High Frequency Switching Noise (continued)
In particular, any such artifacts that make their way into the negative feedback loop of an amplifying stage or power amplifier carry with them the potential to reduce the stability thereof. Specifically, any unintended signal that is sustained or continually repeated in some fashion that makes its way into a circuit's feedback loop alters it such that a new and unintended secondary loop will have been created. In that event, depending on the phase relationship of that signal with respect to whole it is possible and likely that the circuit stage being contaminated thereby will develop some level of oscillation. If the oscillation is only minor, then it may well work to reduce the affected amplifying stage or power amplifier's ability to accurately respond with sufficient speed to the desired input signal - especially signals that contain abrupt changes of amplitude and/or high frequencies, such as those of percussive instruments, the initial hammer strike of a piano note or the pluck of a stringed instrument, etc. In such a case, the resulting sound of these and other instruments will suffer a loss of dynamic impact and realism. In the event that the magnitude of the oscillation is more severe, the circuit can even become unstable and/or worse... eventually fail altogether. This scenario is often the cause behind power amplifier failures, particularly those of the Class-D/Switching types.
Circuit Saturation & Slew Rate:
Going deeper, at the root of the problem is the fact that this noise can lead to circuit saturation effects and instability. The circuits that are most vulnerable to suffer from saturation or partial saturation as a result of this noise are those that are intended to amplify or buffer the audio bandwidth signals. This is because most such circuits are intentionally band-limited so as to be optimized for the audio frequencies they are intended to process, with factors such as maintaining low-noise and distortion performance within its intended band of operation usually being the primary criteria.
Band-limiting of any circuit is always directly associated with a corresponding limitation of its "Slew-Rate," as the two are intrinsically proportional, and for the most part the two are simply different ways of expressing the same electrical property. In layman's terms Slew-Rate is often thought of as the upper "speed limit" of the circuit. which indicates how fast the circuit is able to accurately follow and/or respond to an input signal (called the "demand" signal). Circuits intended to process audio bandwidth demand signals frequently have their upper speed or Slew-Rate purposefully limited, because, as stated above, in most cases doing so improves the circuit's performance over the range of audio frequencies it was designed for. Even circuits that are not purposefully Slew-Rate limited still have an upper limit, as otherwise their Slew-Rate would be infinite - and everything in the universe has a speed limit - with that being the speed of light or (in most cases) well below.
All circuits are then Slew-Rate limited to some upper value either by nature or design, and as we shall see, slew-rate can be directly converted to frequency by means of the following simple formula:
Sr/2*pi*Vp = F
SR = Slew Rate
Pi = 3.1415926
VP = Peak Voltage
F = Frequency
If you are really interested and want to know more about the above equation, we will leave that for you to do your own research. In any case, If nothing else we can clearly see by the formula that Slew-Rate and Frequency are essentially interchangeable terms wherein each is totally dependent on the value of the other.
Establishing the above as the main governing factor, the potential for high frequency noise contamination to degrade and/or limit the dynamic performance of a given audio component is directly related to the Slew-Rates of both the demand signal and the component, circuit or device designed to process it. Succinctly, if the S-R of the demand signal exceeds that of the processing circuitry, a form of "saturation" of one or more components comprising said circuitry can ensue. Saturation is identified by observation of the fact that the affected component and/or circuit does not immediately respond to the demand signal, but rather there is a form of delay or "lag time" in its response causing distortion of desired output signal. During the period of time while the affected circuit or component is said to be in saturation, there is little or no signal change occurring at its output. Therefore, all changes observed in the demand signal during that same period are essentially "lost," as though they were simply being ignored by the circuitry.
In fact, if the saturation event is severe enough, the affected part can "latch" in a permanent or semi-permanent condition wherein its output stage can remain in an unchanged state indefinitely. In cases where a given part or circuit becomes permanently latched, often the only way to restore proper operation is to remove power and then reapply it. In milder cases where latching does not occur, the effect can be considered a form of "hysteresis," which is a type of memory-effect wherein the output state of the saturated part or circuit is based on the conditions of its previous state and the present demand signal. A cursory review of hysteresis as it applies to electrical engineering will show that in circuits designed for linear operation as audio circuits are, that it is a common source of distortion.
In light of the above it then becomes relatively easy to see that if a circuit or component is intentionally band-limited to only process signals in the audio range (DC to 100KHz), that if that same circuit is presented with a demand signal containing significant energy content from above 100KHz up to 1MHz and beyond, there is a very great likelihood that its overall performance may well suffer from temporary periods of saturation. Seeing that most switching power supplies that are often used in modern audio components operate at 100KHz and above and that the switching frequency of most Class-D/Switching Power Amplifiers is set at or above 400KHz, it is obvious that devices employing these technologies can easily become potential victims of high frequency noise contamination and temporary circuit saturation. In addition, virtually all Digital-to Analog Converters (DACs) operate at and make use of Master Clock oscillators with clocking frequencies commonly set as high as 12MHz.,so one can well imagine that their audio-only circuits may too suffer from some degree of clock-generated noise contamination and saturation effects.
