The term "threshold" as applied to audio refers to a specific volume level where a process begins to take effect. In some cases the threshold is variable by the user and in other cases it is at a "fixed' level.
In digital circuitry, the term can apply to the voltage level reached by an input that results in the output of the device in question changing states. This is an important concept in understanding how jitter can result from variations in the waveform input to clock recovery circuitry.
In most cases, a threshold is a specific voltage level, largely because most signals under consideration exist as a voltage waveform. The threshold voltage level is compared to the source voltage level, which may be a waveform in cases where the instantaneous level is important or a processed signal when a more generalized or “average” level is important. When the input voltage level exceeds the threshold voltage level, a signal is generated.
Processing of analog signals for level control typically involves rectifying the AC waveform into a DC waveform and filtering this waveform to create a varying DC level. This level typically represents the average or R.M.S. voltage level of the original audio signal, but can also represent the peak level in applications such as audio limiting.
Digital Clock recovery
Ideal digital electronic signals have infinitely short time intervals for the transition between a “0” and a “1” and vice-versa. This will be referred to as a change of “state” and corresponds to the signal voltage changing between very close to ground or 0 volts and very close to the positive power supply voltage for the circuitry (+5VDC in TTL circuitry). For binary applications, the 0 Volt portion represents a binary “zero” and the 5V portion represents a binary “one.”
Real world signals do not change states in zero time, and this implies that some method must be used to decide when the input signal changes state.
This is accomplished by having a voltage threshold level at the input of the receiving device. Ideally this level would be approximately half of the total voltage range available, but for reasons that are beyond the scope of this discussion, it is typically closer to 25-35% of the total voltage range. As the voltage level of the input waveform slopes upward (or downward) it crosses this threshold, and at this point in time the output of the receiving device changes state.
Anything that changes the shape of the waveform can affect the relative timing of when the threshold is crossed, which is why the primary cause of jitter in signals transmitted between devices is waveform distortion. The single most important source of waveform distortion is reflections in the transmission chain caused by impedance miss-matches. Use of impedance matched cable and connectors as well as proper termination helps to greatly minimize impedance miss-matches in the chain. Interference caused by electromagnetic or electrostatic pickup on the cable or ground loops between equipment can also result in distortion of the waveform.
Analog Dynamic Range Control
In order to "fit" wide dynamic range audio signals into limited dynamic range recordings; some form of volume level control is necessary.
There are two types of volume control:
- Manual- Typically a Volume or Gain control on the recording device or a device which is feeding the inputs of the recording device.
- Automatic- Can be any one or a combination of the following: Compressor, limiter, expander/noise gate, analog soft limiting (or "analog saturation"), digital soft saturation.
In the case of compressor, limiter, or expander/noise gate the threshold is typically user-controllable in the form of a Threshold adjustment or a combination of input and output level controls. Because saturation modes are basically peak level controls; their thresholds are generally fixed near peak volume level.
The basic premise is that for signals with a volume level lower than the threshold the signal is unchanged. As the volume level increases and exceeds the threshold, the process begins. In some cases, the process is applied in a linear manner where every increase of the input level by one dB results in the same amount of change to the output signal, and in other cases the process has a more "progressive" non-linear effect where the effect is minimal just above the threshold and increases as the input level increase. The saturation modes of Lavry AD converters are an example of non-linear processes. Another example is the "soft-knee" compressor or limiter.
- The one exception in the examples provided is in the case of an expander/noise gate; in which case the signal is unchanged above the threshold and processed below the threshold.