Topology Selection by the Numbers Is the topology you selected for your current design the best topology available, or is there another topology that might be more suitable? This is a question that every engineer faces with every new design. It is an engineer’s job to understand the trade offs between different design approaches and how and why one approach is better than another approach for a particular application with its unique set of requirements and operating conditions. Typically we find that there are several different approaches we can take to a design challenge, each with advantages and disadvantages in comparison to the other options. This article looks at some tools that can help the engineer make better design choices. In 1988 a landmark paper (Carsten, B., “Converter Component Load Factors: A Performance Limitation of Various Topologies”, PCIM, 1988, Munich) appeared that provided a simple method for direct numerical comparison of power converter topologies. The Carsten paper provided a description of the concept of component load factor, some examples of its use, and some results and general conclusions. The component load factors for each component type are summed to form a total component load factor for each component type for a specific circuit topology. The total component load factors for one topology can be compared to the total component load factors for other circuit topologies for comparison purposes. Circuits with lower total component load factors can be expected to offer better efficiency than other topologies with higher total component load factors under the specific set of operating conditions in which the component load factors are calculated. In some cases one topology may have lower component load factor for one component type, but higher component load factor for another component type, so the best topology for a specific application may not be obvious. Also, circuit topology A may be superior to circuit topology B for one set of operating conditions, based on component load factor, but circuit topology B may be superior to circuit topology A for a different set of operating conditions. Although the methods available may not provide a definitive answer in every case, it can be said that component load factor and the new numerical method introduced in this article help the engineer to shorten the list of topological contenders for a specific set of operating conditions. This article proposes a new numerical method, which is conceptually similar to, but in the specific details different than, the original concept of component load factor. In order to distinguish the new method from component load factor we will call the new method component stress factor (CSF). The goal of the new method is to provide an apples to apples comparison between topologies which provides numerical scores for each component type for each topology for a specific set of operating conditions. The differences between component stress factor and component load factor lie in how the individual and total component factors are calculated and in the results achieved. In the new method we will assume that all of the candidate topologies have available exactly the same amount of silicon, magnetic winding area, and capacitor volume. We will calculate three different component stress factors. For semiconductors we will calculate a semiconductor component stress factor (SCSF). For magnetically inductive windings we calculate a winding component stress factor (WCSF) and for capacitors we calculate a capacitor component stress factor (CCSF). After calculating a component stress factor for each component we sum component factors of the same type. We get the three sums Total SCSF = ∑ SCSF , i semiconductors Total WCSF = ∑WCSF , and windings Total CCSF = i ∑ CCSF . capacitors i In general we will have an operating condition that specifies a range of line voltages. The CSFs may be different at the high line condition than at the low line condition. In some cases the worst case condition Copyright © Technical Witts, Inc., 2006. All rights reserved. 1 occurs at the minimum line voltage, but in other cases the worst case condition occurs at the maximum line voltage. In a few rare cases the worst case condition will occur at a line voltage in the range between maximum line voltage and minimum line voltage, but not at either the minimum line voltage or the maximum line voltage. Good design practice requires designing for the worst case conditions, since reliability is almost always a primary goal. So it makes sense to calculate a worst case total CSF for each component type and to compare optimal worst case sums for different topologies. We will calculate total CSFs for each line voltage extreme as well as total worst case CSFs for each component type, and we will form the sums Total Worst SCSF = ∑ SCSF WORST ,i semiconductors Total Worst WCSF = ∑WCSF windings Total Worst CCSF = WORST ,i , and ∑ CCSF WORST ,i capacitors , , where SCSFWORST ,i is the worst case SCSF for transistor i , WCSFWORST ,i is the worst case WCSF for winding i , and CCSFWORST ,i is the worst case CCSF for capacitor i . Next we will describe how we form the individual component CSFs. In the case of silicon we can compare circuit A with two transistors to circuits B and C with 8 transistors each, by insisting that circuit B and circuit C must use transistors with one fourth the die area, so that the total amount of silicon used in all cases is exactly the same. Circuit A could be a synchronous rectifier flyback converter, circuit B could be a full bridge forward converter with a full bridge synchronous rectifier, and circuit C could be the same as circuit A, but with four transistors in parallel. We will even the playing field by giving each component a weight and factoring the component’s weight into the component stress factor for each individual component. How can we factor a component’s individual weight into the component’s component stress factor in a way that makes sense and has some validity? For the example of circuits A, B, and C we would want the weighting factor to give a four times advantage to the circuit A transistors over the other two circuits and we will want the total SCSF for circuit A to be equal to the total SCSF for circuit C. Suppose we give each transistor an individual weight of 1 in all three designs. We will want