Audible noise

[Winsu]

Hi All,

I have designed an AC/DC flyback converter that swing the power output from 1W to 19W. The output voltage is 5.2V and 2.4A at maximum power (the 19W comes from poor power factor and it is done on purpose because I have a huge C bulk of 10uf after rectification to improve EMC).

The converter is working fine apart from audible noise at standby mode and light load. At medium and high load there is noise, but it is acceptable when inside the housing box.

I am using the AP3871 and I have a EFD25 transformer former with 54.7 turns on the primary (0.94mH), 11.7 turns on the auxiliary and 3.7 turns on the secondary. I have designed the transformer and chosen an Rcs resistor to work at high frequency when maximum power (the limit of the IC driver is 80kHz). The frequency at maximum power is around 75Khz and no significant noise is noticed. The problem comes when it is at standby mode (1W) or even light load. Using the below formula from manufacture's datasheet and with a Rcs of 0.81ohm that provides an Ipk current of 0.81A the frequency is around 3.2khz. A frequency of 3.2Khz is quite noticeable. I have tried several iterations modifying the transformer but at standby mode the frequency is always very low.

One of the reasons I chose this IC is because of the following feature, that was supposed to keep the frequency at least above 20Khz.

My understanding of this feature if that the Vcs threshold is modified according to the load, therefore the switching frequency is always above 20kHz.

Any comment or help would be very appreciated.

Winsu,
Regards

 

[KlausST]

Hi,

at the beginning you talk about

the power output from 1W to 19W.

but then you talk that the 19W come from poor power factor. Power factor surely relates to the AC input.
I´m confused. Do the 19V refer to input power or output power?
If input power, then shouldn´t it be 19VA instead of 19W .. because you say poor power factor.

If the 10uF is the input DC capacitor then it will have a voltage ripple of about 30Vpp (@ 230V AC, 19W, 5ms). This is huge for my taste. Not impossible.

Noise:
You should look for the source of noise. You may use a small tube as some kind of stethoscope to find out the "noise generator". You know that ceramics capacitors act like piezos: They form the applied voltage into mechanical movements.

Frequency:
I agree with the expectable 20kHz.
Are you sure your schematic and PCB layout is according SMPS design rules. Schematic mistakes as well as bad PCB layout may lead to "discontinous" operation ... which often is the reason for audible noise.

Klaus

 

[cupoftea]

Here is something

https://www.power.com/sites/default/files/PIU-108_Audible%20Noise.pdf

 

[Winsu]

Hi,

at the beginning you talk about

but then you talk that the 19W come from poor power factor. Power factor surely relates to the AC input.
I´m confused. Do the 19V refer to input power or output power?
If input power, then shouldn´t it be 19VA instead of 19W .. because you say poor power factor.

If the 10uF is the input DC capacitor then it will have a voltage ripple of about 30Vpp (@ 230V AC, 19W, 5ms). This is huge for my taste. Not impossible.

Noise:
You should look for the source of noise. You may use a small tube as some kind of stethoscope to find out the "noise generator". You know that ceramics capacitors act like piezos: They form the applied voltage into mechanical movements.

Frequency:
I agree with the expectable 20kHz.
Are you sure your schematic and PCB layout is according SMPS design rules. Schematic mistakes as well as bad PCB layout may lead to "discontinous" operation ... which often is the reason for audible noise.

Klaus

Hi Klaus,

Hi,

at the beginning you talk about

but then you talk that the 19W come from poor power factor. Power factor surely relates to the AC input.
I´m confused. Do the 19V refer to input power or output power?
If input power, then shouldn´t it be 19VA instead of 19W .. because you say poor power factor.

If the 10uF is the input DC capacitor then it will have a voltage ripple of about 30Vpp (@ 230V AC, 19W, 5ms). This is huge for my taste. Not impossible.

Noise:
You should look for the source of noise. You may use a small tube as some kind of stethoscope to find out the "noise generator". You know that ceramics capacitors act like piezos: They form the applied voltage into mechanical movements.

Frequency:
I agree with the expectable 20kHz.
Are you sure your schematic and PCB layout is according SMPS design rules. Schematic mistakes as well as bad PCB layout may lead to "discontinous" operation ... which often is the reason for audible noise.

Klaus

Hi Klaus,

Sorry, I mixed concepts...., The input wattage is 23VA (230V and 0.1A due to low power factor) and the electrolytic capacitor is the DC input capacitor which is 10uF. The 19W figure comes from applying the below formula when the frequency, so it is the power transfer between the primary and the secondary.

I will have a look at what components are generating the noise, but I also think that if I could manage to increase the frequency above 20Khz when it is working in standby mode the problem could be resolved. I think to do that I have to decrease the primary inductance a lot.... .I have done some calculations and if at full power the frequency is 450Khz at standby mode the frequency would be 20khz which is acceptable. Would it be ok to drive an AC/DC flyback converter at 450khz?

Regards,
Winsu

 

[BradtheRad]

From what I know about a flyback, duty cycle is normally designed to be 50 percent. Flux builds during the first half of the cycle, then collapses during the second part.

