Output EMC filter of smps & instability

[eem2am]

hello,

Class D audio amp will draw high current pulses so this is how current will be drawn from the smps that supplies it.

Therefore, for emc sake, the smps will need an output L,C emc filter.

However, if the smps crossover frequency is more than one third of the L,C emc filter resonant frequency, then the smps will go unstable.

so how do you get round this, because for a class d amplifier, we want our smps crossover frequency to be as high as possible so as to get a good transient response.

[johansen]

Originally Posted by eem2am

hello,
so how do you get round this, because for a class d amplifier, we want our smps crossover frequency to be as high as possible so as to get a good transient response.

isn't that a problem for the PSRR of the audio amp?
are they not 90 Db these days?

[eem2am]

sorry but i assumed that a class d amplifier input voltage rail had to be rock solid, and not allowed to overshoot or undershoot more than 1% when experiencing a no-load to full-load transient....am i wrong?

[sawreyrw]

eem2am,

Somehow I missed your post when it was originally posted. Anyway, how did you get the idea that "if the smps crossover frequency is more than one third of the L,C emc filter resonant frequency, then the smps will go unstable"? It is true that a SMPS or class D amplifier feed from a supply with an LC fiter can go unstable. This is because the device is a constant power load, which means the incremental input resitance is negative. In order to avoid this problem the filter must be damped. Dr. Middlebrook was the first one to publish work on this and he developed the Middlebrook stability criteria. Google for it.

[eem2am]

sawreyrw:

Ok Thanks,

Anyway, how did you get the idea that "if the smps crossover frequency is more than one third of the L,C emc filter resonant frequency, then the smps will go unstable"?

....page 685 of the book called switch mode power supplys by Christophe Basso will tell you this........(im pretty sure its page 685)......i've also heard significant contributors from this forum saying this.....basically if the x'over frequency of the smps is more than one third of the output LC resonance frequency, then the smps will try and react to the resonance and will go unstable. (i'm talking about the output LC EMC filter...not the actual output inductor and output capacitor of say a bridge smps)

......i wasnt aware that a class d amplifier is a constant power load......is that what you meant?.......i am aware that an smps is a constant power load.........

There is a rule of thumb for avoiding input filter oscillations......basically the output impedance of the input filter (looking back from the smps input)...should be less than the input impedance of the smps at the smps's crossover frequency.

.....at least i think its that.............

BTW, i once worked at a big RF company where they were building an smps for a RF amplifier..........as usual with everything RF, the RF engineers kept the RF amplifier secret till the last moment, and didnt even let the smps engineers use it as a load for testing...instead insisting that the smps engineers used an electronic load instead..........anyway, nobody told the smps engineers that the RF amplifier had 400uF of capacitance in front of it.......so when finally the smps was used to power the RF amplifier, the smps went badly unstable, since the 400uF capacitance at the input to the RF amplifier reduced the resonance frequency of the LC EMC filter at the output of the smps.....reducing it so that it became less than 3 times the crossover frequency of the smps..........

.....so anyway, the smps went badly unstable, giving a large 50KHz oscillation at its output......this in turn made the RF amplifier give 50KHz sidebands, which completely ruined its operation.

...this was a huge company, and you will have heard of it, but i wont mention the name.

[gmarsh]

Big company, lots of RF engineers, lots of SMPS engineers, crappy corporate culture that has engineering departments keeping secrets from each other... Harris?

Anyway... you'll drive yourself nuts trying to make a power supply work with a high corner frequency - I think you're better off putting extra capacitance on the output of the supply, and moving down the corner frequency to match. Having the extra output C on the supply output also makes it a lot less sensitive to extra capacitance/filtering/etc being thrown at the output.

Forget rules of thumb and start looking at output impedance versus frequency as your sdesign metric. Output impedance at high frequencies will be handled by your output caps, and at low frequencies it'll be handled by the SMPS's regulation. Make sure they overlap well.

And build some spreadsheets, code, etc... that simulate the effect of the second stage LC filter on your overall pole/zero plot.