the weighting factor to be larger by a factor of four for the transistors in circuits B and C. We can do this by forming a weighting factor which is the sum of the weights of all the transistors in the circuit divided by the individual component’s weight. ∑W j Weighting Factor for component i = where Wi is the individual weight assigned to component i , and ∑W j Wi j j , is the sum of the individual weights for all components of the same type in the circuit. For circuits A, B, and C we assigned Wi = 1 for each transistor. For circuit A ∑W j j = 2, and for circuits B and C ∑W j j = 8, so that the weighting factor for each transistor in circuit A is 2 and the weighting factor for each transistor in circuits B and C is 8. This weighting factor technique yields a four times higher weighting factor for each individual transistor in circuits B and C than for circuit A, as needed. We will use the weighting factor, as defined above in the component stress factor for every component, regardless of type. For the case where there is only one appearance of a particular component type, i.e., there is only one winding in a simple buck converter, then the numerator and denominator in the weighting factor is the same number and the weighting factor will have a value of 1. The total CSF for a topology will depend on how weights are Copyright © Technical Witts, Inc., 2006. All rights reserved. 2 assigned to individual components. By adjusting the weights to achieve the lowest total CSFs we gain insight into how to assign silicon, winding area, and capacitor resources to achieve a more optimal design. In making comparisons between designs we must first optimize the designs by finding the combination of weights that yields the lowest total worst case CSFs, in order to make a valid comparison. Another quantity that determines an individual component’s stress factor is the component’s maximum voltage stress. How should a component’s voltage stress effect its CSF? For a transistor, stress is related to power dissipated in the transistor. For the analysis in this article series we will assume that all of the semiconductors are mosfets. Conduction losses in a mosfet can be calculated from the average squared current in the mosfet multiplied by the on resistance of the mosfet, so we will form the CSF for a transistor by using the maximum average squared current, or the rms current squared. The on resistance for a mosfet of a given die size is proportional to the square of the voltage rating of the mosfet. This is because the channel length is proportional to voltage breakdown and, for a given die size, a longer channel length results in a smaller channel cross sectional area. Resistance, R , can be found from R = ρ L , where ρ is A the resistivity of enhanced silicon, L is the channel length, and A is the channel cross sectional area. For a transistor, we include the square of the maximum voltage in the CSF for the individual transistor, because voltage rating increases channel length and decreases channel cross sectional area, so that for an individual transistor we will define the component stress factor to be ∑W SCSFi = j V MAX ,i 2 I RMS ,i 2 , P2 j Wi where SCSFi is the individual transistor’s component stress factor, V MAX ,i is the maximum peak voltage applied to the transistor, I RMS ,i is the maximum rms current in the transistor, and P is the total power processed by the circuit, assuming the converter is 100% efficient. The inclusion of P in the definition of CSF normalizes the CSF by removing any dependence on the power processed and makes CSF a dimensionless quantity. In the calculations of rms current in this article we will assume that the circuit’s inductors are large and that there is no variation in the current during the operating states of the circuit. We can calculate the rms current from I RMS = ∑D i i Ii 2 where Di is the duty cycle of the i operating state, and I i is the transistor current during the i operating state. One operating state may be the state in which the converter’s main switch is on, while a second operating state may be the state in which the main switch is off and the main rectifier is on. For magnetic components we will assume that each circuit has available exactly the same amount of window area for windings. We will not consider the fact that some designs require multiple cores. We will treat each winding as a separate component and assign a weight to each winding and then calculate a CSF for each winding. First we define a CSF for windings ∑W WCSFi = j Wi j V MAX ,i 2 I RMS ,i 2 P2 Copyright © Technical Witts, Inc., 2006. All rights reserved. 3 Conduction losses in a winding are joule heating, or I2R type, losses, so we can readily see why we need to 2 2 include I RMS ,i in the definition. I RMS ,i is calculated in the same way as indicated above for transistors. In order to accommodate applied voltage in a magnetically inductive element we need turns of wire in proportion to the voltage. More turns of wire mean less window area available for each turn and a longer wire. The winding resistance will be proportional to the number of turns and inversely proportional to the area available for each turn, therefore, the winding resistance will increase with the square of the number of turns which will be proportional to the voltage squared. We calculate a worst case V MAX ,i , by the formula, V MAX ,i = ∑ Di |Vi | i under worst case conditions for the winding, i.e., the condition that yields the largest value for V MAX ,i , where Di is the duty cycle of the i operating state of the converter under the worst case condition and |Vi | is the absolute value of the winding voltage in the i operating state. For most converter types V MAX ,i occurs at the maximum line voltage, but for boost chokes V MAX ,i occurs at a duty cycle of 0.5. For boost chokes that operate in a range where the duty cycle is always greater than 0.5, V MAX ,i occurs at the maximum line voltage. For boost chokes that operate in a range where the duty cycle is always less than 0.5, V MAX ,i occurs at the minimum line voltage. For capacitors we calculate CSF in much the same way. We assign a weight to each capacitor in the circuit. We determine the peak voltage stress for the capacitor and we calculate the rms current in the capacitor. We reason that the stress in the capacitor is primarily due to joule heating, so we want