• State
  • Locked Locked
  • Replies 12 replies
  • Subscribers 66 subscribers
  • Views 5912 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.

LP5907 noise

Eric Peters
Prodigy 120 points
Other Parts Discussed in Thread: LP5907

I am having an issue with noise generated by the LP5907SNX-3.3/NOPB LDO on a current project.  This part was selected for its ultra-low-noise characteristic.  However, I am seeing strong periodic noise generated at the output.  Below is a simplified schematic:

 

 In my circuit, VIN is 4.0 V and VOUT is 3.3 V.  The frequency and level of the periodic noise appears to vary from component to component.  In some devices it is a broadband hump and within 10~20 dB of the noise floor.  In the worst devices, it is a fairly narrow peak with strong harmonics.  The periodic noise is load-dependent and roughly inversely proportional with the R and C values.  Thus, I can move the noise out of my band of interest (24 kHz audio) by placing a small parallel resistive load in my circuit.  When I use a ~1 kΩ load, the periodic noise moves completely out of band and I see a relatively flat ~140 dBV noise floor.  I chose a relatively aggressive 1 kΩ for my project, because there is part-to-part variation in the noise and I want to be confident that the solution will work across all devices.  However, that is an extra ~3 mA that is unnecessarily dissipated.

 

The output and input capacitor part numbers are

C1           GRM31CR71C106K

C2           GRM21BR71E225k

 

I did read in the LP5907 datasheet that a very low ESR output capacitor is required for stability, so I tried a variety of lower-ESR capacitors and parallel capacitor combinations, but this seemed to have no effect on the noise.  The datasheet says that the input voltage should be at least 1.0 V higher than the output voltage, so I swept VIN from 4.0 V to 5.0 V, but observed no effect on the noise.  There is also a note suggesting that the input capacitance should be at least as large as the output capacitance, so I increased the input capacitance to 10 μF and observed no effect on the noise.

 

Does someone have any thoughts on why this is occurring and if there are any alternative methods of mitigating the noise?  The total noise I am measuring is about 570 μV, significantly higher than the 10 μV in the datasheet.

  

VIN noise:

 

VOUT noise:

  • 0
    TI__Mastermind 22610 points

    "... strong periodic noise generated ..."

    This does not sound like noise.

    This is probably not from inside the LP5907. It is more likely the LP5907 response to something external. Knowing more about the load and conditions might be helpful.

    The fact that adding 1k load resistor changes the ~noise~ suggests a couple of items:

    1) LP5907 is not happy with only 66uA of load current

    2) The load pole (1 / (2 x pi x Cout x (Vout/Iout)) moves to a higher frequency, improving response. In this case the load pole is so low that this seems an implausable explanation

    3) There is some type of resonance on the ouput and adding the resistor damps out oscillation.

    The LP5907 does not have a Vout spec for less than 1mA load, but we have characterized Vout for a small sample down to 100uA load.

    If you put the LP5907 into the simplified circuit (i.e. just a 50k ohm resistive load) does this behavior persist?

     

  • 0 in reply to Donald Jones
    Prodigy 120 points

    The above measurement was taken with a 50 kohm potentiometer as the load.  The measured load is actually about 53 kohm.

    As I decrease the resistance on the pot, the frequency correspondingly increases.  At 38 kohm, the fundamental is at 10 kHz.

    Interestingly, as I continue to decrease the resistance, there is a point where the frequency seems to "skip" ahead non-linearly.

  • 0 in reply to Eric Peters
    TI__Mastermind 22610 points

    If load R is kept the same (50k) but Cout is changed from 10uF to 1uF (i.e 1/10th), does the frequency go up by 10x?

     

  • 0 in reply to Donald Jones
    Prodigy 120 points

    Decreasing Cout by about half (10uF to 4.7uF) does approximately double the frequency.  I have not checked 1uF, but I will confirm.

  • 0 in reply to Eric Peters
    TI__Mastermind 22610 points

    Then this is probably related to operating at much less than the rated 1mA load current with 10uF Cout.

    Control loop overshoots a little bit then has to wait for the 10uF Cout to discharge through the 50k, then repeats the cycle.

  • 0 in reply to Donald Jones
    Prodigy 120 points

    Is the recommended solution then to just use a 1uF Cout and >1mA dummy load?  Figure 13 in the datasheet leads me to believe that the noise has been characterized at 0mA load.

  • 0 in reply to Eric Peters
    TI__Mastermind 22610 points

     " ... Figure 13 in the datasheet leads me to believe that the noise has been characterized at 0mA load..."

    Possibly, with Cout = 1uF.

    I checked the archives and I didn't see any comments from the original characterization team regarding this chart. (Fig 13), or any design notes regarding issues for output loads near to zero mA.

    At this point the recommended solution would be to use a 1uF Cout.

  • 0 in reply to Donald Jones
    Prodigy 120 points

    I've made some more measurements.  As you can see, the noise is still apparent at light loads, even with Cout = 1uF.  I will move forward with 1uF, 1mA load as my solution.

    Meanwhile, would it be possible to review or amend the datasheet so that a minimum load requirement is made clear?

    10uF 150kohm 

    10uF 60kohm

    10uF 3.3kohm

    1uF 150kohm

    1uF 60kohm

    1uF 3.3kohm

  • 0 in reply to Eric Peters
    TI__Mastermind 22610 points

     " ... amend the datasheet so that a minimum load requirement is made clear ..."
    Yes, I had that discussion with management yesterday. Not sure if we can ~erase~ the 0mA line from the Noise Density chart (aka Figure 13), or if a new chart needs to be created from scratch.
    I set up your conditions (50k & 10uF) with LP5907UVX-3.3 EVM but could not duplicate this behavior. More investigation will be needed on our side.
    Just in case it comes to it, could you please advise me as to which Audio Precision audio analyzer you were using? I certainly don't have one available in the Power Management lab, but the Signal Path Audio group might have something close.

  • 0 in reply to Donald Jones
    Prodigy 120 points

    The analyzer is an AP 2722 ("System Two Cascade Plus").

    My project has four microphone input channels, with one LP5907 per channel sourcing mic/amp bias.  I took similar measurements of all four channels on one board.  The worst-case on this board is channel 2.  If I depopulate the LDOs and replace them on different channels, the noise profile follows the LDO.

    5531.LDO noise.zip

    I've also observed similar behavior on different boards, with each LDO exhibiting different noise profiles.

  • 0 in reply to Eric Peters
    Intellectual 570 points

    Hi,

    Has this been resolved for light load <1mA?  Anyone experiencing similar issue with LP5907?

    Cheers.

    Shanguo

  • 0 in reply to Shanguo Lu
    TI__Expert 8905 points
    Hi Shanguo,

    The notes were added directly into the datasheet, along with requirements on the caps. See the datasheet for the latest recommendations (capacitor requirements table is below the EC table, and more details are available in the apps section.

    -David