« On Semiconductor SMPS Seminar (Ray Ridley) | Main | DCCM Flyback Stability »

May 28, 2001

Power Supply Turn-on Time

Question: How do you calculate the turn-on time of a power supply?

Original Question: I have not been able to determine a general figure for the turn-on time (time between the power on to valid output) of a linear power supply. Is it possible for you to help me with that, as accurate as possible. Email from S. R., 5/28/2001.

Answer: This has a lot of variables. For example, the output will not come up before the input so that everything from the source, circuit breakers, fuses, transformer rectifier set, input filter, etc. may be involved.

If you assume the input voltage is present when you turn on the regulator, there are also many variables.

For example, current limiting, type of current limiting (foldback, etc.) output capacitance, output load including capacitance, soft start circuits, the frequency-shaping circuits in the control loop, etc.

With this many variables, it is best to measure, and then figure out where the delays are coming from if you need to control them.

For example, assume that the power supply current limits at 1 A and there is a 1000 uF output capacitance. Then at no load, the charging time of the output capacitor to 5 V would be T=C*V/I = 5 ms. If there is a load to steal charging current the time would be longer.

However there might be an internal 1 uF control-loop capacitor that needs to be charged to 6 V from a 1 mA source. This might limit the charge time to T=C*V/I = 6 ms independent of load.

The bottom line? You either have to measure it, model it in Spice or make a mathematical model (where the results from both are only as good as your model), or have deep knowledge of the circuit details and do lots of calculations. Then you can never be sure until you validate with measurements.

This theme of measurements and analysis (and measurements and simulation) are themes I will talk about many places on this website -- all leading to the same conclusion. Analysis and simulation are never enough -- you have to measure. And to measure the correct things, you must analyze the problem and needed measurements first.

Posted by smpstech at May 28, 2001 12:00 AM