the CSF to represent a joule heating loss, or I2R loss, where R represents an equivalent series resistance (esr) in the capacitor, which means that the CSF for capacitors will depend on the square of the rms current in the capacitor. Esr is, generally, a function of the volume of a capacitor. The volume of a capacitor is proportional to its energy storage capacity, 1 C V 2 , so that the component stress for the capacitor also 2 depends on the squared peak voltage of the capacitor. For capacitors we define the CSF to be ∑W CCSFi = j Wi j VPEAK ,i 2 I RMS ,i 2 P2 Some circuit approaches combine topologies. Our leading candidates may be a forward converter or a buck converter pre-regulator combined with a dc transformer. We will want to compare the CSFs for the forward converter to the CSFs for the combination of buck sub-converter plus dc transformer subconverter. How do we create a valid comparison in the case where one of the candidate topologies is actually a combination of two topologies? Would it be correct to just add the total component load factors for the buck plus dc transformer and compare that to the forward converter? We decided at the onset that we would insist that each candidate topology would have to rely on exactly the same amount of silicon, window area, and capacitor volume, so we established weighting factors to enforce that rule. We can enforce the equal resource rule by giving each sub-converter a weight for each component type and forming new CSF sums for each candidate topology where we add up the CSF sums but also give each sub-converter a weight for each component type. For the buck plus dc transformer combination we can say that the buck converter, or sub-converter 1, has a set of weights, and that the dc transformer, or subconverter 2, has a set of weights and we can form new CSF sums which combines the CSF sums from the two sub-converters and factors in the sub-converter weights in the new sums. For the total semiconductor CSF (SCSF) for a combination of sub-converters we define Copyright © Technical Witts, Inc., 2006. All rights reserved. 4 Total SCSF = ∑ subconverters ∑W j semiconductors Wi ( Total SCSFi ) , where Wi is the semiconductor sub-converter weight for sub-converter i , ∑W j semiconductors is the sum of the semiconductor sub-converter weights, and Total SCSFi is the total SCSF for sub-converter i . If we are going to compare a buck plus a dc transformer combination to a forward converter we could give the buck sub-converter a semiconductor weight of 1, the dc transformer sub-converter a semiconductor weight of 1, and the total SCSF for the combination would be two times the total SCSF for the buck plus two times the total SCSF for the dc transformer. We calculate winding and capacitor CSFs for the sub-converter combinations in the same way, but the weights for windings and capacitors may be different than the semiconductor weights. We may find that the lowest total CSFs correspond to unequal weights for the sub-converters and we may find that the sub-converter semiconductor weights are different than the subconverter winding or capacitor weights. Knowing the weights for sub-converters that yields the lowest CSFs gives insight into how to assign silicon, winding area, and capacitor resources. For example, if we combine sub-converter A with high semiconductor CSF and low winding CSF with sub-converter B with low semiconductor CSF and high winding CSF it might make sense to give a higher semiconductor subconverter weight to sub-converter A and a lower winding sub-converter weight to sub-converter A compared to sub-converter B so that resources will be assigned where they are most needed. Now that we have CSF defined for transistors, windings, and capacitors let’s apply our new method to some circuit examples and see what kind of results can be achieved. Figure 1. Buck converter with synchronous rectification As a first example consider a buck converter with synchronous rectification, as illustrated in figure 1. The input voltage is 200 volts, the output voltage is 100 volts and the load is 100 watts. In this converter the duty cycle is 0.5. Let’s give each mosfet a weight of 1 for now, which suggests that we intend to make each mosfet the same die size. For MMAIN the maximum voltage stress V MAX , MAIN is 200 volts. We can calculate the rms current to be I RMS , MAIN = DMAIN I MAIN 2 = 0.5 ( 1 amp ) 2 = 2 / 2 amp and the power processed is 100 watts. We can plug the values we know into the formula for SCSF to find the SCSF for MMAIN. ∑W SCSFMMAIN = j j WMMAIN V MAX , MMAIN 2 I RMS , MMAIN 2 ( 1 + 1 ) ( 200 volts ) 2 ( 2 / 2 amp ) 2 = =4 P2 1 ( 100 watts ) 2 For MSYNC the maximum voltage stress is 200 volts, the rms current in MSYNC is the same as in MMAIN, and the power is 100 watts. We calculate the SCSF for MSYNC to be Copyright © Technical Witts, Inc., 2006. All rights reserved. 5 ∑W SCSFMSYNC = j j WMSYNC V MAX , MSYNC 2 I RMS , MSYNC 2 ( 1 + 1 ) ( 200 volts ) 2 ( 2 / 2 amp ) 2 = = 4. P2 1 ( 100 watts ) 2 Now we can calculate a total SCSF for the figure 1 circuit Total SCSF = ∑ SCSF = SCSF semiconductors i MMAIN + SCSFMSYNC = 4 + 4 = 8 . We can calculate a winding CSF for the buck choke and set the weight for the buck choke to 1. The current in the buck choke is a constant 1 amp and the voltage magnitude is always 100 volts, so WCSFCHOKE = 1 ( 100 volts ) 2 ( 1 amp ) 2 =1. 1 ( 100 watts ) 2 Since there is only one winding the total WCSF is also 1. We have two capacitors in figure 1. We will assign a weight of 1 to each of the capacitors and calculate CCSFs. The peak voltage of the input capacitor is 200 volts. The average current into the capacitor is 0.5 amp. During the on time of the main switch there is a net current of 0.5 amp out of the input capacitor and during the off time of the main switch there is a net current of 0.5 amp into the input capacitor, so that the magnitude of the capacitor current is always 0.5 amp. For the output capacitor the input current to the capacitor is always equal to the output current from the capacitor so that the net current in the output capacitor is 0. CCSFCIN = CCSFCOUT = ( 1 + 1 ) ( 200 volts ) 2 ( 0.5 amp ) 2 = 2, 1 ( 100 watts ) 2 ( 1 + 1 ) ( 100 volts ) 2 ( 0 amp ) 2 = 0 , and 1 ( 100 watts ) 2 Total CCSF = CCSFCIN + CCSFCOUT = 2 + 0 = 2 . Figure 2. Isolated flyback converter with synchronous rectification As a second example, consider an isolated flyback converter with synchronous rectification operating at a line voltage of 200 volts, a load voltage of 100 volts, and a load power of 100 watts, as illustrated in figure 2. The turns ratio of the transformer is 2:1 and the duty cycle is 50%. We will give a weight of 1 to both the main switch and the synchronous rectifier. We will give a weight of 1 to each winding, and we will give a weight of 1 to each capacitor. For the main primary side switch the component stress factor is Copyright © Technical Witts, Inc., 2006. All rights reserved. 