先界定更新與供電需求 再選擇 MCU 的持久狀態架構Define Update and Power Requirements Then Select the MCU State Architecture
28 nm 是檢查製程產品組合的節點,不是 eFlash 的普遍物理截止線。先依具名 MCU 或記憶體巨集核對讀寫電壓、更新需求、光罩與資格驗證;再用下方教學模型探索近臨界電壓和修補架構。模型不證明特定產品的功耗、電池壽命或資料保持能力。28 nm is a process-portfolio checkpoint, not a universal physical cutoff for eFlash. Check read/program supplies, update requirements, masks and qualification for the named MCU or memory macro, then explore near-threshold and patch architectures with the teaching models below. These models do not establish product power, battery life or retention.
TSMC-ULP-001
0 Extra零額外光罩Mask adders on a named standard-CMOS option; IP datasheet pending特定標準 CMOS 選項;IP 規格書待補
TSMC-NTV-001
0.5 VNear-threshold cold boot; named node and datasheet pending近臨界電壓冷開機;特定節點與規格書待補
TSMC-AON-001
<10 nADeep-sleep AON leakage; named measurement report pending深度睡眠 AON 漏電流;特定量測報告待補
TSMC-RET-001
15+ Yrs15 年以上Retention at 125°C; named IP datasheet pending125°C 資料保存;特定 IP 規格書待補
These four numbers stay bound to the record ids above. The existing records do not yet identify the specific IP, node and datasheet. They are not AntiFuse or eFlash family guarantees.以上四項數據分別對應所示紀錄編號;現有紀錄尚未列明特定 IP、節點及規格書,並非 AntiFuse 或 eFlash 技術家族的保證值。
PURE LOGIC NVM純邏輯 NVMZERO EXTRA MASKS · 0.5V NTV CONTINUUM零額外光罩 · 0.5V NTV 技術演進
FAIL CLOSED失敗即關閉
Zero extra masks is an option, not a promise零額外光罩是選項,不是承諾
Named-condition figures do not upgrade themselves into a certification.具名條件下的數字,不會自行升級成認證。
Pure-logic AntiFuse OTP can be a zero-adder route in standard CMOS.純邏輯 AntiFuse OTP 可以是標準 CMOS 的零加價路線。
完全相容標準 FEOL/BEOL 邏輯製程,免額外光罩與熱預算衝突。100% standard FEOL/BEOL logic compatible; 0 mask adders and zero thermal budget penalty.
RULE #02 · VOLTAGE CONTINUUM原則 #02 · 電壓演進
0.5 V
近臨界電壓教學輸入;固定活動率、電容與頻率時,動態切換功耗才可用電壓平方估算。Near-threshold teaching input; the voltage-squared estimate assumes fixed activity, capacitance and frequency.
RULE #03 · INSTANT ON原則 #03 · 即時啟動
<10 µs
無高壓電荷泵預充等待,冷啟動瞬時 XIP 執行,休眠漏電 <10nA。No charge pump warmup; instant-on cold boot XIP with <10nA AON deep-sleep leakage.
RULE #04 · RETENTION原則 #04 · 資料保存
15+ YRS15 年以上
125°C、15 年為待查證的資料保持情境;需具名 IP 規格與任務剖面,不能當作電池壽命。The 15-year, 125°C retention scenario needs a named IP specification and mission profile; it is not a battery-life claim.
LENS 01 · THE EXECUTIVE THESIS觀點 01 · 高階決策論述
製程選擇回到可用選項,讀取、寫入與更新分別核對Check Available Process Options: Separate Read, Program and Update Requirements
評估 MCU 的 eFlash、OTP 或其他嵌入式記憶體時,應取得具名製程的元件選項、寫入供電、熱預算與商業可用性。OTP 的單次寫入限制仍存在;是否能以間接跳轉修補程式,取決於保留空間、控制器與更新政策。Evaluate eFlash, OTP and other embedded memories against the named process device options, program supplies, thermal budget and commercial availability. OTP remains one-time programmable; indirection-based code patching depends on reserved capacity, controller support and update policy.
