BIO:

Degang Chen (IEEE Fellow, PhD’1992 UC Santa Barbara) is the Jerry Junkins Chair Professor of Engineering at ISU. Prior experience includes John Pierce Instructor at CalTech, Faculty Fellow with Boeing, Maxim Integrated, and Texas Instruments. Research interests include AMS-IC design and testing. Dr. Chen and his students’ work has received many best paper awards and other honors, including the highly prestigious IEEE Ned Kornfield Best Paper Awards in 2013, 2014, and 2019, the IEEE EJ McCluskey Best Ph.D. Dissertation Awards in 2017 and 2019, and the biannual Karas Outstanding Dissertation Award in 2024.

 ABSTRACT:

Data converters are crucial in the age of AI computing, In-Memory computing, IoT, dense smart sensing, and automotive electronics. They are deeply embedded, causing difficulties in access for testing, trimming and calibration. The need for many converters in one chip also necessitates ultra-small area and low power design, which makes matching and noise performance very difficult to achieve.

We first show that reduced matching produces two types of errors: nonrecoverable errors due to loss of information and recoverable errors that can be tested and calibrated. We then introduce the use of a special redundant bit targeted at preventing nonrecoverable errors. Built-in self-test algorithms not requiring any test instruments or any accurate analog circuitry are introduced. The test results are then used in a low-complexity digital calibration method that will correct all mismatch induced and nonlinearity induced errors. Several DAC topologies with segmented architectures and redundancy will be presented. Analytical and simulation results will be presented to illustrate how the proposed method works. We show that for a 14-bit DAC design, only 7-bit component matching is needed. In comparison to the conventional design requiring 14-bit matching, the area requirement can be reduced by a huge factor. Three DACs, with different architectures, have been fabricated. Measurement results show that after calibration 14-bit linearity has been achieved.