The probability of failure of any complex system follows the so-called bathtub curve where the likelihood of a system problem is relatively high early in its life and diminishes as it passes through infancy. The likelihood of failure then continues at a low rate until the system approaches the end of its life expectancy and then the likelihood of failure increases. Still dealing in generalities: The aging of a system is accelerated in proportion to its operating temperature; the higher the temperature, the quicker the system ages.
The company I work for designs and manufactures several ranges of switch mode power supplies (SMPS) from 20W to 1KW. They typically have a universal AC input 120/240V and a 24VDC output, although there are some significant variations. The SMPS is a complex system and because of the competitive nature of the business and the criticality of their use it was decided to burn in all power supplies. I was not part of the decision to design our own burn-in chamber, but I presume that because we normally test a fairly large number of units and some can be quite bulky and there wasn’t a suitable commercial unit.
There have been three iterations of the chamber: the first when it was built, the second when we couldn’t find spares for our obsolete distributed I/O control system, and the third was when we moved location. In all we have invested many man-years of effort in the project.
Both chambers were built during building expansion or major renovation and are merely a sturdy insulated room (in fact the first room became the tornado shelter for the new building residents when we moved) with suitable ventilation plus a multi-kilowatt electricity supply. The rest gets added. The component parts of the chamber are part commercial products, part hand assembled prototype, and part custom designed — and the system is designed to burn-in up to 200 power supplies simultaneously.
All of the SMPSs we make mount on TS35 rail. Most also have pluggable connectors as well and so the racks were fairly easy to make (as can be seen in Figure 1). The power supply simply clips onto the rail and a quick connection is made using the pluggable connectors.
Figure 1. The racks are built around the TS35 rail and quick connect wires that are then connected to the load cables. A distribution panel is mounted on each rack to centralize the power source and protect the wiring.
Each SMPS is connected to an electronic load which is made up of a MOSFET as described in my earlier blog. Twelve MOSFETs and the associated power resistors are mounted on a large (550mm x 380mm x 15mm) heatsink. Several heatsinks can be slotted into a cart as you can see in Figure 2.
Figure 2a Heatsinks mounted in some racks.
Figure 2b. A close-up of one of the heatsink assemblies. The square object is the MOSFET and the gold devices are the power resistors connected to the MOSFET source. The op-amps are mounted on the PCB in the middle of the heatsink.
We used some hefty cables and connectors to interconnect the SMPSs to the loads. There can be quite a bit of current and so we were minimizing the self-heating and volt drop across the cable. The heat from the assemblies is used to heat the chamber and there are also auxiliary space heaters distributed around.
We used ADAM modules from Advantech for our distributed control, interconnected via Modbus over RS485. On an earlier project I had worked with these modules and I had developed the control in Visual Basic (this is documented in several Circuit Cellar articles Visual Basic 2005 and the Serial Port (December 2006) and Generic Modbus Simulator (March and April 2007)). Since I wasn’t going to do the development and because I wanted an easily understood version of Basic so that both technicians and newly graduated engineers could write and follow the programming, I opted to port the serial driver to BBC Basic and indeed, the whole control algorithm and user interface is now written in this language.
The program allows for setup of which model of SMPS is to be burned-in, and which rack bay is occupied. Parameters include voltage and current settings, time and calibration constants, and the burn-in temperature. The algorithm has to allow for the voltage drop across the cable since the analog to digital converter (ADC) Adam modules are mounted on the racks as you can see in Figure 3. With hindsight, it would have been better to place the ADCs on the racks. It also will supply a report and a count of units and the relevant statistics.
Figure 3. The ADAM control modules are mounted on the cart. Their design actually allows for them to be stacked allowing for smaller rail space, but slightly more complex wiring.
There are various I/O points that allow for the air extraction system to be activated as well as the space heaters in an attempt to keep the chamber temperature constant. It could also be used to follow a temperature profile, but we never pushed the development that far.
Our failure rate is pretty low and I sometimes wonder if it is at all necessary, but then we find a batch which fail and so it is worth it for the fact that we have fewer returns and good reputation in the marketplace.
Do you ever burn-in any of your projects? Do you use a commercial chamber or have you built your own?