[FIRST-PRINCIPLES PHYSICS EQUATION][第一性原理物理方程式] 近臨界電壓動態能耗與等頻功耗模型:Near-threshold dynamic energy & iso-frequency power model:Ecycle = α · Cload · Vdd2,等頻功耗 , iso-frequency power Pdyn = Ecycle · f 當供應電壓由 When the supply voltage decreases from Vnom = 1.2V 降至 to Vntv = 0.5V:: ΔE = 1 - (0.5 / 1.2)2 = 1 - 0.1736 = 82.6%每週期動態能耗淨節省,同時靜態漏電 net dynamic energy saving per cycle, with static leakage Ileak < 10nA (AON 網域)。 (AON Domain).
01 · LOGIC IP01 · 邏輯 IP
零額外光罩0-Mask Adder
100% 採用標準 CMOS 閘氧化層介電質擊穿,免除 8~12 道光罩。100% standard CMOS gate dielectric breakdown; eliminates 8~12 extra masks.
02 · VOLTAGE02 · 電壓
0.5V 近臨界電壓0.5V Near-Threshold
原生支援超低電壓,消除升壓電荷泵,開機冷啟動 <10µs。Native ultra-low voltage support, zero charge pumps, cold boot <10µs.
03 · HOT PATCH03 · 即時修補
向量 CAM 熱修補Vector CAM Hot Patch
硬體即時重定向匯流排程式計數器,實現不可抹除 OTP 的線上熱修補。Hardware bus interception enables zero-cycle firmware hot-patching.
04 · LIFETIME04 · 使用壽命
15年長效15+ Yr Longevity
125°C 閘介電質擊穿再結晶矽微絲零漂移,終生免維護免換電池。Permanent recrystallized silicon filament with zero drift at 125°C, battery-free lifetime.
LENS 02 · INTERACTIVE SCOPE LAB觀點 02 · 互動示波器實驗室
調降供應電壓,動態功耗呈現平方級跳水Step Down Supply Voltage: Dynamic Power Drops by Square Law
滑動調節 Vdd 供電電壓,觀察 0.5V 近臨界電壓 (Near-Threshold) 下開機時序波形與動態功耗指數跳水。AntiFuse OTP 原生支援 0.5V 讀取,無需傳統 eFlash 20~50µs 的電荷泵暖機預充。Adjust the Vdd slider to observe boot waveforms and square-law dynamic power reduction at 0.5V near-threshold voltage. AntiFuse OTP natively reads at 0.5V, bypassing conventional eFlash 20~50µs charge-pump warmup.
極低功耗 IoT MCU 休眠/喚醒能耗預算與電池壽命試算器Ultra-Low-Power IoT MCU Energy Budget & Battery Lifetime Explorer
奈安級待機與瞬間啟動SUB-10nA SLEEP & INSTANT-ON
10 s
5 ms
預估電池使用壽命EXPECTED BATTERY LIFE
—
平均放電電流AVERAGE CURRENT DRAW
—
有效工作週期ACTIVE DUTY CYCLE
—
週期能耗分佈佔比 (Active vs Standby Leakage vs Wake-up):CYCLE ENERGY DISSIPATION BREAKDOWN:運算Active休眠漏電Sleep Leakage喚醒開銷Wake Overhead
—
LENS 02B · SUB-VT LAB觀點 02B · 次閾值極限實驗室
次閾值與近閾值 (0.3V~0.7V) 超低壓 eNVM 讀取能耗與位元漏電比模擬器Sub-Threshold & Near-Threshold (0.3V–0.7V) Ultra-Low-Voltage eNVM Energy & Leakage Ratio Simulator
在無電池環境能量採集 (0.35V~0.5V VDD) 下,MOSFET 驅動電流呈對數擴散崩跌,感測延遲增加誘發靜態漏能反向暴增。透過第一性原理動態 CV² 能耗、Pelgrom 微觀失配與最小能耗點 (MEP) 分析儀,探索 AntiFuse、ReRAM、eFlash 與 10T SRAM 的極限超低壓工作邊界。Under battery-less energy harvesting (0.35V–0.5V VDD), sub-threshold transconductance collapses, and inflated sensing latency triggers a massive leakage surge. First-principles analysis of active CV², Pelgrom statistical mismatch, and Minimum Energy Point (MEP) across AntiFuse, ReRAM, eFlash, and 10T SRAM.
