40 60 80 100 120406080mm14. CAMs, ROMs, PLAsJ. A. AbrahamDepartment of Electrical and Computer EngineeringThe University of Texas at AustinEE 382M.7 – VLSI IFall 2011October 17, 2011ECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 1 / 2940 60 80 100 120406080mmContent Addressable Memories (CAMs)Extension of ordinary memory (e.g., SRAM)Read and write memory as usualAlso match to see which words contain a keyECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 1 / 2940 60 80 100 120406080mm10-Transistor CAM CellAdd four match transistors to 6T SRAM56 × 43 λ unit cellECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 2 / 2940 60 80 100 120406080mmCAM Cell OperationRead and write like ordinary SRAMAdditional “match” operationFor matching:Leave wordline lowPrechargematchlinesPlace key onbitlinesMatchlines evaluateMiss linePseudo-nMOS NORof match linesGoes high if nowords matchECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 3 / 2940 60 80 100 120406080mmRead-Only Memories (ROMs)Read-Only Memories are nonvolatileRetain their contents when power is removedMask-programmed ROMs use one transistor per bitPresence or absence determines 1 or 0Example: 4-word × 6-bit ROMLooks like 6 4-input pseudo-nMOS NORsCan be represented by a “dot diagram”Dots indicate 1s in ROMWord 0: 010101Word 1: 011001Word 2: 100101Word 3: 101010ECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 4 / 2940 60 80 100 120406080mmROM LayoutUnit Cell12 × 8 λ (≈1/10 size of SRAM)Row DecodersROM row decoders mustpitch-match with ROM (only asingle track per word)ECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 5 / 2940 60 80 100 120406080mmROM Layout, Cont’dComplete ROM LayoutECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 6 / 2940 60 80 100 120406080mmPROMs and EPROMsProgrammable ROMs (PROMs)Build array with transistors at every siteBurn out fuses to disable unwanted transistorsElectrically Programmable ROMs (EPROMs)Use floating gate to turn off unwanted transistorsEPROM, EEPROM, FlashECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 7 / 2940 60 80 100 120406080mmBuilding Logic with ROMsROM as lookup tablecontaining truth tablen inputs, k outputsrequires 2nwords × k bitsChanging function is easy– reprogram ROMFinite State Machinen inputs, k outputs, s bitsof stateBuild with 2n+s× (k + s)bit ROM and (k + s) bitregisterECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 8 / 2940 60 80 100 120406080mmExample: RoboAntSensors: Antennae(L,R) = 1 when in contactActuators: LegsForward step FTen degree turns TL, TRGoal: Make the ant smart enough toget out of a mazeStrategy: Keep right antenna on wall(RoboAnt adapted from MIT 6.004 2002 OpenCourseWare by Ward andTerman)ECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 9 / 2940 60 80 100 120406080mmRobotAnt – Determine StatesLost in SpaceAction: go forward until it hits somethingInitial StateECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 10 / 2940 60 80 100 120406080mmRobotAnt – States, Cont’dHits somethingAction: turn left (rotate counterclockwise)Until it doesn’t touch anymoreECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 11 / 2940 60 80 100 120406080mmRobotAnt – States, Cont’dMove a little to the rightAction: step forward and turn right a littleLooking for the wallECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 12 / 2940 60 80 100 120406080mmRobotAnt – States, Cont’dMove a little to the rightAction: step and turn left a little, until not touchingECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 13 / 2940 60 80 100 120406080mmRobotAnt – States, Cont’dDealing with a cornerAction: step and turn right until hitting next wallECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 14 / 2940 60 80 100 120406080mmSimplificationMerge equivalent states where possibleIf state machine is described in a Hardware DescriptionLanguage, the synthesis tool will perform the simplificationECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 15 / 2940 60 80 100 120406080mmState Transition TableS1:0L R S01:0TR TL FLost00 0 0 00 0 0 100 1 X 01 0 0 100 0 1 01 0 0 1RCCW01 1 X 01 0 1 001 0 1 01 0 1 001 0 0 10 0 1 0Wall110 X 0 10 1 0 110 X 1 11 1 0 1Wall211 1 X 01 0 1 111 0 0 10 0 1 111 0 1 11 0 1 1ECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 16 / 2940 60 80 100 120406080mmROM Implementation16-word × 5-bit ROMECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 17 / 2940 60 80 100 120406080mmProgrammable Logic Arrays (PLAs)A Programmable Logic Array performs any function insum-of-products formLiterals: inputs and complementsProducts terms: AND of literalsOutputs: OR of Product termsExample: Full Adders = a¯b¯c + ¯ab¯c + ¯a¯bc + abcc = ab + bc + acECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 18 / 2940 60 80 100 120406080mmNOR-NOR PLAsANDs and ORs not very efficient in CMOSDynamic or Pseudo-nMOS NORs very efficientUse DeMorgans Law to convert to all NORsECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 19 / 2940 60 80 100 120406080mmPLA Schematic and LayoutECE Department, University of Texas at Austin Lecture 14. CAMs, ROMs, PLAs J. A. Abraham, October 17, 2011 20 / 2940 60 80 100 120406080mmPLAs versus ROMsThe OR plane of the PLA is like the ROM arrayThe AND plane of the PLA is like the ROM decoderPLAs are more flexible than ROMsNo need to have 2nrows for n inputsOnly generate the product terms that are neededTake advantage of logic