6 SCSFMAIN = ( 1 + 1 ) ( 400 volts ) 2 ( 2 / 2 amps ) 2 = 16 . 1 ( 100 watts ) 2 For the secondary side synchronous rectifier switch the component stress factor is SCSFSYNC = ( 1 + 1 ) ( 200 volts ) 2 ( 2 amps ) 2 = 16 . 1 ( 100 watts ) 2 For the isolated flyback the total semiconductor component stress factor is Total SCSF = SCSFMAIN + SCSFSYNC = 16 + 16 = 32 For the primary winding the component stress factor is WCSFPRIMARY ( 1 + 1 ) ( 200 volts ) 2 ( 2 / 2 amps ) 2 = = 4. 1 ( 100 watts ) 2 For the secondary winding the component stress factor is WCSFSECONDARY = ( 1 + 1 ) ( 100 volts ) 2 ( 2 amps ) 2 = 4. 1 ( 100 watts ) 2 For the isolated flyback the total winding component stress factor is Total WCSF = WCSFPRIMARY + WCSFSECONDARY = 4 + 4 = 8 . For CIN the component stress factor is CCSFCIN ( 1 + 1 ) ( 200 volts ) 2 ( 0.5 amp ) 2 = = 2. 1 ( 100 watts ) 2 For COUT the component stress factor is CCSFCOUT = ( 1 + 1 ) ( 100 volts ) 2 ( 1 amp ) 2 = 2. 1 ( 100 watts ) 2 For the isolated flyback converter the total capacitor component stress factor is Total CCSF = CCSFCIN + CCSFCOUT = 2 + 2 = 4 . Consider for a moment the total SCSFs we calculated for the buck and the flyback. We found that the total SCSF for the buck was 8 and the total SCSF for the flyback was 32, assuming equal weights for the two transistors in both cases. These results for the buck and the isolated flyback used for the same design problem beg the question. What is it that is worse by a factor of 4 about the isolated flyback converter? If we were to assume that the two transistors in the flyback had the same die size as the two transistors in the buck we might compare the conduction power losses in the mosfets of the two circuits. Let’s do that. We will assume that the channel resistance goes up as the square of the voltage for a given die size. This assumption holds up well at higher voltages, but breaks down at low voltages due to the contribution of Copyright © Technical Witts, Inc., 2006. All rights reserved. 7 lead resistance. For the flyback converter we need a primary side mosfet with a channel that is twice as long to handle the voltage (400 volts) compared to the buck converter with its 200 volt mosfet. Since we stipulated that to make a valid comparison the total die size should be the same for both circuits, the cross section for the channel of the 400 volt mosfet must be half the cross section of the 200 volt mosfet used in the buck converter. The net result is that the primary side mosfet in the flyback would have a channel resistance four times the channel resistance of the mosfet in the buck converter. If the channel resistance for the main switch in the buck converter is R, then the channel resistance for a main switch with the same die size in the flyback is 4R. We can calculate conduction losses. For the buck converter the conduction losses are PBUCK = ( 1 amp )2 R. For the flyback converter we have one mosfet with a channel resistance of 4 R operating with a rms current of 0.707 amp, and one mosfet with a channel resistance of R operating with a rms current of 1.414 amps. The total mosfet conduction losses for the flyback are PFLYBACK = ( 0.707 amp )2 ( 4 R) + ( 1.414 amp )2 R = 4 ( 1 amp )2 R = 4 PBUCK. This result would seem to suggest that the semiconductor component stress factor for the isolated flyback should be 4 times worse than the semiconductor component stress factor for the buck converter, since the total component stress due to conduction losses are 4 times worse for the flyback converter with the same amount of silicon. We defined component stress factor so that it would indicate the relative power losses in the components of the same type, so that we could have a measurement tool that would predict which topology will be more efficient than another topology and by how much. So far the calculated results seem to be consistent with the reality. Figure 3. Full bridge forward converter with synchronous rectification. As a third example consider the isolated full bridge forward converter of figure 3. The isolated full bridge forward converter has a line voltage of 200 volts, a load voltage of 100 volts, and a load power of 100 watts. The full bridge forward converter has a full bridge primary switch arrangement and a full bridge secondary synchronous rectifier arrangement. The transformer turns ratio is 2:1 and each switch operates at 50% duty cycle. We assign a weight of 1 to each mosfet so that the sum of transistor weights is 8. The maximum voltage of the primary switches is 200 volts and the rms current in each primary switch is 2 1 ( 1 / 2 amp ) 2 = amp . 2 4 I rms = For the secondary switches the maximum voltage is 100 volts and the rms current in each secondary switch is I rms = 1 2 ( 1 amp ) 2 = amp . 2 2 For each primary switch the component stress factor is SCSFMi = 8 ( 200 volts ) 2 ( 2 / 4 amp ) 2 =4 1 ( 100 watts ) 2 and for each secondary switch the component stress factor is Copyright © Technical Witts, Inc., 2006. All rights reserved. 8 SCSFMi 8 ( 100 volts ) 2 ( 2 / 2 amp ) 2 = = 4. 1 ( 100 watts ) 2 The total semiconductor component stress factor for the full bridge forward converter is Total SCSF = ∑ SCSF semiconductors Mi = 8 ⋅ 4 = 32 . The total semiconductor CSF for the full bridge forward converter is identical to the total semiconductor CSF for the flyback converter operating under the same conditions of line and load. Compared to the buck converter both the flyback and the full bridge forward converter have 4 times the total semiconductor CSF. The figure 3 circuit has 3 windings. We will assign each winding a weight of 1, so that the sum of the winding weights is 3. The current in the primary winding is 1/2 amp. The current in the other two windings is 1 amp. The voltage of the primary transformer winding is 200 volts, the voltage of the secondary transformer winding is 100 volts, and the voltage of the choke winding is 0 volts. For the transformer primary winding CSF we get 3 ( 200 volts ) 2 ( 0.5 amp ) 2 = = 3. ( 100 watts ) 2 1 WCSFPRIMARY and for the secondary winding we get WCSFSECONDARY 3 ( 100 volts ) 2 ( 1 amp ) 2 = = 3. 