Hissing inductor cores

[akis]

I have a few switched mode supplies: 2 are laptop mains chargers (19.5V, 4A), a few are DC-stepdown converters using the LM2596 ready built, and a few are mine own LM2679s designs.

All of them, with no exceptions, make a high pitch noise which sometimes becomes hissing, when I draw too much power from them (well within the limits). The noise comes from the inductors. Even though I hold them as tight as I can the noise seems to be coming from the windings, even if the windings are individually insulated with "soft" plastic sleeving that you'd think would absorb any kind of vibrations, especially at those frequencies.

Can someone please tell me what it is I am experiencing? Is it "core saturation" ? If yes is there a way to prevent saturation by increasing the number of turns? Air gap? Removing the ferrite core completely and using an air coil?

[IanB]

The inductors are coils of wire in a magnetic field. Another device with a coil of wire in a magnetic field is a loudspeaker. Coils of wire in a magnetic field vibrate when subjected to alternating current. They vibrate more strongly when the current in the wire is increased (e.g. by increasing the load on the power supply).

So there is not much you can do to stop the coils making noise. You could surround them with sound dampening material, but this would most likely interfere with the cooling arrangements. Well made inductors will usually have the windings secured and sealed with varnish or other substance to reduce the vibrations, but they can never be completely silenced.

[minime72706]

If you can isolate it and avoid covering everything with goop, I'd say to try potting just the magnetics in thermal epoxy.

EDIT: The Home Depot over here in The States sells Cree LED incandescent-replacement bulbs for insanely cheap (loss-leader?) but sometimes they buzz, especially when on any kind of dimmer circuit. I opened one up because it had broken anyway and saw the driver board, which looked like a good design, but other than some plastic guides in the base, it was sitting loose. In a semi-successful experiment, I drilled a hole in the side of the base and pumped in thermal epoxy. This lowered the sound considerably but I let some of the epoxy (Black) touch the glass and we also trapped moisture inside the bulb. Just saying it might be worth it.
This is just a fact of reality, though - switchers buzz.

[akis]

I found an old/salvaged toroidal inductor in my drawer, 30mm outside diameter and heavy, probably iron not ferrite, and I used it to replace my own inductor which I made using an RM14 core with N97 ferrite. The toroidal is almost inaudible and produces lots more power.

I have been reading up on coils theory and one of the main questions is: I have a random core in my drawer and want to know how much power I can use it for. I know the Ae, Al, Voltage, frequency, type of wave (eg sine, square, AC or ADC) - but I do not know yet (still reading) how to calculate the power that this core can give me.

[minime72706]

I found an old/salvaged toroidal inductor in my drawer, 30mm outside diameter and heavy, probably iron not ferrite, and I used it to replace my own inductor which I made using an RM14 core with N97 ferrite. The toroidal is almost inaudible and produces lots more power.

I have been reading up on coils theory and one of the main questions is: I have a random core in my drawer and want to know how much power I can use it for. I know the Ae, Al, Voltage, frequency, type of wave (eg sine, square, AC or ADC) - but I do not know yet (still reading) how to calculate the power that this core can give me.

I have the same problem, akis.
I suppose you could experimentally measure at what current level the inductor saturates at and the power limitation probably has a lot to do with the wire diameter used as well as the temperature rise in the core. You can also direct yourself toward a half-correct answer based on the material it appears to be made of. Just talking out of my ass, though.

[NiHaoMike]

Some video cards used open inductors for the GPU core supply. Under certain conditions (most famously when running the Survivor stress test), the card would act as a speaker and emit a buzzing sound at the display refresh rate. Some PSUs exhibit the same issue, especially when both the mains frequency and refresh rate are 60Hz (but not exactly), at which point the two beat together in the PFC inductor and make a very annoying "throbbing" buzz. Ground loops or a sound card with poor EMI filtering can also cause a buzz in the audio.

Switching regulators can also whine or squeal if they're operating within the audible range, either by design or as a result of period skipping or instability.

