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.

TPA3116D2: Filter inductor overheating.

Part Number: TPA3116D2

Hi,

I have been using chinese built 2.1 boards (2x devices) for some time without any problem. Recently I have come across several where the output filter inductor (typ 10uH) overheats to the point of blackening. I now find all the units from different suppliers are doing the same. I have tried replacing with a 10uH bournes 2101-v-rc but this produced no improvement. When supplied from a 24V supply (24.2V) with o/c loads the board takes 400mA (2 devices remember) cutting two (bass one complete device) drops it to 180mA, running on one is 100ma and all removed is 50mA. Checking output frequencies I get 400kHz on one device and 120kHz on the other. Bizarrely, using a known good device the output frequencies are the same (which astonished me) but the no load current is 70mA, about where it should be. When connected to a loudspeaker system astonishingly it seemed to work quite well, until the smell of burning PCB makes you turn it off that is. Anybody got any idea what's going on?

  • Hi,

    There are other threads on this subject.  Such as this one: 

    You need to look at the output with an oscilloscope and make sure you are not seeing anything unusual, such a oscillation of the amplifier.  .

  • Hi, yes I did check it out with a scope hence the info I provided. No parasitics I can see, the output is stable and level and both sides bang on 1/2Vs. I have to say the waveforms were not what I was expecting on no-load. I was expecting the odd spike on each side to charge the filter to 1/2 Vs whilst what I got was a solid square wave on one output and nothing on the other. Most puzzling is the "bass" which is a doubled-up and has a frequency of **120**kHz. This is way off spec although elsewhere its been mentioned that the 1nF between sync and ground on the reference circuit in the datasheet is wrong and should be more like 47pF. whether this is enough to give this odd frequency I'm not sure. It seems to me that what is happening is thermal runaway of the ferrite filter inductors, but getting good information from the manufacturers is not easy. The big inductors I am using are hugely larger than appear on any commercial board yet they are getting warm. They are claimed as suitable for DC/DC converters 10Adc low core loss. Best I could find.

  • oz,

    i will try to separate out the issues with bullet points.

    • If you have a solid square wave on side you have a component issue.  I suspect the capacitor to ground in the output LC filter is open or the wrong value (very small)
    • The 120kHz is puzzling.  This low frequency would not make your filter get hot, but quite the opposite. 
    • My original thought was thermal runaway on the inductors due to saturation.  They may not be able to support the load.  Are they coupled inductors?  Coupled inductors cannot work on our amplifiers.

  • At the actual output (to external) I have DC, completely filtered. The chip side of the 10uH is where I have solid square wave. The inductors are simple toroids here 10uH: 1kHz and 6.3uH @ 10A (+30C) rated for dc-dc conversion should be good. I also considered thermal runaway due to heating of the core increasing core conductivity and this eddy losses, but this seems implausible at no load as there should be no input current, the output being balanced.

  • Oz

    From what I can see, the inductors are made with a powdered iron core, which has a very high saturation point.  From my experience with powdered iron and heating, is that wire may be shorting to the core.  This is rare, but I have seen it.  Need to check it by inspection.  Otherwise, it should not be getting hot. 

  • I agree. There may be a delay whilst I do a bit more digging and try some component swaps, which I need to get.

    In the mean time can you find out why the 'bass' paralleled chip is running at 120khz? I suspect its based on the TPA3116D data sheet design Fig 37 schematic with C41 at 1nF, which does seem high.

  • You can remove the 1nF and still be ok. A 47pF is just to clean up the signal a touch.

  • HI,

    My apologies for the delay. I did a spice analysis on the output filter and found (not entirely unexpected) rather high reactive current (7A with the circuit I used) flowing between the 10uH inductor and the 1uF filter capacitor when excited by a 400kHz pulse train and also a high voltage from Q-multiplication. Not entirely unexpected I guess although the spice model I am using is very simple and doesn't include the impedance characteristics of the output mosfets. Anyway that would explain why the inductors get hot. Obviously the lower the inductor resistance the higher the Q and the higher this current. When the pulsewidth varies in a loaded situation I would expect higher currents to be generated as there will be a greater proportion in a lower frequency (if I have the design right) at the ~1MHz resonant frequency. It seems to me that it would have been better to have the output filter (if in this form) resonant at about 1MHz (above the device clock speed) and then have downstream filters. The original filters only had a ferrite bead inductor which was presumably much less than 10uH and had a resonant frequency of the filter well above the standard design. Can you see any serious disadvantage to such a lower filter inductor?

  • Hi Oz,

    The disadvantages of a lower value inductor are a higher ripple current, less filter capabilities of the PWM, We recommend the values in application schematic so the PWM switching is reduced to be very small compared to the signal.  This also improves the EMI on the speaker wires.  

    You can send me your inductor part number and I will look into the filter design for you.