Skip to Main Content
ADC-based receivers allow for extensive equalization in the digital domain and therefore can easily compensate for channel loss at higher data rates. Digital equalization can be implemented as an FFE or DFE. An adaptive FFE is straight forward to implement, as it relies on magnitudes only (not phases) of the blind samples, however, it enhances the quantization noise of the ADC. A DFE has better noise immunity, but challenges remain in designing an adap tive DFE for receivers with blind clocks. A typical adaptive engine for a DFE com pares the equalized center samples against a desired level to calculate error in each sample. The power of this error is then minimized to guide equalization. In a blind receiver, however, the center samples are not known; the blind samples may deviate from the eye's center and move closer to the zero crossings. As a result, one cannot use a single desired level independent of the blind sampling phase; the desired level must change according to the sampling phase. In this paper, we use a desired waveform, instead of a desired level, to perform the adaptation. Our measurement results confirm that this approach adapts the coefficients within 80μs and opens an otherwise closed eye for a channel atten uation of 13.3dB at 2.5GHz.