Help with bypass/decoupling to remove induced Vin ripple with DC/DC boost controller

[k113]

Need Some help with bypass/decoupling capacitor selection

I'm running into an issue when using a Lm3224 DC/DC boost controller.

When I'm running it, I get a 250mV ripple on my 3V Vin power lines which negatively effects the rest of the board.

The controller itself is a PWM style running at 1.25Mhz.
However I'm activating the controller on a PWM signal where I have a duty cycle of 25-90% at 120hz.

The output wave forms look good, and there is no effective slow start on the controller to mitigate this issue.

How do I effectively remove the ripple on the power lines?

I have been given advice to raise the input capacitor to over 100uF, but before I order parts I'm confused on how high I need to go, whether I should go with a single vs multiple caps, ceramic vs tantalum vs electrolytic, low esr/low esl... etc.

My suspicion is the overall ripple is following the 1.25Mhz cycle, but is being smoothed out by the 20uF input capacitor, but isn't completely removing the dip.

Thanks

 

[RCinFLA]

Two possible reasons for output ripple.

1) Rs of output filter cap. The charging current creates large spikes in current. The output ripple will be the peak coil current times the output cap's Rs. Any old cap will likely have a high Rs at 1.25 MHz.

2) At light loading many switchers go to pulse skipping. This happens when duty cycle gets to a low value. It depends on switcher and frequency but generally when duty cycle gets less the 5%. Output filter cap must be large enough to ride across the longest skip time.

1.25 MHz switcher is pretty high speed. You may have to go to a PI filter on output with addition of L-C filter on output.

 

[k113]

The output side actually looks good.

However, It is the input side (Vcc) that is having the issues with ripple when the controller is active.

Two possible reasons for output ripple.

1) Rs of output filter cap. The charging current creates large spikes in current. The output ripple will be the peak coil current times the output cap's Rs. Any old cap will likely have a high Rs at 1.25 MHz.

2) At light loading many switchers go to pulse skipping. This happens when duty cycle gets to a low value. It depends on switcher and frequency but generally when duty cycle gets less the 5%. Output filter cap must be large enough to ride across the longest skip time.

1.25 MHz switcher is pretty high speed. You may have to go to a PI filter on output with addition of L-C filter on output.

 

[RCinFLA]

Okay so your power source for switcher is having problem with boost switcher peak current.

You can try adding some low Rs capacitance on input.

 

[k113]

The desktop power supply should be more than sufficient for the load, however it dips.
Any suggestions on sizing the caps and type?

 

[ants]

The advice you were given would probably be the best, increase input cap to over 100uF. Vishay sprague tantalum surface mount caps are amoungst the best, they have a low esr and are ideal for switched circuits.

 

[FvM]

Unfortunately, I didn't clearly hear from your verbose description, if you're actually seeing a 1.25 MHz high frequency or a 120 Hz load modulation ripple. Obviously, the means to reduce it would be quite different. In the first case, multiple ceramic capacitors and a series inductor can effectively cancel the ripple. A low frequency ripple respectively "dips" on load current changes are mainly caused by a slow primary supply voltage regulator. The time constant of the voltage transient would determine the dimensioning of additional filters.