LM4871: noise - use your finger to remove it

Part Number: LM4871

Hello TI,

we use your LM4871LD audio amp on ten similar double-layer PCBs on our university. We have a strong noise on some of the PCBs. You can hear a "scratchy sound" when there is no signal input on the IN- Input and you can hear it also when music is played.
To eliminate the noise on that PCBs we have to touch the two 20k resistors on their solder areas next to the IN- Input. The sound output is then (almost) perfect.

We still did a re-design (see last page of the attachment) because we haven't used capacitors exactly as mentioned in the LM4871 datasheet. But the effect is still the same on some PCBs in the re-design.

Do you have an idea or re-design proposal what we can do to avoid that behavior in our actual design and in further designs?
Any help would be great.

All the best

Alexander Wilke

LM4871_noise.pdf

  • Hello Alexander, 

    Our team is looking into this thread and will respond as soon as possible. 

    Best Regards, 

    Justin Beigel

  • Hi Alexander,

    Is there a chance you can get a frequency analysis of the output and the input signals? This would help identify what is the nature of the noise.
    You mentioned some PCBs, does this means you have some of them with noise and some others without this noise?

    I would recommend to avoid right angles on the layout. You seem to have a right angle just at the point were you touch and the noise is reducing.
    Make sure capacitor voltage rating is sufficient based on the supply and signals you're using.
    I see a red area on the other side of the board from your captures, I assume this is GND. What are signals on the nearby connector? I don't see a return path for those signals.

    Best regards,
    -Ivan Salazar
    Applications Engineer - Low Power Audio & Actuators.

  • Hi Ivan,

    thank you very much for your reply and help. I think I can provide the frequency analyses next week. I will add it here.

    We use the LM4871 on some projects (two to four similar PCBs per project). Some of the PCBs show that noise and others (of the same project) do not show that noise.

    You are right, the red area underneath the LM4781 is the GND plane for the audio part. The two signals on the upper right corner belong to a FTDI USB to serial converter which isn't used so far. Their GND return path is outside the capture to the right - I agree that this is bad. Do you think that the (so far unused) traces can affect the LM4871?

    Just to be sure, do you mean that right angle? It seems hard to avoid that due to the two resistors connected to the LM4871 pin. Would a small polygon be better?

    I will prepare some alternative layouts next week together with the frequency analysis.

    All the best
    Alexander

  • Hi Alexander,

    Let's check your updates next week.
    In the meantime, yes I think a small polygon would be better, or pushing the components a bit to the top so the connection from the device is a diagonal itself. Regarding the FTDI signals, if they're not currently used I don't it's a problem now, but I was concerned about possible radiations from those signals to the input of the amplifier.

    Best regards,
    -Ivan Salazar
    Applications Engineer - Low Power Audio & Actuators

  • Hello, 

    Since we have not heard from you in awhile, I am marking this thread as closed. Feel free to respond here or open a new thread if needed. 

    Best Regards, 

    Justin Beigel

  • Hello Ivan, Hello Justin,

    sorry for the late response. You can find the analyze attached with some more information. Here the quick answer:

    1.
    I checked our design and figured out that the main noise is produced by the two capacitors C26 and C27 (see our schematic) which are placed between the LM4871and the speaker connector.
    The primary noise disappears when we remove these capacitors. Hopefully you can give some more insight here.

    2.
    There is still a small noise left which is constant and not amplified with increased sound volume. If the sound volume is loud enough it is inaudible. We tested some additional capacitor placements without any effect on that remaining noise. It is also not detectible via our oscilloscope/FFT. Is it possible that that remaining noise is generated internally by the LM4871?

    3.
    I added the TI design proposal for the Ri and Rf resistor placement to avoid that 90° angle. I also noticed that there is no GND connection between GND PINs and the GND exposed pad (so far we did so...). Hopefully you can give some more insight here too why it is better to avoid that.

    Thanks a lot!

    Alexander

    LM4871_noise_analyze.pdf

  • Hi Alexander,

    I'll take a look at your updates and provide some feedback as soon as possible.

    Best regards,
    -Ivan Salazar
    Applications Engineer- Low Power Audio & Actuators

  • Hi Alexander,

    I have reviewed your analysis, here are some comments:

    • Haven't seen this kind of behavior due to capacitors at the output, although this is not common for analog amplifiers. Actually since this amplifier is analog, you would not need ferrite+capacitor at the outputs, this is usually implemented in switching amplifiers to reduce emissions.
    • Have you tried isolating the device from the input audio source? Perhaps C21 and check the noise. I understand the noise may not come from the input as it is not amplified, but just to double check. There is no oscillating or similar inside the device so I would not expect it to generate this kind of periodic noise/signal. Perhaps the noise is coupling from GND or supply?
    • Actually data sheet does not specify to short pad to GND, or not to do so. I think best would be to follow the EVM design and have similar routing to avoid 90 deg angles and to use an isolated copper area for heatsink, separated from GND.

    Best regards,
    -Ivan Salazar
    Applications Engineer - Low Power Audio & Actuators

  • Hi Alexander,

    It has been some time since last communication. Please let us know if you have any further questions.

    Best regards,
    -Ivan Salazar
    Applications Engineer - Low Power Audio & Actuators

  • Hi Ivan,

    thank you very much for your feedback and help.
    I removed the C21 capacitor but the noise is the same with the two caps (C26, C27) in place. But it disappears when they are removed so I will stick to that.

    So far the main issue is solved. I will follow the EVM design (exactly) in our next version of the PCB to see, if I can further improve the output behavior.

    Best regards,
    Alexander