Advertisement

Article

Understanding noise optimization in sensor signal-conditioning circuits (Part 2 of 4)

Editor's note for this multipart series:

  • Part 1a covers the basics of noise in the signal chain, click here
  • Part 1b covers the various types of noise: white (broadband); pink (1/f); popcorn (burst); shot; Schottky; and resistor noise, click here
  • Part 2a covers op amp selection and passive component selection
  • Part 2b covers bandwidth selection and reviews the lessons
  • As analog-to-digital and digital-to-analog converter resolutions increase and power supply voltages decrease, the size of a least significant bit becomes smaller. This makes signal-conditioning tasks more difficult. As the signal size gets closer to the noise floor, both external and internal noise sources, including Johnson, shot, broadband, flicker, and EMI, must be addressed.


    Noise sources, which are generally uncorrelated, are combined in a root-sum-square (RSS) fashion:




    Correlated noise sources such as input bias current cancellation, on the other hand, must be combined in an RSS fashion with an added correlation factor:




    Figure 1 illustrates all of the noise sources found in a typical signal conditioning circuit, along with a general equation that can be used for inverting, non-inverting, difference, and other common configurations.




    Figure 1: Noise sources include the op amp's input voltage noise and input current noise, plus Johnson noise from the external resistors.

    The right design approach
    Starting with a sensor and its characteristic noise, impedance, response, and signal level, achieving the lowest referred-to-input (RTI) noise will optimize the signal-to-noise ratio (SNR).


    Instead of solving gain and power requirements first, and then struggling with noise issues, it is more effective to approach the problem with a low-noise focus. This is an iterative process. Start by considering the region of operation for the amplifier: broadband or 1/f.


    Next, design for the best noise performance by selecting appropriate active components. Surround the amplifier with the passive components and limit the bandwidth. Then analyze the non-noise requirements, such as input impedance, supply current and open loop gain. If the noise spec is not met, continue this process until you have an acceptable solution.

    Op amp selection
    In some cases, an op amp with 22 nV/√Hz broadband noise may be better than one that specifies 10 nV/√Hz. If the sensor operates at very low frequencies, an amplifier with low 1/f noise might be best. Standard amplifiers, such as the OP177 from Analog Devices, have a noise spectral density that looks like the graph shown on the left in Figure 2 .



    Figure 2: a) standard amplifiers, such as Analog Devices' OP177 and AD707, exhibit 1/f noise at low frequencies;. b) auto-zero (chopper) amplifiers, such as Analog Devices' the AD8551/52/54, have no 1/f noise; c) PSpice correctly models the behavior of the AD8638 auto-zero amplifier.


    (Click on image to enlarge)

    Auto-zero amplifiers, on the other hand, continuously correct any errors that appear at their inputs over time and temperature. Since 1/f noise approaches dc asymptotically, the amplifier also corrects this error. The graph in the middle shows how first-generation auto-zero amplifiers do not exhibit 1/f noise, making them useful for low-frequency sensor signal conditioning. Second generation auto-zero amplifiers, shown on the right of Figure 2, feature lower wideband noise (22 nV/√Hz). The unusual PSpice macro-model correctly simulates the amplifier's voltage noise, showing the elimination of 1/f noise.


    Rail-to-rail inputs
    For low-voltage designs, a rail-to-rail (RR) output and input may be appropriate. As the common-mode input goes from one rail to the other, one differential input pair takes over as the other differential pair stops working. Offset voltages and input bias currents can change suddenly, causing distortion as shown in Figure 3 . For low-noise design, question the need for the RR input feature.



    Figure 3: The input offset voltage of a rail-to-rail amplifier can change dramatically as the common-mode input voltage changes
    (Click on image to enlarge)

    To address this problem, op amps such as Analog Devices' AD8506 use an internal charge pump to eliminate input voltage-crossover distortion. If not designed correctly, the charge pump will produce noise that will appear at the output potentially causing a problem if the noise falls in the frequency band of interest. Use a spectrum analyzer on the output pin to make sure that the amplitude of the clock is much lower than that of the signal.


    Bias current cancellation
    Newer bipolar op-amps use a technique to partially cancel the input bias current. This technique can add uncorrelated or correlated current noise. For some amplifiers, the correlated noise can be larger than the uncorrelated component. With Analog Devices' OP-07, for example, adding an impedance-balancing resistor can improve the overall noise. Table 1 compares two widely used Analog Devices' op amps the OP07 that trades higher voltage noise for lower current noise, compared to the OP27.



    (Click on table to enlarge)

    Choose three to four parts from the available low-noise devices. Consider the process technology. Look for special design techniques such as auto-zeroing, chopping, and bias-current cancellation. Look at die size photos for input transistor area, remembering that large input transistors produce low noise, but large input capacitance. CMOS and JFET amplifiers have much lower current noise than bipolar devices. Low-noise designs use low valued resistors, so the amplifier output drive must be large enough to drive heavy loads.


