Dimensiones Acciones
Objetivo 3: Promover acciones con base en la modalidad mixta como una alternativa innovadora en los procesos de Enseñanza y Aprendizaje de los estudios de Postgrado en el IPMJMSM
The aim of the following experiments is to verify the overall feasibility of the proposed approach and the behavior of the diagnostic parameters. The test is carried out using one open Semikron module (absent of silicon gel). The gate-emitter voltage is frequently checked. In order to determine that the measured voltage rise is caused by the deteriorations of bond wire connections instead of the chip temperature, the chip temperature is indirectly measured using VCE(l)(Tj) as TSEP as discussed in Chapter 4.
The measurement circuit includes two current sources. Isense forces a 100mA current through the chip to determine the junction temperature. A second high current source Itest is also integrated in the inverter and a designated pattern of current pulse trains is injected into the IGBT/diode. The current pulses are controlled by an auxiliary IGBT switch. In order to measure the voltage drop across the bond wire connections a pulse train of 2.5kHz is injected. The pulse train is shown in Figure 6.6.
344us
Figure 6.6 Diagram of pulse trains for bond wire lift-off monitoring
The voltage VCE(h) is measured directly at the end of the high current pulse (Itest+Isense in Figure 6.6). Multiple measurements (3 samples) are made for averaging the readings. In order to link the voltage measurement VCE(h) with the junction temperature, a second measurement takes place shortly after the 344us pulse. The delay time is chosen to be short enough to avoid significant changes in the junction temperature between two sets of measurements. Multiple measurements (3 samples) of voltage drop VCE(l) are recorded and converted into temperature T using the TSEP calibration curve. Figure 6.7
CHAPTER 6 EXPERIMENTAL RESULTS
shows the experimental waveforms. Due to the delay caused by the test current to settle down to the desired current Itest, a pulse train up to 16ms is injected and it can be concluded that a total of 4ms is sufficient for the data logging per device.
Figure 6.7 Voltage and current measurements during pulse train monitoring
The process takes place in the microprocessor and the flowchart is shown in Figure 6.8.
Having associated the junction temperature Tj with the measured voltage VCE(h), VCE(h)(Tj) is now compared with the baseline. So long as the voltage difference between the baseline and the measured voltageVCE(h) is less then 5% at the measured junction temperature no warning is sent to the driver. Once the bond wire health has received the okay-status the selector chooses the next IGBT or diode for measurement.
In the experiment, bond wire lift-off is emulated by individually cutting the bond wires.
Figure 6.9a and 6.9b show, respectively, the drift of the IGBT collector-to-emitter on-state voltage and the diode forward voltage with junction temperature at different stages of the bond wire health. One can notice that the measured voltages have increased by approximately 12mV for the IGBT and 7mV for the diode with one bond wire cut. It must be noticed that the peak-to-peak accuracy of the in-situ circuit for the voltage measurement is less than 1.2mV which is sufficient to detect this VCE voltage rise.
Indeed, the increased voltage is enlarged by a further amount with a second bond wire cut. Both figures show that the state of bond wire health for each device can be determined by evaluating its voltage increase.
CHAPTER 6 EXPERIMENTAL RESULTS
Skip first 5 pulses
return
Measure VCE(l) during low current
Determine mean value of VCE(l)
Convert mean VCE(l) to mean Tj
Generate a array of {VCE(h), Tj, IC}
Start (car stop)
Store {VCE(h),Tj} at Itest in a look-up table Determine mean values of VCE(h) and IC
Measure VCE(h) and IC during high current
Figure 6.8 Flowchart of the bond wire lift-off monitoring process
(a) (b)
CHAPTER 6 EXPERIMENTAL RESULTS
The experimental tests brought forward a design specific impact on the in-situ measurement. The SKM 50GB063D power module has one die per IGBT switch with two aluminium metallization pads. Each pad is connected with the DCB using three bond wires (Figure 6.10). Thus each IGBT has a total of 6 bond wires to handle a rated current of up to 75A (Tcase=25°C) per die. Consequently each bond wire carries a maximum current of about 12A.
Figure 6.10 Bond wires of IGBTs and diodes
When a bond wire fails, the current density in the die does not change dramatically since the current is still distributed between both of the metallization pads. This will lead to a gradual increase in VCE(on) mainly due to the increased voltage drop across the bond wires. However, when one pad loses all its electric connection, the current density in the other pad increases twofold and this leads to a significant jump in the VCE(on) measurement as shown in Figure 6.11 (see at 3 lift-off bond wires in Figure 6.11).
Figure 6.11 Relative errors of VCE(n) and VF(n) as a function of number of lift-off bond wires
CHAPTER 6 EXPERIMENTAL RESULTS
The y-axis in Figure 6.11 shows the relative error for the IGBT and diode. The relative error is given by Equations 6.6 and 6.7:
bond wire lift-off for the IGBT and diode, respectively. VCE(n) and VF(n) describe the on-state voltage and the forward voltage of IGBT and diode, respectively, at the number of lifted bond wires n. Figure 6.11 shows also that the diode does not show a “jump” in the voltage as seen in the IGBT. That is because each diode die has only one pad with six bond wires. Therefore, as the number of lift-off bond wires increases, the current density in the die does not change dramatically and the increased voltage drop across the bond wires corresponding to the lift-off bond wires results in a gradual increase in VF.