Stanley Phillips "Stan" Frankel (1919 – May, 1978) was an American computer scientist. He was born in Los Angeles, received his PhD in physics from the University of Rochester, and began his career as a post-doc student under J. Robert Oppenheimer at University of California, Berkeley in 1942. Frankel helped develop computational techniques used in the nuclear research taking place at the time. He joined the theoretical division of the Manhattan Project at Los Alamos in 1943. While there he organized teams of persons (known as "computers") using electromechanical calculators to divide the massive calculations required for the project into manageable assembly line groups. Even that proved too slow, and Frankel turned to IBM tabulating machines to help process the numbers. This research led to his interest in the then-dawning field of digital computers. In August, 1945 Frankel and Nick Metropolis traveled to the Moore School of Engineering in Pennsylvania to learn how to program the ENIAC computer. That fall they helped design a calculation that would determine the likelihood of being able to develop a fusion weapon. Edward Teller used the ENIAC results to prepare a report in the spring of 1946 that answered this question in the affirmative.
Stan FrankelAfter losing his security clearance (and thus his job) during the red scare of the early 1950s, Frankel became an independent computer consultant. He was responsible for designing the CONAC computer for the Continental Oil Company during 1954–1957 and the LGP-30 single-user desk computer in 1956, which was licensed from a computer he designed at Caltech called MINAC. The LGP-30 was moderately successful, selling over 500 units. He served as a consultant to Packard Bell Computer on the design of the PB-250. His last computing project was the SCM Marchant Cogito 240SR electronic calculator introduced in 1965.
Frankel published a number of scientific papers throughout his career. Some of them explored the use of statistical sampling techniques and machine driven solutions. In a 1947 paper in Physical Review, he and Metropolis predicted the utility of computers in replacing manual integration with iterative summation as a problem solving technique. As head of a new Caltech digital computing group he worked with PhD candidate Bernie Alder in 1949–1950 to develop what is now known as called Monte Carlo analysis. They used techniques that Enrico Fermi had pioneered in the 1930s. Due to a lack of local computing resources, Frankel travelled to England in 1950 to run Alder's project on the Manchester Mark 1 computer. Unfortunately, Alder's thesis advisor was unimpressed, so Alder and Frankel delayed publication of their results until 1955, in the Journal of Chemical Physics. This left the major credit for the technique to a parallel project by a team including Teller and Metropolis who published similar work in the same journal in 1953.
In September, 1959, Frankel published a paper in IRE Transactions on Electronic Computers proposing a microwave computer that used travelling-wave tubes as digital storage devices, similar to, but faster than the acoustic delay lines used in the early 1950s. Frankel published a paper on measuring the thickness of soap films in the Journal of Applied Physics in 1966
出版作品一览表编辑本段回目录
Frankel, S Phillips, “Elementary Derivation of Thermal Diffusion,” Physical Review, Volume 57, Number 7, April 1, 1940, p 661.
Frankel, S and N Metropolis, “Calculations in the Liquid-Drop Model of Fission,” Physical Review, Volume 72, Number 10, November 15, 1947, p 914-925.
Frankel, Stanley P, “Convergence Rates of Iterative Treatments of Partial Differential Equations,” Mathematical Tables and Other Aids to Computation, Volume 4, 1950, p 65-75.
Frankel, S P, “The Logical Design of a Simple General Purpose Computer,” IRE Transactions on Electronic Computers, March 1957, p 5-14.
Frankel, S P, “On the Minimum Logical Complexity Required for a General Purpose Computer,” IRE Transactions on Electronic Computers, December 1958, p 282-284.
Frankel, Stanley P, “A Logic Design for a Microwave Computer,” IRE Transactions on Electronic Computers, September 1959, p 271-276.
Frankel, Stanley P and Karol J Mysels, “On the ‘Dimpling’ During the Approach of Two Surfaces,” Journal of Physical Chemistry, Volume 66, January 1962, p 190-191.
Frankel, Stanley P and Karol J Mysels, “Simplified Theory of Reflectometric Thickness Measurement of Structured Soap and Related Films,” Journal of Applied Physics, Volume 37, Number 10, September 1966, p 3725-3728.