1 ( 100 watts ) 2 For the output choke winding we get WCSFCHOKE = 3 ( 0 volt ) 2 ( 1 amp ) 2 = 0. 1 ( 100 watts ) 2 We calculate the total winding CSF to be Total WCSF = WCSFPRIMARY + WCSFSECONDARY + WCSFCHOKE = 3 + 3 + 0 = 6 . Despite the fact that the full bridge forward converter has 3 windings the total winding CSF for the full bridge forward converter is lower than the total winding CSF for the flyback. We can achieve even lower total WCSF by choosing different winding component weights. We achieve the optimal result of 4 by assigning a weight of 1 to each transformer winding and a weight of 0 to the choke winding. The figure 3 circuit has two capacitors. We will assign a weight of 1 to each capacitor so that the sum of capacitor weights is 2. The peak voltage for the input capacitor is 200 volts and for the output capacitor the peak voltage is 100 volts and the rms current in each capacitor is 0 amp. We calculate the capacitor CSFs to be CCSFCIN = 2 ( 200 volts ) 2 ( 0 amp ) 2 = 0 and 1 ( 100 watts ) 2 Copyright © Technical Witts, Inc., 2006. All rights reserved. 9 I )SY C N sSY ,(rm =− I1 D C N CCSFCOUT 2 ( 100 volts ) 2 ( 0 amp ) 2 = = 0. 1 ( 100 watts ) 2 We calculate the total capacitor CSF to be Total CCSF = CCSFCIN + CCSFCOUT = 0 + 0 = 0 . Compared to the flyback converter the full bridge forward converter has much lower total capacitor CSF. By comparison to the flyback converter the full bridge forward converter has the same total SCSF, but lower total WCSF, and lower total CCSF. That would suggest that for an application with a very limited line voltage range a full bridge forward converter would have an efficiency advantage over a flyback and that the efficiency advantage would appear in the form of higher efficiency magnetics and capacitors. Alternatively, for a given efficiency in the subject application we would expect that the flyback converter would require larger magnetics and capacitors than the full bridge forward converter. Figure 4. Buck converter with synchronous rectification. We will now consider the effects of variable line voltage. Figure 4 illustrates a buck converter with variable line voltage. We first consider the semiconductor CSF. We will assign a weight of 1 to each transistor. For both transistors the voltage stress is 800 volts. At high line the duty cycle is 12.5% and at low line the duty cycle is 50%. We will assume that the inductor is large and the inductor current is a constant 1 amp. We need to calculate the rms currents in the switches at high line and low line. For the main switch the rms current is found from I rms , MAIN = D I MAIN 2 so that at 800 volts Total CCSF = CCSFCIN +CCSFCOUT = 0+0 = 0 and at 200 volts C C S F C O U T = 2 1 ( 1 0 0 v o lts ) 2 ( 0 a m p ) ( 1 0 0 w a tts ) 2 2 = 0 . For the synchronous rectifier we use the formula so that at 800 volts Copyright © Technical Witts, Inc., 2006. All rights reserved. 10 ,5.=watts rm − = MAIN 8SCSF 1Y vs== F otalSC T IN A 20()M =+ = a206418)(Y Sw p7+ 5/.m volts00 F C N 1600 lts and at 200 volts . The CSF for the main switch at 800 volts is . For the synchronous rectifier the CSF at 800 volts is . The total SCSF at 800 volts is . The CSF for the main switch at 200 volts is . For the synchronous rectifier the CSF at 200 volts is . The total SCSF at 200 volts is . The total SCSFs are the same for the buck converter at high line and at low line. For the buck converter the total SCSF does not vary with line voltage, but the component stresses of the two transistors vary over the line range, so that, as the line voltage changes, the stresses shift from one transistor to the other transistor. In the figure 1 buck converter the total SCSF was 8 and the ratio of maximum line voltage to minimum line voltage was 1. In the figure 4 buck converter the ratio of maximum line voltage to minimum line voltage is 4 and the total SCSF is 128, which is 16 times 8, or 4 squared times 8. Can we say, in general , that the total SCSF is proportional to the square of the ratio of maximum to minimum line voltage? It seems to be true for the buck converter and it will prove to be true for other converters too with a few caveats. For comparison purposes we must calculate a total worst case SCSF which takes into account the fact that we have to design for worst case conditions. The worst case CSF for the main switch in the buck converter of figure 4 is Copyright © Technical Witts, Inc., 2006. All rights reserved. 11 (|∑ Y= == T R O N ,=+ V + X A aM w SF C 817W si).IvoltD 205rm − = vSM I+ A F T R O W C N p871236olts,4Y m 05(.)a . For the synchronous rectifier the worst case CSF is . The total worst case SCSF for the buck converter of figure 4 is . The reader will notice that over the line range the synchronous rectifier has CSF that is greater than or equal to the CSF of the main switch. This suggests that it would make sense to use a larger synchronous rectifier and a smaller main switch for this application. If we adjust the relative weights of the two semiconductors we can achieve a lower total worst case SCSF. We achieve a minimum total worst case SCSF when the synchronous rectifier has a weight of 57% and the main switch has a weight of 43%. The total worst case SCSF with optimal weights is 172.7. When we compare the buck converter to the other topology contenders we will use the 172.7 number for the buck converter and compare that number to the optimized worst case numbers for the other contenders. We calculated SCSFs above assuming that the two transistors had equal weights, but we have found that unequal weighting of the transistors will improve the worst case condition. We have not illustrated the calculation of SCSFs for figure 4 with the unequal weighting, but we show the results of the SCSF calculations below in table 1. The results shown in table 1 are not precisely consistent with the hypothesis that the total SCSFs are proportional to the square of the ratio of maximum line voltage to minimum line voltage, but as an approximate rule of thumb the hypothesis holds up, and it would hold up