⚠️ 示意模擬模型與技術邊界說明:⚠️ Schematic Model & Technical Boundary Notice:本工作台之次閾值動態能耗 (CV²)、靜態漏電與最小能耗點 (MEP) 為基於典型 40nm/28nm ULP PDK 標稱參數與 Pelgrom 失配高斯模型之數值示意,非實際矽晶圓跨製程角 (Corner) 全溫全壓實測數據。超低壓感測裕度需依據晶圓代工廠極限失配蒙地卡羅 (Monte Carlo) 模擬 sign-off 判定。The sub-threshold active energy (CV²), static leakage, and Minimum Energy Point (MEP) metrics in this workbench are schematic numerical estimations based on typical 40nm/28nm ULP PDK nominals and Pelgrom Gaussian mismatch models, not multi-corner silicon wafer measurements. Ultra-low-voltage sense amplifier margins must be validated through foundry sign-off Monte Carlo statistical analysis.
⚠️ 示意模擬模型與技術邊界說明:本模型基於 Pelgrom 晶體管失配公式與臨界電荷 (Qcrit) 近似,模擬次閾值感測裕度與 Alpha 粒子軟錯誤趨勢,非特定 MCU 晶片之重離子加速器輻照實測。⚠️ Illustrative Simulation Model & Technical Boundary Note: Grounded in Pelgrom mismatch and critical charge (Qcrit) approximations; illustrative for subthreshold margin planning, not heavy-ion accelerator radiation test reports.
當 IoT 節點降壓至近閾值或次閾值區 (0.4V~0.8V VDD) 以追求極致微瓦級節能時,NVM 感測電流從數十微安急降至數十皮安,差分感測裕度極度收窄;同時臨界電荷 (Qcrit) 驟降使 Alpha 粒子與中子引發的軟錯誤率 (SER) 飆升 10~50 倍。本模擬器評估不同 NVM 物理單元在極端低壓下的耐受性。When IoT nodes scale voltage into near-threshold or subthreshold regimes (0.4V-0.8V VDD) for microwatt energy harvesting, NVM read currents collapse into nanoamperes, severely squeezing differential sense margins; simultaneously, plummeted Qcrit spikes alpha-particle soft error rates (SER) by 10~50x.
次閾值感測電流 (I_read)Read Current (I_read)650.0 nA
淨差分感測裕度Net Sense Margin45.2 mV
翻轉臨界電荷 (Qcrit)Critical Charge (Qcrit)0.25 fC
軟錯誤率 (SER)Soft Error Rate (SER)0.05 FIT/Mb
零軟錯誤免疫:無電荷儲存節點,次閾值下抗 Alpha 粒子與中子翻轉率達 99.9% 以上,極致適用於長期部署之物聯網感測節點。Soft Error Immune: Zero floating storage node; >99.9% alpha/neutron immunity under subthreshold, ideal for edge IoT sensors.
LENS 03 · PROCESS INTEGRATION CHECKS觀點 03 · 製程整合條件
eFlash 製程整合:四項需由具名實作回答的問題eFlash Process Integration: Four Questions for a Named Implementation
節點名稱本身不能判定電荷陷阱或浮閘 eFlash 是否可用。以下是供應商與 PDK 查核問題;缺少實作文件時,溫度、電壓、光罩、成本與面積均保持未知。A node label alone does not establish charge-trap or floating-gate eFlash availability. These are supplier and PDK review questions; temperatures, voltages, masks, costs and areas remain unknown without implementation-specific documents.
01A · THERMAL CONFLICT01A · 熱預算衝突
HKMG 熱預算衝突HKMG Thermal Budget
核對記憶體模組的沉積與退火順序、允許溫度及邏輯元件相容性;不能由家族名稱推定固定熱預算。Check the memory-module deposition/anneal sequence, allowed temperatures and logic-device compatibility; the family name does not define a fixed thermal budget.