《别闹了，费曼先生》相关章节编辑本段回目录
不输IBM的人力计算机
我曾经着手研究过另一个问题，当时，我们要处理很多计算，而我们使用的是玛灿特计算机。让我顺便谈谈那时罗沙拉摩斯的景况：玛灿特计算机是手摇式的。你用力摇，它就能加减乘除——当然没有现在的计算机那么方便。
它们全是机械装置，经常发生故障，坏了要送回原厂修理；而隔没多久，所有计算机都在厂里，我们就无机可用了。
于是我们有些人便开始把机盖掀开，自己动手修理。按照规定这是不行的，他们说：“自行掀开机盖者，后果概不负责……”但我们自行把机盖掀开，而且还学会了怎样修理这些计算机，修得愈多，手艺愈精。碰到一些太复杂的状况时，我们才把计算机送回原厂去，一切计算工作才得以继续进行。最后我发现，所有计算机都是我在修；负责机械修理的那位仁兄，只修打字机。
总之，后来我们觉得最大的问题——准确算出原子弹爆发时究竟会出现什么状况，从而知道释出多少能量等等——所需要的计算工作，远超过我们的能力。有个名叫弗兰科（Stanley Frankel）的聪明小伙子想到，也许可以使用IBM计算机来进行这方面的计算。那时IBM制造了用在商业上的计算机，像把数字加起来并列出总和的“加数机”，或者是从你插入的资料卡片上，读出其中两个数字来相乘的“乘数机”；此外还有“校勘机”和“分类机”等。
弗兰科想出一套很好的方案：我们可以在同一房间内放很多这类机器，然后让卡片逐一通过这些机器。今天，任何需要做数字计算的人，都会明白我在说什么，但在当时这还是很新的想法，还没几个人想到用机器做大量计算。
之前我们试过利用加数机做过类似的计算，例如放一堆加数机在那里，加完一些数字后传到另一个加数机那里，进行下一步的计算，所有事情都自己来。
但新方案是首先你走到加数机那里，再走到乘数机，再用加数机……我们都觉得这是个很好的方法，于是弗兰科设计好整套程序，跟IBM订了机器。
这些机器经常需要维修，军方也会派专人来修理机器。
但他们总是姗姗来迟，而我们永远是急急忙忙，每件事都十万火急，这次也不例外。我们已经设计好所有计算程序，乘这数，然后这样，再减那个数等，也弄清楚需要哪些工具，但我们没有任何机器来测试这些想法。终于，我们找了一些女孩子来帮忙。我们给她们一人一部玛灿特计算机：这个负责乘数，下一个是加数，另一个负责立方——她的工作就是算出卡片上数字的三次方，再交给下一个女孩。
我们把整套程序从头到尾一遍一遍地演练，直到正确无误。结果发现，这种分工计算的方法：要比单独一个人从头算到尾的方式快了不知多少倍！而我们这套作业方式的速度，等于使用IBM机器的速度了，唯一的分别是，IBM机器不会疲倦，一天能连续三班不停工作，可是我们雇来的女孩没多久就全累倒了。
总之，我们用这方法把作业系统内的缺点，全纠正过来；最后机器也送到了，但维修工人还是没有出现。这些机器属于当时的最新科技，结构十分复杂，体积庞大，是拆开分件装箱送来的，还附了很多电线和说明如何安装的蓝图。弗兰科、我以及另外一个家伙，一起把它装起来，其中碰到不少困难，但最大的困难，是那些大人物不停地跑进来说：“你们会把它弄坏！”
小心电脑病
我们继续把机件装置好，它们有时操作良好，有时候却因为什么弄错了，而出问题。后来当我在弄一部乘数机时，注意到里面有一个零件弯了，但我不敢把它弄直，因为害怕把它弄断——而他们一直都在唠叨，说我们早晚会把什么东西搞砸。终于，维修工人出现了，立刻把我们没有弄对的机件一一装妥，一切就都运作良好。除了那部我一直没法弄好的乘数机，三天之后，他还在跟那最后一部机器挣扎奋斗。
我跑去看他，说：“哦，对了，我注意到这里有点弯曲。”
他说。“噢，当然，就是它了！”他用力一扭，机器全好了，就那么简单。
至于弗兰科呢，这个“程序”是他发明的，但这时候他却跟所有后来的电脑使用者一样，患上了电脑病。这是种很严重的病，甚至干扰到正常工作的进行了。电脑的麻烦，在于你会跟它“玩”。它们是那么的有趣——所有的按钮都在你掌握之中，你这样弄得到某个双数，那样弄就是单数。不久之后，只要你够聪明，能计算的东西便愈来愈多。
可是不久之后，我们的系统也崩溃下来了，因为弗兰科无法专心工作，更没用心督导其他人。计算系统运行得很慢很慢，他却坐在房间内，思索如何能让列表机自动算出角度的反正切值。好了，列表机开始动作，画出一行行的线，发出“嗖！嗖！嗖！”的声音，一边画一边计算积分值，然后把所有角度的反正切值列出来，一次完成。
这绝对是没用的事情，因为我们早已有反正切函数表了。但如果你用过计算机，你就会充分了解这种病——发现自己有多能干的喜悦。这是他第一次感染上这种病症，好笑的是，那套系统却刚好是这个可怜虫创造出来的！
终于，他们要我停下手边工作，负起督导IBM小组的责任，我很小心不让自己染上那种病。虽然9个月以来，他们只解决了3个问题，小组成员的素质却很高。真正的问题是，从来没有人告诉他们任何事。军方透过称为“特遣工程师”的计划，从全美各地挑出具有工程才能的高中生，送到罗沙拉摩斯来，安排他们住在营房里，却什么也不告诉他们。
脱胎换骨
这些青年就这样开始上班了，他们的工作呢，却是在这些IBM机器的卡片上打洞，计算一些他们不知所谓何来的数字；因此他们的进度非常慢。当下我建议，这些技术人员必须知道我们究竟在做什么。于是奥本海默跑去跟安全人员商讨，获得特别许可，我便给他们好好上了一堂课。
