Advertisement

Blog

ADC Noise: Where Does It Come From?

In my last few blogs, we took a pretty good look at interleaving ADCs. We have seen what advantages interleaving brings to the table as well as some of the drawbacks that come along with all that nice speed and bandwidth. Let's now move on to another topic that several readers have commented on at various points.

That question revolves around the noise contributors for an ADC. What are the things we need to consider when evaluating what the noise of the ADC will be? Noise can enter the ADC in a variety of ways. Over the next few blogs we'll take a look at all the doorways through which noise enters the ADC and can appear in an FFT of the output data. First we'll begin by identifying the doorways.

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. As Figure 1 demonstrates, noise can enter the converter through the power supply inputs, the analog inputs, and the clock.

Figure 1

ADC Noise 'Doorways'

ADC Noise “Doorways”

Since noise is quite a loose term here, let's put a little more meaning to it based on which input (doorway) to the ADC is being discussed. We'll start at the top of the diagram and work our way around counter-clockwise.

The power supply inputs are pathways for noise to make its way into the ADC and appear in the FFT of the output data. In this case, there are a few ways to evaluate this noise and its impact on the ADC performance. The ADC should be designed in such a way that the device itself attenuates noise input coming from the power supply. The measurements here to evaluate noise on the power supply are the power supply rejection ratio (PSRR) and the power supply modulation ratio (PSMR). Measuring these two parameters gives us an idea of how well the ADC will handle noise entering through the power supply inputs. We'll take a more detailed look at this later. For now, let's continue looking at the noise doorways.

Next, let's take a look at the analog input of the ADC. From this perspective, noise must be considered in two ways. First, there is general broadband noise that enters the converter through the analog inputs and is generally from the components preceding the ADC in the signal chain. We can select a very low noise driver amplifier for the ADC, but there will still be a finite amount of noise that is amplified and input to the ADC.

To help combat this, an anti-alias filter (AAF) is typically used at the inputs to the ADC. This helps to filter much of the broadband noise that might enter the ADC. This ultimately shows up in the signal to noise ratio (SNR) of the ADC. In addition to broadband noise, spurious content and harmonics can also enter the ADC through the analog inputs. The AAF helps to filter these as well. This will be reflected by the spurious free dynamic range (SFDR) of the ADC. It is important to have a good AAF design to help with both of these aspects. Again, we'll look at this in more detail in later blogs.

The last doorway that we see as we move counter-clockwise around our ADC is the clock input. This input, similar to the analog inputs, can allow both broadband noise as well as spurious and harmonic content to enter the ADC and appear in the FFT of the output data. It is important to make sure an appropriate clock input driver is selected that provides a clean, low jitter input clock to the ADC.

This clock signal should be routed to the ADC in such a fashion that it does not couple in noise that can find its way into the ADC. Similar to the analog inputs, a filter may be used on the clock input to help filter out noise that may have otherwise entered the ADC through the clock input. Once again, as in the case of the analog inputs, the noise mechanisms through the clock input can manifest themselves in the SNR and the SFDR performance of the ADC.

All of these doorways must be considered when designing a system using an ADC. We see that we should treat the ADC as a mixer which mixes the various noise content from any one of these doorways onto the output data in the FFT. Obviously, a system designer would like to have the desired signal only at the output of the ADC. In order to do so, we must take the proper steps on each of these inputs to make sure that noise is minimized and does not enter these doorways. Stay tuned as we take deeper dives into each of these inputs and evaluate in more detail how noise couples into the ADC and what can be done to help prevent it.

Related posts:

16 comments on “ADC Noise: Where Does It Come From?

  1. Davidled
    October 22, 2013

    Instead of looking noise in the board, engineer would like to remove noise as much as possible in the circuit. As this blog comments, and anti-aliasing filter would be treated in S/W.  Power and ground would be a different layer and high current signal and low current signal might be separated in addition with ground isolation.

  2. samicksha
    October 23, 2013

    you are right Daej, using low-pass filter prior to the ADC, including by-pass capacitors and using a ground plane will also eliminate this type of noise.

  3. Netcrawl
    October 23, 2013

    @Daej I agree with you. anti-aliasing its probably most effecive way of removing noise, its proven and very reliable. Designing a 12-bit ADC board is not an easy task , there's some challenges that need to be address- conductive noise and radiated noise. Addressing noise problem is very important because it can affect the overall ADC resolution. 

  4. Netcrawl
    October 23, 2013

    @samicksha you're right those two options are very useful in removing noise. We got two type of noise here- conductive and radiated. Its not hard to eliminate these two, its addressable.

    In order to improve ADC performance and make it immune against radiated and conductive noise we need to know some electromagnetic mechanisms , its a “must do things.”  

  5. jonharris0
    October 23, 2013

    Thanks for all the great comments.  We'll continue to look at these noise doorways in my upcoming blogs and see how we can address each of them.  Keep these great comments coming! 🙂

  6. Bonnie Baker
    October 23, 2013

    Jonathan,

     Nice starter article. Blogs are hard because you are not given enough words to really delve into the topic. I noticed that there is on very important pin that is missing from your diagram; the voltage reference pin. The fact that this is a path for noise is often overlooked and misunderstood. Actually, I find in reading data sheets and articles that this pin function is poorly defined if not ignored.

