Troubleshooting Audio Noise in SMPS at 50KHz: Tips and Solutions

[ElectroPhysics]

I have a problem with my switch mode power supply, smps. It creates audio sound when I power it up, especially at minimum load. As I increase the load then sound disappears and in some designs it is still there. I don't know what to do. It is a bottleneck for me. Anyone who knows what to do about it. I'm operating it at 50KHz which is well above audio frequency.

 

[berkeman]

I have a problem with my switch mode power supply, smps. It creates audio sound when I power it up, especially at minimum load. As I increase the load then sound disappears and in some designs it is still there. I don't know what to do. It is a bottleneck for me. Anyone who knows what to do about it. I'm operating it at 50KHz which is well above audio frequency.

Most likely it is magnetostriction noise in your main switching magnetic component. What topology is your SMPS? What transformer/inductor are you using? The general fix/prevention is to put some epoxy material in the gap of your transformer. Who is your transformer/inductor supplier?

 

[ElectroPhysics]

I'm using it in forward topology. etd49 as main transformer and etd39 with gape for output inductor.
I have measured the waveshape at the drain of mosfet. It looks like distorted when audio noise comes otherwise it is clean waveshape.

 

[berkeman]

I'm using it in forward topology. etd49 as main transformer and etd39 with gape for output inductor.
I have measured the waveshape at the drain of mosfet. It looks like distorted when audio noise comes otherwise it is clean waveshape.

Can you find the source of the distortion/noise? Maybe a feedback stability problem?

 

[Genius_Engg]

Hey ElectroPhysics, did you solve the audible noise issue in your SMPS. Actually, I have exactly the same problem. I do not know how to get rid of this noise. Please help me out.

 

[ElectroPhysics]

@Genius
Yes partially solved. the noise is due to instability in current sensing circuitry due to leading edge spike across current sense resistor in current mode control (also in voltage mode.) there are chips which have feature of leading edge blanking circuitry built in. how ever you have to keep in mind that such chips require dummy loads as minimum load. if you can, please use cores kool u and high flux. for example from Magnetics inc.

 

[ElectroPhysics]

@Genius Engg
Yes now i think completely solved. the other part of noise was due to loose core parts. use some adhesive to fix the core parts. at the cost of damaging my designed power supply i found that even current sensing circuitry was not the only culprit. i removed the current sensing circuitry in my voltage mode control SMPS. then after giving voltage to SMPS i found that noise was still there. this proved that it was due to loose EE core parts. second time when i repeated the same experiment, something blow up due to high current in mosfets (as i have removed the current sensing/limiting circuitry), i think it was primary windings :)

 

[ElectroPhysics]

@Genius Engg and others
Now I'm in better position to tell the right reason. At light loads the current sensing circuitry measures the low current demand and lowers the duty cycle. In real scenarios this duty is so small that almost diminishing. Thus in next cycle a high primary current flows through mosfet and current sensing circuitry. This produces more than necessary power at output and thus in next cycles the cycle is skipped or duty cycle is lowered too much. One thing to be noted that this primary current when flowing at low output power is still at peak primary current level. Also this current is at random cycles. This high level of random current produces audible noise. The solution is that we can remove audible noise almost if we can keep primary current level to a lower level instead of at peak cutoff level at low output power. Then if possible we can also operate this SMPS in burst mode without danger of audio noise.

 

[ElectroPhysics]

Hi,

Here is an article on acoustic noise in SMPS based on my research uploaded at my website link,

https://sites.google.com/site/electrophysics/articles/acoustic-noise

 

[ElectroPhysics]

Hi,
May be I don't maintain my website for coming years so I have posted the full article written so far as below. I got these results practically doing the design of SMPS so they would work!

Acoustic noise can be a troublesome thing for a novice designer. However in order to remove it from a prototype, better understanding and sound knowledge of magnetic components is required. While troubleshooting such cases the reasons behind acoustic noise generation that I found in flyback and forward type topologies are as below.



At light loads the current sensing circuitry measures the low current demand and lowers the duty ratio. In real scenarios this duty ratio is so small that almost diminishing. Thus in next cycle a high primary current flows through mosfet and current sensing circuitry. This produces more than necessary power at output and thus in next cycles the cycle is skipped or duty ratio is lowered too much. One thing to be noted that this primary current when flowing at low output power is still at peak primary current level determined and limited by current sensing circuitry. Also this current is at random cycles either because of this reason or due to false triggering of current sensing circuitry by leading edge spike. This high level of random current produces high peak flux in transformer core which causes audible noise due to magnetostriction. The solution is that we can remove acoustic noise almost if we can keep primary current level to a lower level instead of at peak cutoff level at low output power. Then if possible we can also operate this SMPS in burst mode without danger of acoustic noise.



To rectify the problem I redesigned a flyback SMPS on vero board. It is operating at 50 KHz but no audible noise from transformer detected at no load to full load. First of all I did as before and there was audible noise of around 1.6KHz. It was designed using UC3844, an old traditional IC. I used the same type of snubber and clamp "ceramic" capacitors as were in previous design and it was not the culprit though I don't recommend them. The things that I did to solve this problem were as below.


1) Don't use single op-amp part as feedback with opto-LED cathode side connected to its output side (and anode connected to VCC side) to save another op-amp part. Otherwise there might still be enough voltage differential acorss opto-LED to let it turn on at input turn off command. This issue can be seen frequently in op-amps that don't provide rail-to-rail operation. Even if used two op-amp parts so that signal level could be inverted again the acoustic noise would be still loud there and next suggestions should be followed.


2) The feedback op-amp and connected opto-LED should be given power from much higher than 5V or other intelligent methods should be used so that it could occupy and thus sweep full range of resistor limited current from this opto-LED. This will produce more shades of opto-LED light intensity instead of just on and off. This will prevent cycle skipping command.



3) Correct the opto-transistor current according to current transfer ratio, CTR.



4) Add a small value non-polar capacitor in parallel with current sense resistor. This will further reduce the acoustic noise.



5) Add a small value high voltage capacitor above 1 KV working volts between primary and secondary ground.



6) You may now hear light acoustic noise and that would be at no load. By pressing core parts to each other this should vanish. Use some varnish and epoxy to fasten the two core parts and problem solved.



7) If your PWM IC does not support leading edge blanking option then you can implement at current sense pin with few external parts. This will almost eliminate the false triggering at this pin and help reduce acoustic noise further.



8) Operate feedback error amplifier output in current source mode by using a transistor in opto-LED path.



9) Another trick that works is to keep PWM chip error amplifier at unity gain on primary side after opto-transistor with compensation done on secondary side and primary error amplifier input and feedback resistor near minimum set value as shown in datasheet. As an example it is 8.8 K for UC384X series and 30 K for SG3526 type ICs.



10) In order to eliminate cycle skipping and hence acoustic noise slightly lower the reference voltage used to power opto-transistor collector pin.


Some other ICs like NCP120x series have built in adjustment technique for peak level of current at current sense resistor to reduce or forbid the acoustic noise completely. They lowers the peak level of current at low power demand. This way peak level of flux is lowered which reduces magnetostriction effect so that no acoustic noise can be heard from transformer core. The advantage is that we can operate the power supply in burst mode at light loading without danger of acoustic noise. Burst mode operation can also be implemented with UC384X chips using few external parts.