Copyright IMEC Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs) Submission Title: [A/D Converters for 60 GHz Radio] Date Submitted: [14 January, 2007] Source: [André Bourdoux, IMEC] Address [Kapeldreef 75, 3001 Leuven, Belgium] Voice:[+32-16-288215], FAX: [+32-16-281515], E-Mail:[bourdoux@imec.be] Re: [] Abstract: [] Purpose: [Information on Analog-toDigital converter technology] Notice: This document has been prepared to assist the IEEE P802.15. It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein. Release: The contributor acknowledges and accepts that this contribution becomes the property of IEEE and may be made publicly available by P802.15. André Bourdoux
A/D converters for 60GHz radio André Bourdoux Geert Van der Plas André Bourdoux
A/D research: innovative A/D solution for 60GHz radios Copyright IMEC A/D research: innovative A/D solution for 60GHz radios Radio Receiver ADC Analog FE Digital FE Energy-efficient, reconfigurable analog-to-digital converters in deep sub-micron CMOS technologies for 60GHz wireless systems ADC is converting analog continuous time to discrete time discrete amplitude, between analog FE and digital baseband The mission is: 60 GHz in the next few slides I will specify what the highlighted words mean in more detail André Bourdoux
Copyright IMEC The ADC FoM, the energy per conversion step EQ, is the energy efficiency measure P is the power consumption empirically EQ has been found to be almost constant for a given technology independent of fsample or effective number of bits (ENOB), hence a FoM. A value of 1 pJ/step is/was considered good current best public known value is 0.16pJ/step This FoM is also our measure for reconfigurable A/D: halving speed should halve power, removing one bit also what is energy efficieny for an ADC. Explain Eq in two steps say it is an empirical measure which is found to be almost constant independent of fsample and enob, for a given tech 1pJ/step, best value ever will be presented by Jan at ISSCC feb next year and is 65 fJ/step This defines our targets for the AD André Bourdoux
Deep sub-micron CMOS enables better energy efficiency Copyright IMEC Deep sub-micron CMOS enables better energy efficiency FoM is dependent on technology example: CMOS scaling has enabled reduction from 8.2 pJ/step (0.25u) to 0.16 pJ/step (90nm) for GS/s 4-6b flash converters Target technology is 65/45nm invent new or modify existing architectures to exploit the potential energy efficiency enabled by scaling You might have noticed that I said the energy per conv step is constant for a given technology ! André Bourdoux
60 GHz radios require 4-6 bit, GS/s A/D: Energy Efficiency is at best 0.16 pJ/step Design Power [mW] Resolution [bit] (ENOB) fs [GS/s] ERBW [MHz] Proc. [μm] FoM (EQ) [pJ] Jiang (ISSCC03) 310 6 (4.7) 2 750 0.18 3.5 Scholt. (JSSC02) 340 6 (5.7) 1.6 550 4 Varzagh. (VLSI04) 70 5 (4) 0.6 N/A 7.3 Park (ISSCC 06) 89(+dig) 4 (3.84) 2000 1.55 IMEC (DAC 06) 10.4 4 (3.7) 1 5800 0.8 Chen (VLSI 06) 105 6 (5.0) 400 Okada (ESSC03) 182 4 (N/A) 0.13 2.8 Sandner (JSSC05) 160 6 (5.5) 1.2 700 2.2 Schvan (ISSCC 06) 3000 5 (4.4) 22 14 Chen (ISSCC 06) 5.3 6 (N/A) 4000 0.22 Draxel. (ISSCC04) 10 6 (4.9) 300 0.09 0.54 Figuerei. (ISSCC 06) 55 6 (5.3) 500 1.4 IMEC (ISSCC 06) 2.5 1.25 3300 0.16 Park (CICC 06) 227 5 (3.6) <1000 18 Ginsburg (VLSI 06) 6 0.5 250 0.065 0.77 André Bourdoux
In 65/45 nm CMOS, A/D consume ~10 mW if we can maintain energy efficiency @ 0.1pJ Assume a 0.1 pJ/step energy efficiency a 4b, 4GS/s A/D ~6.4 mW a 5b, 2GS/s A/D ~6.4 mW a 6b, 2GS/s A/D ~12.8 mW etc. 1 mW 10 mW 100 mW André Bourdoux
@ IMEC we have A/D technology that can reach the 0.1pJ target in 65nm two power supplies buffer & core (both 1.8V for 0.18u) single-ended time multiplexed outputs (2) serial input data & clock Chip: 4b 1GS/s 0.8pJ in 0.18u, 4b 1.25GS/s 0.16pJ in 90nm single-ended clock input differential input André Bourdoux
Designs in the pipeline two 90nm designs are planned for early next year (Q1 2007) 5b ~2GS/s target power 10mW (0.15 pJ/step) 6b ~1GS/s target power 10mW (0.15 pJ/step) 65nm designs Q3 2007 increase sampling speed and improve energy efficiency (0.1pJ/step) André Bourdoux