他们全都兴奋极了：“原来我们在参加作战！我们明白这是怎么一回事了！”现在这些数字对他们都别具意义了。
如果计算出来的压力值较高，那么被释出的能量也相应增加……等。他们充分明白自己在做什么了。
他们简直是脱胎换骨了！大家开始发明新方法把工作做得更好，也改良了整个系统。他们更自动自发，晚上加班，完全不需要任何监督：事实上，现在他们什么也不需要了，因为他们明白一切，后来还发明了几套很有用的程序。
这批小伙子真的变得很了不起。而从头到尾，我要做的只不过是告诉他们，这究竟是怎么一回事。结果，虽然前面他们花了9个月，才完成3个问题；我们后来却在3个月内解决了9个题目，效率几乎提升了10倍之多！
不过，我们有很多秘密武器，其中之一是利用不同颜色的卡片。我们的作业方式，是一大叠卡片需要绕场一周。
先加、再乘，就那样走遍房间内的每一部机器，一圈又一圈地绕，很慢。因此我们想到，将另一组不同颜色的卡片放进计算循环中，但这组卡片跑的比前面一组稍为慢一点。
这样一来，我们可以同时进行两三项计算。
不过这也带来了麻烦。举个例子，战争接近尾声，就在原子弹在阿布奎基正式试爆之前，大家面对的问题是：究竟爆炸时会释放出多少能量？不错，我们计算过各种不同设计所释出的能量，可是从来没有就最后采用的那种设计，计算到底会有多少能量释放出来。克利斯蒂跑来跟我说：“我们要知道这东西会怎样爆发，希望能在一个月内拿到计算结果。”——确切的时限记不得了，也许是3周，总之是很短的时间。
我说：“这是不可能的事。”
他说：“看，现在你们一个月差不多交出一两个问题。
那等于说两三星期便可以解决一个问题啦。“
我回答说：“我知道。不过我们实际花在一个题目上的时间，没有那么短，只不过我们用平行的运算方式而已。
整个操作过程很费时，我们没办法跑得更快了。“
他离去后我开始想，到底有没有办法加快运算速度呢？
假如我们全力处理一个问题，所有机器不受其他干扰，结果会怎样？我在黑板上写“我们做得到吗？”向这些小孩下挑战书。他们开始高喊：“可以，我们多轮一班，我们加班工作！”他们不停地叫：“我们要试！我们接受挑战！”
于是我们约法三章：其他计算一概暂停，我们全力以赴，只处理这个题目。大家立刻开始行动！
面对阿琳的离去
那时候，我太太阿琳正患了肺病，病情实在严重，看起来随时会出什么状况。因此我预先跟宿舍里的一个朋友商量好，有急需时便借用他的车，好能够迅速赶到阿布奎基去看阿琳。那位朋友名叫福斯，后来发现原来他是一名间谍。他就是用他的车子把罗沙拉摩斯的原子弹机密带到圣塔菲（Santa Fe）去，但当时没有人知道这些事。
紧急情况发生了。我开了福斯的车，路上还载了两个搭便车的，以防途中车子出了什么问题，也可有个帮手。
果然，我们才开到圣塔菲时，一个轮胎就破了，他们两人帮我一起把备胎换上。而当我们要离开圣塔菲时，另一个轮胎也破了，我们只好把车子推到附近的加油站。
加油站的人正在修理另一辆车，看来要等很久才会轮到我们。我根本没想到要说些什么，但这两位乘客跑去跟加油站的人说明了我的状况。很快地，他就替我换上新轮胎。但我们再没有备胎了——在战时，车胎是稀有物资，取得不易。
离阿布奎基还有30英里，第3个轮胎也爆了。我干脆把车子停在路边，大家一起拦便车到目的地。我又打电话给修车厂，请他们把车子拖去修理，一方面赶去医院看阿琳。
在我抵达医院数小时后，阿琳去世了。护士进病房来填写死亡证明书，然后离开。我陪着阿琳又过了一会儿，无意中看到我送给她的闹钟。那是7年前的事情了，当时她才刚感染上肺病。在那些日子里，这种数字钟算是很精巧的东西，它利用机械原理，能够显示数字。由于它结构极为精巧，因此很容易故障，隔不多久我便须动手修理一下；但多年来我还是没把它丢掉。这次它又停摆了——停在9点22分上，刚巧是死亡证明书上记下的时间！
记得在麻省理工念书时，有一天在兄弟会宿舍里，无缘无故地心电感应，觉得祖母去世了。紧接着电话铃声突然响起，不过电话不是打给我的，祖母还健在。这件事让我印象深刻，经常惦着也许有一天，别人会告诉我结局相反的故事。我想那也很可能碰巧发生，毕竟那时祖母已经很老了。当然，如果真有那样的事，很多人会认为是种超自然的现象。
阿琳生病期间，一直把那只钟放在床边，它却刚好在她去世的那一刻停顿。我明白，那些对这类事情疑信参半的人，在这种情况之下，不会立刻去研究事情的真相；他们会认定没人碰过那时钟，事情无法解释；而钟确实停了，确实可以算是一件惊人的超自然案例。
不过我注意到房间的灯光很暗，我甚至记得护士曾经拿起钟来，迎着光以看清楚一点，那很容易就把它弄停了。
我到外面走了一会。也许我在骗自己，但我很惊讶，自己竟然没有感觉到一般人在这种情况下应有的感觉。我并不愉快，也没有觉得特别难受，也许那是因为7年来已有心理准备，这件事早晚会发生。
我不晓得如何面对罗沙拉摩斯的朋友。我不想别人愁眉苦脸地跟我谈这件事。回去之后——路上又爆了一个轮胎——他们问我发生了什么事。
“她过世了。工作进行得怎么样？”
他们立刻明白，我不想镇日沉埋在哀伤里。很明显，我对自己做了心理建设：正视现实是那么重要——我必须慢慢厘清发生在阿琳身上的是怎么一回事——以致于一直到好几个月之后才哭出来。那时我在橡树岭，刚巧路过一家百货公司，看到橱窗内的洋装，心想阿琳一定会喜欢其中一件，就再也按捺不住了。
小鬼当家
