# -20dB slope at crossover frequency in smps?

1. ## -20dB slope at crossover frequency in smps?

why, in smps feedback loop, ....why does the overall closed loop gain have to be going down at -20dB per decade at the crossover frequency?

I would have thought -40dB would have been ok too?

Surely, as long as the closed loop gain is <1 for frequencies above the crossover frequency then its fine.........(as long as the phase is always >45 degrees above the crossover frequency too.)

so why does the slope of the closed loop have to be going down with a gradient of exactly -20dB per decade at the crossover frequency? •

2. ## Re: -20dB slope at crossover frequency in smps?

I've never heard anyone say that it has to have a certain slope. In most cases (like with voltage mode converters) it will be going at -40dB or even -60dB at and above the crossover. In general I'd consider a higher slope to be preferable (more filtering of ripple).

1 members found this post helpful. •

3. ## Re: -20dB slope at crossover frequency in smps?

The loop phase is directly linked to magnitude slope by elementary network theory laws, respectively can only get worse by introducing additional delays. For a minimal phase negative feedback system, 20 dB/decade slope refers to 90° phase margin and 40 dB to 0° respectively perfect oscillation condition.

You should additionally consider, that a switched mode system usually involves a typical delay of 0.5 switching period.

1 members found this post helpful. •

4. ## Re: -20dB slope at crossover frequency in smps? Originally Posted by FvM The loop phase is directly linked to magnitude slope by elementary network theory laws, respectively can only get worse by introducing additional delays. For a minimal phase negative feedback system, 20 dB/decade slope refers to 90° phase margin and 40 dB to 0° respectively perfect oscillation condition.

You should additionally consider, that a switched mode system usually involves a typical delay of 0.5 switching period.
I am wondering if phase margin is big enough, do we still need magnitude slope to -20db/dec at at crossover frequency? --[[ ]]--