If data or clocks are sent down long lines, these must be terminated to prevent ringing. Ringing is generated on the transitions of digital signals when a portion of the signal is reflected back down the line due to a mismatch between the line impedance and the terminating impedance. A similar mismatch at the driving end will re-reflect a further portion towards the receiver, and so on.
The amplitude of the ringing depends on the degree of mismatch at either end of the line while the frequency depends on the electrical length of the line. But if the transition time (rise or fall time) of the edge of the digital signal is substantially slower than the round-trip time down the line and back again, the edge will still be in transition when the reflection arrives to add to it, and a fully-formed ring oscillation will not occur. A digital driver/receiver circuit should be treated as a transmission line, and therefore properly terminated, if
2 x tPD x line length > rise time
where tPD is the line propagation delay in ns per unit length; for FR4 fibreglass PCB, tPD is typically 0.056ns/cm for surface microstrip and 0.07ns/cm for embedded stripline
Termination is achieved by adding series or parallel resistors so that the output and input impedances at either end of the line are matched to the impedance of the line itself. This in turn implies that the PCB structure which makes up the line must have a “controlled impedance”, so that its dimensions and the relative permittivity of the PCB material are specified. At the driver end, the output impedance of the driver plus the series resistance should make up to the transmission line characteristic impedance Z0. At the receiver end, given that a CMOS receiver has generally a very high input impedance, the combination of terminating resistors should be set to Z0.
The picture below demonstrates how varying the serial termination resistor value of a transceiver can reduce overshoot and ringing.
Ringing and EMC
Emissions
Ringing is to be avoided in digital circuits because of its implications for signal integrity: if the ringing amplitude is greater than the threshold of the receiver, spurious transitions will be registered and the circuit's operation will suffer. But aside from this effect, ringing may also be a source of interference in its own right. The effect of ringing on any clock or other switching signal, is to create a concentration of RF energy around the ringing frequency, modulating nearby harmonics of the carrier clock or data. These harmonics will be emphasized in the emissions spectrum and can be mistaken for a resonant coupling peak. They will be resistant to diagnosis and treatment until the relevant transmission line or group of lines is isolated and matched. Using a near-field probe around a PCB and an RF current probe on the cable to look for a matching spectrum peak, you can detect the area of the lines that need to be dealt with.
Because the energy in clocks is concentrated at their harmonics, the relative increase in amplitude due to ringing is not that great; a more significant increase can be seen when ringing occurs on data lines which inherently by themselves are creating more broadband emissions.
Immunity
Severe ringing will affect the data transfer since it can exceed the device’s input noise margin. Ringing on a data or clock line automatically implies that the logic decision threshold margin has been eroded for the period of the ring. That is, at the peak of the ring the input voltage to the receiving logic node is much closer to its switching threshold than during the quiescent logic 1 or 0 state. If the ring amplitude is high enough, then the threshold will be exceeded for a time after the correct edge has been received, and this will be registered as a false edge, leading to incorrect operation. This is well understood and should be trapped and corrected in functional testing, since it doesn't of itself have any relationship to external EMC. It is an example of a classical signal integrity failure.
But it is also possible that the ringing is not at a high enough level to cause such an SI failure. The ringing, although present, remains at a level below the logic threshold during normal operation and therefore there is no degradation of function. This may not be detected, or treated, in functional tests. But if it is augmented by an external signal at a close frequency to the ringing frequency, the ringing and the external signal can add in-phase to a level sufficient to push the output voltage over the threshold. This then appears as a susceptibility to external RF, worst at or around the ringing frequency. In other words, the product's immunity has been compromised by the ringing.
Radiated emission (RE) and radiated immunity (RI) have become much more pervasive issues in recent years. There are several reasons for this, which include the shift to more compact design, more portable products, as well as the fact that noise margins for digital logic levels continue to be reduced as supply voltage decreases (The popularity of using ARM based architecture means 3.3V systems replace 5V systems).
So it is good practice to deal with ringing from an EMC viewpoint, even if the circuit's normal operation is unimpaired and its emissions remain below limits. As well as the consequences outlined here, it is possible to show that any differential circuit interacts least with external high frequency fields when its conductors form a matched transmission line. Both emissions and susceptibility are controlled by this technique.
|