4. CONDUCTION LOSSES - Purdue Engineering

1 54 4. CONDUCTION LOSSES In this chapter, CONDUCTION LOSSES for a power switching transistor and power diode are measured and characterized. A static procedure for measuring IGBT and power diode CONDUCTION LOSSES is set forth in section section describes the development of the proposed CONDUCTION loss model. CONDUCTION Loss Measurement The hardware configuration used to measure CONDUCTION LOSSES is set forth in Figure Fig. Hardware Test Configuration for IGBT Static LOSSES For the purposes of demonstration, the transistor and diode considered herein are the upper IGBT and diode of the U leg of a 30A, 600V device (Fuji Electric 7 MBR30SA060) [18] commonly used for inverter motor drives. The IGBT is turned on with the gate driver, and the current through the device is then adjusted through its full range.

2 At the same time, the corresponding voltage across the device is recorded. The power analyzer (Voltech PM3000A [19]) was utilized to measure the current through the device and the voltage across it. The product of the current through the device and the voltage across it 55yields the power loss ( CONDUCTION loss). The junction temperature can be approximated as cjjcjTPRT+= ( ) where jcRdenotes the junction-to-case thermal resistance, jPdenotes the device power , and cTdenotes the case temperature. The junction-to-case thermal resistance for the transistor and diode are obtained from the published data sheet [18]. These resistances are C/W and C/W respectively for the transistor and diode. CONDUCTION LOSSES are computed as a function of the current through the device and the junction temperature.

3 This experiment was conducted at different case temperatures using the proportional plus integral control discussed in Chapter 3. Details pertaining to the control and thermal model are included in Appendix A. The resulting IV characteristics for the IGBT are shown for different case temperatures in Table In Table , V denotes voltage, I denotes the current, and cTdenotes the case temperature. Figure displays plots of the IV characteristics for each case temperature used. Note that as the case temperature increases, the CONDUCTION LOSSES decrease. 56 Table Measured IGBT data CTco25= CTco35= CTco45= CTco55= CTco65= V (V) I (A) V (V) I (A) V (V) I (A) V (V) I (A) V (V) I (A) (V)Current (A) Tc=25Tc=35Tc=45Tc=55Tc=65 Fig.

4 Measured IV Characteristics for IGBT Identical procedures are followed for the diode. The hardware configuration used to measure the CONDUCTION LOSSES for the diode is set forth in Figure Fig. Hardware Test Configuration for Diode Static LOSSES The resulting IV characteristics for the diode at each temperature (case temperature) are shown below in Table Figure displays the IV characteristics for each case temperature used. As in the case of the IGBT, the LOSSES decrease with temperature. 58 Table Measured Diode Data CTco25= CTco35= CTco45= CTco55= CTco65= V (V) I (A) V (V) I (A) V (V) I (A) V (V) I (A) V (V) I (A) (V)Current (A) Tc=25Tc=35Tc=45Tc=55Tc=65 Fig.

5 Measured IV Characteristics for Diode CONDUCTION Loss Characterization In the previous section , data was collected for the purpose of measuring the CONDUCTION LOSSES of the power transistor, in this case an IGBT. For the purposes of mathematical modeling, the voltage drop was assumed to obey the relationship ==31,,,jctbjtjttcdjtjtiTav ( ) where ti is the current through the switch, tcdvis the voltage drop across the switch, jtTis the transistor junction temperature ( C), andta,tb, and tc are parameters of the model. In order to determine the model parameters, the function ),,(101),,(6ttttttcbaecbaf+= ( ) was maximized where == =KkmktjcmktbkjtjttttviTacbaejtjt12,31,,, ,,),,( ( ) 60where mktv,and mkti,denote the thk measured voltage and current points respectively for the transistor, and Kis the total number of measurement points described in section kjtT,is computed using ( ) and the measured voltage, current, and case temperature data described in the previous section .

6 In order to solve this optimization problem, a genetic algorithm is used. The search space for each parameter to be identified is shown in Table Table Search Space for IGBT parameters Gene Minimum Value Maximum Value Unit 1,ta -3 3 bcCAo1 1,tb -3 3 1,tc -3 3 2,ta -3 3 bcCAo1 2,tb -3 3 2,tc -3 3 3,ta -3 3 bcCAo1 3,tb -3 3 3,tc -3 3 Figure depicts statistical information on the evolution of the optimization. The lowest trace depicts the Best {B (in blue)}, the Median {Md (in green)}, and the Mean {Mn (in red)} fitness values of the population versus generation. The upper trace illustrates a histogram of parameter values over the population. NumberNormalized 104050100150200f:B(b) Md(g) Mn(r)Generation Fig - Genetic Algorithm Evolution for IGBT Parameters The resulting model parameters are displayed in Table Table IGBT Parameters Parameter Value Unit 1,ta bcCAo1 1,tb x 10-1 1,tc x 10-2 2,ta bcCAo1 2,tb x 10-1 2,tc x 10-1 3,ta x 10-2 bcCAo1 3,tb x 10-2 3,tc x 10-1 62 The measured IV characteristic curve as well as the fitted curve ( ) is depicted in Figure As can be seen in Figure , the fitted curve closely resembles the measured characteristic.

7 (A)Junction Temperature (oC)Voltage (V) Fig. IGBT Static Loss Measurements and Fit Curve For comparison, the percent error between the measured curve and fitted curve is plotted in Figure The maximum percentage error was while the average percentage error was 630102030050100150-4-3-2-1012 Current (A)Junction Temperature (oC)Percent Error (% ) Fig Percent Error between Measured and Fitted IGBT Characteristics Similar procedures are followed to determine the CONDUCTION LOSSES of the diode. The voltage drop across the diode is assumed to obey the same relationship as the one ( ) used for the transistor. In particular, the diode loss is assumed to obey ==31,,,jcdbjdjddcdjdjdiTav ( ) where di is the current through the diode, dcdvis the voltage drop across the diode, jdTis the diode junction temperature ( C), andda,db, and dc are parameters of the model.

8 A fitness function of the form shown in ( ) is also used for the diode. Again, a genetic algorithm is used to solve the optimization problem. A search space identical to the one used for the transistor is used for the diode. Figure depicts statistical information on the evolution of the optimization. The lowest trace depicts the Best {B (in blue)}, the Median {Md (in green)}, and the Mean {Mn (in red)} fitness values of the population versus generation. The upper trace illustrates a histogram of parameter values over the population. NumberNormalized 1040100200300400f:B(b) Md(g) Mn(r)Generation Fig. Genetic Algorithm Evolution for Diode Parameters The resulting parameters determined by the genetic algorithm are displayed in Table Table - Diode Parameters Parameter Value Unit 1,da x 10-2 bcCAo1 1,db x 10-2 1,dc 2,da bcCAo1 2,db 2,dc 3,da bcCAo1 3,db x 10-1 3,dc x 10-1 65 The fitted IV characteristic curve as well as the measured curve is depicted in Figure As can be seen in Figure , the fitted curve closely resembles the measured characteristic.

9 (A)Junction Temperature (oC)Voltage (V) Fig. Diode Static Loss Measurements and Fit Curve The percent error between the fitted curve and measured curve is plotted in Figure The maximum percentage error was while the average percentage error was (A)Junction Temperature (oC)Percent Error (% ) Fig Percent Error between Measured and Fitted Diode Characteristics Note that in retrospect it would have been better to represent CONDUCTION LOSSES for the transistor and diode using a different form. This is because at 0 C, the form used predicts that the CONDUCTION LOSSES of the transistor and diode are infinity and zero respectively, which would not be the case. The utilization of a revised form is left as future work.