    Passive component selection
    After selecting the amplifier, surround it with the appropriate resistors and capacitors. These too have noise. Figure 4 shows the effect of using the wrong resistor values. The output noise goes up as the resistors, used for setting the gain, go up. In all three situations, the gain is 1000.



    Figure 4: Use low value resistors to maintain low output noise
    (Click on image to enlarge)

    Understanding the sensor's characteristics is important. Neglecting the noise contribution of R1 and R2 and focusing on the noise of the source impedance, R, Figure 5 shows that the voltage noise dominates with small values of R, the Johnson noise dominates with medium values, and the current noise dominates with large values. Therefore, sensors with low output impedance should use small resistors and op amps with low voltage noise.



    Figure 5: Voltage noise dominates with small resistors, Johnson noise dominates with medium values, and current noise dominates with large resistors.

    (Click on image to enlarge)

    In addition to resistors, capacitors are also used for compensation and noise reduction. Reactive components do not add any noise, but noise current flowing through them will develop noise voltages that enter your calculations. In summary, it is important to use low impedances around the amplifier to minimize the effects of current noise, thermal noise, and stray pickup of EMI.


    Bandwidth selection
    After choosing an amplifier and the associated resistors and capacitors, the next step is to design for best bandwidth (BW). Be careful not to over-design for a wide bandwidth. The bandwidth should be wide enough to pass the fundamental frequencies and important harmonics, but no wider. Select an amplifier that has enough BW and follow it by an RC filter. The amplifier itself is also a single pole filter. Amplifiers and resistors have noise over each Hertz of BW so the greater the BW, the greater the output noise and the lower the SNR.


    Figure 6 shows the effects of the amplifier's BW vs. noise for the same circuit configuration as before, but using amplifiers with different BW. To limit the added noise, the BW should be narrowed as much as possible.



    Figure 6: Output voltage noise increases with increasing amplifier bandwidth


    (Click on image to enlarge)

    To narrow the bandwidth, use an RC filter after the sensor. This can create loading problems that can be overcome by using a buffer as shown in Figure 7 .



    Figure 7: Add a buffer between the sensor and filter to avoid loading problems.
    (Click on image to enlarge)

    An amplifier and ADC having the specs and configuration shown (amplifier BW 350 MHz) will have 166 μVrms noise. Adding an RC filter after the op amp, creating an effective BW of 50 MHz, reduces the noise to 56 μVrms.


    Narrowing the BW by using the correct RC will greatly improve the SNR as shown, but the resistor itself can add noise. A better way to reduce the BW is by using the circuit shown in Part C of Figure 8 . This places the resistor inside the op amp's feedback loop, reducing its effect by 1 + loop gain. Don't forget to reduce power supply noise from the signal path by using adequate decoupling caps at the supply pins.



    Figure 8: a) simple RC filter; b) buffer reduces loading; c) putting resistor inside feedback loop minimizes noise.


    (Click on image to enlarge)

    After going through these steps, review the other system requirements. Below are some examples:

    • Do the selected components meet the other target specs?
    • Does the amplifier require dual supplies?
    • Is there a positive supply?

    • Does the amplifier consume too much power?
    • Are the components too expensive?

    If necessary, return to Step 1 and repeat the process.

    Conclusion
    Every sensor has its own noise, impedance and response characteristics, so matching these to the analog front-end is critical. A well defined low-noise design process is needed to overcome many challenges in today's applications and to get the best SNR. This iterative process will produce a signal conditioning solution that is most suitable for today's challenging applications.

    References
    (note: all from Analog Devices, Inc, except the book)

    1. Online noise seminars on this topic listed at: http://www.analog.com/en/content/0,2886,759%255F786%255F109826%255F0,00.html#seminars
    2. Application Note AN-202, An IC Amplifier User's Guide to Decoupling, Grounding, and Making Things Go Right for a Change ,
      http://www.analog.com/static/imported-files/application_notes/135208865AN-202.pdf
    3. Application Note AN-358, Noise and Operational Amplifier Circuits , http://www.analog.com/static/imported-files/application_notes/5480117281535838576388017880AN358.pdf
    4. Application Note AN-940, Low Noise Amplifier Selection Guide for Optimal Noise Performance , http://www.analog.com/static/imported-files/application_notes/AN_940.pdf
    5. Bryant, James Bryant and Lew Counts, Op Amp Issues — Noise , Analog Dialogue, Volume 24 Issue 2, 1990
    6. Motchenbacher, C. D., and J. A. Connelly, Low-Noise Electronic System Design . New York: John Wiley & Sons, Inc.1993

    About the author
    Reza Moghimi is an applications engineering manager for the Precision Analog Products group at Analog Devices, Inc. He holds a BSEE and MBA from San Jose State University (SJSU), CA. In addition to Analog Devices, Reza has worked for Raytheon Corp., Siliconix Inc, and Precision Monolithic Inc. (PMI). He enjoys traveling, music and soccer.

    0 comments on “Understanding noise optimization in sensor signal-conditioning circuits (Part 2 of 4)

    Leave a Reply

    This site uses Akismet to reduce spam. Learn how your comment data is processed.