等我重新投入计算工作时，发现情况一团糟。那里有白色的、黄色及蓝色的卡片。我说：“你们不是应该只做一个题目吗？只能做一个题目！”他们说：“出去，出去。
等一下，让我们说明一切。“
原来事情是这样的。卡片通过机器时，它们有时会出错，又或者数字打错了。从前碰到这种情况时，我们都得重来一遍。可是他们发现，在某一轮的计算中出的错误，只会影响到邻近的数字，但下一轮计算中它会影响到某些数字，以此类推，例如，你一共要处理50张卡片，第38张发生错误，而影响到第37、38及39这3张卡片。到了下一循环，受影响的卡片是第36、37、38、39及40等5张。然后，错误就像瘟疫般蔓延开来。
有一次他们发现前面出了错误，想到一个办法，那就是只重新处理在错误前后的10张卡片。10张卡片通过机器所需的时间，要比50张少多了，因此当那有“病”的50张卡片还在跑的同时，他们让这10张快速通过，然后再把正确的卡片插回去，一切便回复正常了。十分聪明。
他们就用这种方法加快速度。事实上也别无他法了，如果他们碰到错误就停下来补救，进度一定落后。当然，你知道就在他们忙得不可开交时，发生了什么事，他们在蓝色的一叠卡片内发现有错，因此他们加进去一叠黄色的卡片，它们比蓝色的一叠运行快多了。而在紧要关头——弄完这个错误他们还要处理白色的卡片——我这当主管的跑进来了。
“不要来烦我们，”他们说，我再也没去烦他们。一切顺利，我们如期缴出答案。
费米、冯诺曼
刚开始时，我只是个无名小卒，后来我当了小组长，因此见过一些伟大人物。一生之中最令我振奋的经验之一，就是碰到这些光芒四射的物理学家。
当然，其中包括了费米（Enrico Fermi）。有一次他从芝加哥南下；那时我在研究一个题目，也得到了一些结果。可是牵涉到的计算十分复杂困难。通常我是这方面的高手：我总是能预测答案会是什么，又或者解释为什么会得到某些答案。可是这个题目太复杂了，我简直无法解释为什么得到那样的答案。
我们举行了会议，告诉费米我的困难，然后开始描述我得到的结果。他说：“等一下，在你告诉我答案之前，让我先想想。它应该是如此这般（他对了），然后因为这样跟这样，答案便变成这样这样，最明显的解释是……”
他做的就是我最在行的事，但他比我高明十倍。那真是印象深刻的一课！
还有就是伟大的数学家冯诺曼。我们经常在星期天一起散步——通常在附近的峡谷中，同行的还有贝特及巴查，那是很愉快的经验。冯诺曼教会了我一个很有趣的想法：你不需要为身处的世界负任何责任。因此我就形成了强烈的“社会不负责任感”，从此成为一个快活逍遥的人。大家听好了，我的不负责任感全都是由于冯诺曼在我思想上撒下的种子而起的！
不怕大人物
我也跟波耳（Niels Bohr）会过面。那时候，由于受到德国纳粹的威胁，他化名为贝克，跟他一起的是儿子吉姆。贝克，本名是艾殊。波耳（Aage Bohr）。他们从丹麦跑来，都是大大有名的物理学家。对很多大人物而言，老波耳就像上帝一般伟大。
他第一次来时，我们开了一次会。大家都想一睹伟大波耳的风采，因此很多人都来了，我们讨论了原子弹的问题，我坐在后面的某个角落。他开过会后又走了，而我从头到尾都只能在众多脑袋瓜的缝隙间看到一点点波耳的影子而已。
他第二次要来开会的那天早上，我接到一个电话。
“喂，费曼吗？”
“我就是。”
“我是吉姆。贝克。”是他儿子，“我父亲和我想跟你谈谈。”
“跟我谈？我是费曼，我只是个……”
“没错了。8点钟可不可以？”
于是，就在早上8点，大家都还没起床之际，我跑去跟他们会面。我们跑进技术区的一个办公室，他说：“我们在思索怎样可以令原子弹威力更大，我们想到这些这些。”
我说：“不，这行不通，这没有效……哗啦哗啦等等。”
他又说：“那么这跟这呢？”
我说：“听起来好像比较像样，但这里头包含了这个笨主意呢。”
我们反复检讨很多想法，反复争论。伟大的波耳不断点他的烟斗——它却不断熄灭。他讲的话很难听得懂——咕哝咕哝的不容易明白。小波耳讲的就易懂多了。
“好吧，”他最后说，一边又在点烟斗，“我想我们可以把那些大人物请进来了。”波耳父子把其他人叫来，一起讨论。
后来小波耳告诉我究竟发生了什么事。上次他们来访后，老波耳跟他儿子说：“记得坐在后面那小伙子的名字吗？他是唯一不怕我的人，只有他会指出我的荒谬想法。
下次我们要讨论什么时，单找这些只会说‘是，波耳博士’的人是不行的，让我们先找那个小子谈谈。“
在这方面我总是笨笨的。我总是忘记在跟谁说话，而一味担心物理上的问题。如果对方的想法差劲，我就告诉他那很差劲。如果他的想法很好，我就说很好。就那么简单，这就是我的处事方式。我觉得那样很好，很愉快——大前提是你要做得到。我很幸运正是这样的一个人。
炸弹婴儿出世
我们的计算做完之后，接下来就是试爆了。那时候阿琳刚去世不久，我请了个短假在家，有一天收到通知：“某某日，婴儿便要出生……”
我立刻坐飞机回去，抵达营区时，巴士正要离开，于是我直接跟大家到离试爆地点20英里的地方等候。我们有一具无线电，而理论上他们会告诉我们，原子弹将在什么时候爆炸。可是无线电坏了，因此我们根本不知道外面发生什么事。不过就在试爆前数分钟，对讲机又好了，他们说对我们这些离得较远的人来说，大约只剩20秒了；其他人在较近的地方，只有6英里。
我们每人发了一副墨镜，以供观测试爆之用。墨镜？
