I do get many questions about how to know if the power supply is stable, reliable and meets the specification of your application.

You can do a lot of math to proof it your power supply is meeting everything you specified but to be absolutely sure about everything there is just one way to proof and that is testing. There is not just one test to do, there are many and here is a list of everything you should consider when doing power supplies (for analog and digital power supplies its the same)

Start Up Behavior

1. No Load: Here you test how the power supply behaves when powering up with literally no load except the output capacitors it has. The result should be a calm but fast startup with no or little overshoot. If the overshoot is too high it can damage your circuit after the power supply. Help in this case is to slow down the startup by either a soft start condition (analog its a cap on a certain IC pin) or limiting the control loop gain but that is no thing I can recommend except you’re doing battery charging or LED lighting with no PWM dimming.
2. Full Load: When starting up at full load the power supply has to deliver more than nominal load at the startup because it has to charge all the output capacitors also. Things that can happen here is that the power supply runs into an overload condition. For this case it helps to implement a primary or secondary max current that the power supply can deliver. Analog IC’s often have a certain pin to set this value.

UVLO a.k.a. Under Voltage Lockout

• This very simple test is to ensure that the power supply switches of when the input voltage drops under a certain voltage level. This is important because the power supply tries to keep the output voltage stable and for this it will increase the current through the switches and the magnetics till they get a thermal overload and get damaged.

Load Regulation

• This test needs a little bit more effort. A thing you will need is a load you can adjust reliably and repeatably. This test should prove if your control loop is able to keep the output voltage at the desired level over the complete load range. This means literally from infinity Ohms to the resistance value that represents the load. This tests your control loops gain. Due to the difference in the output load the system you want to control changes. An effect you can have is that with nominal load the converter regulates perfect but with decreasing load the converter starts to fade away the output voltage. This means you don’t have enough gain at low frequencies in your control circuit or the integral character of your control loop is gone and the regulation error is not as small as defined anymore. If you have peak current controlled analog converters like almost all of the Linear Tech‘s are you should not see such problems. If you have a digital system controlled with a PI controller you are going to see that and you have to make your control loop adaptive this means it needs to change parameters according to the load.

Line Regulation

• Line regulation means the capability of the controller to react onto changes of the input voltage. Ideally if the input voltage changes the converter is able to cover that and this change cannot be seen at the output. In practice it always be measurable because the loop gain of your control loop is not infinity. At this test you attach your power supply to your lab power supply and step through the input voltage range to see what happens. This test changes the duty cycle of the power supply and with this it changes its system parameters for your control loop. The effects you can see can look quite similar to the load regulation test. With peak current controlled systems there should not be a problem in your system but with digital systems you can see a change and you also need to adopt your parameters here – but in this case not maintaining the gain, more changing the timing constants.

Feedback Loop / Step Response

• A properly adjusted feedback loop is the spine of your power supply so this needs to be tested properly. In some (more simpler) cases a step response is enough to see if the feedback loop is set up properly. Step response means you have a defined minimum load like 5-10% and your nominal load and you switch between these loads hard with MOSFets for example. If you have overshoots in your step response it means the gain is high but the phase margin is too low and the system tends to oscillation. If the system takes a lot of time to recover it means your feedback bandwidth is too narrow.
  This step response test gives you a raw estimation of your feedback loop but no detailed information. The best way to test your power supply would be with a Bode100 from my friends of Omicron Labs. This system gives you a detailed view of gain and phase margin of your system. This enables the designer to know exactly where to change the system and where it is fine as it is. This is in any case a good investment for a power supply designer because it gives you a good insight in the control loop health of your power supply.

Temperature Test

• Temperature tests consume the most of the time during a power supply development process. Reason for this is that there are a lot of components built in your power supply that need time to settle their final temperature. A part that needs time to settle is the transformer. It takes often hours (depending on the size) for this part to settle to its final temperature.
  The testing itself happens in 3 steps.
  1. Testing with the IR – cam. This test reveals the hot spots on the board. It is important to paint the board matte black before the test otherwise there will be reflections and things that put errors in your measurements. ATTENTION: Dont use Graphite paint! Its conducting! Always use Acryl painting.
  2. Testing the hot spots with thermo couples. The IR-Cam test puts a certain error on your measurements and to overcome them there needs to be a second measurement with thermocouple measurements. This measurement is more accurate than the IR-Cam test.
  3. This part is just about calculations. You now know the case temperatures of your power supply or complete design now you can recalculate to the junction temperatures and see if you violate these temperatures.
• Important is to repeat this test with multiple temperatures because the temperature of some components may not behave linear over the temperature range.
• Transformers operate at 85-90°C the most efficient ways – don’t be too scared about this temperature.
• Pay attention to your electrolytic caps – the warmer they get they sooner they fail!

With this tests you should be able now to develop a reliable and secure power supply – so happy testing!