What is the point of varying the switching frequency in SMPS?

[treez]

We are designing a 60W offline flyback.

We notice that the L6566B pwm controller, in fixed frequency mode, can vary the switching frequency from 82KHz to 88Khz, if we wish it to. (ie the switching frequency is constantly going up and down between these two frequencies, at about 1Khz)
It is said that this offers ease in passing EMC.

But how can this be so?

The main problem for EMC in SMPS is the harmonics of the switching frequency, not the switching frequency itself. It is the high di/dt and high dv/dt currents and voltages that create the problem, and these are not eased by varying the switching frequency as described. So what is the point of varying the switching frequency as described?

L6566B datasheet:
https://www.st.com/st-web-ui/static/active/en/resource/technical/document/datasheet/CD00167474.pdf

 

[ravindragudi]

Having the variation in switching frequency 'may' help in improving the EMC performance but may not actually make a failing unit pass. It depends on how much power spreading happens over the frequency band and in what time duration.

The EMC, i.e, the radiated emission tests are conducted to measure the 'peaks' in the RF emissions from the device. If these are not within the limits then 'quasi peak' detection is done and also the 'average detection'. These last two measurements calculate the RF energy coming from the DUT over a period of time. That's the reason, if the energy is spread out over time over frequency band, there are chances that DUTs emissions will be improved compared to fixed switching frequency.

 

[mtwieg]

Frequency dithering is a common method to passing EMC. For a fixed frequency converter the EMI will show up as several narrow and strong peaks at harmonics of the switching frequency. Dithering spreads the energy of all those harmonics over a wider bandwidth with lower peak PSD, so it can pass EMC despite the overall power in the emissions being the same.

 

[treez]

thanks, but the switching edges are the same regardless of whether or not the switching frequency is dithered slightly...thus there will surely be no effect on EMC test results?
Also, RF emissions are surely nothing to do with the switching frequency, but are due to harmonics of the switching frequency, far higher in frequency than the switching frequency.

The switching edges, the di/dt and the dv/dt are not changed by dithering the switching frequency.

I think its just a hoax to flog more pwm controllers.?

 

[FvM]

Your post sounds like you didn't actually read the frequency modulation paragraph in the datasheet. Or at least don't understand the relation of measurement bandwidth and detected emission level in case of a frequency modulated signal. It's worth to think about it.

- - - Updated - - -

EMI measurement is using 120 kHz IF bandwidth and quasi-peak detector above 30 MHz. In so far, it's not completely following the frequency modulation, you get definitely a certain attenuation of detected peak levels.

 

[treez]

Whatever the measurement bandwidth, the emissions from the switching edges will be the same, I fully admit that the emission at the exact switching frequency will be less...but that's not really the difficult thing to filter out.....surely the difficulty in passing emc is in dealing with the high harmonics.?

I read page 31 of the datasheet, and it does nothing to say that the actual di/dt of the switching edges changes when the frequency is dithered.

 

[FvM]

The emission spectrum is comprised of discrete lines at integer multiples of the switching frequency. The spectrum envelope doesn't basically change, but the lines are shifted with frequency modulation. As an additional effect, the emitting "antenna" structure may involve resonances with a bandwidth smaller than the EMI measurement bandwidth of 120 kHz.

 

[dick_freebird]

This is about fooling spectrum analyzers and gaming the
specs as written without regard to total RF energy output.
Anything that is edge or "spike" sensitive will benefit none
by dithering frequency. Narrowband radios will see a "noise"
which comes and goes at a rate slower than data, perhaps
making a more difficult to debug / assign blame situation.

 

[FvM]

This is about fooling spectrum analyzers and gaming the
specs as written without regard to total RF energy output.

Most spectrum analysers will have difficulties to evaluate an emission level exactly according to EMI regulations, because they don't implement the required quasi-peak detector.

I agree that spread spectrum clock generators and SMPS frequency modulation are questionable from an emission minimization viewpoint, but you should blame existing EMI regulations.

 

[treez]

thanks, you have now influenced me to believe that its a good few of the lower harmonics which are reduced by "dithering" the switching frequency, but the higher harmonics, the harmonics involved with the switching edges, which are the ones which cause most emc problem, I still don't think are significantly affected by "dithering" the switching frequency.

- - - Updated - - -

thanks, you have now influenced me to believe that its a good few of the lower harmonics which are reduced by "dithering" the switching frequency, but the higher harmonics, the harmonics involved with the switching edges, which are the ones which cause most emc problem, I still don't think are significantly affected by "dithering" the switching frequency.

 

[mtwieg]

Of course all of the harmonics are dithered equally, they're all frequency locked to each other. That's what it means to be a harmonic. You are basically convolving each of those spectral peaks with a PSF related to your modulation.

 

[Velkarn]

but the higher harmonics, the harmonics involved with the switching edges, which are the ones which cause most emc problem,

why you think it is?

the most emc problem is switching frequency, i think. why you think in all simple power supply, designers sets the switching frequency at value lower than 150 kHz?
i have designed and provide emi tests of flyback converter, llc-converter, active clamp forward - main peak is at switching frequency.

 

[dick_freebird]

While the main peak is at fSW, the nuisance is at the front end
of radios and the like.

DC-DC EMI consists of both fSW synchronous harmonics, and
a whole mess of ringing in the output, filter and supply loops.
The really high frequency content "could" be the 1953rd
harmonic of your 1MHz switcher, but is more likely some
passives / board ringing that is only "kicked" by the switch
events and then only changes relative phase, not amplitude,
as frequency shifts.

Looking at the supply and switch nodes for ring-cycles that
attend the edge, will show you things intra-cycle that
dithering won't touch.