在20英里之外，再戴上墨镜能看到什么鬼？我在想，一般亮光是不会伤害眼睛的，唯一能伤害到眼睛的大概只有紫外线。我坐在卡车的挡风玻璃后面，觉得这样便能看得清楚又能兼顾安全；因为紫外线是穿不过玻璃的。
时间到了，远处出现的强大闪光亮得我立刻躲下来，在卡车的地板上看到一团紫色的东西。我对自己说：“不对，这只是眼睛内出现的视觉暂留现象。”再度抬起头来，看到一道白光转变成黄光，又再变成橘光，在冲击波的压缩及膨胀作用下，云状物形成又散去。
最后，出现了一个巨大的橘色球，它的中心是那么的亮，以致成了橘色，边缘却有点黑的，慢慢上升翻腾。突然我明白，这是一大团的烟，充满了闪光，火焰的热力则不断往外冒出。
前后大约过了一分钟。
这个从极亮变成黑暗的过程，我全都看见了。我大概是唯一真正看着那鬼东西——后来称为三一角试爆——的人。其他人都戴上墨镜，而在距离6英里处的人根本什么都没看，因为他们都依指示趴在地上。我大概是唯一用肉眼直接看着那次试爆的人。
大约一分半钟以后，突然传来“砰！”的一声巨响，紧接着是打雷般的隆隆声。那声巨响比什么都有说服力。
在整个过程中，从头到尾都没有人讲半句话，大家只默默地观看，可是这些声音使所有人都如释重负——特别是我，因为从远处传来的声音是那么的厚实，证明它已完全成功。
站在我身旁的人问：“那是什么？”我说：“那就是原子弹了。”
这个人名叫劳伦斯（William Laurence），他的目的是要写文章报导整件事情。按照原定的安排，我要带他四周参观，可是许多东西对他来说都太技术性了。后来史迈斯来访，我便改当他的向导。我们曾经跑进一个房间，里面有个瘦瘦长长的支架，上面陈列了一个镀银的小球。把手放在上面，你会感觉到一阵暖意，事实上它具有放射性，是个钚球。我们站在房门口聊天，谈论这个小球的意义。
这是由人类制造出来的一种新元素，之前在地球上从没出现过，顶多在地球刚形成时出现过一下子——而眼前就有完全分离出来、具备辐射等特性的钚。这是我们制造出来的，它可说是个无价之宝。
我们一边谈话时，下意识会做一些动作。当时他无意间轻踢门垫（防止门猛然撞上墙壁的衬垫），我就说：“是呀！这个门垫跟这扇门实在很配。”门垫是个直径10英寸的黄色金属半球——事实上，这是纯金的。
事情是这样的。我们需要了解中子打到不同物质上后，有多少会被反射回来。我们测试过许多材料，像白金、锌、黄铜，也测试过黄金。实验结束后留下了好些碎金块，也不知是谁出的聪明主意，把碎金合成一个大金球，做为钚球陈列室的门垫！
只是白费工夫？
试爆成功以后，罗沙拉摩斯充满了兴奋的气氛，到处都有聚会，大家跑来跑去。我还坐在吉普车后座，一边打鼓。但只有威尔逊独自坐在那里闷闷不乐。
我说：“你干吗这么忧郁？”
他说：“我们造出来的怪物太可怕了。”
我说：“但这都是你开的头，你还把我们拖下水呢。”
你看，对我来说——对我们来说——开始时，我们都有极充分的理由说服自己参与这工作，然后拼命努力完成使命。这是一种快乐、一种刺激，你会停止思考，明白吗？
很单纯地不去想其他事情。在那一刻，只有威尔逊在思考整件事情的冲击。
以后不久，我又回到文明世界，在康奈尔大学教书。
刚开始时我有一种很奇怪的感觉，我不太能够理解为什么会那样，但当时的感受非常强烈。我坐在纽约一家餐馆里，看着窗外的建筑物，就开始想：投在广岛的原子弹炸毁的半径有多大……从餐馆到34街又有多远？那么多的建筑，全都化为灰烬——不停地想。在路上走着时，看到有人在盖桥、筑路，我又想：他们都是神经病，什么都不懂，干嘛还要盖新的东西？一切都是白费工夫而已。
而白费工夫的日子又继续了差不多40年了，对不对？
事实上我的想法错了，盖桥并不是白费工夫的事，我很高兴这些人有此远见，继续往前迈进。
Stanley P. Frankel, Unrecognized Genius编辑本段回目录
Most of this Web site covers the exploits of people who directly contributed to the development of Hewlett-Packard’s desktop calculators and computers. However, a relatively unknown computer pioneer named Stanley Phillips Frankel made exceedingly strong indirect contributions to these developments although he never worked at or consulted for HP. Instead, Frankel worked on the early development of nuclear weapons at the laboratories in Los Alamos, New Mexico. As a post-doc student under J. Robert Oppenheimer in 1942, Frankel along with Eldred Nelson used an electromechanical calculator to run the calculations that indicated that a uranium fission chain reaction