SMPS SEPIC Compensation design: R vs C ratio

I'm designing a SEPIC converter. I have everything done, power components picked, board laid out, but I'm stuck on the compensation network.

The SC4501 (Warning! The datasheet ONLY covers BOOST, not SEPIC) says that it needs an R and C in series to form the network. These components are adjusted to form a zero at the same frequency as the output (dominant?) pole to cancel it out.

The output pole frequency is at F=1/(2*PI*R_Load*C_Output), which is calculated to be at 181Hz. (20 Ohm load, 44uF output capacitance)

Therefore, the R and C in the compensation network should have a corner frequency of 181Hz. For example, I could pick a 10nF capacitor and a 88K resistor

Assuming I have done all the previous steps correctly, how does the ratio of R to C matter? I could increase C and decrease R, and the network would have the same corner freq, but the datasheet says the gain would change. How do I pick my components?



I'm way out of my depth with all this analog stuff :???:

How hard did you look?



They suggest 5 ~10 mΩ non-inductive caps

Both Sepic and Cuk reduce ripple by design so lower C values are needed.

I think he's talking about the error amp compensation network, not the output bypass capacitor.

To determine the correct values for a compensation network, you should actually analyze the open loop transfer function of the converter. Choose a crossover frequency, and a phase margin. Then that will tell you what phase and gain the compensation network will need at the crossover frequency. Design should be straightforward from there on.

From this suggested working model design , I would simply follow their lead and choose a single pole RC feedback at just above the operating frequency so phase shift is < 45deg and adequate harmonic attenuation with the feedback voltage bias ratio,N set to Vout= N * 1.25V They run at 1.5Mhz here.

analyze the open loop transfer function of the converter

Click to expand...

Are you talking about building the circuit and testing the gain and phase in real life? Or do you mean finding the location of all the poles and zeros and plotting them?

Choose a crossover frequency, and a phase margin

Click to expand...

This paper says that to choose the crossover frequency, find the RHPZ and the resonant frequency of the output cap and L2, take the lowest one, and divide by six.

The RHPZ freq is 58.2 kHz.
The resonant freq is 5.11 kHz

So my crossover frequency is 850Hz

choose a single pole RC feedback at just above the operating frequency so phase shift is < 45deg

Click to expand...

By operating frequency, do you mean the output (dominant) pole? How do you verify the correct phase shift? I have no idea, since all the papers I've read seem to skip phase entirely

pia32 said:

Are you talking about building the circuit and testing the gain and phase in real life? Or do you mean finding the location of all the poles and zeros and plotting them?

Click to expand...

Either will work, assuming your circuit model is accurate. And you need to find the overall gain of the transfer functions, not just the pole and zero locations.

If you go the model calculation route, make sure you keep in mind that the transfer functions vary significantly based on whether you are operating in DCM or CCM.

This paper says that to choose the crossover frequency, find the RHPZ and the resonant frequency of the output cap and L2, take the lowest one, and divide by six.

The RHPZ freq is 58.2 kHz.
The resonant freq is 5.11 kHz

So my crossover frequency is 850Hz

Click to expand...

A crossover frequency well below the RHPZ frequency is a good idea, but it's often feasible to have it be well above the resonant frequency of the LC, especially when using a current mode controller like this. So if you want you can aim for a much higher bandwidth than that.

Does a 10kHz crossover sound reasonable? Once I have these numbers, how do I size my components? I haven't been able to find a formula or anything.

I feel like I'm missing something...

my friend i am cognisant that you do for 7W .

so why you not just do DCM sepic, and then you set your creossover frequency with much more freedom, because in dcm, there no rhpz.

also, you should use coupled inductor, because you get less LC resonantions at switch on or transient load change.

my friend i am cognisant that you do for 7W

Click to expand...

Yes, I forgot to mention it earlier, but I am aiming for 12V at 600mA, or 7.2W

so why you not just do DCM sepic, and then you set your creossover frequency with much more freedom, because in dcm, there no rhpz.

Click to expand...

The thing that I'm designing this for doesn't have very tight requirements for transient response, but if I could improve it, its worth investigating. How do you make a SEPIC that goes into DCM more? Reduce inductor size? Aren't there significant disadvantages, like reduced efficiency? If I were to do this, is there any different way that I have to design the compensation network?

As you guys have probably guessed, I have no formal training in electronics at all

pia32 said:

Does a 10kHz crossover sound reasonable? Once I have these numbers, how do I size my components? I haven't been able to find a formula or anything.

I feel like I'm missing something...

Click to expand...

10KHz should be feasible. For a good beginner's guide to feedback compensation, read this app note: https://www.venable.biz/tp-03.pdf
If you find that helpful, you should read all of venable's literature. It's all quite helpful.

pia32 said:

How do you make a SEPIC that goes into DCM more? Reduce inductor size?

Click to expand...

Just increase the ripple current amplitude to be greater than the max average inductor current (in both inductors). Can be done through decreasing frequency or decreasing inductance.

Aren't there significant disadvantages, like reduced efficiency?

Click to expand...

Maybe, it depends on a lot of variables. DCM can actually be beneficial in low power supplies, but at 7W, CCM should give better efficiency.

If I were to do this, is there any different way that I have to design the compensation network?

Click to expand...

You can use a similar procedure as in the above application note for designing with DCM converters, but the results will differ somewhat, since a DCM converter has a somewhat different open loop transfer function.

you will find that a dcm sepic is INTRINSICALLY STABLE........it cannot go unstable because its phase cannot go above 180 degrees.

but it can ring like mad if you dont watch the phase margin.............7W is nothing, just go DCM.

ccm SEPIC is known for its difficult feedback loop, i think its true to say that sepic has never been fully analysed mathematically in its feedback loop.

...and remember coupled inductor, otherwise your L's and C will be resonating like billy-O