precisely had we not altered the component weights for optimization. In the buck converter there is only one winding so the weight assignment for the winding is arbitrary and inconsequential. Let’s give the choke winding a weight of 1. The worst case winding voltage occurs at maximum line voltage for the buck converter. The maximum average choke voltage is The choke current is a constant 1 amp. We can calculate the winding CSF to be . Since there is only one winding and the winding current does not change with line voltage 3.0625 is the total winding CSF and also the total worst case winding CSF. We have two capacitors in the buck converter of figure 4. Let’s assign an initial weight of 1 to each capacitor. For the input capacitor the peak capacitor voltage stress is 800 volts. When the line voltage is 800 volts the rms current for the input capacitor is At the minimum line voltage the rms current for the input capacitor is . Copyright © Technical Witts, Inc., 2006. All rights reserved. 12 volts8w ()SF IN 201C == 1032()ats4 For the output capacitor the peak voltage is 100 volts and the current is 0 amp. For the input capacitor the CSF at 800 volts is . At a line voltage of 200 volts the CSF for the input capacitor is . For the output capacitor, since the output capacitor current is 0, we can say that the CSF for the output capacitor is 0. The total CCSF for the buck converter of figure 4 will just be equal to the CCSF for the input capacitor. The total worst case CCSF for the buck converter is the CCSF for the input capacitor at 200 volts, or 32. Since the CSF for the output capacitor is 0 this suggests that we can eliminate it or replace it with a very small capacitor, very much smaller than the input capacitor. The lowest total worst case CCSF that we can achieve corresponds to the case where the output capacitor is assigned a weight of 0 and eliminated. In that case the optimal total worst case CCSF is 16 for the figure 4 buck converter. Now let’s try to calculate CSFs for a different topology operating under the same conditions of line and load. What would be the alternate circuit topologies to solve the same non-isolated problem that might yield lower total worst case CSFs? A tapped inductor buck converter comes to mind as a possibility. We will evaluate the tapped inductor buck converter, as illustrated in figure 5. Figure 5. Tapped inductor buck converter with synchronous rectification. In the tapped inductor buck converter we have a coupled inductor. The turns ratio of the inductor affects the switch voltages and currents and the CSFs of all of the components. Since we have already illustrated how to calculate all of the quantities we need for CSF, instead of illustrating all of the calculations again we will summarize the results for both the buck converter of figure 4 and the tapped inductor buck converter of figure 5 in table 1. The tapped inductor buck converter with a turns ratio of 0 reduces to the simple buck converter of figure 4. As the turns ratio increases the semiconductor CSF and the capacitor CSF decreases while the winding CSF increases. One of the major shortcomings of the simple buck converter for wide line ranges is that the semiconductor stress becomes very large because, at the low end of the line range, a relatively high voltage switch with relatively high on resistance must operate at the relatively large currents required at relatively low voltage. By tapping the inductor we need a slightly higher voltage switch operating at much lower current for the main switch, but the synchronous rectifier becomes a much lower voltage switch that can more readily handle the higher currents. The minimum semiconductor CSF occurs at a turns ratio of about 8/3. At this turns ratio the capacitor CSF is also about half that of the simple buck, but the winding CSF is almost 4 times larger, so we have traded off semiconductor CSF and capacitor CSF for winding CSF. Notice also that when the turns ratio is 2 or higher the highest total CSFs for the tapped inductor buck Copyright © Technical Witts, Inc., 2006. All rights reserved. 13 converter corresponds to the minimum line voltage and all the CSFs decrease at higher line voltages. All things considered the tapped inductor buck converter with a turns ratio of 2 seems like a more suitable topology for the application than the simple buck converter. The calculation of CSFs in this case yields some useful numbers to indicate which topology might be better, but not a definitive answer. In the case of CCSFs the capacitor requirement may be driven by other considerations, such as filtering, hold up, or transient requirements. Also, the two topologies illustrated are candidates for multi-phasing so that, if the capacitor stress factor is a major liability for one topology, but the other CSFs are superior, then the CCSFs can be reduced by the square of the number of sub-converters in a parallel multi-phase converter system. The simple buck converter would benefit most from multi-phasing since it has the highest total CCSFs. For the particular problem under consideration in those cases where efficiency is a primary goal, the tapped inductor buck converter illustrated in figure 5 or one of the other tapped inductor buck topologies available are likely to be preferred, since in many, if not most, situations the semiconductor losses account for a significantly greater share of the losses than either the winding or capacitor losses. Where tapped inductor converters are clearly superior are in applications with limited line voltage range and where the step up or step down ratio is either very large or very small. For very small step up or step down ratios with limited line voltage range there are tapped inductor converters that provide CSFs less than 1! Table 1 Figure # Turns Ratio Weight Main Switch Weight Sync Switch Weight L2 Weight L1 Weight Input Capacitor Weight Output Capacitor Total SCSF( 800 volts ) Total SCSF( 200 volts ) Total Worst Case SCSF Total WCSF( 800 volts ) Total WCSF( 200 volts ) Total Worst Case WCSF Total CCSF( 800 volts ) Total CCSF( 200 volts ) Total Worst Case CCSF 4&5 0 43 57 0 1 1 0 117 131 173 3.06 3.06 3.06 7 16 16 5 1 1.23 1 0.596 1 8 1 54.8 88.9 99.5 4.53 4.36 6.5 4.43 10.1 10.1 5 2 1.66 1 1.08 1 4 1 40.7 83 85.7 5.46 7.07 9.7 3.65 8.33 8.33 5 2.66 1.91 1 1.36 1 3 1 36.7 83.9 83.9 5.9 8.78 11.7 3.4 7.76 7.76 5 4 2.24 1 1.89 1 2 1 32.3 90.5 90.5 6.56 12 15.3 3.15 7.2 7.2 What other topologies might be worth considering for the subject application of figures 4 and 5? A boost converter pre-regulator with a non-isolated dc transformer has lower total SCSF and lower total CCSF than the buck, but with higher total WCSF and it does not compare favorably with the tapped inductor buck for this particular set of line and load conditions. Figure 6. Full bridge forward converter with wide line voltage range. Copyright © Technical Witts, Inc., 2006. All rights reserved. 