天才少年三年自制兩代芯片:晶體管密度設計與英特爾第一款處理器相同

{"type":"doc","content":[{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Sam Zeloof 已經不能用簡單的“天才少年”來形容了。2018 年,他在車庫裏製造了自己的第一塊光刻集成電路。在製造這款名爲 Z1 的芯片時,他還是個高三的學生;如今大四的他,在掌握了更多芯片知識與工藝技能之後,做出了更進一步的嘗試。"}]}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"Sam Zeloof 的第二代芯片"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"2018 年,高三的 Sam Zeloof 自己研發的第一代芯片 Z1 只包含 6 個晶體管,屬於一塊純粹的製程與設備測試芯片。這一次的 Z2 則計劃在 10 微米多晶硅柵極製程上容納 100 個晶體管,這樣的晶體管密度設計與英特爾推出的第一款處理器完全相同。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/ea\/ea671288f56fff83d7408e6df6b4a9a6.gif","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這塊芯片包含一個簡單的 10 x 10 晶體管陣列,用於測試、表徵及調整製程,也作爲邁向更先進 DIY 計算機芯片的重要一步。當初的英特爾 4004 包含 2200 個晶體管,而 Sam Zeloof 現在已經能夠在同一塊芯片上佈設 1200 個晶體管。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/fb\/fb0ff14e4f364b417ff421aa9863f37e.jpeg","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/7d\/7d710d106dc4d091c54c9e04d7c1e876.jpeg","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"與行業發展速度相比,Sam Zeloof 的芯片晶體管增長速度明顯更勝一籌。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"之前,Sam Zeloof 使用金屬柵極工藝製作芯片——鋁柵極會與下方的硅溝槽之間出現很大的功函數差,由此實現高閾值電壓(>10 伏)。在一些有趣的項目中使用了這些金屬柵極晶體管,例如吉他的失真跳板與環形振盪器 LED 閃光燈。但由於電壓閾值較高,所以這兩個項目都需要一到兩節 9 伏電池才能讓電路運行起來。而轉爲多晶硅柵極工藝之後,Sam Zeloof 的新一代芯片性能猛增(這種自對準柵極能大大減少電容重疊),包括更低的電壓閾值。如此一來,這些芯片就能與 2.5 伏與 3.3 伏邏輯電平相兼容。總之,這款新 FETs 擁有以下優異特性:"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"NMOS電氣特性:"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Vth             = 1.1 V"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Vgs MAX         = 8 V"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Cgs             = <0.9 pF"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Rise\/fall time  = <10 ns"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"On\/off ratio    = 4.3e6"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Leakage current = 932 pA (Vds=2.5V)"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"超低的泄漏電流令人頗爲驚訝。此值在室內正常照明環境下會上升約 100 倍。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/2f\/2ffd5ced0e2364da5653885de3a8d80f.jpeg","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":"center","origin":null},"content":[{"type":"text","text":"NMOS, 0.5V Vgs 步長"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/9f\/9f305e19ce17d5bd3c45c9e3ed859f9c.jpeg","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":"center","origin":null},"content":[{"type":"text","text":"二極管曲線"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/11\/1116b7fd895eb9ebddc566e638d83095.jpeg","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":"center","origin":null},"content":[{"type":"text","text":"C-V 指示 Vth = 1.1V"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"現在,我們已經證明可以使用普通的化學品、非無塵室以及自制設備製造出良好的晶體管了。當然,整個產量與工藝可重複性仍然不高。Sam Zeloof 會進行更多測試以收集關於 FET 屬性的統計數據與波動性信息,但就目前來看一切還算進展順利。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/a9\/a99aa56aaccc872cdd3a71b954fab909.jpeg","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":"center","origin":null},"content":[{"type":"text","text":"1MHz,50Ω負載"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/a5\/a5ce395bf7bfe3747a0c3e76c11a2b33.jpeg","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":"center","origin":null},"content":[{"type":"text","text":"20MHz,50Ω負載"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/38\/38deae28d1d327eb7ba95d6d647a7c62.jpeg","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這塊芯片很小,大約是上一塊芯片面積(2.4 平方毫米)的四分之一,所以檢測工作很難進行。而每塊芯片上都有一個簡單的 10 x 10 N 溝槽 FET 陣列,可以爲 Sam Zeloof 提供大量特性數據。因爲設計非常簡單,所以 Sam Zeloof 直接用 Photoshop 就完成了佈局設置。每列 10 個晶體管共享一個公共柵極連接,每行與相信的晶體管串聯在一起,共享同一源極 \/ 漏極端子。整個結構類似於 NAND 冷豔,但我這麼設計只是爲了讓金屬焊盤能足夠大,降低探測難度。