Buck converter feedforward capacitor

[theleakydiode]

Hi,

What is the purpose of a feedforward capacitor in a buck converter across the upper feedback resistor? and what does it mean for the output of the converter?

At the moment I am not using one and my LM2576 adjustable supply I built seems to be working correctly, but I am not sure of what value to use and how to measure if it is actually doing anything.

Here is the schematic I am following:



Thanks.

[tom66]

It feeds some ac ripple from the output into the amplifier. Changes the gain and phase margins, slows the transient response. Can improve stability.

[Phoenix]

slows the transient response. Can improve stability.

I'm pretty sure it has the opposite effect, as in it will improve transient response and decrease stability. In control theory it would create a derivative feedback term.

[theleakydiode]

Ok thanks for the explanation. What areas of the buck converter should I be probing when trying out different capacitor values?

This is my feedback pin waveform with no feed forward capacitor, does it look like it needs one adding?


200mV/div 5us/div

When I add a 2.2nF across the upper resistor those spikes diminish.

It feeds some ac ripple from the output into the amplifier. Changes the gain and phase margins, slows the transient response. Can improve stability.

Can these be measured with just an oscilloscope?

Thanks.

[Alex30]

I don't really have enough expertise to explain this very well but if you look at the datasheet
http://www.onsemi.com/pub_link/Collateral/LM2576-D.PDF

Check figure 25 and the blurb around it as it discusses this

[Richard Head]

It all starts with analysing the circuit and predicting the control - output transfer function. From this you can determine how to counter the gain and phase so that you cross over the 0dB line at a  -20dB/decade slope with 60 degrees of phase to spare.
To test you have to place a dynamic load on the output and switch between 10% and 90% load. Using a scope (AC coupled) you monitor the transient response of the output voltage.
This will tell you whether the control loop is underdamped or overdamped. Standard control theory applies.
Try to aim for a phase margin of about 60 degrees. Gain crossover frequency can be about 5-10% of switching frequency.

Dick

[Alex30]

EDIT:
Sorry I thought you were talking about the output cap.

[DanielS]

slows the transient response. Can improve stability.

I'm pretty sure it has the opposite effect, as in it will improve transient response and decrease stability. In control theory it would create a derivative feedback term.

In a linear application, adding feed-forward/difference-terms in the loop adds noise and can indeed be problematic.

In a PWM control loop though, having some of the output voltage rise in the feedback loop when the resistive feedback term is just about to pass reference voltage allows the error amplifier to end the on-time before overshoot becomes significant. The feedback network has all the off-time to settle in its new state before the next on-time, which makes the extra difference noise far less bothersome than it would be in a pure-analog/linear circuit.

[theleakydiode]

I don't really have enough expertise to explain this very well but if you look at the datasheet
http://www.onsemi.com/pub_link/Collateral/LM2576-D.PDF

Check figure 25 and the blurb around it as it discusses this

I am using a low ESR (38 milliohm) 1000uf 25v capacitor and a 220uh toroidal inductor that is good for 10 amps (all I had), also I have made sure to keep the layout tight and was careful about the current paths.

All of the input and output sections share a star ground at the IC's ground pin where there is another low esr input capacitor in parallel with a .1uf film cap for bypassing.

It all starts with analysing the circuit and predicting the control - output transfer function. From this you can determine how to counter the gain and phase so that you cross over the 0dB line at a  -20dB/decade slope with 60 degrees of phase to spare.
To test you have to place a dynamic load on the output and switch between 10% and 90% load. Using a scope (AC coupled) you monitor the transient response of the output voltage.
This will tell you whether the control loop is underdamped or overdamped. Standard control theory applies.
Try to aim for a phase margin of about 60 degrees. Gain crossover frequency can be about 5-10% of switching frequency.

Dick

Thanks, so I am looking at the brief drop that occurs when the load current changes?

If so then this is what I have got so far (no feedforward capacitor).

50mV/div 100uS/div

I am not sure if this is considered bad or acceptable. That ripple can be seen as this chip only switches at about 50khz.

[David Hess]

That capacitor adds phase lead to the feedback loop.  In practice it can be used to compensate for one of the poles in the output LC network and prevent 180 degrees of lag before the gain drops to 1.

Usually the output capacitor ESR provides a zero which helps compensate the loop but the phase lead capacitor would be necessary if the output capacitor ESR is very low.  The LM2576 datasheet mentions that this can be an issue with an ESR below 0.03 ohms.

[Alex30]

I would also experiment with adding an output ripple filter as shown in figure 34 of the datasheet. This may even improve the transient response