would indeed release considerable energy in an exceedingly large explosion and they made the first calculations to determine the critical amount of uranium needed for a fission bomb. Three years later, Frankel ran other sorts of calculations on ENIAC, the world’s first electronic computer, which laid the computational groundwork for developing thermonuclear (hydrogen) weapons.
Stanley P. Frankel
Photo from IRE Transactions on Electronic Computers
September, 1959
This theoretical work caused Frankel to invent computational techniques that drove the rapid development of computers immediately following World War II. The work also infected Frankel with “the computer disease” (a term coined by the famous Nobel-winning physicist Richard Feynman specifically for Frankel). Frankel became enraptured with computers but he was cruelly cut off from the main force driving advanced computer development (weapons research) when he lost his security clearance during the Red scare of the early 1950s, just as the development of digital computers really started to take off.
In response to this lost opportunity, Frankel became an independent computer consultant. He is directly responsible for designing some significant computers during the 1950s and for indirectly triggering the development of a calculator that was the pre-prototype of HP’s 9100A desktop calculator and the development of the BASIC programming language, which HP used for it’s top-of-the-line desktop calculators and computers throughout the 1970s and into the 1980s.
First There Was Thermal Diffusion
Stanley Frankel was born in Los Angeles in 1919. His first scientific paper appeared in 1940, two years after he earned his BA. The 1-page paper was published while Frankel was a grad student at the University of Rochester in New York. It’s titled “Elementary Derivation of Thermal Diffusion” and it appeared in the Physical Review’s letters section. This paper established Frankel’s expertise in the mathematical modeling of physical phenomena, which will become the bedrock foundation of all theoretical work associated with nuclear weapons development for the next 60 years and beyond. However, in 1940, there’s nothing remotely “nuclear” about Frankel’s work.
