It’s the mid-1990s, and I’m a design engineer for a company that designs and manufactures custom measurement systems and high-power electronics. Our customers range from operators of experimental fusion reactors to electric utilities.
One day my boss told me to join him in the conference room to meet some people from a transit car manufacturer. He says they are having trouble with the product from one of their vendors and they are asking for our help.
We meet the project manager of the metro car manufacturer and an engineer from one of their suppliers. The product is a new red tail light installed on subway cars. Contrary to its name, the rear light is installed at both ends of the cars because most subway cars run alternately in both directions. The problem is that the light works fine several times and then fails.
The light assembly contains both the LEDs and the electronic power supply to power it. They show us the schematic, parts of it are here:
Part of the original diagram, drawn from memory.
Input power is 74 volts DC and comes from a battery. The unit is a non-isolated DC-to-DC step-down flyback converter that powers the on-board LEDs. It uses a controller IC, U1, with a separate power MOSFET, the MOSFET is not shown in my schematic. The MOSFET switches the primary of the transformer-inductor, T1. There are two secondary windings on the transformer. The main winding goes to a rectifier, D2, and a filter, not shown, then to the LEDs, not shown. The auxiliary winding is rectified by D2 and filtered by C1. It then powers the controller IC. The main load is a series-parallel arrangement of red LEDs consuming approximately 15 watts.
The controller cannot be powered directly from battery power because the voltage is higher than the maximum power IC value. Also, since it sends a gate signal to the MOSFET, it cannot be powered more than 20V or it will destroy the MOSFET.
The taillight designers estimated that if they connected the Vcc pin of U1 to the power supply via a series resistor shunt Zener circuit, the power lost to power the IC would be over 2 watts and reduce the overall efficiency by about 15%.
So they decided to use a start-up circuit that would momentarily power the IC and that once the converter was running, the IC would be powered from the output of the converter. They must also cut off this starter circuit. They used a series resistor, R2, connected to the collector of a small high voltage BJT signal, Q1, with sufficient voltage and current. The emitter of transistor Q1 is connected to the IC’s power pin array, as is the second secondary winding that supplies power. The base of Q1 is connected to a Zener, D1, biased through a high value resistor, R1, from the input power. After the converter is started and running, the auxiliary winding voltage is rectified by D2 and filtered by C1. The Vcc rises above the D1 voltage and skews the base-emitter junction of Q1. This cuts the current and reduces the losses of the starting circuit to almost zero. All in all, half a dozen small signal components cost a lot less than one or two high-powered components, so the solution they chose for the starter circuit makes sense.
The problem is that Q1 repeatedly fails. The converter starts quickly and the designer assures me that none of the parts overheat.
I immediately see the source of the problem, but I am silent. If I tell them the solution now, they will pack up and not pay us for our expertise, arguing that the time we spent is not worth the paperwork to pay us. My boss says we will work on it urgently. They leave us a unit of work and the diagrams.
Once they leave the building, I talk to my boss. The start circuit design is a series of compromises: select a Zener value higher than the minimum operating voltage of the IC, calculate the number of turns on the transformer to produce a voltage higher than the Zener but not too high, otherwise it will damage the MOSFETs. The problem is added to the fact that the converter regulates the main output current and not the auxiliary output voltage and must operate over the entire range from the minimum voltage to the maximum voltage of the battery.
If the operating voltage of the auxiliary winding is much higher than the Zener voltage, the reverse-biased base-emitter junction of Q1 may fail and destroy the transistor. The maximum Vebo is typically 6.0V for the MPSA42, a transistor similar to Q1, and is typical of many BJTs . If the reverse current level of the emitter-base junction is low, the junction behaves like a Zener diode. With higher current and longer duration, beta drops and noise increases when resuming normal operation. If the reverse current is excessive, the transistor fails  and that is why Q1 fails in this product.
I tell my boss the solution is simple, add a series diode, D3, between the emitter of Q1 and the Vcc pin of the controller, see the modified schematic:
Part of the modified schematic, taken from memory.
He agrees with me. I change the unit. I have tested it several times with on and off cycles, over the entire battery voltage range and the device works perfectly.
Late the next day, my boss calls our client to tell him we’ve identified the problem and have a simple, inexpensive solution for him. After a week, our client informs us that the manufacturer has confirmed our solution to the problem and both are very satisfied with our work.
The lessons learned are:
1- beware of reverse biasing the base-emitter junction of a BJT, 2- implement rigorous design reviews and use a checklist, the list here is a good start 3- think before giving answers to customers, there is always money at stake, your boss pays your salary with the customer’s money, do not give it away, and above all 4- never let the customer pass for a fool.
 ONSEMI, technical sheet MPSA42: NPN small-signal bipolar transistorFebruary 2013 − Rev. 8. https://www.onsemi.com/pdf/datasheet/mpsa42-d.pdf [Page retrieved 2022-06-21].
 Pease, Robert A. Troubleshooting Analog Circuits. Butterworth-Heinemann, 1991. Page 77.
 Motchenbacher CD and JA Connelly. Design of low noise electronic systems. John Wiley & Sons, Inc. 1993. page 133.
 Wallace, Hank. Electronic Design Checklist, http://www.jldsystems.com/pdf/Electronics%20Design%20Checklist.pdf [Page retrieved 2022-06-19].
Daniel Dufresne is a retired engineer and has worked in telecommunications, mass transit, consumer products and high power electronics design. He was also a teacher at Cégep de Saint-Laurent. He received his bachelor’s degree from École Polytechnique de Montréal. He lives in Montreal, Canada and still works on electronic projects and repairs electronic test equipment.