On Partial Response Signaling for MIMO Equalization on Multi-Gbit/s Electrical Interconnects

Abstract

Because of its ability to deal with intersymbol interference (ISI) and crosstalk (XT) over mutually coupled electrical interconnects, multiple-input multipleoutput (MIMO) decision feedback equalization (DFE) has proven to be a promising low-cost solution for achieving multi-Gbit/s wireline communication on-and off-chip. However, not only does the channel become very sensitive to manufacturing tolerances at very high symbol rates, the latency in the feedback loop becomes prohibitively large as well. Whereas the former issue has been addressed by adopting a stochastic MIMO approach where (part of) the equalization filters depend on the channel statistics rather than on the actual channel, we tackle in this paper the latency issue by setting to zero the first N taps of the feedback filters. Moreover, we show that precoded partial response (PR) signaling can improve the performance of the resulting scheme, although the achieved gain is smaller than in the case of single-input single-output (SISO) equalization.

Robust Spatio-Temporal Partial-Response Signaling over a Frequency-Selective Fading MIMO Channel with Imperfect CSI

“…Including additional feedback from either previous decision at the RX (MIMO decision feedback equalizer [2]) or previously transmitted symbols at the TX (MIMO Tomlinson-Harashima precoding [3]) can further enhance the performance of these equalizers. Partial-response signaling (PRS) offers a particularly interesting alternative to the traditional full-response signaling (FRS) as the former can handle large amounts of ISI [4], [5]. PRS is characterized by a target impulse response (TIR) with integer coefficients, enabling to perform the symbol-by-symbol detection in a periodically extended symbol set.…”

Optimized Precoded Spatio-Temporal Partial-Response Signaling Over Frequency-Selective MIMO Channels

“…Contrary to THP, this precoder does interestingly not induce any power loss. A common example of PRS is duo-binary signaling, where the TIR in z-transform notation is 1 + z −1 [10], [11].…”

“…Although PRS has already been described in the literature for the FS SISO channel [9] and for the flat-fading MIMO channel [12], no general ST PRS equalization scheme has been investigated for the FS MIMO channel. Indeed, the TIR in [12] and [10] is confined to spatial-only components and temporal-only components, respectively. Moreover, most contributions assume that the TIR is given [9], [10], [11], and no optimization with respect to this TIR is performed at all, resulting in an inferior trial-and-error selection of the target response.…”

“…Indeed, the TIR in [12] and [10] is confined to spatial-only components and temporal-only components, respectively. Moreover, most contributions assume that the TIR is given [9], [10], [11], and no optimization with respect to this TIR is performed at all, resulting in an inferior trial-and-error selection of the target response.…”

“…Interestingly, the TIR of the PRS contains both spatial and temporal components that are complex-valued as the baseband channel can be complex-valued. Contrary to the trial-and-error TIR selection often found in literature [9], [10], [11], the ST-TIR is jointly optimized with the ST-DFE according to the MMSE criterion. First, the optimal equalization coefficients can readily be expressed as a function of the TIR.…”