Jitter
Overview
The term "jitter" is used to describe variations in a periodic signal, which can be in the frequency, amplitude, or phase of the signal in relationship to the idea waveform or the waveform at the point of generation. In digital audio; one of the most problematic issues involving jitter is in "clock recovery" from signals transmitted between equipment.
- If jitter in the recovered clock signal affects the clocking of conversion, even small amounts of jitter can effectively reduce the resolution far below the theoretical limits of 16-24 bit conversion.
- In purely “digital” transmission such as the digital audio connection between an AD converter and a recording computer jitter has no effect on the quality of the encoded digital audio, unless it is so extreme that it results in data corruption. When the digital audio data becomes corrupted, the effects are not subtle and are typically in the form of mutes or loud “pops.” However, if the clock recovered from the digital audio is used to clock a DA converter, the conversion quality of the DA may be degraded. This may mislead the listener to conclude that the jitter affected the recording, when in fact it is only affected the reproduction of the recording at that time.
Basics
Because even "digital" signals are actually very high frequency analog signals, the receiving device must reconstruct the signal by means of some form of amplitude "threshold." Ideal digital signals have a "zero rise time and zero fall time" which means the change in voltage level between a "1" and a "0" requires no time. In digital circuitry, a "1" is represented by a voltage level very close to the power supply voltage; which is typically a value like 5 Volts or 3.3 Volts. In the same circuitry, a “0” is represented by a voltage level very close to “0 Volts” or in other words ground. In order to reduce power dissipation in newer designs, power supply voltages are being reduced further; but that is beyond the scope of this discussion.
Because of the difference between the ideal signal and the real signal, the transition between a “0” and a “1” (between 0 Volts and 5 Volts) takes a finite amount of time; which is referred to as the “rise time.” The reciprocal change from 1 to 0 is referred to as the “fall time.” Thus a threshold is needed to decide when the signal has changed from a “0” to a “1” (or vice-versa) and this is a voltage threshold which is fixed, by design. If either the amplitude of the signal or the “slope” (rise/fall time) of the signal changes; the signal will cross the threshold at a different time relative to the waveform at the source; thus effectively changing the amplitude variation into a timing variation.
Although the voltage threshold is “fixed,” variations in the voltage of the power supply on which the digital circuitry operates or ground noise can affect the absolute voltage level of the threshold at any instant in time. Whether the variation is in the signal waveform or caused by this type of non-ideal circuit characteristic, the result is still a timing variation of the recovered clock signal.
There are a number of sources of jitter. In signal transmission between equipment the primary cause is “reflections” of the signal caused by an impedance miss-match in the signal path. A miss-match can result from things like the impedance of a connector differing from the impedance of the cable, connector to PC board wiring, or circuit board conductors in series with the connector. The transmitting and receiving IC’s also contribute a finite amount of jitter to the signal.
Much like ripples on the surface of a pond, the “waves” add or subtract from each other in a complex changing pattern that results in changes to the shape of the transmitted waveform. These variations cause the waveform to cross the voltage threshold at different (relative) times than the transitions of the original signal; resulting in a time-domain distortion in the “recovered” output signal regardless of whether the distortion of the shape of the waveform was in the time domain or the amplitude domain.
Cables and Connectors
An example of an impedance matched connector is a 75 Ohm BNC connector typically used for Word Clock connections. When installed on cable such as RG-59 with a characteristic impedance of 75 Ohms, the impedance miss-match is minimized.
“Characteristic” impedance is the impedance of the cable at the frequency or range of frequencies typically transmitted through the cable. It is different than the DC resistance of the cable, which is typically in the range of 0.25 to 2 Ohms (depending mostly on cable length).
Termination
Proper termination is important to achieving low-jitter transmission. The termination at both the sending end and receiving end must match the characteristic impedance of the cable to minimize reflections. Most Word Clock cables are made with 75 Ohm coaxial cable, so the termination would typically be a form of 75 Ohm resistor. For BNC connections, a BNC Terminator can be used in conjunction with a BNC T to provide termination on inputs with no internal or selectable termination.
For more information on jitter click here
For more information on the effect of jitter in conversion click here