Design of Low Power Memory Controller with Adaptive Eye Detection Algorithm

Abstract

and the read margin was enhanced from 0.30UI and 76mV without AF-CTLE to 0.47UI and 80mV to with AF-CTLE. The power efficiency during burst write and read were 5.68pJ/bit and 1.83pJ/bit respectively.

A 4266Mb/s/pin LPDDR4 memory controller with an asynchronous feedback continuous-time linear equalizer and an adaptive 3-step eye detection algorithm is presented. The asynchronous feedback continuous-time linear equalizer removes the glitch of DQS without training by applying an offset larger than the noise, and improves read margin by operating as a decision feedback equalizer in DQ path. The adaptive 3-step eye detection algorithm reduces power consumption and black-out time in initialization sequence and retraining in comparison to the 2-dimensional full scanning. In addition, the adaptive 3-step eye detection algorithm can maintain the accuracy by sequentially searching the eye boundaries and initializing the resolution using the binary search method when the eye detection result changes. To achieve high bandwidth, a transmitter and receiver suitable for training are proposed. The transmitter consists of a phase interpolator, a digitally-controlled delay line, a 16:1 serializer, a pre-driver and low-voltage swing terminated logic. The receiver consists of a reference voltage generator, a continuous-time linear equalizer, a phase interpolator, a digitally-controlled delay line, a 1:4 deserializer, and a 4:16 deserializer. The clocking architecture is also designed for low power consumption in idle periods, which are commonly lengthy in mobile applications. A prototype chip was implemented in a 65nm CMOS process with ball grid array package and tested with commodity LPDDR4. The write margin was 0.36UI and 148mV