• State
  • Locked Locked
  • Replies 4 replies
  • Subscribers 30 subscribers
  • Views 420 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.

LM49450 unstable / random audio

dh
Prodigy 220 points
Other Parts Discussed in Thread: LM49450

We are using the LM49450 Audio DAC in a new product that we are having problems with.

We are using the device in I2S Slave mode and playing a 44.1 KHz 16-bit stereo track with an approximate 12 MHz MCLK.  We are not using any headphone functions or Analog Audio Inputs.  We are only using I2S with the Right & Left loudspeakers.

The problem we are having is that after a playing an audio track correctly for a while (can be a few minutes or it can be a few hours) via the I2S interface, the LM49450 suddenly generates what only can be described as very loud random audio. 

When the problem happens the ‘real’ I2S audio track can just about be heard, although the very loud ‘random’ audio virtually drowns it out.

Once this has happened the LM49450 is ‘stuck’ in this mode.  If we stop the I2S audio track, and in fact if we stop all I2S clocks, the loud random audio continues.

Whilst in this error state, if we look at the Right & Left speaker outputs, we see the waveforms switching randomly, without any correlation to the usual 300 KHz. 

Whilst in this error state, we can read the I2C registers and they are correct to the values that they have previously been set to.

If we disable the LM49450, by clearing Bit0 of Reg 0x00, and then re-enable the LM49450 by setting Bit0 of Reg 0x00, then the problem is cleared, and correct audio is heard.  Although, subsequently the problem will re-occur sometime later.

If we repeat the test but use the headphone output instead of the Class-D speaker output then the problem does not occur.

This implied the problem is somehow related to the Class-D ouputs ??

Any thought or help would be appreciated.

  • 0
    TI__Guru 55556 points

    Is this dependent on the power output (does this happen regardless of volume)?

  • 0 in reply to Dominik Hartl11
    Prodigy 220 points

    Hi D.Hartl.

    Thanks for showing an interest. 

    Your question regarding the power output (volume) is one that we have asked ourselves. 

    We don't have a definitive answer yet, as sometimes the units take hours to fail.  We are still testing multiple units to get a better picture.

    We can state the following:

    a) The LM49450's SPEAKER VOLUME CONTROL is always set to 0x1D (e.g. 19.5 dB). The volume is not varied using this register.  The volume is then controlled using WIndows Media Player volume control. 

    b) The speakers are 8 Ohms.

    c) We have had failures at low volumes.  WMP is set to a volume of 10%.

    d) We have had failures even when the speakers are disconnected.  Nothing at all connected to the LM49450's speaker pins.

     

    What prompted your question?  We had explored this question with a theory that the failures 'could' be some form of 'hiccuping' protection mechanism, although we find it strange that this causes such loud random audio.  The datasheet mentions 'short circuit' and 'thermal' protection, BUT it provides no specific details.  Nothing to say under what conditions it kicks in, or if it self-recovers. 

    Dave

  • 0 in reply to dh
    TI__Guru 55556 points

    From your description, it sounds as if something stresses the device (or the system - for example the power supply) resulting in the class-D modulator misbehaving grossly.

    The rationale for my question was that when you stop the I2S interface, the noise continues, which points to a rather drastic failure mode. If this happens at full volume settings, the part will draw fairly high currents for loud audio signals, which may cause short drops in the power supply, which may result in unpredictably behavior. The fact that it recovers when you disable/enable the part also supports this assumption.

    However, because this also happens when you disconnect the speaker altogether, it does not appear to be related to the device stressing the power supply.

    In any case, this sounds like there's something wrong on the system level or that the device is damaged. Please check that all power supplies are within the required limits (all the time - even short glitches are a problem).

  • 0 in reply to Dominik Hartl11
    Prodigy 220 points

    Hi D.Hartl,

    We are pretty confident it is nothing to do with the device's power supplies.  This was the first thing we looked at when we started investigating the problem.

    The devices are not damaged when the problem occurrs.  We have left them in this state for hours and as soon as we reset them the audio is correct again.

    Our standard product has a ferrite bead on each speaker output close to the LM49450. 

    Over the weekend we have modified x2 units by adding a 680 pF capacitor to GND on each speaker output (x4 in total), on the speaker side of the ferrite bead.  In other words we have the LM49450 then the FERRITE BEAD then a 680 pF to GND, all close to the LM49450.  We tried this having run out of ideas and because a number of Class-D amplifier data sheets always show this arrangement - even though the (filterless) LM49450 doesn't.  We understand that this is normally required to improve EMI issues.  We have already passed compliance testing and did not have any EMI issues, hance we had no need for any additional filtering. 

    Surprisingly none of the modified units have failed after 3 days of continuously playing the same audio track!

    We have now modified a further x2 units, and continue to test.

    We do not understand why these capacitors make a difference? Is there a feedback-loop on the Class-D amplifiers within the LM49450 that is somehow jamming itself?  Remember that we can get the problem to occur when we have nothing at all connected (no ferrite bead, no speaker, no 680 pF cap) to the LM49450!