Efficiency is the name of the game in the design of power supplies. This is a major reason why switch-mode power supplies (SMPSs) are chosen for power-supply designs. However, even SMPSs have their own losses, which are associated with their switching transistors, be they MOSFETs, bipolar transistors or insulated-gate bipolar transistors (IGBTs). Understanding these losses via accurate measurements can minimize the losses and increase efficiencies even further.
This is the first part of a two-part article series on understanding saturation losses in SMPSs. Part one discusses the contribution of saturation losses to power-supply inefficiency, how those losses are a function of a transistor's saturation voltage, and techniques for measuring saturation voltage. The second part, to appear in the next issue, describes in detail how to build a novel low-cost tester for accurately measuring saturation-voltage losses in the presence of high switching voltages or noise.
Forms of Inefficiencies
These inefficiencies generally take three forms: switching losses, saturation losses and the power required to drive the transistors. Let's examine each in more detail.
Switching losses result from the time it takes the transistor to switch from a high to a low level and back again, and are affected by the parasitic capacitances of the circuit and transistor. When switching, the transistor is in a linear mode and hence highly inefficient. These inefficiencies are typically controlled by making sure the switching device transitions quickly and cleanly from one level to the other, and by keeping the duty cycle of the transitions low versus the switching period. Parasitic capacitances are those present in the circuit, but do not contribute to the power output from the power supply. These are controlled by careful layout and selection of the switching transistor and power-supply topology.
Saturation losses are the main sources of inefficiency of the three types of losses mentioned. The saturation voltage is one of the key items that determine the efficiency of SMPSs. Saturation losses are a function of the voltage dropped across the transistor due to the on-resistance of the switching transistor when conducting and the current flowing through the switching transistor. Saturation losses are controlled by selecting a switching transistor with the lowest saturation or on-resistance possible given other constraints such as cost, and by properly driving the switching device.
The power required to drive the switching transistor does not contribute to the overall power output from the power supply and thus lowers the overall efficiency of the power supply. These drive losses vary depending on the specific transistor used, the operating frequency of the supply and the type of transistor. Driving losses associated with MOSFETs and IGBTs are largely due to the input and Miller capacitances of these transistors. Thus, these losses are a function of the required switching and drive voltage, and power-supply switching frequency. There is no constant bias required to keep MOSFETs or IGBTs conducting.
Bipolar junction transistors (BJTs) are different. They require a constant bias and specialized drive waveforms to turn them on and off efficiently. This is the reason MOSFETs and IGBTs have largely displaced BJTs in recent years. This trend has slowed recently and, in some cases, reversed itself as competition has forced many manufacturers to reduce the cost of their power supplies.
BJTs are the lowest-cost switching transistors on the market. They also offer better saturation performance than MOSFETs, especially at high voltages, assuming that comparable transistors are evaluated. BJTs thus offer good value but with increased drive losses and complexity.
Interestingly, drive losses often can be fully offset by the power savings realized from the improvement in saturation performance and by using drive circuits that maximize drive efficiency (e.g., proportional drive). Note that the increased complexity is generally due to the lack of power-supply chips specifically designed to drive BJTs. Some newer power-supply chips offer high drive efficiency with minimum complexity.
An SMPS's switch can have a high voltage across it in comparison to the saturation voltage, which makes it difficult to get an accurate saturation-voltage measurement. This is especially true in off-line supplies. If you use an oscilloscope and set it to a range that does not overload the oscilloscope's input, then the 0.5-V to 3-V saturation voltage is lost in comparison to the 200-V to 500-V (or more) signal across the switching transistors in off-line supplies. If the oscilloscope is set to a low range to get the needed resolution, then the input is overloaded and may be damaged, or will simply produce an inaccurate answer (Fig. 1).
Fig. 1 shows a typical input circuit for an oscilloscope. If a large voltage is applied, causing the clamp diodes to conduct, the input attenuator (all-pass network) will take a charge and produce an input offset voltage on the waveform. Of course, this assumes that the input voltage is not so large that it damages the input attenuator and that the clamp adequately protects against continuous overload. A saturation probe provides a way to ignore the high-voltage switching waveform, while producing an accurate look at the saturation voltage (Fig. 2).
As shown in Fig. 2, the voltage across the collector to emitter, or drain to source (MOSFET) of the switching transistor, is sampled by diode D1, a UF4007, which has a 75-ns maximum switching time and will withstand 1 kV. It also has relatively low capacitance (17 pF at 4V). D1 is forward biased by a 2-mA current source when the input voltage is below 5.2 V. The diode disconnects the circuit from the switching transistor when the voltage across the switching transistor is greater than 5.2 V, protecting the probe circuit.
