• State TI Thinks Resolved
  • Locked Locked
  • Replies 20 replies
  • Answers 1 answer
  • Subscribers 49 subscribers
  • Views 2845 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

LMV321: In the process of production, the IC LMV321 used for amplification function is 16% bad, and there will be noise.

2018
Expert 2630 points
Part Number: LMV321
Other Parts Discussed in Thread: , LMP7707, OPA314, TLV6741
This is a wired headphone product, which uses two LMV321.
One LMV321 is used for comparison and the other for amplification function . Noise problems will arise when it is used for amplification function .
After checking, the problem appeared in the LMV321 amplifier.  
In the process of production, the IC LMV321 used for amplification function is 16% bad, and there will be noise.  
   
Remove the problem after replacing a good LMV321   
     
Install the bad LMV321 on the good product, and the good product will have noise as well.        
Impedance measurement (After remove from the board)
the pins vs. GND comparison between failed IC and good IC.
Bad waveform

Good waveform:

  • 0
    TI__Guru**** 165166 points
    Hello Clark,

    The signal level in this application is low and the amplifier gain is high. So noise can be a significant factor. There will be some improvement by using the next generation LMV321A that is a low noise CMOS process. If even lower noise is needed, I can look for a device with a maximum noise specification.

    Alos be sure that VCC pin is greater than 2.7V as some voltage will be lost across R245.
  • 0
    Guru 140200 points

    Hi Clark,

    the inverting OPAmp circuit you are using does not allow a very low noise operation. And the LMV321 isn't a low noise OPAmp either. You should use a non-inverting OPAmp circuit. This will give you the chance to choose much smaller and by this much less noisy feedback resistors in the amplifier circuit. And you should use a low noise OPAmp.

    Kai

  • 0 in reply to kai klaas69
    TI__Guru**** 165166 points
    Clark,

    If the response from myself or Kai was helpful, then mark the thread resolved. If you have further questions then add anew reply to this thread.
  • 0 in reply to kai klaas69
    Expert 2630 points
    Which op amp is more suitable? Is there any recommendation?
  • 0 in reply to 2018
    Guru 48395 points
    Hello Clark,

    It is hard to tell what your noise gain really is without the Telecoil impedance. But, lets assume the gain is high. Couple of points

    1. Even though your nominal design is probably pretty high output noise, what you are saying is some parts "work" and others do not? That is more of maybe a part to part low frequency noise or popcorn issue perhaps.
    2. If you gain is in fact high, the fastest way to lower noise for your solution (along with lower resistor values) would be to use a decomp part. Not sure if the LMV family as a decomp part, but take a look at the LMP7707 - might be too high a power for your app but maybe Ron can suggest a better decomp low power, low cost option.
  • 0 in reply to 2018
    Guru 140200 points

    Hi Clark,

    I think the OPA314 is ideal for your application. It offers very low noise and very low supply current at the same time. And it offers a bit more speed.

    Kai

  • 0 in reply to kai klaas69
    TI__Guru**** 165166 points
    Clark,

    OPA314 is a good choice. I prefer TLV6741; reducing noise was a key factor in its design. R243 and C217 are set to 73kHz, if you need less bandwidth then C217 can be increased to reduce total noise level.
  • 0 in reply to Ron Michallick
    Expert 2630 points
    we tested it tgis way,and it didn't work.
  • 0 in reply to 2018
    TI__Guru**** 165166 points
    Clark,

    Which way did not work? Are you referring to increasing C27 to reduce bandwidth and total noise did not having any change or not enough noise level change?
  • 0 in reply to 2018
    Guru 140200 points
    Hi Clark,

    I would decrease R245 to 100R.

    Kai
  • 0 in reply to kai klaas69
    Expert 2630 points

    When R245 is changed to 0R, the problem persists           
    When R245 is changed to 4.7K, the voltage on IC is 3.17V. The problem persists.

  • 0 in reply to kai klaas69
    Expert 2630 points
    Do you mean that this LMV321IDCKR is not suitable for this application?
    So what's the solution to this noise problem in this application?
    We tried the above method without any effect.
  • 0 in reply to 2018
    Guru 140200 points

    Hi Clark,

    have you tested the OPA314 I recommended you here?

    e2e.ti.com/.../2902306

    The "telecoil" in your schematic, what is it? Can you give a link to the datasheet?

    Kai

  • 0 in reply to kai klaas69
    Expert 2630 points
    LMV321IDCKR is already in mass production. Is this model not suitable for this circuit? If the IC model is not replaced, is there any other way to improve it?
  • 0 in reply to 2018
    Guru 140200 points
    Hi Clark,

    if the circuit is already in mass production what do you want to change then??

    Kai
  • 0 in reply to kai klaas69
    Expert 2630 points
    In the process of production, the IC LMV321 used for amplification function is 16% bad, and there will be noise.
    Can the noise problem be solved by changing the peripheral resistance or capacitance?

    Is this IC not suitable for this application? Please confirm.Thanks!
  • 0 in reply to 2018
    Guru 48395 points

    Hello all,

    There might be some semantic mistakes here - if you look at the scope waveforms, the upper one (bad channel?)  looks like it has an oscillation superimposed on the "good" part sine wave down below it. So random noise may not be the issue. Cannot do a complete simulation without the Telecoil model, also, would really need the load on the other side of the 100nF C215.

    The LMV321 does not appear to have a TINA model in the TI web folder, but it was in the TINA  V11 library I have. Not sure the op amp stability itself is an issue - here are some files

    But, the point is 84% parts work apparently - so something is right on the edge that is introducing that superimposed oscillation in 16% of the channels. Either that, or there is enough variation in whatever op amp parameter is related to this to cause fallout - so, you have to isolate where that loop is coming in and fix it, or use a different part. For a bad channel, one way to is probe with a leaded resistor on the circuit with your finger capacitance being added to the loops - if you probe where it changes the frequency, you are in the loop. Another interesting test would be to take a "bad" channel into a spectrum analyzer to see what unwanted frequencies might be showing up?

    LMV321 fallout discussion.docx

    LMV321 closed loop.TSC

  • 0 in reply to Michael Steffes
    TI__Guru**** 165166 points
    Clark,

    Can you get waveforms with no input signal. It'll be much easier to see the oscillation or noise if the input signal is not there.
  • 0 in reply to Michael Steffes
    Guru 140200 points

    Hi Michael,

    hmm, maybe I'm wrong, but I don't think that this is some sort of oscillation. The LMV321 looks like it is not able to properly reproduce the sine signal. As if it makes mistakes.

    Many years ago, when handsoldering of SMD parts failed from time to time due to overheating I have seen something very simlar with a cracked ceramic capacitor in the signal path. It looked like a signal with huge and random distortion.

    So, I think the LMV321 or some of the passive components became damaged due to an overheating during the soldering.

    Kai

  • 0 in reply to kai klaas69
    Guru 48395 points
    Hard to say but a reliable 16% fallout in production does not sound like a random soldering issue. I have seen many different kinds of oscillation, some that go through a kind of FM effect where it is not a single tone. Others that ramp up in amplitude, self heal, and then start over. This may not be a classic single tone oscillation, but something is going on that does not look like noise. These are tough ones if there is a very slight part to part variation in a circuit hanging on the edge already. Normally you would get suspect parts back for re-test. They all go out tested, so if they fail something has hurt them after they shipped. And that happens and would be worth checking - so this is really a return part for FA task also. They will visually inpect first then run through ATE - normally, they pass or they would not of shipped - if they fail, something is stressing them and that should be fixed first.

    Another long shot (that has happened) is production mixing TI with counterfeit parts at the 16% level?