Noise Analysis Help

Discussion in 'The Projects Forum' started by Eoin_oc, Dec 21, 2015.

  1. Eoin_oc

    Thread Starter New Member

    Oct 24, 2015
    23
    0
    Hi all,

    I have a current meter design using a 2-Ohm sense resistor and 2 op-amps.
    The op-amps each have a gain of 11 for a total overall gain of 121. I am basically using the design from EEVBLOG current meter.(http://www.eevblog.com/projects/ucurrent/)

    My PCB is powered from 5V USB, this 5V input is first regulated by a 5V LDO (incase a higher supply voltage is supplied).
    This 5V LDO drives a 4.0V low noise LDO LP2985 that powers the MAX4239 op-amps and a 3.3V LDO for some digital stuff.

    The final voltage output from the MAX4239 drives an AD7741 Voltage-Frequency converter, the frequency output is counted by a PIC24F to measure the current. I'm using the V-F approach as I want an average current measurement not an instantaneous one.

    The system works very well and seems to show great linearity but there is a few uA of noise(+- 3-5uA) on the reading. My question is how can I begin to analyse the noise of the system? I have at my disposal a digital scope with built in FFT and an RF spectrum analyser that can measure down to 2Hz.

    What is the best way to find out where this few uA of noise is coming from?
     
  2. Lestraveled

    Well-Known Member

    May 19, 2014
    1,957
    1,215
    OK, we are going to need a schematic of your circuit. Also a picture that shows layout, grounding and shielding.
     
  3. Eoin_oc

    Thread Starter New Member

    Oct 24, 2015
    23
    0
    I have uploaded my layout files.
    I don't really want to upload my schematic as its a mess firstly and and I would prefer not to share some of it.

    I have 4 separate planes on the PCB:
    GND
    DIGITAL_GND
    VDD_ANALOG
    ANA_GND

    DIG and ANA GND connect back to GND through inductor chokes.

    The AD7741 uses the 4V regulator as its REFIN voltage and I'm using a 6MHz XTAL.
     
  4. Lestraveled

    Well-Known Member

    May 19, 2014
    1,957
    1,215
    Since you have not provided any details, I will make some general statements. Noise enters circuits via two paths, inherent noise energy (noise figure) or the result of poor design.

    Noise figure - All matter generates energy (noise) when it is above zero degrees Kelvin. The formula, KTb describes that energy. Do you understand this formula and how it applies to electrical design?

    Bad design - There is universe of ways to make a quiet circuit noisey.

    So, it is impossible to assist you with your project without you "sharing" the schematic and more construction information. You took the basic design from EEV, so, why so secretive?
     
  5. ronv

    AAC Fanatic!

    Nov 12, 2008
    3,292
    1,255
    What happens if you do 10 times more averages?
     
  6. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    2,689
    2,750
    So you've got 15 nA RMS (or 30nV on 2 ohms) input referred noise. How does this compare to your expectations?

    What is the full-scale input current?
     
  7. Eoin_oc

    Thread Starter New Member

    Oct 24, 2015
    23
    0
    The full scale input is around 15mA at the moment.

    I have attached a zip file with a hand drawn schematic.

    I forgot to include the 6MHz XTAL on the V-F converter. There is also a noisy LCD and PIC24F hanging off the VDD Digital regulator, it seems to have some kind of charge pump for producing some voltage needed for its operation
     
    Last edited: Dec 22, 2015
  8. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    2,689
    2,750
    120dB signal-to-noise ain't all that bad!

    Again, what are you expecting?
     
  9. Lestraveled

    Well-Known Member

    May 19, 2014
    1,957
    1,215
    OK
    Your amplifiers do not have enough gain (41.6 db) to be bothered with noise figure.

    The op-amp input noise density appears to be a bigger player. The MAX4239 has a input noise density of 30nV/(square root of HZ). You are doing little to limit the bandwidth so lets do some math:
    30E-9 * SR of 6.5E6 = 30E-9 * 2.55E3 = 76.5 microvolts. This is the equivalent input noise at 6.5 Mhz bandwidth.
    Multiply by the total gain: 76.5E-6 * 121 = 9.3 milli-volts. This is the noise output (RMS) of your amplifiers with the input terminated at a bandwidth of 6.5 Mhz.

    Do you really need 6.5 Mhz bandwidth? Cut the bandwidth in half and your output noise is cut in half.

    What is the input voltage range of your V2F converter? Does 9.3 milli-volts (RMS) account for the noise you are seeing?
     
  10. Eoin_oc

    Thread Starter New Member

    Oct 24, 2015
    23
    0
    This level of noise referred back to the input should give me more noise than I am seeing on my readings. I estimate im seeing around 1mV at the output. But its in the ball park region.
    I would like to keep a Bw of around 1mhz to pick up fast changes in whatever load I'm measuring. Can I put a parallel C on the op amp feedback to limit the Bw? Also can somebody suggest a lower noise op amp for this application?
     
  11. Lestraveled

    Well-Known Member

    May 19, 2014
    1,957
    1,215
    Your first op-amp sets your noise floor, so you could add an active low pass filter right before your V2F without adding to the noise.

    Going from 6.5 Mhz to 1Mhz bandwidth would reduce the noise by a factor of 6.5.

    The MAX4239 are very very low noise. Stick with them.
     
  12. Eoin_oc

    Thread Starter New Member

    Oct 24, 2015
    23
    0
    I will add a 1MHz low pass to the very input and post some plots from the spectrum analyser later.
    Should I use another MAX4239 for the active filter on the output?
     
  13. Lestraveled

    Well-Known Member

    May 19, 2014
    1,957
    1,215
    Yes, the MAX4239 is a good amp for that.

    BTW, what V2F chip are you using?
     
  14. Eoin_oc

    Thread Starter New Member

    Oct 24, 2015
    23
    0
    AD7741
     
Loading...