Is Noise Always Unwanted?

We have to admit that, as analog designers, we tend to get scared when the word “noise” pops up in front of us. As an almost immediate side effect, the concepts of denoising, cleaning, and filtering come to our perfectionist minds. But noise, though by definition unwanted signals, can also be among our most precious debugging tools arsenal — and a valuable source of information for diagnostic purposes.

There are plenty of examples where noise signals are relevant for the application. They range from very simple cases, like injecting a 50/60 Hz signal into an audio amplifier using the tip of the finger, to very complex and math-filled ones, like detecting, analyzing, and locating the source of cosmic rays.

True random numbers play a key role in the field of communications security. Unfortunately, most number generation algorithms produce only pseudo-random series. How do we generate a true sequence of random numbers? We have to rely on stochastic physical processes for ensuring randomness. Thermal and atmospheric noises have found their way into this process.

White and colored noise generators are particularly useful in ambient noise blocking and radio jamming. In cry analysis, the presence or absence of specific formants and distortions may provide clues about a baby's health. Fluctuations of current or potential associated with corrosion processes are useful for determining corrosion types and speed.

Vibration analysis has proved a very effective technique for early detection of bearing and structural damages due to mechanical stress. In my years working as a designer and programmer for smart card manufacturing machines, I've realized that, most of the time, having the best-in-class software debugging tools and being a skilled and experienced programmer are not enough when addressing situations like the one illustrated below.

Where did the pipette go?

Where did the pipette go?

When it comes to debugging mechanically complex systems, one very valuable skill is the ability to discriminate noises that deviate from the expected, normal-operation noise. Having a trained, noise-friendly ear is a real advantage and a must-have component of the machine hardware debugging arsenal.

For example, two conditions in a paper/chip ticket converting machine are shown below — normal operation noise and noise produced when paper strip breaks.

When paper breaks, the machine noise changes shape and level.

When paper breaks, the machine noise changes shape and level.

Noise can be our ally and not just our most feared analog enemy. Have you ever found applications where noise has not been merely an unwanted signal? What other types of noise would be signals of interest?

8 comments on “Is Noise Always Unwanted?

  1. David Maciel Silva
    February 3, 2014

    Hello Victor,

    Here is an account of a great villain … 've Been good bits in welding machines, used when the spark is no filter that lasts lol ..

    Basically we have developed a fan cirucito to act in a system of cold, with a comparison that checked if the torch was thrown and so connected the cooler, but to arm the torch, the spark was also fired, causing time to enter the system time no …

    The solution was supresssores some noise, were few pieces too, which did not justify a more “expensive” circuit …

  2. Victor Lorenzo
    February 3, 2014

    Good point @Maciel, noise can come from many sources and fortunately some of them allow proper and effective handling with just a few components and simple circuits.

  3. Victor Lorenzo
    February 4, 2014

    One note about what could be called The disappeared pipette case . The first pipette we used was made of alluminium. I was standing by the machine looking at the card milling station when several artifacts in the background noise strongly called my attention: a missing sound (the pneumatic piston moving the pipette up) and two new sounds, a metallic crack followed by the horrifying sound of a metal tube hiting the wall (the pick-and-place header was moving really fast).

    For testing the hypothesis of a software bug in a states machine we made another pipette (this time a plastic one, weakened in the upper part so it could easily break), placed a webcam and activated all debug messages logging functions.

    As these machine types work following what could be considered a rithm (like an orchestra) it is feasible to use noise, vibration and sound analysis techniques for early malfunction detection.

  4. Hughston
    February 7, 2014

    I can thing of a few times at least when you want noise.  In some audio systems you want a small amount of noise known as comfort noise so you know the system has not dropped audio. You can also use white noise and average to look at the frequency response of a system. A third example is in A/D conversion, when noise can be added to help you extract a value near the LSB to determine if it is closer to 1 or 0. A fourth example is in some speakerphone audio tests, you use pulsed pink noise to determine a detection threshold. And of course with noise cancelling systems, you would want noise resembling what you want to cancel, to test them.

  5. Victor Lorenzo
    February 7, 2014

    Thanks @Hughston for your comment. I find particularly interesting the usage you mention for intentional sub-1LSB noise. We usually tend to increase resolution,  reduce dynamic range or add one PGA (or even all of them) when the application requires to discriminate lower input values.

  6. Victor Lorenzo
    February 7, 2014

    As @Hughston mentions, white noise input signals can be used for determining the frequency response of one particular system, it is an alternative method to frequency sweep based methods.

    Also very short in time impulses are used for similar purposes in accoustic characterization for rooms and theaters where echoes and reverberation degrade the overall listener experience.

  7. Hughston
    February 9, 2014

    I always wondered if a person has noise in their system, how do you troubleshoot it and eliminate it? I have often been confronted with this problem. I have gotten rid of noise many times, but what are fast ways to find it and eliminate it.

  8. Victor Lorenzo
    February 9, 2014

    @Hughston, “how do you troubleshoot it and eliminate it? “. In my case I start analysing the noise signals for identifying and locating the source, or sources. Time domain waveform shapes give some clues and spectrograms give some other clues. Next steps have always been dependant on noise(s) source(s), level and affected frequency bands.

    Mixed systems (analog+digital in the same PCB) with high gain input stages have often been difficult to troubleshoot, many times due to PCB layout influence.

Leave a Reply

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