14 Total dcX + = W SCSF Boost 2o9 r s t .9 04 32 124 + = 1 Figure 6 illustrates a full bridge forward converter similar to figure 3, but it operates over a wide line voltage range. The application is the same as the application for figures 4 and 5, except that the line and load are isolated in figure 6. The isolation requirement adds to the design problem, and, generally, adds to the component stress factors. The results for the full bridge forward converter are illustrated below in table 2. Figure 3 illustrated a full bridge forward converter with a ratio of maximum line voltage to minimum line voltage of 1 and it had a total SCSF of 32. If our hypothesis that the total SCSF increases with the square of the line voltage ratio holds up we should have a total SCSF for figure 6 of 512 which is 16, or 4 squared, times 32. At the minimum line voltage the total SCSF is 512, but at maximum line voltage the total SCSF is only 224. Figure 7. Combination of boost pre-regulator sub-converter with full bridge dc transformer sub-converter Figure 7 illustrates a boost pre-regulator plus dc transformer combination that seeks to solve the same problem as the figure 6 circuit. In the figure 7 approach a boost converter provides a regulated 800 volts to the input of a full bridge dc transformer, which provides 100 volts to the load. This two stage approach might seem like an unlikely candidate topology for the problem, because all of the power is processed twice, once by the boost sub-converter then again by the dc transformer, but the numbers suggest otherwise. Table 2 summarizes the results of the CSF calculations for figures 6 and 7. Notice that the total SCSF for the boost converter varies significantly over the line voltage range. The total SCSF for the boost converter is more than 10 times smaller at the maximum line voltage than at the minimum line voltage, unlike the simple buck converter whose total SCSF varies not at all over the line range, or only very little if the transistor weights are altered for worst case optimization. If the boost converter of figure 7 had equal transistor weights then its total SCSF at high line would be 2 and its total SCSF at low line would be 32, which is 16 times, or 4 squared times, larger than the high line SCSF. In general the highest total SCSF corresponds to the minimum line voltage, regardless of the topology. We will illustrate the calculation of the total SCSF for the combination of boost sub-converter with dc transformer sub-converter according to the equations and method described earlier. We will use the subconverter weights indicated in table 2. From table 2 the boost sub-converter semiconductor weight is 1 and the dc transformer sub-converter weight is 1.04. The total worst case semiconductor CSF for the boost converter is 29.9 and the total worst case semiconductor CSF for the dc transformer is 32. The total worst case semiconductor CSF for the combination is . Copyright © Technical Witts, Inc., 2006. All rights reserved. 15 Table 2 Figure # Transformer Turns Ratio Weight Main Boost Switch Weight Sync Boost Switch Weight Primary Switch Weight Secondary Switch Weight Sub-converter Boost Semiconductor Weight Sub-converter dc Xfmr Semiconductor Weight Boost Choke Weight Forward Choke Weight Primary Winding Weight Secondary Winding Weight Sub-converter Boost Winding Weight Sub-converter dc Xfmr Winding Weight Input Capacitor Forward Weight Output Capacitor Forward Weight Input Capacitor Boost Weight Output Capacitor Boost Weight Input Capacitor dc Xfmr Weight Output Capacitor dc Xfmr Weight Sub-converter Boost Capacitor Weight Sub-converter dc Xfmr Capacitor Total SCSF( 800 volts) Forward Total SCSF( 800 volts ) Boost Total SCSF( 800 volts ) dc Xfmr Total SCSF( 800 volts ) Boost + dc Xfmr Total SCSF( 200 volts ) Forward Total SCSF( 200 volts ) Boost Total SCSF( 200 volts ) dc Xfmr Total SCSF( 200 volts ) Boost + dc Xfmr Total Worst Case SCSF Forward Total Worst Case SCSF Boost Total Worst Case SCSF dc Xfmr Total Worst Case SCSF Boost + dc Xfmr Total WCSF( 800 volts) Forward Total WCSF( 800 volts ) Boost Total WCSF( 800 volts ) dc Xfmr Total WCSF( 800 volts ) Boost + dc Xfmr Total WCSF( 200 volts ) Forward Total WCSF( 200 volts ) Boost Total WCSF( 200 volts ) dc Xfmr Total WCSF( 200 volts ) Boost + dc Xfmr Total Worst Case WCSF Forward Total Worst Case WCSF Boost Total Worst Case WCSF dc Xfmr Copyright © Technical Witts, Inc., 2006. All rights reserved. 6 2:1 1 1 1.5 1 1 7 8:1 1 1.73 1 1 1 1.04 1 1 1 1 1 1 0 0 1 1 1 1 0 224 2.73 32 68.3 512 29.9 32 124 512 29.9 32 124 7 0.25 4 8.5 12.25 4 4 16 12.25 4 4 16 Total Worst Case WCSF Boost + dc Xfmr Total CCSF( 800 volts) Forward Total CCSF( 800 volts ) Boost Total CCSF( 800 volts ) dc Xfmr Total CCSF( 800 volts ) Boost + dc Xfmr Total CCSF( 200 volts ) Forward Total CCSF( 200 volts ) Boost Total CCSF( 200 volts ) dc Xfmr Total CCSF( 200 volts ) Boost + dc Xfmr Total Worst Case CCSF Forward Total Worst Case CCSF Boost Total Worst Case CCSF dc Xfmr Total Worst Case CCSF Boost + dc Xfmr 16 3 0 0 0 0 3 0 3 3 3 0 3 We can see from table 2 that the boost plus dc transformer has much lower SCSFs than the forward converter for the wide line range condition. The CCSFs for the two are the same and the WCSFs are just slightly better for the forward converter. For this wide line range application and for isolated wide line range applications in general, the boost plus dc transformer is the topology of choice for high efficiency. For limited line ranges the forward converter has lower CSFs. When the line voltage maximum is about 1.5 times the line voltage minimum the total SCSFs for the forward converter are about the same as the total SCSFs for the boost plus dc transformer combination. Are there other topologies that we should consider for isolated wide line range applications? For the power level specified the most likely topologies that one would likely find in common use are single ended flybacks and forward converters. Single ended flybacks and active reset forward converters may have lower total SCSFs than the full bridge forward converter of figure 6, but they will have larger total WCSF and larger total CCSF. The preference of the single ended flybacks and forward converters at lower power levels is a result of the lower circuit complexity which will likely translate into lower cost, although there is a price to be paid for the simplicity and cost savings in higher CSFs. What conclusions can we draw from CSF calculations? 