Decision Feedback Equalization

Figure 1: DFE filter output is based on
a linear combination of previous bit
decisions

In debugging high-speed serial links, one must be cognizant of various forms of equalization that might be used in the link to compensate for signal degradation in the channel. Inter-symbol interference (ISI), attenuation, impedance mismatches, and insertion losses can all contribute to this loss of signal quality. To combat these effects, designers implement techniques such as continuous time linear equalization and feed-forward equalization.

Another such technique, known as decision feedback equalization (DFE), can deliver much better noise performance in that it doesn't amplify the noise level at the receiver. DFE is a physical implementation that compensates the signal by feeding back the detected signal (1 or 0) after weighting it (Figure 1). The feedback then shifts the signal before the detector depending on the detected bit. DFE operates much like de-emphasis, except that the feedback signal is binary. It's also non-linear because the feedback signal is discrete.

Let's say you're using one-tap DFE. It will feed back what the previous bit was, in increments of 1 or 0. That enables the bit slicer to make a decision as to where to have the transition level for the bit. It's a very effective technique for compensating for post-cursor ISI, but not so good for pre-cursor ISI. Basically, it's a non-linear system implemented as a digital filter.

One drawback of DFE is that if the bit slicer makes one wrong bit decision, it can lead to a burst of errors: the feedback of an incorrect decision shifts the next bit in the wrong direction, resulting in a subsequent error that's then fed back again, and so on. However, the receiver will quickly recover from such situations, especially if the data sequence is random.

Overall, DFE is a good solution in terms of noise performance and is better in this regard than either CTLE or FFE. Also, because it's a physical implementation, it is relatively inexpensive, making it the preferred choice in many scenarios.

Our next post will discuss some of the limitations of the various equalization techniques.

Previous posts in this series:

Introduction to Debugging High-Speed Serial Links
A Look at Transmission-Line Losses
How Much Transmission-Line Loss is Too Much?
Inter-Symbol Interference (or Leaky Bits)
Rise-Time Degradation and ISI Jitter
Introduction to Channel Equalization
The Effects of De-Emphasis on Eye Diagrams
Serial-Data Channel Emulation and S Parameters
Continuous Time Linear Equalization
Feed-Forward Equalization