In the spring of 1942, Frankel was a post-doc student living on the other side of the US, at the University of California at Berkeley, working for Dr. J. Robert Oppenheimer who is a brilliant theoretical physicist teaching nuclear physics at Berkeley and CalTech. Oppenheimer’s good friend and colleague, Dr. Ernst O. Lawrence, was also a professor at Berkeley and the world’s foremost authority on cyclotrons. He’s very good at both building ever bigger cyclotrons and at raising the funds to build these machines from both private and government sources.
Because of his work on the cutting edge of nuclear physics and his close association with powerful people in government and industry, Lawrence was heavily involved with the Briggs Advisory Committee on Uranium, which was created by order of President Franklin D. Roosevelt. During 1941, this committee was trying to decide whether or not the development of an atomic bomb was critical to the immediate war effort. Lawrence agitated energetically and repeatedly for the start of a uranium bomb project. He recommended that his friend Oppenheimer be brought in to study the feasibility of building an atomic weapon. The urgency of the topic explosively escalates when the Japanese attack Pearl Harbor on December 7, 1941, thus bringing the United States into World War II.
Oppenheimer, Robert Serber (formerly one of Oppenheimer’s postdoctoral students now at the University of Illinois), and graduate students Eldred Nelson and Stan Frankel performed the calculations on “neutron diffusion” (how neutrons move in a critical assembly of uranium during a nuclear chain reaction) and hydrodynamics (how the energy released by the chain reaction will produce an explosion) that are required to determine the feasibility of an explosive weapon based on uranium fission and the amount of fissionable material needed to build a weapon.
These calculations allowed a group of blue-ribbon theoretical physicists convened by Oppenheimer in June of 1942 to conclude that a fission weapon can be made to work and President Roosevelt subsequently authorizes the creation of the Army’s Manhattan Engineering District (better known as the Manhattan Project) to develop the atomic bomb. In Frankel’s words, the Manhattan Project’s mission was “to put genocide on a paying basis.” However, there’s no evidence that Frankel objected to this mission.
Box 1663
The Manhattan Project charter included the development of an atomic-weapons lab. General Leslie Groves convinces Oppenheimer to become director of the project and Oppenheimer recommends the Los Alamos mesa near Santa Fe, New Mexico for the Lab site because he loves its views of the mountains. Oppenheimer has spent a lot of time in the Los Alamos region of New Mexico and fell in love with the entire area. He’d owned a ranch called “Perro Caliente” (Hot Dog!) located near the Los Alamos mesa since the 1920s.
The Los Alamos weapons lab is constructed during 1943. To the rest of the world, the lab on the mesa is nothing but an anonymous post office box (Box 1663, Santa Fe, New Mexico). Oppenheimer took along many of his associates including Bob Serber, and his doctoral students, including Eldred Nelson and Stan Frankel, to Los Alamos. They became part of the laboratory’s theoretical division (the T division).
Nelson and Frankel had organized a computing service for Lawrence’s cyclotron-based electromagnetic isotope-separation project at Berkeley in 1942. At Los Alamos, they purchased a quantity of mechanical calculators (Marchants, Fridens, and Monroes) like the ones they used at Berkeley to help perform the many calculations required by the project physicists.
Although the Los Alamos physicists performed many calculations on their own, the sheer volume of calculations for the uranium bomb project proved too much, which led to the use of computers (the female kind). The wives of Los Alamos scientists—including Frankel’s wife Mary—were recruited to become, literally, human or “hand” computers that operated the mechanical calculators like industrial production equipment to crank out numbers. These human computers became the T-5 hand-computing group. Nick Metropolis and future Nobel Laureate and expert-on-pretty-much-everything Richard Feynman learned how to fix the mechanical calculators and became the T-5 handymen, working closely with Frankel and Nelson.
Frankel and Nelson did a superb job of organizing the calculations into an assembly line. Frankel’s deep understanding of the mathematical simulation of physical phenomena allowed him to decompose complex calculations into many simpler calculations that were distributed among the human computers. These simpler calculations incur fewer errors and the system worked well, until the humans started to tire of the endless, repetitive calculations.
Before 1943 ended, the endless bomb calculations proved more than the T-5 hand-computing group could handle and the T division decided to automate. There are no electronic computers to be had in 1943 (they didn’t exist yet, except in Iowa) so Los Alamos ordered IBM punched-card tabulation equipment including the IBM model 601 multiplier to help with the calculations. The tabulation equipment could add, subtract, and multiply using IBM punched cards for data input. The IBM punched-card tabulation group was designated T-6 and the tabulating machines were run by enlisted soldiers from the Army’s Special Engineering Detachment.
