similar ratings from different manufacturers. While a detailed parameter-by-parameter comparison is technically the best way to make the selection, it is not the most practical approach.

Designers resort to making their first cut based on three or four simple parameters. Among these are package, voltage and current ratings, R_DS(ON), etc. Here, we will take a look at current rating. For this illustration, we will focus on MOSFETs in low- and medium-power packages, but the considerations can be applied to other technologies as well.

RATED CURRENT: SEVERAL VERSIONS OF THE TRUTH

The definition of voltage rating is well accepted; it is measurable with a high degree of consistency under conditions that are not far removed from the real world. It is reasonable to expect that the device can be subjected to its rated voltage continuously without causing failure. The situation is different for current ratings. There is no such thing as a measured value of rated current; it is always arrived at by indirect calculation.

As a result, several versions of rated current can exist. The most common is specified at case temperature, T_C = 25°C, and another at T_C = 100°C. SMT packages are also specified at ambient temperatures, such as 25°C and 70°C. We will focus on T_C-based definitions for leaded packages; the conclusions can be easily extrapolated to other packages.

The usable current of any device is mainly limited by the heat it generates within the die and the maximum permissible junction temperature, T_J(max), for the silicon. Continuous dc current is assumed, so rated current follows:

The inequality has its reasons. With low-voltage MOSFETs heading towards sub-mΩ values of R_DS, at T_C = 25°C the formula can yield hundreds of amperes as rated current in a tiny DFN package. In these cases, the limiting factor is not the silicon, but the mechanical construction of the package — such as the bond wires and pin size.

Even with the equality sign, this may not indicate the device's usable current. Most quick references give the value for T_C = 25°C, which either requires a massive heat sink to maintain, or is theoretically impossible when the ambient is in excess of 25°C. Even if you look at the T_C = 100°C definition, it assumes continuous dc current and zero switching losses, excluding the majority of today's applications. Most companies also set strict derating standards and limit junction temperature to 125°C, even if the manufacturer claims a T_J(max) of 175 °C. Usable current rating is then only 57.7% of the specification.

While ignoring absolute value, many users are tempted to use current rating as a comparative benchmark for different devices. If R_DS(ON) values are close enough, it reduces to a comparison of package thermal resistance. Technically, this is a meaningless exercise, as a number of variations are possible within the definition. Different manufacturers assume different tolerances on the thermal resistance for the same package; the number can vary from 15% to 50%. Some do not even provide a maximum value and use the typical value to boost the current rating.

If that is not enough, many manufacturers use a creative version of steady-state and limit it to 10 s. Power is applied and junction temperature is measured after only 10 s to calculate R_θJC. Not all data sheets are clear about exactly which definition of steady-state is being used. As a result, for the same R_DS(ON) and case temperature, rated current can vary as much as 20% among essentially identical products. Given a specific application and operating current, the assumption that selecting a device touting a higher rated current somehow leads to better performance or reliability is misguided, to say the least.

THERMAL RESISTANCE

It is, of course, possible for devices with the same RDS(ON) and package to have thermal resistances that are actually and significantly different. The reason is that over the years, MOSFET technology has been evolving and the silicon area required to achieve the same RDS(ON) has been shrinking.

As an example, the RDS-Area product, which indicates the area required for achieving a given value of RDS at a specified voltage, has been shrinking by a factor of two every five years or so. For a given package and lead frame, a larger die will invariably result in a lower RθJC and higher current rating. It then follows that if the current rating of a MOSFET is higher than that of identically rated competitors' devices, the reason could very well be an older technology platform that demands a larger die size to realize the same RDS(ON).

Whatever the reason, so long as the thermal resistance for one of them is genuinely lower, can it be used at a higher current? The answer will most likely be in the negative. Remember, the ultimate destination of heat generated in the junction is not the case but the ambient air surrounding the entire system. The amount of acceptable current in the device is not defined by its current rating but thermal management of the system. If the device is mounted on a heat sink, thermal resistance, RθHS, of the case temperature is given by:

For simplicity, assume that RθHS includes a heat-sink-to-case interface and PD is the power dissipation in the device. In other words:

Fig. 1 shows the thermal resistance chain from junction to ambient of the power MOSFET.

Combining Eq. 1 and Eq. 2, we can write the system-limited current rating of the same device as:

POWER DISSIPATION

At one time, power dissipation was assumed to occur only due to conduction. To illustrate the issue further, we assume a typical high-voltage MOSFET in a TO-220 package with an RDS(max) of 1 Ω and RθJC(max) of 1°C/W. The data sheet would show a current rating of 7.07 A dc at a TC = 100°C, resulting in a power dissipation of 50 W in the device. Since industrial products must work at an ambient of 50°C at the least, we also need a heat sink with an RθHS of 1°C/W to maintain the case at 100°C. But a heat sink of this thermal resistance would be quite large.