換句話說,如果每個 FET 都對應 3 個焊盤,那尺寸就太小了。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"下圖所示,爲一個 10 微米 NMOS 晶體管,金屬層上有輕微的未對準痕跡(左側有一部分觸點未能覆蓋)。紅線部分爲多晶硅,藍色則是源 \/ 漏柵極。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/f0\/f0fb7709bb297d9ffdb2f3282f2b8482.jpeg","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":"center","origin":null},"content":[{"type":"text","text":"單 NMOS 晶體管"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/e7\/e7ada5a28a1a32c00efdc7d9d4536076.jpeg","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":"center","origin":null},"content":[{"type":"text","text":"單 NMOS 晶體管"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"到這裏,Sam Zeloof 已經制作出一款運算放大器(Z1)與一個類似於內存的陣列(Z2)。儘管晶體管密度極低,但整個項目仍然非常有趣。具體制程還需要一點調整,但現在已經能夠以穩定的方式製造高質量晶體管,接下來也應該可以設計出更復雜的數字與模擬電路。每個芯片的測試工作非常乏味,所以 Sam Zeloof 還打算自動化完成測試流程,並在成功後發佈更多數據。Sam Zeloof 已經制作了 15 個芯片(共 1500 個晶體管),而且其中至少有一個芯片實現了“功能完備”,至少兩個芯片能實現“基本功能”——即至少約 80% 的晶體管能夠正常工作。但目前 Z2 的產量數據還不樂觀,最常見的缺陷出現在源或漏柵極同主體硅溝槽之間的短路上。順帶一提,Z1 的主要缺陷集中在柵極漏電或短路上。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"有關 Z2 的“先柵極” 工藝可以閱讀 Sam Zeloof 的博客:"},{"type":"link","attrs":{"href":"http:\/\/sam.zeloof.xyz\/","title":"","type":null},"content":[{"type":"text","text":"http:\/\/sam.zeloof.xyz\/"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在 Sam Zeloof 的博客中寫道:整個製程工藝還有不少改進的空間,特別是應該與 CMOS 相兼容;但目前的方案至少簡化了製造流程,也把必要工具控制在最低限度。在理想情況下,1 微米厚的介電層(橙色)應該使用 CVD SiO2(可以在家中加工的 TEOS 氧化物),但我實際使用的是光刻膠。大多數光刻膠能夠在 250°C 左右的烘烤下形成堅硬的永久介電層,可以直接替代 CVD 或 PECVD 氧化物。當然,這裏也可以使用旋製氧化硅 \/ 溶膠凝膠。SiO2 蝕刻則是在由除鏽劑或 RIE 製成的緩衝 HF 溶液中完成的。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"下面就是拼合之後的晶片整體圖了:"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/09\/0940b0d7ba4554729ffe3cf194663d93.jpeg","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":"center","origin":null},"content":[{"type":"text","text":"Sam Zeloof 的第一代自制芯片"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Sam Zeloof 經常在博客和 Twitter 上定期發佈他的進展,他在 2018 Superconference 會議上向所有人講授半導體物理課程,以及他如何成爲第一位使用光刻工藝生產集成電路的業餘愛好者。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"2018 年,Sam Zeloof 第一代自制芯片 Z1 有 66 個單獨的製造步驟,完整運行大約需要 12 小時。與此同時,他還製作了涵蓋半導體制造理論 和分立 MOSFET 製造的 Youtube 視頻。儘管當時他沒有用於將其安裝到引線框架上的焊線機,但該芯片已經過測試並且可以正常工作。在當時的博客中,Sam Zeloof 詳細介紹了 Z1 芯片的製造過程。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Sam Zeloof 用了三個月的時間複製 Ellsworth 的晶體管,目標是在 Jeri Ellsworth 自制硅晶體管的基礎上製造實際的集成電路。”到目前爲止,他只用了一些組件來製造簡單的集成電路,但他的目標是建立一個烏克蘭微處理器的克隆,他嘗試在車庫裏製作出英特爾在 1971 年發佈的著名處理器 Intel 4004。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/32\/329ffc1d9c851bc3aea742f6625c0a34.jpeg","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":"center","origin":null},"content":[{"type":"text","text":"當時的工作臺"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"爲了在芯片上設計電路,他使用了一種在 20 世紀 70 年代所沒有的技巧:通過增加小型光學平臺來修改數字視頻投影機。然後,可以創建一個數字圖像的掩模,並將其投影到晶圓上,以曝光光刻膠。利用他目前的設置,可以創建分辨率約爲 1μm 的摻雜特徵,而無需創建物理掩模的時間和費用 (但是,如果沒有潔淨室設置來防止污染,他認爲 10μm 是獲得合理的限制工作裝置產量)。整個設計使用簡單的邏輯元器件——74 系列元器件從零開始製作,全程總共用了 28 塊 74 系列芯片。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"2019 年 7 月 8 日,Sam Zeloof 更新博客:FET 柵極長度(特徵尺寸)減小到 <5μm,使該項目可以媲美 1975 年左右最先進的項目,並允許晶體管以更好的特性運行。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/ed\/ed63dd575f7c84ecf81aebc88f774321.jpeg","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Sam Zeloof 近幾年的所有重要工作包括整個芯片製造過程的想法都更新在了自己的博客和 Twitter 上,歡迎感興趣的開發者圍觀。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"博客地址:"},{"type":"link","attrs":{"href":"http:\/\/sam.zeloof.xyz\/%EF%BC%88%E6%9C%AC%E6%96%87%E9%85%8D%E5%9B%BE%E6%9D%A5%E6%BA%90%E4%BA%8E%E5%8D%9A%E5%AE%A2%EF%BC%89","title":"","type":null},"content":[{"type":"text","text":"http:\/\/sam.zeloof.xyz\/(本文配圖來源於博客)"}]}]}]}
發表評論
所有評論
還沒有人評論,想成為第一個評論的人麼? 請在上方評論欄輸入並且點擊發布.
相關文章