Design Article
Why consider a power BJT rather than a MOSFET?
Yong “Perry” Li, Ph.D., Director, Power Systems and Applications, iWatt Inc.
10/25/2010 3:34 PM EDT
The power MOSFET has been the major semiconductor switching device in the modern power-electronics and power-management industry, and is widely studied and used by power-system designers. Conversely, the power bipolar junction transistor (BJT) is generally considered an outdated device that has poor switching performance and is unfamiliar to today’s power system engineers.
Recently, however, electronics market leaders in low-power AC/DC off-line power-supply designs have adopted BJT control ICs in applications such as high-volume cellular-phone chargers. This paper explores the questions: Why consider a Power BJT rather than a MOSFET? What are its advantages? Can BJT-based solutions achieve high efficiency?
Low cost is key
The power BJT technology matured in the mid-1960s, while the MOSFET did not become practical until the late 1970s. Fundamentally, BJTs cost less than power MOSFETs because their fabrication involves fewer layers and simpler processes than the power MOSFET, in particular for high voltage (>700V) and low power (below 5 watts) applications.
BJTs normally have less switching di/dt and dv/dt, easing electromagnetic interference (EMI) design with no-Y capacitor, no common-mode choke, and simpler transformer construction for easy, low-cost manufacturing. Also, due to slow di/dt at turn off, some energy in the transformer leakage inductance can be dissipated at the BJT turn-off transitions.
Therefore, without causing voltage stress to exceed the safe operation region, the snubber circuits can possibly be removed in flyback power supply designs of less than 3 W. Removing the snubbers helps reduce component count and bill-of-material (BOM) costs and eliminates power loss associated with the snubber circuits.
For example, a 3-watt, BJT-based, complete universal-AC input power supply design can use only 21 components. Furthermore, very-high-voltage (900 V and above) BJTs are economically available, making BJT-based designs attractive in offline power supplies for the industrial market (white goods, motor control and smart meters), and in regions with widely-varied utility voltages.
Overcoming challenges in EMI design
Switching power converters are tremendous EMI sources. The challenge is to pass EMI regulations as efficiently as possible in terms of size and cost. EMI design necessitates filters, normally in both differential mode (DM) and common mode (CM), at the input or output interface of the system.
CM filters consist of so-called “CM choke” and Y capacitors. CM chokes are bulky and expensive, and therefore unacceptable in cost- and size-sensitive applications such as cell phone chargers. Y capacitors can lead to leakage current from the input lines to the loads at the output sides, so eliminating them makes it easier to pass leakage current safety tests. Furthermore, Y capacitors hinder cell phone communications, as they provide the path for the EMI noises generated in the transformer primary side to flow to the secondary side, where the phone is connected. Consequently, modern cell phone charger manufacturers forbid the use of Y capacitors, greatly challenging EMI design.
To try to meet this requirement, MOSFET-based solutions will often require very complicated transformers, which increase complexity and cost, and still don’t guarantee EMI results. The switching turn-on transitions greatly affect CM noise. Comparing a power MOSFET to a BJT, the MOSFET has much faster switching speed and much shorter turn-on time than the BJT, and therefore generates higher dv/dt and more CM noise. BJT-based solutions directly reduce the sources of EMI noise, such as the switching di/dt and, in particular, the dv/dt which is the major source of CM noise.
Figure 1 compares the source of CM noise between BJT and MOSFET drives, which are the measured CM voltages between the transformer primary and secondary windings in 5W/5V/1A flyback converters. The BJT-based design has about 2.7 peak-to-peak CM voltage, while the MOSFET-based design (with the same 5V/1A rating) has about 3.6V peak-to-peak CM voltage. In this case, the source of CM noise is reduced by about 25% in the BJT-based design.

Figure 1: CM noise source comparison: the measured CM voltage between the transformer primary and second windings.
Digital Control Technology Enables High-Efficiency BJT Design
Admittedly, there are drawbacks to the BJT. It is generally considered to have high switching losses because of its slow switching speed. Also, as a current-driven device, BJT driver design is more complicated.
However, innovative digital-control technology has enabled development of new BJT-control-IC-based solutions with very high efficiency [Reference 1]. There are some key features that previously were not achievable:
The digital IC not only controls the on and off of the BJT, but it also has a built-in, digitally-controlled driver that dynamically adjusts the BJT base-current amplitude in real time based on the load change. The digital real-time control not only works in cycle-by-cycle, it is also even finer in resolution, since it can modify the base-drive current in several steps over the duration of a single switching period. This is a significant advantage that ensures the BJT always operates in optimal switching conditions, with minimum switching and conduction losses.