Best solution I have found for noisy toroid inductors (if replacing is not an option) is to squirt a little hot glue into the hole.

[akis]

I have the same problem, akis.
I suppose you could experimentally measure at what current level the inductor saturates at and the power limitation probably has a lot to do with the wire diameter used as well as the temperature rise in the core. You can also direct yourself toward a half-correct answer based on the material it appears to be made of. Just talking out of my ass, though.

Is there a tool, or a theoretical test so then I can make the tool, to allow me to measure a core's saturation point ?

Also, it seems those toroidal irons, outside diameter 30mm, are what I need for my switching circuits. Is there a place to buy them bulk/cheap so then I can wind them ? Farnell sell them for £1.90 each http://uk.farnell.com/jsp/search/productdetail.jsp?sku=1929736

[BravoV]

Is there a tool, or a theoretical test so then I can make the tool, to allow me to measure a core's saturation point ?

Provided you have a scope, and an understanding on the "basic" inductor's theory and it's common formulas, read here -> Inductor Saturation Tester , credit goes to Jay_Diddy_B as the circuit designer, really simple to build using just common discrete components and dirt cheap as well.

Also just noticed on the salvaged unknown ferrite cores, coupled with above inductor saturation tester and inductance meter if its available, then read here -> Question : Profiling an unknown power inductor or power transformer

[akis]

I think I am getting a better understanding of what is happening now that I have read a bit more.

First, here is a datasheet for a series of Bourns inductors, showing the reduction in nominal inductance with increasing current.

In my switcher circuit the problem is that as the current and voltage increase the inductance drops. At some point the inductance has dropped so much that it's like a short to the load and the switcher cannot maintain the square wave anymore.

[AndyC_772]

All inductors make a noise. However, the fact you can hear it tells you it's making a noise with at least some frequency components in the audible range, and that's often a sign of a problem.

If, say, the controller is switching at 500kHz and is stable, then that's the frequency at which the inductor should be oscillating. You can't hear either the fundamental or the harmonics, and the design works silently.

However, if it's unstable, you could have much lower frequency oscillation going on, which you'll see if you probe the PWM signal. If you have groups of pulses together separated by gaps, or the widths of consecutive pulses are different, then there's a source of audio frequency noise which you can only fix by improving the stability of your supply. The sound from your inductor is a warning that this is going on.

Check your feedback network carefully, and look with a scope on the regulator's feedback pin. Normally it should be a DC voltage at whatever the IC's reference voltage is; if it's oscillating or noisy, your regulator is unstable.

[fcb]

Sub-harmonic oscillation.

Your largely hearing the effect of the control loop and control loop stability.

[penfold]

Every time this subject comes up my blood always starts to come up a light simmer.  There's far too much confusion that inductors will make audible noise regardless of excitation frequency.  And they don't seem to realise that if a 100 kHz transformer makes audible noise, there's nothing that can't be fixed with a bit of varnish.

More often than not the effect is actually the two halves of the core clattering together, which is why toroids are much quieter, the remaining noise is the much lesser effect of the windings doing what they do.

So, much respect to AndyC and fcb for clearing it up.  Whilst I'm in ranting mode however, its not sub-harmonic oscillation, it is definitely sub-harmonic and it is oscillation, but sub-harmonic oscillations are still related to the fundamental and normally only appear with crazy things like bang-bang and limit cycle controllers.

[Kremmen]

While it could be nit-picking, let me add this anyway:
True sub-harmonic oscillations should of course be related to the fundamental frequency by some simple ratio. Often what are called subharmonics are not, but they exist nonetheless. They are a phenomenon of chaotic period bifurcation; a rather complex thing to explain but perhaps surprisingly, one that obeys the same basic behavior in widely differing systems. A typical issue in this context is the behavior of an uncompensated buck switcher when the D value grows - the loop begins to exhibit irregular switching cycles that are often called subharmonics, even though they are chaotic in nature. Another, totally different context is the dynamics of animal populations where you see _exactly_ the same bifurcation effect. It is a strange world... http://en.wikipedia.org/wiki/Bifurcation_diagram

[IanB]

I think there is also the question of under damped and over damped systems. If a system is under damped it will be able to resonate at a particular frequency that may be lower than the exciting frequency. If the resonant frequency of a coil is in the audible range then it may emit noise even if the electrical frequency in the coil is higher. This is partly why securing the coil with mastic or hot glue may help, since it will tend to damp the system and suppress the resonance.

