TPS61094EVM-066: The measured output ripple is larger than theoretical calculated value

Hi Ethan,

The major deviation comes from the difference between the marked capacitance and the effective capacitance. 

Consider the capacitance loss from the DC biasing and AC ripples, the total effective output capacitance is around 50% left, that is, 33μF, not 66μF.

And regarding the self-oscillating point, although close the the 1MHz, it's still bigger than 1MHz and the the capacitor model is still there.

Regards

Lei

Hi Lei

Thanks for the quick response! I agree about your opinion about the 50％ cap loss(when operating at DC & low frequency). But when comes to fsw frequency, which happens to be or close to the self-oscillation point of the cap, the reactance is quite small due to ESL and cap are cancelling each other's contribution to reactance, although it still could be considered as a cap model C' with Z=ESR-j/wC', the equivalent capacitance value C' is much more larger than the original value 33uF(when at the oscillation freq, the value is +infinity and cap is now a pure resistor).

So in my opinion, when cap operating near self-osc frequency, vout,ripple that contributed by cap is small due to the large equivalent cap value, and the deviation still exist. And I do think that choosing the self-osc frequency of output capacitance near the switching frequency of converter is a good thing for design, since the AC impedance is so small.

Correct me if I'm wrong, and thank you for your time.

Best Regards

Ethan

Hi Ethan,

Here is my understanding for your reference.

In all frequency ranges, the cap can be modeled as cap + resistor + inductor.

So even at the self-oscillation point, for a output capacitor, it's capacitance is still there. There will still be charge and discharge current on the cap, so is the ripple caused by the cap. 

Regards

Lei

Hi Lei,

Again I want to thank you for the response.

Here is my understanding of this question.

Surely the current flow through the cap cause charge & discharge ripple, the phase of ripple,c is 90° lagging the current --- Vripple,c=I(t)/jwC. But the current also flow through the serial parasitic inductor and cause ripple,L, which is 90° leading the phase of current --- Vripple,L=I(t)*jwL. The output ripple Vout,ripple=Vripple,c+Vripple,L+Vripple,R, the Vripple,c and Vripple,L are cancelling each other due to the 180° phase shift, and the summation of these two reaches to 0V when at self-oscillation frequency, only Vripple,R left in the equation.

Thank you for your support!

Best Regards,

Ethan

Hi Ethan,

Let me check and reply to you a little later.

Regards

Lei

Hi Ethan,

Thank you for your information.

I totally agree with you! You calculation seems the one what we're very familiar with. Well, the real Vout ripple should be much bigger even when the switching frequency is right the self-oscillation frequency. Just let me try to find out where is going "wrong". It's an interesting question. 

Regards

Lei

Hi Lei,

Glad you find this topic interesting. Looking forward to your reply!

Best Regards,

Ethan

Hi Ethan,

Let's further discuss early next week.

Regards

Lei

Hi Lei,

Totally agree with you on this. The impact caused by higher harmonic frequency is not considered in my previous analysis. I set up the equivalent output node circuit in PSpice, simply a current source to simulate the current flow to output, and three parallel C(model inside connected with a 3mOhm ESR and a ESL to set self-oscillation freq at 1MHz), then I ran the transient sim to get the Vout waveform.

When I,input is single-frequency sinewave, the value of output ripple is basically ESR*I(p-p), that's when C can be considered as a pure resistor.
When I,input is a triangle wave(that's a good approximation for Buck, but the current is more complicate than a triangle wave for Boost), the output ripple is much larger than sinewave scenario due to harmonic frequencies are introduced here, the impedance is now dominated by ESL, and spikes occurs(I think it's also caused by ESL).

Appreciate the support! Thanks for joining the discussion of this trivial question.

Best Regards,

Ethan

Hi Ethan,

Welcome!. Thanks for providing the simulation results.

I'm glad that we can make this interesting question clear together

Regards

Lei