How do I determine DC-DC converter start-up rise time to meet a microcontroller power rising slope spec?

Question

Below, I have waveforms showing power-up on the 3.3 V rail of two otherwise-identical microcontroller boards, the only difference being two different 3.3 V DC-DC switching regulators. In the blue, the V7803-1000 and in the yellow, the VX7803-1000. According to their respective datasheets, both regulators have nearly-identical specs, but the VX takes significantly more time to stabilize after power-up, and in fact exceeds the rising slope requirement of the microcontroller (min 1.9 V/ms), badly enough to cause boot failure.

More data points: I tested the K7803-1000R3 which resulted in a curve almost identical to the yellow curve below. After a few hours reviewing other regulators, I only found two companies that list rise time in a datasheet: TI's TPSM84203 intentionally provides a 5-ms soft start (0.66 V/ms), and the new-to-me XP Power sells the JCA0624S03 which has comes closest at 0.94 V/ms.

What causes the significant difference between rising slope on two apparently "identical-spec" DC-DC converters? Is there any way, short of purchasing and physically testing potentially hundreds of equivalent converters, to determine which will output a rising slope that meets the spec, since few datasheets appear to provide information about start-up rise time? Is there a better way for a 3.3 V power rail to meet a minimum 1.9 V/ms rise time, that renders the choice of regulator moot?

Answer 1

A better approach might be to design your microcontroller circuit to accept any rise time, which might involve using an external supervisory reset chip.

Or if that's not possible use a power supply sequencing chip to switch the power after a worst-case delay time.

Answer 2

One option is to use a load switch that you turn on after the DC-DC output is stable.

Many DC-DC converter ICs will show a typical startup waveform on the datasheet, or you can see it in the documentation for their "demo" boards. Of course that's only valid for the test circuit and conditions given.

Others have adjustable soft-start times, so you could potentially pick a converter with that feature.

Another option is to download the simulation model provided by the manufacturer and see what the startup waveform looks like in simulation before committing to hardware.

Of course most of the above applies if you use a DC-DC converter/controller IC. If you want to use a module you may not have as many choices.

Since the CUI modules don't give schematics showing which controller/converter is used internally, it's difficult to say why the startup times are different between the two.

Answer 3

exceeds the rising slope requirement of the microcontroller (min 1.9 V/ms)

That "requirement" is in essence an admission of poor performance of the POR circuit. The MCU won't get damaged with lower rise time, and you can test that of course. Just keep it in reset with an external part until the power is good. If the SMPS has a power-good output, so much the better :)

Answer 4

Typically, microcontroller power-on reset circuits are designed to hold the CPU in reset for a certain amount of time after the input voltage reaches a certain threshold. Often, both delays and logic thresholds, as well as the minimum voltage at which a controller can operate reliably, may be affected by process variations, but it may be possible for a manufacturer to influence one in response to changes in the other. If the spec says 1.9V/ms, that means that the startup delay, reset threshold, and minimum operating voltage will be such that multiplying the startup delay by 1.9V/ms and adding the start-timing-the-reset threshold will yield a voltage which reaches the minimum required for reliable operation.

On many parts, the power supply voltage may vary harmlessly in arbitrary fashion within the range specified by absolute maximum ratings, while the reset pin is asserted, provided that the voltage is within the valid operating region when the reset pin is released, and remains there unless/until the pin is reasserted or a situation arises where nothing a chip might do would be considered unacceptable (e.g. because the device is being powered down, and nobody would care if the device turns on lights, beeps, etc. for a brief moment).

If the supply voltage might take awhile to stabilize when powering on a device, some means external to the microcontroller will likely be needed to ensure the device doesn't start until the supply is stable, but adding such means would avoid the need to clean up the startup voltage ramp.