01B · VOLTAGE CLIFF01B · 電壓限制
寫入與抹除供電條件Program and Erase Supplies
分別取得讀取、寫入與抹除的供電範圍、元件額定值及可靠度條件;不可用讀取 VDD 代替寫入電壓。Obtain separate read/program/erase supply ranges, device ratings and reliability conditions; read VDD does not specify the program voltage.
02 · MASK ADDERS02 · 額外光罩
光罩與成本逐案核對Check Masks and Cost per Process
確認具名製程的額外模組、光罩與報價;光罩道數本身不能推導固定成本百分比或流片時程。Confirm added modules, masks and quotations for the named process; mask counts alone do not establish a fixed cost percentage or schedule.
03 · CHARGE PUMP03 · 電荷幫浦
電荷幫浦與周邊面積Charge Pump and Peripheral Area
由目標容量、寫入負載、供電與版圖核對幫浦及周邊面積;不得用通用比例代替巨集資料。Check pump and peripheral area against capacity, programming load, supplies and layout; a generic percentage cannot replace macro data.
第一性原理物理推導:First-Principles Physics: 深入查閱超薄氧化層直接穿隧漏電模型與 4 階段介電質崩潰物理,請參見 記憶體物理基底導覽 ↗。For direct tunneling leakage equations and 4-stage dielectric breakdown microphysics, see Memory Physics Foundations ↗.
物聯網 eFlash 架構演進 · 低功耗與供電突破IOT eFLASH ARCHITECTURAL ADVANCES · LOW-POWER & SUPPLY VOLTAGE LEAPS台灣專業 NVM IP 創新體系Taiwan Specialized NVM IP Ecosystem
在物聯網與超低功耗微控制器設計中,傳統 eFlash 長期面臨兩大工程瓶頸:一是傳統 NMOS 熱電子寫入電流大(mA 級)引發的電荷泵面積與熱應力;二是讀取電壓門檻多高於 1.8V,無法與 0.9V~1.2V 超低壓邏輯核心共軌。台灣 NVM IP 產業發展出兩條獨特的架構革新路徑:In IoT and ultra-low-power MCU designs, conventional eFlash faces two historic bottlenecks: high NMOS hot-electron programming currents (mA-class) burdening charge pumps, and a rigid ≥1.8V read voltage threshold conflicting with 0.9V–1.2V ultra-low-voltage logic cores. Taiwan's specialized NVM IP ecosystem addresses these via two distinct architectural paradigms:
下列節點名稱是製程產品組合的查核入口,不是本站驗證過的 OTP 可用性清單。每一節點都須取得具名 IP 版本、供電、容量與資格驗證文件;一代可用不能推定下一代可用。These node names are process-portfolio review entries, not a Hub-verified OTP availability list. Each node needs a named IP revision, supplies, capacity and qualification evidence; availability in one generation does not establish the next.
NODE 01 · 22ULL節點 01 · 22ULL
TSMC 22ULL (Planar HKMG)
超低漏電微縮平面:延伸 28HPC+ 成熟架構,邏輯核心電壓降至 0.6V~0.8V,靜態待機漏電削減達 90%。支援 BEOL 後段整合之 STT-eMRAM(4~6 道光罩,不破壞前段高熱預算)與 0 額外光罩 AntiFuse OTP(用於安全啟動密鑰、類比 Trimming 與 PUF)。 待核對:可用記憶體選項、供電與容量;與其他節點面積比較需在同一設計與版圖條件下進行。Ultra-Low Leakage Planar: Extends 28HPC+ with logic operating down to 0.6V–0.8V, slashing standby leakage by 90%. Supports BEOL STT-eMRAM (4–6 mask adders without disturbing FEOL thermal budget) and 0-mask AntiFuse OTP (for secure boot keys, analog trim, and PUF). Verify memory options, supplies, and capacity; cross-node area comparisons require identical layout constraints.
