- 1.1 Introduction
- 1.2 System Architectures
- 1.3 Line Coding of Digital Signals
- 1.4 Electrical Signaling
- 1.5 Summary
- 1.6 References
This chapter is from the book
This chapter is from the book
This chapter is from the book
1.4 Electrical Signaling
1.4.1 Single-Ended Signaling
Single-ended systems use a shared reference rail for both the transmitter and the receiver that provides the reference level for zero/one bit decisions and also carries the return current (Figure 1-28). The reference rail (or ground) needs to be sufficiently low in impedance; otherwise, the return currents will cause a voltage drop that reduces the signal across the receiver inputs. Other concerns are noise coupling and electromagnetic interference.
)
Figure 1-28 Single-ended system. The receiver makes bit decisions based on the difference between the signal and a reference voltage.
1.4.2 Differential Signaling
Differential signaling uses pairs of wires: One of the wires carries the signal, while the other wire carries the inverse of the signal (Figure 1-29). The receiver makes its bit decisions based on the difference between the two wires and thereby removes the dependency on the signal ground; a differential signal essentially carries its own reference.
)
Figure 1-29 Differential system. The receiver makes bit decisions based on the difference between two signals; the reference voltage is irrelevant.
We call the two signals signal A and signal B. The differential signal is then the difference between the two signals:
Equation 1-14
The common signal is the average of the two signals:
Equation 1-15
Figure 1-30 shows an example. The two single-ended signals with 400 mV amplitude (peak to peak) and an offset of 1.0 V can be decomposed into an 800 mV peak-to-peak differential signal and a 1.0 V constant common signal.
)
Figure 1-30 Differential signaling example: 100 MHz digital clock signal, single-ended amplitude of 400 mV (peak to peak) with a 1.0 V offset. Top to bottom: Signal A, signal B, differential signal, common signal.
A differential system is fully described either by the two signals A and B or by the differential and common signals. If the differential and common signals are known, the signals on the two lines can be recovered as the sum (for VA) and difference (for VB) of the common signal and half the differential signal:
Equation 1-16
Equation 1-17
If the two wires are routed in parallel, they receive nearly the same interference; this makes the system almost immune to most types of common mode noise coupling and electromagnetic interference. Figure 1-31 shows an example. Note how the sinusoidal interference signal shows up in the common mode of the signal, while the differential signal is not affected at all.
)
Figure 1-31 Differential signaling example: 100 MHz digital clock signal, single-ended amplitude of 400 mV (peak to peak) with a 1.0 V offset. Interference signal is a 5 GHz sinusoidal, with amplitude of 200 mV peak to peak. Top to bottom: Signal A, signal B, differential signal, common signal.
Because of its noise immunity, differential signaling is the method of choice for almost all high-speed serial data transmission systems. There are associated costs, however: Two wires per signal require twice as many pins and double the routing space. Because differential pairs need to be routed close together, and ideally with low skew, the routing complexity increases. However, this additional complexity is offset because we don't need nearly as many ground signals as for single-ended signaling.
1.4.3 Preemphasis and Receiver Equalization
High-speed serial signaling suffers from bandwidth limitations in transmission channels. Especially in cost-sensitive applications, where less expensive board materials are used, signal distortions introduced by the channel can render a signal unusable.
Signal distortions due to high-frequency cutoff can be moderated to some degree by signal-processing techniques in the transmitter. Two options are available: We can either amplify the higher-frequency spectral components of the signal or attenuate the low-frequency content of the signal. The former is called preemphasis, the latter deemphasis. Signal-processing methods can also be applied to the receiver. Several types of receiver equalization can extract a meaningful signal out of an apparently random signal.