21

We designed a switching regulator circuit using a 1.5MHz, internal-switch, switching regulator (semtech.com/images/datasheet/sc185.pdf). Vin is 5V, Vout is 3V3. We have an input capacitor (47uf), an output capacitor (47uf) and an inductor (1uH). The problem is that we hear a high pitch sound coming — presumably — from the inductor when we turn the system on. It appears that the sound is more noticeable when the circuit is drawing very small amounts of current. As the current demand increases, the sound usually becomes unnoticeable, but not always.

Any ideas what we might have done incorrectly? Is there any other information I can provide to be more specific? I've looked at the regulator output, just before the inductor, and I see some ringing, but I can't tell whether the ringing is normal or not.

JYelton
  • 32,302
  • 33
  • 134
  • 249
SomethingBetter
  • 2,260
  • 7
  • 26
  • 33
  • What is the controller? Is it fixed frequency or variable, is it possible that under very small loads the operating frequency drops into the audiable range? – timrorr Jun 30 '12 at 10:06
  • 2
    Possibly the same cause as here: http://electronics.stackexchange.com/q/14256/930 What IC or control circuit are you using? Does it go into burst mode when the charger enters a light-load or no-load condition? – zebonaut Jun 30 '12 at 10:11
  • 1
    Related question: http://electronics.stackexchange.com/questions/19210/which-electric-components-could-make-a-high-pitch-sound-when-defected – m.Alin Jun 30 '12 at 10:15
  • @timorr: it is a fixed 1.5Mhz controller, from Semtech (http://www.semtech.com/images/datasheet/sc185.pdf). – SomethingBetter Jun 30 '12 at 10:41
  • I agree with zebonaut's presumption. Check his answer to the question he refers to, and PSAVE mode in the SC185's datasheet. Voting to close this question as exact duplicate. – stevenvh Jun 30 '12 at 10:52
  • 1
    @stevenh: I know about PSAVE mode. As you can see in the datasheet, PSAVE mode can be disabled. Even with PSAVE off, we have the same problem. Also, even with higher loads currents, we can still hear a noise, just not as loud. – SomethingBetter Jun 30 '12 at 10:56
  • @SomethingBetter as I closed your comment here changed my mind, it did not load off the bat. You also have provided a datasheet and such as part of your question, sorry for the momentary confusion. – Kortuk Jun 30 '12 at 14:32

2 Answers2

20

The usual places sound comes from in electronic circuits is inductors and ceramic capacitors.

The cross product of current and magnetic field is a force. Forces always work on two things, which in the case of a inductor are the core and individual segments of wire that make up the windings. At the right frequency, this can make the winding vibrate a bit, which you hear as sound.

Ceramic capacitors exhibit piezo-electric effect to varying degrees. The more efficient ceramics capacitance-wise are also more susceptible to this. If I remember right, barium titanate is particularly good at this since the titanium atom in the lattice changes between two energy states, which also cause it to change its apparent size. Yes, the ceramic is actually shrinking and growing very slightly as a function of voltage.

I just recently had a problem with this in prototypes of a new product. A power supply capacitor was subjected to 5-10 kHz ripple, which causes the whole board to make a annoying whining sound. I test five different models from different manufacturers, but all the ones that had sufficient capacitance had the noise problem. I have now reluctantly switched to a aluminum electrolytic for that part.

In your case your switching frequency of 1.5 MHz is way too high to be audible, so it can't be the switching frequency directly. Most likely your power supply is meta-stable and you are hearing the control fluctuations. There may not be much output ripple at the audible frequency, but you can probably see a little difference in the duty cycle at that frequency. At very low currents the control loop may be causing bursts of pulses with some dead time between bursts, which could have a strong component in the audible range. At higher currents the system is probably running in continuous mode and is more naturally damped, which is why the control response in the audible range decreases.

Also look at the current draw of whatever the power supply is driving. That may be in the audible range, forcing the power supply control response into the audible range too.

Olin Lathrop
  • 310,974
  • 36
  • 428
  • 915
  • 1
    Called "magnetostriction" in the inductor. I'd try covering it in glue goop. – tyblu Jun 30 '12 at 18:18
  • @Olin, thanks for the writeup. I did more debugging today. We have 3 of these regulators on board. I got a new board populated, with only one of these regulators. At its output, connected an artificial load with a resistor, such that it can draw anywhere between 30ma and 300ma. (Did not populate any other components on board such as the microcontroller, ram, etc, just one regulator). I don't hear anything in this case. So maybe the problem is not with a single regulator/inductor, but all of them working simultaneously? – SomethingBetter Jul 02 '12 at 12:04
  • Olin, have you tried C0Gs? They're more expensive, but show [no piezoelectric effect](http://www.kemet.com/kemet/web/homepage/kfbk3.nsf/vaFeedbackFAQ/A4EF6B58E22D378685257418004656B1/$file/2008-03%20Advances%20in%20Class-I%20C0G%20MLCC%20and%20SMD%20Film%20Capacitors.pdf). – stevenvh Jul 02 '12 at 13:32
  • @stevenvh: Yes, CoG would be good, but that ceramic is much less efficient and capacitors of the size and voltage I need aren't available, at least for prices I'm willing to pay. The whole board should cost about $18 to produce in lots of 1000. The one problem cap in particular was 10 uF and 35 V, and I would have liked a bit more capacitance. I'm replacing with a 22 uF 35 V electrolytic. – Olin Lathrop Jul 02 '12 at 14:13
  • 1
    @tyblu: Actually magnetorestriction is yet another effect than what I described. That is where the magnetic material changes mechanically due to applied magnetic field, much like the piezo effect when a electric field is applied. – Olin Lathrop Jul 02 '12 at 14:15
  • @SomethingBetter Wow! It has been nearly a decade, have you actually fixed your problem? – John Cortex May 15 '21 at 04:08
2

Your regulator is probably going into a low switching frequency mode at small loads to improve efficiency. This puts the capacitor vibration within our hearing frequency range. The other reason is that at low switching frequencies, the capacitor ripple voltage is higher thus increasing the amplitude of vibrations. It is hard to get around ceramic capacitors as they provide good density at low enough cost and good ESR, frequency characteristics. A good way to avoid this effect is to have 2 of these capacitors placed on opposite sides of the PCB. If you need 100uF capacitance, all you need to do is to place 47uF on top and 47uF exactly on the opposite side of the PCB. The effect of these capacitors' gets counteracted and the PCB doesn't make sounds any more. Way cheaper than using C0G or some other specialized capacitors.

Subhash SV
  • 31
  • 1