NODE 02 · 12FFC+節點 02 · 12FFC+
TSMC 12FFC+ (FinFET ULP)
立體鰭式微縮突破:跨越 28nm 平面微縮斷崖,3D Fin 幾何提供強大閘極靜電控制與近理想次閾值擺幅 (~68 mV/dec),原生支援 0.5V 近閾值邏輯。傳統 eFlash 完全終止,全面轉向純邏輯 0-Mask AntiFuse 與 16/12nm 車規級 eMRAM。 待核對:邏輯與記憶體各自的最低電壓、讀取路徑、待機模式及漏電量測條件。FinFET Scaling Leap: Overcomes the 28nm planar scaling cliff; 3D fin geometry delivers tight electrostatic gate control with near-ideal subthreshold swing (~68 mV/dec), natively enabling 0.5V near-threshold operation. Replaces discontinued eFlash with 0-mask logic AntiFuse and automotive-grade 16/12nm eMRAM. Verify separate logic and memory voltage limits, read paths, standby modes, and leakage test criteria.
NODE 03 · N6e節點 03 · N6e
TSMC N6e (IoT EUV)
EUV 邊緣智慧物聯網:專為超低功耗常時監控 (AON) 與邊緣 AI 打造,結合 EUV 微影簡化雙重曝光光罩道數,邏輯密度較 12FFC+ 提升逾 100%。採用 0-Mask AntiFuse 搭配硬體 CAM 向量熱修補,替代高昂大面積嵌入式 Flash。 待核對:特定製程版本的 IP 可用性、可配置容量、修補空間與公開量測資料。EUV Edge AI & IoT: Tailored for ultra-low power always-on (AON) sensing and edge inference, leveraging EUV to eliminate multi-patterning masks while doubling logic density over 12FFC+. Pairs 0-mask AntiFuse with hardware CAM vector hot patching to eliminate costly embedded Flash. Verify IP availability, configurable capacity, patch reserves, and silicon-proven measurement reports.
NODE 04 · N4e節點 04 · N4e
TSMC N4e (ULP FinFET)
極致能效次半伏特節點:延伸 4nm 家族之超低功耗衍生節點,提供次 0.5V 供電能力與頂級 SRAM/Logic 能效比。在先進奈米下完全由 0-Mask AntiFuse 充當硬體信任根 (RoT/PUF),代碼儲存則外掛極低接腳數 xSPI PSRAM 或 Chiplet 封裝。 待核對:供應商產品組合、記憶體整合方式與設計支援;節點命名不證明已具備目標產品的 eNVM。Sub-0.5V Ultra-Efficiency Node: Extends the 4nm family with sub-0.5V supply capabilities and peak SRAM/logic efficiency. Pure-logic 0-mask AntiFuse anchors hardware Root of Trust (RoT/PUF), while application code offloads to low-pin-count xSPI PSRAM or Chiplet SiP. Verify supplier portfolios, memory integration interfaces, and design enablement; node naming alone does not establish eNVM availability.
硬體 CAM 比對 CPU 取指位址;命中後,MUX 可將存取重新導向修補向量。實際延遲依記憶體、匯流排、管線與快取配置而定,不能由 OTP 的不可抹除特性推論零等待。A hardware CAM matches instruction-fetch addresses, allowing a MUX to redirect matching accesses to patch vectors. Actual latency depends on memory, bus, pipeline and cache configuration; OTP immutability does not establish zero-wait operation.
向量補丁 CAM 查找與微控制器 ROM 熱修復延遲能耗試算器Vector Patch CAM ROM Replacement & Hot-Patching Latency / Energy Workbench
Arm FPB/自訂 RISC-V 修補模型Arm FPB / Custom RISC-V Patch Model
⚠️ 示意模擬模型與技術邊界說明:⚠️ Schematic Model & Technical Boundary Notice:本工作台參考具名 Arm FPB 功能,並使用本頁自訂 CAM 教學假設,試算查找延遲、修補命中率與能耗;這不是 RISC-V Zicbom 規範所定義的修補模型,也不是商用 MCU 的邏輯閘級時序報告。實際延遲仍受匯流排仲裁、記憶體與管線清空影響。This workbench references named Arm FPB features and uses custom CAM teaching assumptions to estimate lookup latency, patch hit rates and energy. It is not a patch model defined by RISC-V Zicbom or a commercial MCU gate-level timing report. Bus arbitration, memory and pipeline flushes still affect actual latency.