Diodes D2 and D3 clamp any leakage currents or transients. Note that the 5.2-V input limit was chosen to be high enough to capture most reasonable saturation voltages, while low enough to limit the maximum output swing to allow 0.5-V/div input scaling on the oscilloscope without overloading the oscilloscope's input.
When D1 conducts, the voltage at the base of Q1 is approximately one diode junction above the actual sampled voltage. Output amplifier transistor Q1 (emitter-follower) drops the voltage at its base-one junction to restore the sampled voltage at the output. Both the sampling diode and the output amplifier transistor are biased with current sources so that the current and offset for both devices remain constant as the sampled voltage changes. This gives good X1 dc coupled performance from -7 V to5.2 V. Each device also thermally compensates the other to give good thermal tracking over laboratory temperature ranges. Note that the 2-mA sampling current flows through the switching transistor. This is well below the current that normally flows through the switching transistor in most applications and can be ignored. The current was chosen to give moderate speed while conserving battery power.
Notice that this circuit is intended for use with ground-referenced power-supply switch transistors and directly interfaces the input of an oscilloscope. It is run by two 9-V batteries for safety and isolation. If measurement of the high-side transistor is required, the output from the saturation probe can be connected to the input of a high-voltage differential probe. The differential probe output is then connected to the oscilloscope (Fig. 3). The top of this figure shows how the saturation probe can be used with a high-voltage differential probe to make high-side measurements.
A few of the possible measurements that can be made with the saturation probe are shown in Fig. 4. Remember, the measurements can be performed on MOSFETs, IGBTs, BJTs or diodes. Fig. 5 and Fig. 6 are some pictures taken from a lamp ballast evaluation.
Fig. 4a illustrates how the probe can be used to make collector-base voltage measurements on BJTs. Fig. 4b shows forward- and reverse-voltage measurements on the base-emitter junction, and Fig. 4c illustrates how the probe might be used to measure both saturation voltage and base-emitter (and for FETs, drain-source) reverse voltage.
Fig. 5 shows both the forward and reverse base-emitter voltage for a lamp-ballast switching BJT transistor. The reverse-voltage measurement is important with BJTs because many manufacturers do not properly limit this voltage as per most application notes that specify no more than -6 V or so (n-p-n devices). This allows some margin for overshoot.
In fact, the reverse voltage can be limited to -0.7 V with excellent results. Beyond -6 V, there is evidence that the safe operating area (SOA) of the BJT starts to degrade. The reason for the reverse voltage is to rapidly turn off the transistor and to be able to use the reverse-bias SOA breakdown voltage (VCBO) rather than the forward-bias SOA (VCEO).
If the emitter-base is zenered, the transistor can avalanche, producing very high current flow and catastrophic failure in most applications. This is especially true for half-bridge applications.
Fig. 6 illustrates the reverse collector-emitter voltage of a lamp-ballast BJT plus the saturation voltage. As noted, the saturation voltage is one of the main inefficiencies in an SMPS. The lower the saturation voltage, the better the power-supply efficiency. Note that the saturation-voltage waveform is not a flat slope as would normally be expected, because the lamp ballast is semi-resonant.
Measurement of the reverse collector-emitter (or drain-source) voltage is important in BJTs or FETs. With BJTs, the reverse voltage indicates the possibility that the collector base could be forward biased during switching. If this happens, the forward current can extend the time it takes to turn off the transistor (extend storage time) and could potentially cause transistor failure if the discharge time constant is large enough to allow the transistor that's off to still conduct when the alternate transistor turns on.
Under these conditions, the full power-supply voltage would be shorted through the half-bridge transistors, resulting in very high instantaneous power and possible catastrophic failure in microseconds. This issue could be of particular interest if a lamp-ballast output is opened momentarily due to a bad connection or simply when lamps are changed out without turning off the power, as can occur in normal lamp-fixture maintenance. This is the reason that freewheeling diodes are placed across the collector-emitter junction in most applications.
With both MOSFETS and BJTs, freewheeling or reversal diodes provide a path for an inductive or resonant tank current during transistor turn-off. These diodes provide a discharge path to avoid forward biasing the collector-base junction of BJT switching transistors (reverse current protection), or gate breakdown and possible catastrophic failure with MOSFETs. The diodes should be very fast-acting types.