I have built a small, isolated DC-DC supply. It's basically a flyback with optically coupled feedback. The transformer is planar type, and runs primarily in CCM. The problem is that the core rattles, sings, at light loads - its audible and easily heard. The PWM chip has a light-load feature where it skips as needed....this is when the core rattles. Without skipping (either under a heavier load or fooling the chip with a high value current sense resistor), the loop is very stable, and the core does not rattle at all.
My ears are not great, but listening to the pitch seems to be at about the same frequency as the bursts during light-load. So - I'm pretty sure the skipping feature is causing the core to rattle away.
Does anyone have an idea on how to fix this?
My ears are not great, but listening to the pitch seems to be at about the same frequency as the bursts during light-load. So - I'm pretty sure the skipping feature is causing the core to rattle away.
Does anyone have an idea on how to fix this?
I suspect you hear the core rattling when the skipping is at an audible frequency. The switching frequency itself may be high, but the skipping, discontinuous mode frequency, may fall within the high treble range or even lower.
Rectangular cores are more susceptible to rattle/make noise when used in a switched mode power supply. My limited experience is, toroidal cores are less susceptible, although you might end up with audio waves produced as a result of magnetostriction (core shrinking in diameter as a result of stress from magnetic forces).
A permanent solution would be to make sure the discontinuous mode frequency is ultrasonic but this may not be possible with the circuit components you used. In other words, the suspend - resume cycle frequency has to be ultrasonic. While this is not necessary for the circuit to work, it is necessary for you not to hear the sound produced from the core.
Rectangular cores are more susceptible to rattle/make noise when used in a switched mode power supply. My limited experience is, toroidal cores are less susceptible, although you might end up with audio waves produced as a result of magnetostriction (core shrinking in diameter as a result of stress from magnetic forces).
A permanent solution would be to make sure the discontinuous mode frequency is ultrasonic but this may not be possible with the circuit components you used. In other words, the suspend - resume cycle frequency has to be ultrasonic. While this is not necessary for the circuit to work, it is necessary for you not to hear the sound produced from the core.
Last edited:
Make sure you load it so it does not go into skip mode. Not only the transformer rattles, but also caps, so you will hear it when it skips. The nature of SMPS, you must never over size SMPS for audio b/c they will go into skip mode.
Add an external parallel load until it runs continuously, then figure out either if you just have a dump-load in parallel, or adjust your SMPS for lighter duty.
Add an external parallel load until it runs continuously, then figure out either if you just have a dump-load in parallel, or adjust your SMPS for lighter duty.
Firstly, this isn't an audio power supply - just an ordinary supply for something else. Power output is from very low, perhaps 0.25 Watts, to about 21 Watts on the high end of things. Some growing pains for me on this one....
I did try to run transformer in DCM at light loads, and then CCM at heavier loads. This required a pretty large gap in the core, and the core got very hot. As well, the efficiency was in the toilet. I have some other power losses that I don't yet see - will keep at it, though. Anyway, I redesigned the core such that it is almost always in CCM operation. The ripple current is very low, the gap is very small, the core runs nice and cool (except when loaded at the higher power levels), and efficiency is at least better. But - this is where the rattling occurs.
A side effect of the skipping is that the output requires a lot more capacitance in order to hold the output at reasonable levels until the next burst comes along. And - the rattling core. It just occurred to me - I have some Mag-Inc toroidal cores handy, and could try winding some to try out. I really wanted to avoid toroids because of the cost to manufacture...planar is pretty great in this regard.
I did try to run transformer in DCM at light loads, and then CCM at heavier loads. This required a pretty large gap in the core, and the core got very hot. As well, the efficiency was in the toilet. I have some other power losses that I don't yet see - will keep at it, though. Anyway, I redesigned the core such that it is almost always in CCM operation. The ripple current is very low, the gap is very small, the core runs nice and cool (except when loaded at the higher power levels), and efficiency is at least better. But - this is where the rattling occurs.
A side effect of the skipping is that the output requires a lot more capacitance in order to hold the output at reasonable levels until the next burst comes along. And - the rattling core. It just occurred to me - I have some Mag-Inc toroidal cores handy, and could try winding some to try out. I really wanted to avoid toroids because of the cost to manufacture...planar is pretty great in this regard.