CPU 時脈頻率:CLOCK FREQ:48 MHz
核心供電電壓 VDD:CORE VDD:1.2 V
CAM 平行比對動態功率CAM SNOOP POWER
—
重定向攔截延遲INTERCEPT LATENCY
—
修補指令動態能耗PATCH INSTR ENERGY
—
晶圓成本節省 (vs eFlash)WAFER COST SAVING
—
剩餘安全修補年限裕度SECURITY LIFESPAN
—
硬體嗅探與重定向分析儀Hardware Snoop & Redirect Analyzer
Loading Vector Patch CAM calculations...
微架構解析 · Arm FPB 與 RISC-V 自訂修補模型Microarchitecture: Arm FPB and a Custom RISC-V Patch Model
證據與假設:Arm Cortex-M3 的完整 FPB 實作可將指令或資料存取重新映射;下方 RISC-V CAM 則是本頁的教學架構提案。平行位址比對可能縮短修補路徑,但零等待仍須以具名晶片的時序、匯流排與快取行為證明。Evidence and assumptions: The full Cortex-M3 FPB can remap instruction or data accesses. The RISC-V CAM below is a teaching proposal. Parallel address matching may reduce patch latency; zero-wait operation requires timing, bus and cache evidence for a named chip.
階段 1 · 位址比對Stage 1 · Address Matching
以取指位址比對修補槽位。槽位數與比對延遲屬實作參數;本頁不假設所有核心都能在半個週期內完成。Compare the fetch address with patch entries. Entry count and match delay are implementation parameters; a half-cycle match is not assumed for every core.
階段 2 · 替換與取指同步Stage 2 · Remapping and Fetch Synchronization
命中後選取替換位址或指令。等待週期、仲裁、管線清空及快取一致性需納入延遲;RISC-V 的 FENCE.I 處理同一 hart 的指令可見性,多核心還需跨 hart 同步。A hit selects a replacement address or instruction. Include wait states, arbitration, pipeline flushes and cache coherence. RISC-V FENCE.I synchronizes instruction visibility on the local hart; multicore systems also need inter-hart synchronization.
階段 3 · 驗證與防回滾Stage 3 · Authentication and Rollback Control
教學方案可用 OTP 保存版本下限或驗證信任錨點,並在啟用修補前驗證簽章與版本。不可逆儲存本身不保證更新流程原子性,也不取代故障注入與金鑰保護評估。The proposal may store a minimum version or verification anchor in OTP and authenticate the patch before activation. Irreversible storage alone does not guarantee atomic updates or replace fault-injection and key-protection evaluation.
窄螢幕可橫向捲動比較表;鍵盤聚焦後可用左右方向鍵閱讀。On narrow screens, scroll the table horizontally; focus it and use the arrow keys with a keyboard.
具名硬體功能與教學自訂介面的界線Named Hardware Features and Proposed Custom Interfaces
項目Item
Cortex-M3 FPB
RISC-V 自訂 CAM 提案Proposed RISC-V Custom CAM
規範依據Specification
Cortex-M3 TRM 第 7.3、8.3 節;功能隨完整/精簡配置而異。Cortex-M3 TRM Sections 7.3 and 8.3; full and reduced configurations differ.
自訂修補單元;Zicbom 定義快取區塊管理,未定義此 CAM 或修補 CSR。Custom patch unit; Zicbom defines cache-block management, not this CAM or patch CSRs.
控制暫存器Control Registers
FP_CTRL, FP_REMAP, FP_COMPn 完整配置為六組指令與兩組 literal 比較器。The full configuration has six instruction and two literal comparators.
mpatch_ctrl, mpatch_match[i], mpatch_target[i] 本頁假設名稱;不是 RISC-V 標準暫存器。Illustrative names, not standard RISC-V registers.
重新映射Remapping
完整配置可將 code space 存取映射至 system space;比較器也可用於 BKPT。The full configuration can remap code-space accesses to system space; comparators can also generate BKPT.
