The Impact of Analog-to-Digital Converter Architecture and Variability on Analog Neural Network Accuracy
نویسندگان
چکیده
The analog-to-digital converter (ADC) is a key component in analog in-memory computing accelerators, but also bottleneck for the efficiency and accuracy of these systems. While trade-offs between power consumption, latency, area ADC design are well studied, it relatively unknown which implementations optimal algorithmic accuracy, particularly neural network inference. We explore space with focus on investigating sensitivity outputs to variability inside how this depends architecture. Compact models pipeline, cyclic, successive-approximation-register (SAR) ramp ADCs developed, used system-level simulation Our results show complex image recognition benchmark (ResNet50 ImageNet) capacitance mismatch, comparator offset, effective number bits (ENOB) each four architectures. find that robustness variations strongly design, inference sensitive value-dependent error characteristics cannot be captured by conventional ENOB precision metric.
منابع مشابه
signal specific successive approximation analog to digital converter
چکیده: در میان انواع متفاوتی از مبدل های آنالوگ به دیجیتال که تا کنون معرفی شده اند، مبدل های آنالوگ به دیجیتال تقاریب متوالی(sar ) به علت سادگی ساختار و همچنین توان مصرفی کم، همواره یکی از پرکاربرد ترین مبدل های آنالوگ به دیحیتال در کاربرد های بایومدیکال بوده اند. به همین دلیل تاکنون روش های متعددی برای کاهش هرچه بیشتر توان مصرفی در این مبدل ها ارائه شده است که در اکثر آنها توجهی به مشخصات سی...
15 صفحه اولDesigning an Integrated All-Optical Analog to Digital Converter
We present the procedure for designing a high speed and low power all-optical analog to digital converter (AO-ADC), by integrating InGaAsP semiconductor optical amplifier (SOA) with InP based photonic crystal (PhC) drop filters. The self-phase modulation in the SOA can shift the frequency of the Gaussian input pulse. The two output PhC based drop filters are designed to appropriately code the f...
متن کاملTrimming a digital-to-analog converter to improve accuracy
More and more of today’s modern industrial systems are requiring high accuracy. Traditionally, the test and measurement market has been the primary driver of highly accurate signal chain components, but now this trend is seeping into other markets such as factory automation, optical networking, and medical. Applications such as automated test equipment (ATEs), data acquisition cards (DAQs), and...
متن کاملAnalog-to-Digital Converter Testing
Analog-to-digital converters are essential building blocks in modern electronic systems. They form the critical link between front-end analog transducers and back-end digital computers that can efficiently implement a wide variety of signal-processing functions. The wide variety of digitalsignal-processing applications leads to the availability of a wide variety of analog-to-digital (A/D) conve...
متن کاملOptimal fast digital error correction method of pipelined analog to digital converter with DLMS algorithm
In this paper, convergence rate of digital error correction algorithm in correction of capacitor mismatch error and finite and nonlinear gain of Op-Amp has increased significantly by the use of DLMS, an evolutionary search algorithm. To this end, a 16-bit pipelined analog to digital converter was modeled. The obtained digital model is a FIR filter with 16 adjustable weights. To adjust weights o...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: IEEE Journal on Exploratory Solid-State Computational Devices and Circuits
سال: 2023
ISSN: ['2329-9231']
DOI: https://doi.org/10.1109/jxcdc.2023.3315134