[Kremmen]

What you describe is the definition of sub-harmonic oscillation. If the excitation frequency is an integral multiple of the resonant frequency of a system, then energy will be transferred to that system and it will oscillate provided its Q factor is sufficiently high (or conversely, the damping factor low). The frequency of the resulting oscillation is a true sub-harmonic of the excitation frequency. Often you will also see significant beat frequencies when the excitation is close but not quite precisely an integral multiple. The causal mechanism of chaotic period variation is totally different however.

A true sub-harmonic will usually make a "singing" noise, while a chaotic period bifurcation sounds more like white noise.

[IanB]

I understand. It looks like I was saying the same thing in less technical words. I was just trying to address the implications earlier in the thread that "a 100 kHz coil can't emit audible noise" and "it must be controller instability".

Sure, it could be controller instability--but isn't that likely to occur at low loads, rather than at high loads as indicated in the OP?. Also, if it is controller instability in an off the shelf supply like a laptop power brick, what practically can you do about it?

[AndyC_772]

I do more often hear noise from switchers when they're on a light load, more often than not because they've gone into a power saving 'pulse skipping' mode, in which the switching frequency is effectively reduced.

At high loads it could be electrical noise coupling into the feedback node. If the layout is poor, and there's noise at the feedback pin with respect to the analogue ground at the controller, then there will be variation in pulse width from one switching cycle to another. The spectrum of this variation can be very broad, and contain audible components.

If it's a bought-in PSU module then there's an easy fix: use a different one.

[akis]

I do more often hear noise from switchers when they're on a light load, more often than not because they've gone into a power saving 'pulse skipping' mode, in which the switching frequency is effectively reduced.

At high loads it could be electrical noise coupling into the feedback node. If the layout is poor, and there's noise at the feedback pin with respect to the analogue ground at the controller, then there will be variation in pulse width from one switching cycle to another. The spectrum of this variation can be very broad, and contain audible components.

If it's a bought-in PSU module then there's an easy fix: use a different one.

This is a good summary verified by the scope too.

1) There are pulse skipping modes, and the scope cannot lock, and even then the frequency is half what it should be. You increase the load and it the square wave becomes more regular and at the right frequency.

2) As you continue to increase the load the duty cycle increases. The noise now starts usually very high pitch and a single frequency. As you increase the load, I believe due to the coil becoming more saturated, the switcher has real trouble trying to pulse it until the point where it simply cannot anymore and the square wave disappears and becomes a mess. At the same time the noise now becomes white noise like hissing or frying.

3) Changing coils has a profound effect in what power we can draw. To reduce the flux density we need a lot of turns, but the inductance then increases outside of the range that the switcher can use. The requirements are: F=260KHz, VDC square wave (ie unipolar) of amplitude of 39V, output voltage up to 30V and 5A (150W).

4) Example calcs are as follows for an RM14 core with N97 ferrite (Ae=200mm2 and AL=6000nH)

Code: [Select]

VDC current (one direction)
magnetic flux density 0.020000 T
cross sectional area (window size) 200.000000 mm2
cross sectional area (window size) 0.000200 m^2
magnetic flux in one direction 0.000004 Wb
Volts rms 20.000000 V
frequency 260000.000000 Hz
seconds full cycle 0.000004 sec
seconds semi cycle 0.000002 sec
turns 9.615385
Al 6000.000000 nH
L 57.692308 uH

I have tried this and I get about 3A before it comes crashing down. I have also tried RM12 core, and ETD29 core, all with N87 or N97 materials. Nothing can match the 30mm iron toroid which is salvaged and I know nothing about it

So I am annoyed that I have a core which is larger than the toroid but still cannot get the power out (150W).