替換與跳躍由自訂單元決定;CBO.INVAL 的快取無效化不等同指令替換。Replacement and branching are custom-unit behavior; CBO.INVAL cache invalidation is not instruction replacement.
NVM 結合NVM Integration
由系統設計決定修補映像儲存與載入;FPB 不規定 AntiFuse。Patch storage and loading are system choices; FPB does not prescribe AntiFuse.
可評估 OTP 信任錨點與可更新修補儲存區;容量、壽命與更新政策分開驗證。OTP anchors and updateable patch storage are options; validate capacity, lifetime and update policy separately.
四方非揮發記憶體,在物聯網微控制器上的取捨全景Four eNVM Technologies: Trade-Off Landscape for Edge MCUs
評估邊緣 MCU 時,將光罩、讀取供電、啟動、待機與保持條件列成具名實作的待查資料。缺少來源時保留未知,不用家族平均值代替。For an edge MCU, request masks, read supplies, boot behavior, standby and retention conditions for the named implementation. Missing evidence stays unknown rather than being replaced by family averages.
比較邊界:下表是查核清單,不是產品規格或排名。每格需具名 MCU/巨集、容量、PVT 與工作負載;可追溯的實際數值另見兩個具名 MCU 的三項指標。Comparison boundary: This is a review checklist, not product specifications or a ranking. Each cell needs a named MCU / macro, capacity, PVT and workload; traceable values appear in three metrics for two named MCUs.
邊緣 IoT MCU 儲存分層與神經形態研究:特定產品範例(須逐 SKU 核對)Edge IoT Storage Tiering & Neuromorphic Research: Named Product Examples (Verify per SKU)
Weebit Nano 類比神經形態運算 (CIM)Weebit Nano Analog Neuromorphic CIM
V14 的供應商公告支持 SkyWater S130 具名 ReRAM 巨集資格揭露,並不證明類比 CIM 已在該平台量產,或達成特定關鍵詞喚醒功耗。CIM 仍需另附具名電路、容量、精度、工作負載與量測來源。The supplier announcement in V14 supports qualification disclosure for a named ReRAM macro in SkyWater S130. It does not establish production analog CIM on that platform or a particular keyword-spotting power result. CIM needs separate named-circuit, capacity, precision, workload and measurement evidence.
車規 28nm eNVM 破局:Infineon eRRAM & ST ePCMAutomotive 28nm eNVM: Infineon eRRAM & ST ePCM
Infineon 與 ST 的公開資料提供 AURIX RRAM 與 Stellar ePCM 整合案例。需依確切型號核對軟體介面、更新方式、溫度與資格驗證;不能把記憶體機制直接當成位元覆寫、無停機更新或認證保證。下列總帳保留供應商來源與限制。Infineon and ST disclose AURIX RRAM and Stellar ePCM integration examples. Check software interfaces, update behavior, temperatures and qualification for the exact part; the storage mechanism alone does not establish bit-level overwrite, zero-downtime updates or certification. The ledger below retains supplier sources and limitations.
LENS 07 · STATE CONTRACT & VERTICALS觀點 07 · 狀態契約與垂直應用
MCU 狀態契約三層架構:不可變晶片上的彈性邊界Three-Tier MCU State Contract: Agile Boundaries on Immutable Silicon
透過分層契約,MCU 設計者兼得純邏輯 AntiFuse 的 0 光罩加價與近臨界電壓極致低功耗,同時打破「OTP 無法更新程式碼」的傳統刻板印象。Through tiered state contracts, MCU architects gain 0-mask CMOS cost advantages and 0.5V efficiency while disproving the myth that OTP cannot support code updates.
01 · SMART METERS01 · 智慧電表
智慧水電氣表Smart Meters
<10nA 休眠漏電結合瞬態開機,確保表計 15~20 年免換電池。<10nA sleep leakage enables 15~20 yr continuous life on non-rechargeable cells.
02 · WEARABLES02 · 穿戴裝置
醫療穿戴裝置Wearables
0.5V 近臨界電壓微能耗,微型能量採集即可驅動感測與 BLE 廣播。0.5V near-threshold operation allows micro-energy harvesting to power telemetry.