As one may suspect though, any such saturation events occurring within a given audio component must be of minor severity, as otherwise its performance would suffer such that it would generate obvious levels of distortion and therefore be of little marketable value. We fully recognize this fact, but we still maintain that such devices and components may well suffer from sufficient self-noise contamination at subtle levels that, if additional treatment is employed to reduce it further, additional audio performance gains will be realized. Therefore, any such contamination must be subtle by nature, so the question arises as to what area of performance is actually being affected? Again, based on the obvious constraints just outlined, logic dictates that any such saturation events must be very short-lived so as to be difficult to detect, as otherwise the manufacturer of the affected component would likely have already remedied the problem before commencing with its production.
Saturation, Dynamic Compression & Dynamic temporal smearing:
In music we find that short-lived events are otherwise defined as "transient" and/or "dynamic" events, and that their audible contribution is one of imparting the perception of dynamic impact and/or a sense of "liveliness" and/or "realism" to a given musical passage. Transient/dynamic events are what define the sound of percussive instruments such as drums & cymbals and/or the percussive portions of certain instruments, like the sound of the piano hammer first striking the piano string ,or the initial sharp "attack" sound that occurs when a stringed instrument is first plucked. These and other similar musical events are all relatively short-lived by nature and occur in a time range anywhere from about 50-microseconds (20KHz) up to around 10-milliseconds (100Hz). In most cases it would likely be in the region from 200-microseconds (5KHz) and above where any circuit saturation events would have a reasonable statistical chance of occurring. This is mainly because the resulting distortion artifacts generated by them would all reside at and above 10KHz, making them considerably less audible than those at lower frequencies. Thus, they are also less likely to be given much attention or concern by the product's manufacturer.
In order to fully grasp the process being discussed here, a good comprehension of how musical waveforms are constructed would be helpful. As an example, we will attempt to describe the makeup of a typical piano note being produced from beginning to end, as the piano hammer strikes the string until the note has finally ended and is no longer sounding.
To begin, if one were to place a microphone nearby the piano string and feed the electrical signal produced by it into an device that is able to visually project it onto a screen or graph such that the amplitude is represented on the vertical axis and time is displayed on the horizontal axis, one will have a visual representation of the sonic waveform that is produced. Such a device is commonly referred to as an oscilloscope and is a common tool used in electrical engineering and repair service laboratories..
Now, at the moment the hammer makes initial contact with the string, it does so abruptly and with relative high velocity. This translates to a sudden large displacement of the string from its normal rest position. Likewise, on the oscilloscope we will observe a large vertical displacement of the oscilloscope trace at that exact moment, which represents the primary dynamic peak of the waveform. Please keep this in mind as that peak is the main object of this discussion.
Next, as the string begins to vibrate we will see the oscilloscope trace begin to cycle repeatedly about the vertical axis. Furthermore, as we continue to watch it will do so with decreasing amplitude as the trace continues to move from left to right along the horizontal time axis. If the note is held by the pianist such that it is allowed to continually decay in amplitude until it is no longer detectable by the microphone, we will observe the analog equivalent of that same decay process on the oscilloscope represented as a continually decreasing signal waveform until it is no longer visible on the screen.
OK, from the initial hammer strike to the point in time where the piano string has completely stopped vibrating we have what is called the total event "envelope," in that it represents the entire period of time that the piano note exists from beginning to end. What we are concerned with is that each section of the envelope (beginning, middle & end) be reproduced by our audio device with as much accuracy as possible so as to fully preserve every aspect thereof. With respect to the dynamic performance of audio devices or components, iour concern is that the magnitude of the peak that is produced during the initial hammer strike not be artificially reduced or attenuated with respect to the amplitude of the remaining envelope in any way. If any attenuation were to occur we would call that process "compression" of the dynamic peak, or simply peak dynamic compression. If that does occur, then from a listener's perspective the sonic impact of the piano note will be reduced such that the resulting sound will exhibit less of a "live" or realistic quality. In addition, if the dynamic peak in question is somehow lengthened along the time axis with respect to the rest of the envelope, we would say that a type of temporal "smearing" of the dynamic peak has occurred. To the listener, temporal smearing of the dynamic peak has a similar effect as peak dynamic compression, in that a certain amount of liveliness and realism will have been lost in the sound quality of the piano note.
With the above in mind and going back to our original discussion regarding audio device feedback loop contamination,, assuming the contaminating signals only cause a mild degree of oscillation or "quasi-oscillation" in the affected circuitry... or even simple, temporary saturation thereof as described previously, Then if for simplicity's sake, we totally disregard any measurable harmonic distortion artifacts that would likely be simultaneously generated during these events, the effect on the affected circuit or audio device in question could be either peak dynamic compression and/or peak dynamic temporal smearing. Actually, in most such cases both will occur at the same time to varying degrees along with the distortion we just said we would ignore. As stated above, in virtually all practical cases the entire process occurs in matters of milliseconds or even more likely, microseconds... so it is not directly obvious as a source of distortion to the listener. Rather, the effect is commonly noticed as a "softening" or :dulling" of the affected sounds and instruments with respect to their dynamic impact.