Edit: on a couple of occasions I cheated and put two small slides of paper in between the two core halves so as to keep the inductance low around the 15uH-27uH that the switcher needs. Still did not improve the power handling.

[Kremmen]

I understand. It looks like I was saying the same thing in less technical words. I was just trying to address the implications earlier in the thread that "a 100 kHz coil can't emit audible noise" and "it must be controller instability".

Quite so. There are a number of reasons why audible noise may be generated by a switcher whose operating frequency is far above hearing range. As AndyC mentions, hiccup mode is one such reason. I guess chaotic period variations qualify as "controller instability" and then there are the true sub-harmonics as well.

Sure, it could be controller instability--but isn't that likely to occur at low loads, rather than at high loads as indicated in the OP?. Also, if it is controller instability in an off the shelf supply like a laptop power brick, what practically can you do about it?

That depends. The stability margins behave differently in different topologies. Of course a properly designrd PSU is supposed to be stable but YMMV. And there isn't much you can do in case of a power brick with "no user serviceable parts".

[AndyC_772]

I think you're barking up the wrong tree with the inductor saturation idea.

The controller doesn't switch the inductor, it switches a MOSFET. The inductor current is monitored, either by being sensed directly or by being reconstructed by a separate circuit, and when it reaches a predefined threshold, the MOSFET is switched. There's no sense in which the same MOSFET can become "difficult" to switch.

What can become "difficult" is determining the proper time to switch, in a noisy environment.

In an ideal world, monitoring (or, in some circuits, modelling) the inductor current results in perfectly uniform switching cycles, so the only frequency components present will be the fundamental (say, 500kHz or whatever), plus higher harmonics. If the inductor starts to saturate, it'll heat up, the supply becomes less efficient, and the output voltage will sag. The controller compensates by increasing the PWM duty cycle, which may work up to a point, but the circuit is now working beyond its realistic upper limit. It still shouldn't be making audible noise, though.

What happens in a real circuit is that switching noise from the MOSFET couples into the voltage feedback and/or current sense nodes. This results in random, chaotic variation in the PWM duty cycle. It still averages out to the proper value over time, thanks to the closed loop control, but it means there are audible components in the frequency spectrum of the switching waveform.

The noise scales with current, and so, therefore, does the magnitude of the chaotic variation.

I suspect that changing inductor changes the noise because of each part's mechanical characteristics, but it's not really making any difference to the real reason why the noise exists in the first place.

[akis]

It cannot be just circuit noise causing all these issues. If that was the case we would simply remove the inductor altogether or could replace the inductor with a bent piece of wire, and blame all problems on the PCB layout.

My *theory* (just theory) is that the inductance drops as the core saturates, not because it is getting hot over time, but instantaneously, the reactance then drops, and the switch sees a short at the output (well, a short bar the load) - at that moment the current limit kicks in or the IC gets fried - all in a matter or less than 1-2 seconds.

I have fried thus 6 ICs. 4 of them have failed badly: they have shorted input and output and present the full unregulated voltage to the downstream system thus destroying that too (one of the IRF9540s then reached at least 150C and started to smoke and blister before I realised that the DC switcher had failed....)

I also accept that there is noise. The application note mentions that the switch output can oscillate into the MHz area, depending on layout, and they suggest a Zobel network at the switch output to dump these oscillations. I have been forced to dumpen the circuit and therefore I accept that there must be a physical layout that will not give rise to these oscillations. This dumping makes a huge difference, it allows me to draw the full power, without it probably 50% of the power.

However I am not sure I can design the PCB layout such that it does not make noise which might affect the voltage and current feedback. I have tried huge ground areas and very thick traces (4mm) for the input and output, caps almost touching the IC, I have followed datasheet's suggestions as much as possible but still I need a 470pF capacitor over the switching diode or else I cannot draw full power. I have also noticed that the 470pF cap (or 10R + 470pF) must be placed physically over the switching diode, not somewhere else in the circuit for it to work. I attach the PCB layout .