As an example, consider Aavid Thermalloy's 530002B02500G heat sink with a PCB footprint of 1,000 mm2. We will need 166 mm of this extrusion, making the overall volume impractical for any system designed with this MOSFET. You can choose a more practical piece, like the 513002B00000G heat sink with a 1-in. height, but the RθHS will be 13.4°C/W. Substituting these values in Eq. 3 gives an IDS system rating of less than 2.63 A.

These calculations are, of course, quite elementary and known to every power designer. But the point here is that with the heat sink fixed by the thermal design of the system, the usable current of any 1-Ω MOSFET is limited to 2.63 A no matter what its current rating says. In Eq. 3, the temperature differentials are split as TJ - TC and TC - TA instead of the single term TJ - TA. The reason is that thermal design of the system could also restrict the permissible Tc. Operating just one device at a higher case temperature to utilize its lower RθJC is invariably ruled out when sharing a heat sink or PCB copper as a common coolant.

When all power devices share a common heat sink, no single component will be allowed to dump too much heat and raise TC for all the others. This is even truer in the case of point of load (POL) converters built on PCBs. The PCB copper is shared by rest of the electronics, including sensitive ICs. Maximum PCB temperature is collectively set by the ICs and, in turn, limits the allowed dissipation for power devices. In our example above, if the heat-sink temperature has to be clamped at 100 °C, the device current gets limited to 1.9 A. In Fig. 2 the usable current of small packages is determined by the heat sink they are mounted on.

A common misconception is that for the same operating current, selecting a device with a higher tag somehow builds up a higher degree of derating in the design. In switching applications, the ratio of operating current to the so-called rated current is an illusion and has no relevance to reliability. The key parameters that define robustness are power loss and case temperature, neither of which is predicated by the current rating. Losses are set by measurable device parameters such as the RDS(ON), gate charges, and body diode characteristics. Case temperature is dictated by the rest of the system, independent of current rating. Because of the lower RθJC there may be a marginal reduction in the junction temperature, which has little influence on the overall reliability.

Manufacturers often design a standard silicon die and offer it in different packages. For example, our 1-Ω MOSFET above could be sold in TO-220, TO-220F, and D-Pak, all marketed under the same data sheet. Since typical RθJC for these packages can vary in a 1:4 ratio, we would expect their rated currents to also vary in the 1:2 range. However, the common data sheet may show identical values of rated current for all of them.

Working backward from RθJC, one may even prove that there is no way some of the packages will carry the specified current continuously under any condition without exceeding TJ(max). The data sheet is not misrepresenting the facts; it is just that the term “rated current” has a different connotation. It should neither be equated nor confused with the current-carrying or power-handling capabilities of the device.

IS CURRENT RATING MEANINGFUL?

At this point, the end user might justifiably ask why this number is provided at all. Experienced power designers are known to dismiss the rated current as the single most useless piece of information printed on the datasheet (SOA curves are also strong contenders for that distinction). There are cases where current rating is a useful indicator in device selection. The inequality of RθHS >> RθJC(max) in Eq. 3 is valid for relatively low-power and/or low-current applications. In industrial applications where a module can dissipate hundreds of watts, relatively massive heat sinks are used, for which RθHS and RθJC(max) are quite comparable. The rated system current is responsive as much to RθJC(max) as it is to RθHS. A current rating of 150 A could be a meaningful indicator for an IGBT module intended for a multi-kW inverter system. But when a low-voltage MOSFET in a D-PAK flaunts the same number, designers need to run a reality check.

In many consumer motor-control applications, the switching device is subject to very high currents for short durations. Designers cannot and/or do not implement safe and precise current limits in the circuit, and the device has to absorb all the power losses associated with the spike and survive. The higher current rating, which is synonymous with lower RθJC for a given RDS(ON), implies that temperature rise will be lower under the pulse. TJ(max), of course, could be way above 150 °C in all the cases, but the device with lower RθJC has a better survival rating.

Manufacturers also use rated current as a point of reference for specifying other parameters such as the UIS, forward drop of the body diode, transconductance, and so on. But just because the forward drop of the body diode is measured at a certain current does not mean that the diode, or the MOSFET for that matter, is capable of continuously carrying that current in the real world.

When comparing two power devices the correct question is not which one of them has a better current rating, but which will contribute less heat to the system. The answers to the questions are not related and can be contradictory. In a high-current synchronous buck switching 15 A at 500 kHz, a low-side MOSFET with a 70-A rating could very well be more efficient than an 80-A device and, therefore, the better choice. Unfortunately, it is easy to look up the answer to the wrong question in a short-form catalog but there are no shortcuts to answering the right question. The losses have to be calculated separately for each application using several parameters distributed all over the data sheet. And, rated current most certainly is not one of them.