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.

TPA3126D2: Amplifier distortion in output waveform.

Part Number: TPA3126D2

I'm working on an amplifier using the TPA3126 in BTL operation.  I'm running it off of a 24VDC rail.  At all times the input signal and power rail look clean.  If I put in a 700mVrms input signal to the amp, the output waveform looks clean.

I

If I bump up the input to about 800mVrms I get the following waveform:

The schematic for the amplifier is identical to the schematic of the EVM board. 

I will also note that the amplifier is going through a thermal cycle, that's set up for self recovery while both waveforms were created.  

Any ideas?

  • I forgot to mention its connected to 4ohm loads to simulate speakers, and the amp chip is set for 20dB gain, and no power limit.

  • Hi 

    I think TPA3126 was OT shutdown. You can add one fan to decrease the heat-sink temperature.

    Regards,

    Derek

  • Hi Derek,

    Thanks for the reply.  I don't believe so.  The Fault line is tied to the SDZ line, so when I say its going through thermal cycle, the fault line goes low during over temperature, then the SDZ line clears it when its pulled low allowing the unit to recover (turn back on) after the die temperature drops.  This does not explain the waveform from what I understand of the part.  In fact, I've used this chip in other designs and when similar thermal cycling occurs I don't see this waveform.  Any other thoughts?

  • Hello,

    Do you have a TPA3126D2EVM, can you try running the same input signal through that and seeing the behavior of the output.

    Can you Try disconnecting the Fault line and the SDZ line from each other and seeing the behavior of the output.

    Can you try resoldering the IC, a faulty soldering connection could cause this issue.

    Best Regards,

    Luis

  • Hi Luis,

    As a matter of fact I do, and I did this very thing yesterday.  I pulled my amp supply and signal lines from the board in question and ran them into the EVM board, and everything appeared normal.  Ran the unit to the max output.  As for re-soldering the chip, I originally thought it was a bad amp.  It was producing the same waveform.  I replaced that chip on the target board and still had the same issue, so if its a bad component in the amp circuit, its a different component.  I wanted to avoid replacing each part one at a time, so was hoping to get some idea of what part of the circuit to operate.  Also, yesterday I replaced the EVM board with a second amp assembly with the same layout.  It too operated as expected when powered from and signal from the board in question, so this appears to be a failure of this specific board.  I still need to understand exactly what the failure is though.

    Thanks,

    Michael

  • Hello Michael,

    Can you upload your layout file.

    Best Regards,

    Luis

  • Hi Luis

    Here's a cut of the bottom layer with the amp chip on it:

    Note:  This is a 4 layer board and the via stitching ties through the ground plane on all layers.  Here's what the opposite side of the board looks like:

  • Hi,

    The Boost pin trace should be wider.

    Did you try to use EVM board inductors on your board?

    Regards,

    Derek

  • Hi Derek,

    The board is using the exact same inductors as those on the demo board.

    Thanks,

    Michael

  • Hi,

    Did you solder other devices on your board? Do the other devices have same phenomenon?

    I checked the PCB layout, the Boost trace should be wider. 

    If every device has same phenomenon, I thins the issue is related to the PCB layout.

    Derek,

    Regards

  • I've only seen this issue on this one board.  So in looking at the demo board, the boost trace appears to be the same width (or close anyway).  What should the boost trace width be?

  • Hi,

    Did you see other chips have same issue on the same board?

    Maybe the issue chip was damaged by other accident. I did not see same issue before.

    Regards,

    Derek

  • I haven't noticed this before on my other amps, and my first thought was that the amp chip was damaged.  The first thing I did was replace it and still got the same waveform, so something else in the circuit has failed.  The waveform exists on each of the 4 outputs, so its not a failure of the boost coupling cap as near as I can tell.

  • Hi,

    Firstly, when the input signal increases to 800mV, the device go into OT Fault, Fault will pull-low the SDZ pin, then the output goes to Hi-Z. And you can see the Hi-Z output on waveform.

    When the temperature decrease, the Fault disappeared, and Fault pin goes to high. The device will go to work again. 

    However, if you did not handle the MUTE pin (MUTE pin always low), at the SDZ pull-high, there will be pop-noise. So the abnormal waveform is the Hi-Z to pop-noise happening. We suggest that before SDZ pull-high, device should be mute, MUTE pin should be high to avoid pop noise.

    The EVM board does not have the issue, because EVM board has triode on MUTE pin. You can see Q1 on EVM board.

