ADC Noise: A Second Look, Part 1

As I read over the comments from my last blog, ADC Noise: Where Does It Come From?, it became apparent that I had not covered all the potential doorways for noise in an ADC. There are a few more doors that we must consider. Thanks to folks for pointing out a few and for helping me to get my thoughts in order to look at a few more places I had missed previously.

As I commented last time, when considering noise in an ADC, one can almost consider the ADC as a mixer. If there is noise entering the ADC from any one of the various doorways, then it can manifest itself in the FFT of the output data. Let's now take a look at a modified version of the figure I introduced in my last blog.

Figure 1 shows the noise doorways that we discussed previously, which included the power supply inputs, the analog inputs, and the clock. However, there are a few more doorways that I missed that we should definitely consider when working with an ADC. The first is the common mode voltage (Vcm) output that supplies the common mode level for the analog inputs. Next there are the digital inputs and outputs (I/Os) which can be a pathway for noise into the ADC. Last, there is a doorway that is probably one of the most overlooked — the ground or circuit-common.

Figure 1

Additional ADC noise 'doorways'

Additional ADC noise “doorways”

The Vcm output is used by many high-speed ADCs today to provide the common mode reference voltage for the ADC analog inputs. It's a midpoint voltage within the minimum-to-maximum range of the ADC's inputs. This pin usually requires a decoupling capacitor of about 0.1μF. This provides a dominant pole for output stability as well as filtering of higher frequency noise. It is important to have proper decoupling, as this node provides a potential direct line for noise into the ADC analog inputs. And even though it's an output, noise can force its way in to the internal bias circuitry of the ADC.

In addition to the capacitor, many ADCs with two or more channels also require a small amount of resistance in series with each connection from the Vcm output to each of the channels. This is also a form of noise reduction as it typically helps with reducing the crosstalk between the channels of the ADC. Another way to say this is that the additional series resistance helps improved channel-to-channel isolation so that the signal from one channel doesn't make its way to another channel.

In the next part of this blog, we'll look at the digital I/O lines (including the SPI) of the ADC and consider how they can also provide a doorway for noise. And we'll finish by looking at the ground connections.

Related posts:

5 comments on “ADC Noise: A Second Look, Part 1

  1. amrutah
    November 15, 2013


      Thanks for the Blog and also updating th e sources of Noise.  I think the noise from reference or common mode is very less due the huge capacitors on that node.  The main culprits I consider are the noise from Supply and Ground.  We may have star connections to these nodes but they are still significant.  Also the inherent devices noise will add a lot to the output noise.  The network used for switching, sampling will introduce new set of noise.

  2. yalanand
    November 30, 2013

    The main culprits I consider are the noise from Supply and Ground.  

    @amrutah, thanks for sharing your analysis. I am curious to know how can we reduce noise from ground and supply ? What techniques can be used to reduce noise from ground and supply.

  3. amrutah
    December 1, 2013

    @Yalanand: There are many techniques to reduce the noise from power supplies and depends on the scope of application, few of them that I know

    1> Reduce the effect of noise by using cascode devices, avoid any high switching digital and sensitive analog sharing the same supply (this can be done to some extent by proper laying out the powerlines)

    2> bypassing the power supplies at the load so that the peak current (di/dt), which usually leads to noise, can be taken care to some extent

    3> Low pass filters can be used to reduce the effect of noise at the load (but this may cause functional problems).

    4> Snubber circuits to isolate the noisy elements.

       As far as ground is concerned, which is a reference node for the entire application, I think making it strong (wide routings) is the trick that suffices.


  4. amrutah
    December 1, 2013

    @Yalanand:  Another important point while bypassing is to place several capacitors in parallel instead of lumping it into a single capacitor.

  5. Victor Lorenzo
    December 1, 2013

    Another important point while bypassing is to place several capacitors in parallel instead of lumping it into a single capacitor “.

    We could add also that not only to place several capacitors in parallel, it should be taken into account the capacitor's behaviour at high frequencies when selecting the capacitors that will 'filter' the high frequency noises (usually ceramic, commonly 0.01uF-to-0.1uF).

Leave a Reply

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