1. For many topologies there is a relationship between total SCSF and line voltage range that depends on the square of the ratio of line maximum voltage to line minimum voltage. This would suggest that any topology will be more efficient if it can be operated over a limited line voltage range. There are many converters that are designed to operate over a line voltage range of 85 volts ac to 265 volts ac. One approach to this problem is to use a rectifier that functions as a voltage doubler rectifier at the low end of the line voltage range and as a full bridge rectifier at the high end of the line voltage range. The change from doubler rectifier to full bridge rectifier would occur near mid range and can be implemented automatically with a triac and line voltage sensing circuit or manually with a switch or jumper. With a switchable rectifier the effective line range seen by the power conversion circuits is much reduced in comparison to a full range full bridge rectifier circuit. The power conversion circuits can be expected to be significantly more efficient with the switchable rectifier. 2. For non-isolated power conversion tapped inductor bucks, boosts, flybacks, and SEPICs do better than the simple ( untapped ) bucks, boosts, flybacks, and SEPICs for applications with very small step up or step down ratios and for applications with very large step up or step down ratios. These results are not proven in this paper, but can be proven by the reader by applying the methods taught in this paper. Another approach that has been proposed for wide line range non-isolated applicationsis are the quadrature converters. The advantage of the quadrature converters is that they require only a single modulator to operate over a wide line range, but they have high CSFs. With a dual modulator approach much lower CSFs are achieved. Copyright © Technical Witts, Inc., 2006. All rights reserved. 17 3. For some wide line range applications the combination of a boost sub-converter plus a dc transformer sub-converter may provide results that are as good or better than what can be achieved by a single converter topology. 4. Most isolated topologies are most efficient when operated in a range near 50% duty cycle. For full bridge forward topologies efficiency is greatest when the duty cycle is 50% at the minimum line voltage. For single ended flybacks and active reset forward converters greatest efficiency (lowest total SCSF) will occur when they operate over a duty cycle range centered around a duty cycle of 50%. It should be recognized that some converters, such as full bridge forward converters and some two switch flybacks and forward converters, cannot operate above 50% duty cycle. 5. Boost converters will have efficiency greater than or equal to buck converters in any application where either type of converter can be practically used. Buck converters have a total SCSF that varies little over the line range. Boost converters have total SCSF that is about equal to the buck converter at minimum line voltage but boost converters have much lower SCSF at maximum line voltage. 6. For non-isolated topologies ratios of line voltage to load voltage near 1 provide low CSFs, near zero in some cases, while ratios much different than 1 provide relatively high CSFs. For isolated topologies the relationship between line voltage and load voltage is inconsequential. Isolated topologies will always have CSFs that are equal to or larger than what can be achieved with a non-isolated topology. 7. The assumptions made in the analysis for SCSF assumes a relationship between transistor voltage rating and transistor on resistance that holds up well at medium and high voltages, but does not hold up well at low voltage. The reader should re-examine any conclusions drawn from the analysis for applications with operating voltages less than about 50 volts. In many situations there are alternate circuit topologies that are roughly equivalent but operate at higher voltage and lower current. These should be preferred at lower voltages. 8. The analyses based on CSF calculations provides the user with a set of three numbers that can be compared to sets of three numbers for alternate candidate topologies. The analysis also yields a set of compoents weights that specify how resources should be assigned to achieve optimal results. In some cases one topology will have a set of CSFs wherein all three CSFs are equal to or lower than the other candidates, but, more often, one topology will be superior for one component type and inferior for another component type, so that the choice of the best candidate may not be obvious. In many cases the set of three numbers achieved will depend on the parameters of the design such as the turns ratio of a transformer or coupled inductor and varying the parameters enables the designer to improve CSF for one component type at the expense of another component type. This paper has not provided a method to achieve a single number for a topology for comparison purposes, although it is easy to imagine how the numbers might be combined. The three CSF numbers could simply be summed. There are many other ways that a single number can be achieved. The question remains as to how the three numbers can best be combined into a single number. This subject will remain a topic for a future paper or article. 9. The methods taught in this article are methods that can be readily performed by a computer. The creation of a computer program to calculate CSFs and to optimize designs intelligently based on CSF and other considerations will be a valuable tool to the power supply design engineer. Copyright © Technical Witts, Inc., 2006. All rights reserved. 18 Copyright © Technical Witts, Inc., 2006. All rights reserved. 19