The digital IC can adjust internal parameters in response to varying line, load and temperature conditions. It allows adaptive multi-mode control schemes that flatten the efficiency curve from full-load to no-load condition by switching between PWM to PFM (pulse frequency modulation) to DPWM (deep PWM) to DPFM (deep PFM) modes, without compromising other system performance, such as audible noises, voltage ripples or regulations. Moreover, placing various sections of the IC in sleep mode greatly reduces standby or no-load power consumption.
The digital IC seamlessly integrates improved quasi-resonant switching schemes that achieve valley-mode turn-on for every switching cycle in any PWM/PFM mode, including the fixed-frequency PWM. This further reduces the switching losses and EMI, without causing modification or compromise to the main control schemes.
Conclusion
For low-power AC/DC power supplies, BJT-based solutions offer low cost and easy EMI design. The cost savings extend to both the power device and the total system BOM due to simple transformer manufacturing and possible elimination of snubber circuits for low-power designs. Moreover, innovative digital control technology has enabled BJT solutions to achieve even greater efficiencies than MOSFET-based solutions in the same power levels. Therefore, new digital-BJT-control-IC-based solutions are viable and preferable in low-power AC/DC applications.
Reference
1. Y. Li and J. Zheng, “A Low-Cost Adaptive Multi-Mode Digital Control Solution Maximizing AC/DC Power Supply Efficiency,” in IEEE 2010 Applied Power Electronics Conference and Exposition (APEC), pp. 349 – 354.
About the author
Dr. Yong “Perry” Li is the director of power systems and applications at Los-Gatos based iWatt Inc. Since joining iWatt in 2007, he has been the chief inventor, system architect, and IC design team lead of several digital AC/DC controllers, including the iW1690, iW1696, iW1697 and iW1698. These digital controllers are designed to meet the demanding energy-efficiency regulations for AC/DC power supplies while reducing system cost, and have been widely adopted by portable, networking, and solid-state lighting electronics device manufacturers. From 2002 to 2007, as a senior design engineer at International Rectifier, Li made substantial contributions to the development of the IRMCF300 series digital control ICs and iMOTION Design Platform.
Dr. Li received his Ph.D. from the Center for Power Electronics Systems (CPES) at Virginia Tech, and his MSEE and BSEE from Tsinghua University, Beijing. He also completed the VLSI Engineering Certificate Program at the University of California Santa Cruz. Li has over 20 technical publications in referred international journals and conferences, and holds 9 U.S. patents, with several still pending.


Les_Slater
10/26/2010 4:53 PM EDT
I like seeing a fresh look at trying to meet the latest in power requirements. Why not use the relatively slow switching characteristics of the BJT to thermally dissipate an appreciable amount of the pesky noise? The most impressive though is the clever controller IC technology to keep efficiencies at competitive, and even superior, levels while at the same time reducing complexity and cost.
Sign in to Reply
http://www.lulu.com/spotlight/poconoarmchairreview
10/26/2010 8:32 PM EDT
Lower EMI in power supplies will be important for communications devices that employ greater frequency range and autonomy in choosing frequencies. I'm referring to cognitive radio solutions, where the device will have to sample multiple frequency bands to see if they are in use. A noisy power supply would make the device think a band was in use when it is not.
Sign in to Reply
Neetin
10/28/2010 4:07 AM EDT
One of the main reasons for BJT to MOSFET shift was the elusive Secondary breakdown at high voltages. Are the new BJTs free from this problem?
Sign in to Reply
Simon7382
11/1/2010 1:38 AM EDT
Secondary break down can be dealt with with properly designed base drive. MOSFETs took over mostly because they were much faster, enabling higher switching frequencies and appropriately smaller DC-DC converters. However, in small off-line DC-DC converters the switching frequency is typically low (around 50-100 kHz) due to the conducted EMI limits, hence the speed advantage of MOSFETs is not an advantage but a disadvantage.
Sign in to Reply
PetJh
7/16/2011 5:57 AM EDT
Comparing just noise as on the picture is .. not correct.
One of the transformers (on the left) is completely compensated - no switching waveform is available.
The picture on the right (even maybe the same transformer) shows completely uncompensated mode, maybe MOSFET capacitance.
It should be stated where on the 50 Hz the noise has been measured as it is completely different during conduction/nonconduction of the bridge.
Both switches must be the same case, pinout etc. as ANY small change in the collector/drain are will make big effect over the CM noise.
Sign in to Reply