    You have definitely covered some good basics, but the issue of EMI is missing. Actually, this could be an entire series that would last for a year or two or three. Many of the comments are trying to fill in that blank.

     

  7. Davidled
    October 23, 2013

    I wonder where conductive and radiated noise come from in what system.  I heard that there are so many different type noises in the circuit with interface board.

  8. samicksha
    October 24, 2013

    I agree you Netcrwal, conducted noise is already in the circuit board by the time the signal arrives at the input of the ADC.

  9. jonharris0
    October 24, 2013

    Thanks for the comments Bonnie.  I am assuming the Vref circuitry is internal as is the case on many high speed ADCs today.  And I agree wholeheartedly that the space is limited with the blog, but I hope to split this out and cover many of the topics over several blogs.  I hope all can enjoy the ride as we look over these many doorways! 🙂

  10. Scott Elder
    October 24, 2013

    @DaeJ, “I wonder where conductive and radiated noise come from in what system.”

    Consider an ADC system with a switching power supply and voltage reference.

    An example of conductive noise is the power supply ground return current flowing through a shared return path also used by the ADC reference.  In this case, the switching ground current will generate a switching noise voltage that will conduct directly on to the reference voltage thereby reducing the precision of the reference voltage.

    Using the same system example, radiated noise can likewise couple into the reference ground if the reference ground return signal path is sufficiently parallel to the switching power supply return path such that the two paths are electromagnetically coupled as in an antenna.

    A “star” ground system along with large ground planes and shields is usually adequate to solve these noise issues.

  11. yalanand
    October 27, 2013

    All analog-to-digital converters have a certain amount of input-referred noise exhibited as a noise source associated in series with the response of a noise-free ADC. Input-referred noise is not to be chaotic with quantization noise, which is only of notice when an ADC is handling time-varying signals.

  12. David_O
    October 30, 2013

    Jonathan

    Your ADC seems to be missing its ground pin, a very important signal to consider. I suppose you sort of covered it with with the other power lines, but it is often an analgue reference point as well. While it may be less important with differential devices, as shown, the absense of this pin can only help reinforce the assumption that all grounds are the same and provide a safe and stable refernce point. Unfortunately, this is often not the case.

     

     

  13. jonharris0
    October 30, 2013

    Aha, yes, David, you are correct.  What an oversight I have made here!  I shall have to update that in the next blog.  As I once heard Bruce Archambeault say giving a conference paper, “Ground is a place for potatoes and carrots!  We have current return paths!”  It is definitely a potential pathway for noise into the ADC.  As you correctly point out, it is not always a stable reference point.  Thanks for the comment!

  14. Victor Lorenzo
    December 1, 2013

    @Netcrawl, if you let me, regarding your phrase “We got two type of noise here- conductive and radiated. Its not hard to eliminate these two, its addressable “. I don't agree with the bolded part of it. Yes, it is most of the times a hard design task to end up with a correct PCB design showing a high EMI noise inmunity and at the same time featuring low EMI emissions (conducted and radiated). It specially holds true when working with synchronous (clocked) digital signals switching at clock speeds over 50MHz. From my experience on the area, newcomers to high speed PCB design almost always fail on that at least the first time.

    we need to know some electromagnetic mechanisms “, I agree on that and I would add, we should also be prepare to think that a piece of wire is not a near zero ohm 'wire' at high frequencies, it is a complex transmission line with many parasitic effects to take into account.

  15. Victor Lorenzo
    December 1, 2013

    @Daej, “I wonder where conductive and radiated noise come from in what system ” Just a small experiment that could surprise you. Fo this you'll need one high-bandwidth oscilloscope with the highest resolution you can find (10 or more bits per sample).

    Radiated Emissions: make two loop antennas, one about 10cm in diameter and another about 3cm, both with about 10 turns. Hold the bigger antenna with your hand at about 50cm from a fluorescent light and look at the 'scope. Take notes. Now take the small antenna and place it over the chips (MCU, FLASH, ADC, Switching regulator, etc) of your board. Once again, take notes. Of course you'll need to adjust the vertical resolution for the 'scope and gain some experience in differentiating the incoming noise from the quantization noise inherent to the 'scope's ADC.

    Conducted Emissions: take a piece of wire about 20cm in length and wind it around one of the DC power cables for your board. Take a look at what the 'scope reveals. Now repeat the same experiment with different positions between the power/ground wires (putting them apart, putting them together, twisting them, putting a ferrite clamp, etc.)

  16. Victor Lorenzo
    December 1, 2013

    @Scott Elder, “In this case, the switching ground current will generate a switching noise voltage that will conduct directly on to the reference voltage thereby reducing the precision of the reference voltage “. You're right, but we should always consider that as part of the design that situation must be avoided. Two points to address here, the high current switching paths should be as short as possible, separated from the analog part of the circuit (they use to have less effect on the digital circuits) and occupy the minimum current path area as possible to minimize emissions (perhaps the most importan rule).

    The voltage reference should be placed as close as possible to the reference pin and provided with its own return path (once again, it should occupy the minimun possible area).

    Most digitized signal noise I've found has come from other sources, far more difficult to circunvent than the reference voltaje noise.

Leave a Reply

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