    When Fault happens, Mute pin will be high and device is on mute condition. After Fault disappears, the Mute pin goes to low, device is on normal condition.

    So on your board, when OT Fault happen, you must mute the device at once. When Fault disappears, un-mute the device.

  • In fact, my board is handled the exact same way as the demo board  The pull up on the fault line is also pulling up on the FET that is holding the mute low (or off).  When Fault occurs, th e Mute is automatically applied..

  • Hi

    Please check the sequence between SDZ, Falut and MUTE pin.

    Try disconnect the connection SDZ an Fault, make sure the device is mute condition before SDZ pull-high.

  • I guess I don't understand what you are going for here.  First, the operation is identical to the demo board.  The MUTE line is automatically exerted when the Fault occurs (goes low) and remains muted until the Fault is cleared.  I am not sure why you think this would be associated with a "pop", because it occurs on every positive half cycle (or just about every cycle) until the next thermal shut down.  It does not occur at slightly lower levels (see original post).  I'd expect a "Pop" to occur on the first couple of cycles only.

  • Hi 

    As you know, I do not have your board in my hand, I can not find the root cause at once only according the waveform. I need more information.

    Maybe your design is almost same as demo board, but we must make sure the sequence is same. 

    Please check the SDZ, Fault, Mute pin sequence difference between demo board and your board. Capture the sequence waveform.

    We can not find the root cause just according to the information you provide before. 

  • As I have previously stated, the SDZ, FAULT and MUTE operation is not almost the same, but exactly the same as the demo board (as I showed in the schematic snippet I attached).  The Mute line is activated upon a FAULT condition.  The MUTE line is released upon clearing of the FAULT condition.  I have other circuitry that can cause the MUTE to become active, but the over riding condition requires a NON FAULT state for that circuitry to work since it is all fed by the pull up on the Fault line.

    But as you asked, the top trace is the FAULT line, while the Bottom Trace is the MUTE line.

  • Hi,

    Today I checked on EVM board and also talked with design team.

    Firstly, I did not reproduce the issue on EVM board. And the abnormal waveform is not from auto recovery. I added comments on the abnormal waveform picture which you provided to us before.

    Secondly, the suggestion test points are below:

    1) Does the audio input connection of your board is same as EVM board? Please check the audio input waveform once the issue happened. I doubt that the audio input disappeared suddenly.

    2) Capture the waveform of audio input, PWM output, audio output at same time.

    3) change the BST caps to check if any improvement. You can mount the BST caps of EVM boards on your board.

        

  • Ok, so let me say that this poor board has a lot of hand add changes to the front end circuitry prior to the amplifier.  Outside of me changing out the amplifier chip, that section all remains as originally built.  I can tell you that the first thing I did when it started showing this waveform was to check the signal into the amplifier and it was normal.  Today I brought up the board to get the waveforms you requested, and the amp symptoms have totally changed.  It looks like it is on its last legs.  I have six new boards I should receive tomorrow.  I'll check them out (since they won't have these hand adds) and report back if any of those boards have these symptoms.  Thanks for the help so far.

  • Ok, I've completed the start up of the latest build of boards and have not seen a repeat of this issue.  I'm going to close this for now and attribute it to a one off issue on this specific board.  Thanks for all the input and advice.

  • Update:  The new boards developed an extremely similar issue.  The waveform is clean until you hit a level, then starts to generate the below waveforms.  The top trace is the actuall output across a 4ohm Load, the yellow is the output from the chip itself. 

    This second screen shot shows the input waveform on the bottom with the output on the top.

    Any thoughts?

  • Ok, I finally actually found the smoking gun.  The design has a Power loss mute circuit that monitors the switched input power.  When the input power disappears, the PVCC line starts to fall out.  At that point, we mute the amp to eliminate any noise during rail collapse.  Well, apparently, as you increased the input signal, the draw of the amp was causing a voltage ripple on the switched input power line.  This was actually triggering the amplifier mute line partially during the individual cycles causing the above waveform.  I did not realize that the mute line would be that quick to as mute partial cycles.

  • Hi,

    Thanks for providing the root cause to us. It also is helpful for us.

    Thanks again.

    Regards,

    Derek

  • Hi Derek

    As a user of forums like this to get information on designs, I do understand the importance of closing the issue out, even if it ends up in a totally different direction than originally thought like this one did.  You are welcome for the information, and I hope it helps others.  My only comment on the part is that I would have expected that the mute would be implemented at a zero cross to avoid any pop.  That is part of the reason I hadn't considered a mute issue.  In any case, thanks for all your help and pointers.