If your converter were buck type...
Then light load should be handled by shortening duty cycle. This reduces power through the system. (There is a formula for buck converter which derives Vout given Vin, based on duty cycle.)

However when you change only frequency, it changes waveform shapes but it affects average power just slightly.

Can your flyback controller IC be adjusted so it handles a change in load by changing duty cycle? Then frequency can remain constant in your preferred kHz range.

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
As for noise coming from a homebrew unit, sometimes I think it's a good idea for me to be able to hear it running so I have a clue that it's operating properly. Or alternatively, to get clues it's in abnormal mode.

 

[cupoftea]

no dont drive a mains powered flyback at 450khz....switching losses will kill it, and emc coudl be problematic.
Did you wind tightly?
Current mode in light load is known for its noise.
Sometimes you can varnish it...so the coils dont vibrate about.
Did you wind nice and tight and put tape between layers to tighten it still ( the coils)

Even if you stay above 20khz you can still get audible noise in light load, since there may be a kind of "surge" every 4khz even though fsw is >20khz.
You can always switch to a voltage mode controller....but is it worth it?...power integrations are the only ones i know that really put an effort into making them (their flyback controllers) quiet in light load.

I remember doing a cheapo dustbin lorry led warning light which absolutely screeched in light or heavy load......nobody cared as it just acted as like an alarm to people who then avoided it running them over, and also you couldnt hear it most of the time due to engine and motor noise.

BTW, is your current sense resistor voltage right up near the sense upper threshold when you are on max load?...if it isnt, then you will get more noise in light load

 

[Easy peasy]

Whatis the gap in the Tx and the core type ?

 

[Taihung]

Hi
Audible - noise mean something vibrated with frequency < 20kHz. Yes, it is not from Fsw >= 20kHz.
It is from control loop.
As you known, all PWM IC has its minimum on time, example 50ns is good.
Why need to take care of this ?
It related to minimum load can regulated with PWM.
If minimum load made PWM duty < minimum, control loop will jump in AM mode.
It will run with few cycle minimum duty then off in short time, or ping 01 pulse output and skip few next pulses. This repeat with frequency < 20kHz. And then make audible noise.
Next, the problem can be your transformer with leakage inductance too high and RCD clamp to low value. This will absorb more energy that make auxilary coil don't have enough voltage to supply IC. That make IC restart multiple time. Can change RCD clamp to TVS diode to test.
Need a minimum on load to make duty != 0, that make axilary coil have voltage to supply IC. Add a small at output resistor < 3% max load is acceptable. It belong to IC operation current and leakage inductance of transformer.

 

[Winsu]

From what I know about a flyback, duty cycle is normally designed to be 50 percent. Flux builds during the first half of the cycle, then collapses during the second part.

If your converter were buck type...
Then light load should be handled by shortening duty cycle. This reduces power through the system. (There is a formula for buck converter which derives Vout given Vin, based on duty cycle.)

However when you change only frequency, it changes waveform shapes but it affects average power just slightly.

Can your flyback controller IC be adjusted so it handles a change in load by changing duty cycle? Then frequency can remain constant in your preferred kHz range.

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
As for noise coming from a homebrew unit, sometimes I think it's a good idea for me to be able to hear it running so I have a clue that it's operating properly. Or alternatively, to get clues it's in abnormal mode.

Hi, The IC controller work on PFM so I can change the duty cycle. To change the power transfer, I need to change the value of Rcs and to control the voltage I need to modify the value of FB1 and FB2. Primary inductance and Rcs value affect to frequency...

--- Updated Nov 6, 2022 ---

no dont drive a mains powered flyback at 450khz....switching losses will kill it, and emc coudl be problematic.
Did you wind tightly?
Current mode in light load is known for its noise.
Sometimes you can varnish it...so the coils dont vibrate about.
Did you wind nice and tight and put tape between layers to tighten it still ( the coils)

Even if you stay above 20khz you can still get audible noise in light load, since there may be a kind of "surge" every 4khz even though fsw is >20khz.
You can always switch to a voltage mode controller....but is it worth it?...power integrations are the only ones i know that really put an effort into making them (their flyback controllers) quiet in light load.

I remember doing a cheapo dustbin lorry led warning light which absolutely screeched in light or heavy load......nobody cared as it just acted as like an alarm to people who then avoided it running them over, and also you couldnt hear it most of the time due to engine and motor noise.

BTW, is your current sense resistor voltage right up near the sense upper threshold when you are on max load?...if it isn't, then you will get more noise in light load

Indeed, 450Khz could be too much but I could try to see if the IC can handle it...

The coil turns are tight but of course it is done by hand as it is a prototype..., I guess pushing the transformer with something made out of plastic while it is being noise could help me to assess if the noise comes from the transformer or from something else. I also need to look at capacitors...

At maximum power the voltage at the source of the mosfet is 600mV which is what it should be at CS pin so at least that is not happening

--- Updated Nov 6, 2022 ---

Whatis the gap in the Tx and the core type ?

Hi,

The gap is 0.22mm and it is a EFD25...