03 · INDUSTRIAL03 · 工業應用
工業物流追蹤Asset Trackers
125°C 長效留存搭配 Vector CAM,支援偏遠地區低頻韌體熱修補。125°C retention with Vector CAM for remote low-bandwidth security patching.
04 · BODY MCU04 · 車身 MCU
車身區域控制器Body Zone MCUs
門禁與車燈模組在 28nm/22nm 替代額外光罩 eFlash,可對照 AEC-Q100 Grade 1至 Grade 0 任務載記 車規環境;配合硬體 SECDED ECC 與安全降級架構,可映射 ASIL-B至 ASIL-D 任務載記詞彙 (SEooC) 功能安全需求。Architecture-class option to replace costly mask-adder eFlash in door and lighting modules across 28nm/22nm under AEC-Q100 Grade 1 to Grade 0 mission-profile vocabulary (target-dependent); hardware SECDED ECC can support ASIL-B through ASIL-D (SEooC) systematic-capability language (not a certification claim).
第一性原理微能耗建模 · 斷續計算架構FIRST-PRINCIPLES MICRO-ENERGY MODEL · INTERMITTENT COMPUTING
常時關閉 (Normally-Off) 能量採集與非揮發狀態保留電源自給試算器Normally-Off Energy Harvesting & Non-Volatile State Retention Simulator
微瓦級無電池自給運作ZERO-BATTERY AUTONOMOUS
在微瓦級環境能量採集(弱光、壓電、微弱 RF、人體溫差)應用中,傳統 eFlash 高昂的冷啟動電壓(≥1.6V)與電荷泵暖機突波電流(>5mA)常導致儲能電容塌陷與死鎖循環。本工作台計算微型儲能電容充放電、冷啟動邊界、Normally-Off 狀態保存/恢復能耗,以及完全斷電零漏電下的自給自足平衡點。Under micro-watt ambient energy harvesting (solar, piezo, RF, TEG), conventional eFlash cold-boot thresholds (≥1.6V) and charge-pump inrush surges (>5mA) collapse tiny reservoir capacitors into reboot brownout loops. This workbench calculates capacitor charge cycles, cold-boot safety, Normally-Off state save/restore overhead, and zero-leakage autonomous energy balance.
⚠️ 示意模擬模型與技術邊界說明:⚠️ Schematic Model & Technical Boundary Notice:本工作台之常閉運算能量採集、休眠漏電與冷啟動延遲為基於一階電容儲能積分與簡化狀態機之數值示意,非微型能量採集晶片在非理想環境光照或熱電波動下之實測數據。實際斷電保留可靠度需搭配電源管理單元 (PMU) 掉電偵測 (BOD) 電壓閾值驗證。The normally-off energy harvesting, sleep leakage, and cold-start latency calculations in this workbench are schematic numerical models based on first-order capacitive integration and state-machine abstractions, not bench measurements under fluctuating ambient energy sources. Real-world retention requires validation against PMU brown-out detector (BOD) thresholds.
儲能電容容量:RESERVOIR CAPACITOR:47 µF
運算工作占空比:ACTIVE DUTY CYCLE:0.50%
採集輸入功率HARVESTED POWER
—
環境能量產生率Ambient intake
可用儲能電量USABLE STORED ENERGY
—
電容有效儲能Cap buffer energy
冷啟動回充時長COLD BOOT CHARGE
—
開機最低充能時間Initial charge time
系統平均總功耗AVERAGE POWER
—
含待機與狀態保存Active + sleep + state
能量平衡係數ENERGY RATIO
—
採集功率 / 平均功耗Generation / load
電源自給狀態OPERATION STATUS
—
自律運行判定Autonomy verdict
Intermittent Energy Harvester Monitor
Loading Normally-Off Energy Harvesting calculations...
PRIMARY SOURCES · FOUNDRY & INDUSTRY EVIDENCE原始來源 · 晶圓代工與產業證據
技術證據原則:Evidence Principles:下列連結為各段主張的特定一次來源;未列出的數字屬教學假設或需另附產品/量測邊界。Links below are named primary sources per section; unlisted figures are teaching assumptions or require separate product/measurement boundaries.