TPS53513: noise suppression methods

kazuhiro.ozawa 

kazuhiro.ozawa said:

Is the only way to reduce this noise is to filter the input stage of the TPS53513 or increase the switching frequency? Or should I use another IC? 

Buck switching regulators operate with discontinuous input current.  They draw the full load current from the input during the On-time, then no current from the input during the Off-time.  The On-time is approximately VOUT/VIN, so that the input current is approximately IOUT x VOUT/VIN.  That way the input power (VIN x IIN) is approximately equal to the output power (VOUT x IOUT)  This discontinuous input current causes the input to have some switching noise at the switching frequency.

This is generally reduced by adding low ESR / low ESL ceramic bypassing capacitance from the input voltage to the converter, to it's power ground.  Generally, input ripple can be reduced to a few hundred millivolts by just ceramic bypass capacitance.  If further input voltage ripple reduction is required, it is common for designers to include a ferrite bead in series with the input.  The bead should be rated for at least 1.3x IOUT_max x (VOUT/VIN_min) to ensure the input current, including efficiency, does not exceed the bead's rating.  Adding some additional higher ESR electrolytic capacitance after the ferrite bead and before the ceramic capacitance at the input to the BUCK converter can help minimize the possibility of an L-C resonance by using the ESR as a damping resistor.

kazuhiro.ozawa said:

It generates 0.9V from 12V, but 12V has switching noise of about 100kHz.

100kHz is low for the switching frequency of the TPS53513 unless it is operating at light load with autoskip enabled.  If this is at higher load or if Forced Continuous Conduction Mode (FCCM) is selected, then I would recommend checking the switching frequency at the SW node of the IC to see if it is switching at 100kHz or if the 100kHz input ripple might be a sign of a different issue.

Possibly a resonance frequency between the input capacitors and some source inductance, insufficient output capacitance, or excessive power-path propagation delay between the inductor and the senses output voltage.

kazuhiro.ozawa said:

In terms of switching noise, is it more advantageous to generate from 5V than to generate 0.9V from 12V?

The longer on-time generated by the conversion from 5V to 0.9V (18% to Tsw) verse 12V to 0.9V (7.5% of Tsw) will generally cause a 5V input to have more ripple current and need more input bypass capacitance to reduce the input ripple voltage, though it may be easier to achieve the required input bypass capacitance for a 5V input than a 12V input.  Due to the DC bias effect on ceramic capacitors reducing their effective capacitance, it is generally recommended that input capacitors be rated for at least 2x the input voltage, so a minimum of 10V for a 5V input and 25V for a 12V input.

kazuhiro.ozawa said:

In terms of line regulation and load regulation, is it more advantageous to generate from 5V than from 12V? 

The TPS53513 uses Ti's D-CAP3 constant on-time control, which has very good line and load regulation from either 5V or 12V.  The efficiency will be slightly higher from a 5V input, but the input source current will be much higher for the same output power.

If you try the suggestions above, and it does not help you reduce the ripple, or you have SW waveforms to share for additional assistance, please let us know.

Hi, Peter.

Thanks for your quick response. Also, thanks for the polite explanation.

I understanded this matter. However, I don't have the waveform you say because I don't have a board at the design stage yet.

I'm sorry.

Good regard.