• State Resolved
  • Locked Locked
  • Replies 6 replies
  • Subscribers 62 subscribers
  • Views 1563 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LM2674 - Audible Inductor Noise

Byron Scheerer
Prodigy 120 points

I'm using two different LM2674M-ADJ regulators to provide 12V and 3.3V respectively. I get a very audible wine from the 3.3V regulator that does not come from the 12V part. I've done comparisons between the two in an effort to understand why one is fine while the other is audibly noisy. Please be aware that both are using a 680 uH inductor from Bourns (SRR7045-681M). It has a SRF = 3 MHz which appears to easily be high enough for the 260 KHz switching frequency involved. I have had customer returns because the audible noise coming from the inductor on the 3.3V regulator is so bad (seems to get worse on some units over time). Inductor was made common to both to help reduce unique component count.

Current drawn from the 12V part is 75 mA, while the current drawn from the 3.3V part is 145 mA. I can see from the switching waveform across the flyback diode that it does not change, ie: constantly 260 KHz. The duty cycle of the switching waveform on the 12V part is 52% and the 3.3V is 16%. I've scoped the switching current waveform and know for sure the inductor is not saturating. I've also tried adding extra loading to the 3.3V part thinking that might have something to do with it (since I know other switchers vary the switching frequency based on loading). This particular part seems to switch at 260 KHz regardless of load. My best guess is that the duty cycle is more related to the required output voltage then anything else.

Am I missing something about the design that would explain why one behaves very different than the other? At this point the only conclusion I can draw is that inductor construction is to blame. I have swapped in other manufacturer's parts which do seem to behave better. Before I pin my hopes on this being the solution, I wanted to make sure there is not some other little detail I'm missing that could explain it.

Schematic looks like this:

Any insight would be appreciated.

  • 0
    Prodigy 120 points
    Just to be clear, the audible noise is coming directly from the 3.3V output inductor and not the regulator itself.
  • 0 in reply to Byron Scheerer
    TI__Expert 4365 points

    Hello,

    Referring to the inductor value selection table on page 14 of the data sheet; the range of recommended inductor values for a 12V output is 22 to 220 uH and for 3.3V output from 22uH to 150 uH. So it should be possible to select a common value for both output voltage designs. However the selection of an inductance value that is substantially higher than the recommended value (680uH in your example) will cause the inductor ripple current to be too small and the corner frequency of the output filter to be too low for proper stability based on the internal compensation circuitry. This will be a greater problem with the lower output voltage case which is the situation you describe. It is common for regulators to buzz or hiss when they are operating in an unstable condition. This situation can be remedied but following the guidelines for component selection detailed in the data sheet.

    I would suggest lowering the inductor values to something in the 100uH or 150uH range and also refer to the recommended capacitor values associated with each output voltage. The recommended capacitors are shown also in the table on page 14. Keep in mind that the selection of capacitors shown in the data sheet was chosen many years ago when the IC was originally introduced.

    Alan Martin - Simple Switcher Applications manager

  • 0 in reply to Alan Martin
    Prodigy 120 points
    Thank you very much for the feedback Alan, it's greatly appreciated. The 680 uH value was arrived at based on Webench approximately 1 year ago. I don't recall what the input parameters where that I provided at the time, but I obviously didn't look close enough at the datasheet to know what was a reasonable range in comparison to what Webench was suggesting.
    I had been experimenting with a 100 uH inductor last week (since I had one in stock) which has an alternate construction. It has been behaving very well from an audible noise perspective, however the output ripple voltage went from 20 mV PK-PK to 80 mV PK-PK. This was not un-expected, and I have added 100 uF of capacitance on the output to get it back down to 20 mV. Duty cycle remains un-changed, but I'd appreciate if you could comment on this......is there any concern with the duty cycle being down at 16%? I want to be able to understand how to properly evaluate the operation so I know it is stable beyond not hearing inductor hum.
  • 0 in reply to Byron Scheerer
    TI__Expert 4365 points

    Hello again,

    If you are referring to the switch duty cycle it is always approximately Vout / Vin in conventional non-synchronously rectified buck converters. [ 3.3Vout / 12Vin = 27.5% or more like 31% if you add the diode drop loss term to Vout in the equation.  (Except at lighter loads when the inductor current goes discontinuous.) If you are seeing only 16% then either are at light load or there might be an error in how it is being calculated.   Also this is a constant frequency device so the frequency is always the same unless the load current is so low that it starts skipping cycles. I have to tell you that the most important information present in the inductor current waveform. This should be done with an ac-dc current probe compatible with your scope. I wish current probes were low cost but sadly they never are, even when you risk buying one on the used market.

    Send us a scope capture and we may be able to resolve the question regarding 16% switch duty cycle. Is this at precisely 12V input, or is the input voltage higher?

  • 0 in reply to Alan Martin
    Prodigy 120 points
    The input voltage is 24 VDC max, but since it is being powered from our central equipment, this goes down proportionally based on the power loss along the line. It could go down as low as 13-14 VDC, however all my tests where done connected to a 24 VDC source externally. Therefore 3.3V/24V works out to 13.75%. There are some drops across some FET's etc inside that would take the input voltage a bit lower that the regulator sees so this would explain the difference between 13.75% and measured 16%.Likewise on the 12V regulator (12/24 = 50%). I didn't realize until you pointed it out that it is dependent upon input voltage but it does stand to reason.
    Unfortunately we don't have the type of probe for the scope you describe. We do the poors mans method of adding a current sense resistor (0.1 ohm) in series and scope across it to view the current waveform. Either way, I think I'm in good shape with the information you provided. Now it's down to a simple matter of the appropriate output capacitor(s) as per the datasheet to go with the 100-150 uH inductor.
    Thanks for your insight!
  • 0 in reply to Byron Scheerer
    Prodigy 120 points
    One final question. How do you determine what is too low a corner frequency for the output filter? I'm finding that the ideal output capacitance to keep the ripply voltage down to a respectable 20 mV Pk-Pk is 150 uF, but going by the datasheet, using table 8 and 9 it would suggest 100 uF. This doesn't change the corner frequency much, but it does take it lower......... and I don't understand how low is too low. If I use 100 uF (2 x 47 uF populated) the output ripple is still 40 mV Pk-Pk.