IBM card-tabulation equipment is brought in to supplement the calculation capabilities of the Los Alamos T-5 group. Photo from Los Alamos Science
Stanley Frankel and Eldred Nelson were put in charge of the IBM-equipped calculations group. Unfortunately, Frankel wasn’t really cut out to lead the group. He reportedly had a temper and verbally abused the soldiers (who were working for peanuts and didn’t have a clue as to why they were running the calculations). Frankel also doesn’t appear to have worked well under pressure and the tensions inevitably rose as the first atomic test (Trinity) loomed.
In addition, Frankel essentially became enraptured with the IBM tabulation equipment. He ignored his supervisory duties (which he didn’t execute well anyway) and focused all of his energy on the wonderful technical abilities of the IBM tabulating machines. He started working on personal projects such as programming the equipment to print nicely formatted trigonometric tables and neglected the bomb calculations.
Hans Bethe, head of the T division, responded to this problem by replacing Frankel with Feynman and Nicholas Metropolis (who would later pioneer the development of computers specifically designed for nuclear weapons development at Los Alamos). Feynman explained to the soldiers that they were helping to develop a weapon to win the war. Life in the calculation group got a lot better and productivity soared. Meanwhile, Frankel was out of a job and needed something else to do. He ended up working on thermonuclear fusion for Edward Teller in Fermi’s F Division.
Teller and the Super
Teller was a brilliant and erratic physicist from Hungary. He had participated in the blue-ribbon symposium Oppenheimer held in June, 1942. By that time, Teller along with Emil Konopinski had determined that it might be possible to initiate thermonuclear fusion in deuterium (heavy hydrogen) using the energy released by an atomic (fission) bomb. The concept gripped Teller and he could think of little else after that. He named the fusion weapon “The Super” for super-bomb. If Frankel had “the computer disease,” Teller had “the Super” disease.
Until the spring of 1944, Teller worked mostly on fission-related calculations but his heart really belonged to thermonuclear fusion research. Oppenheimer had allowed Teller to continue work on the Super during the early years at Los Alamos but certain discoveries about plutonium made during those early years caused the lab’s research to emphasize a plutonium fission bomb, which required a complicated ignition mechanism called implosion. Teller’s group was supposed to be working on the complex implosion calculations but Teller and his group probably spent half of their time on fusion calculations instead.
Hans Bethe reorganized the T Division in March, 1944 and put Teller in charge of a group that was charged with performing the calculations to produce a mathematical description of the hydrodynamics of implosion. The group included Konopinski, Nick Metropolis, and mathematician John “Johnny” von Neumann. Although Teller successfully determined the equation of state for highly compressed uranium and plutonium that was expected to result from an implosion, Teller wouldn’t supervise the detailed calculations for that equation of state because he wanted to work on the Super. In June, 1944, Teller got Oppenheimer to separate Teller’s group so that Teller to devote himself to the Super. Stanley Frankel joined this group and started working full time on thermonuclear fusion.
By 1945, Teller and his Super group had begun to realize just how immense the job of calculating the Super’s physics was. The complexity of the task was far beyond the reach of mechanical calculators run by a corps of female hand computers. It also proved beyond the reach of the punched-card tabulating machines. However, Johnny von Neumann had another way.
A chance meeting on a train platform in the summer of 1944 between von Neumann and Herman Goldstine led von Neumann to find out about another top secret military project: ENIAC, an electronic computer that was being developed for the army at the Moore School at the University of Pennsylvania in Philadelphia. Goldstine was co-directing the project for the US Army. ENIAC’s mission was to compute more accurate ballistics tables for artillery. The existing hand-computed tables were riddled with errors and the army desperately wanted to automate the generation of tables to weed out the errors. By early 1945, von Neumann started suggesting that ENIAC, when it became operational, might be the answer to the problem of running the Super calculations.
Trinity, Hiroshima, and Nagasaki
The Trinity test went off successfully on July 16, 1945. It proved the design of the plutonium implosion bomb. No such test was needed for the uranium bomb. The physicists had known that the rifle-like uranium bomb design would work since 1942. Two atomic bombs were dropped on Japan in August and World War II ended with Japan’s unconditional surrender on September 1, 1945. Because the war had ended, ENIAC’s primary mission, to compute artillery ballistics tables, became far less important. Consequently, von Neumann was able to arrange for Los Alamos to use the machine even before it was delivered to the army.
On July 16, 1945, the Trinity test verifies that the calculations made by Stanley Frankel in 1942 are correct. Photo from Los Alamos Science
Although ENIAC was not yet fully operational, Frankel and Nick Metropolis had traveled to the Moore School in Pennsylvania in August, 1945 to learn how to program ENIAC and what the machine was capable of doing. Herman Goldstine and his wife Adele gave Frankel and Metropolis a complete ENIAC course including the machine’s theory of operation and the rudimentary programming concepts of the day. Adele Goldstine eventually created the one and only ENIAC programming manual.
The ENIAC of 1945 was not a stored-program computer but an unconnected collection of function units including 20 accumulators that performed addition and subtraction, a multiplier, three function tables, a divider/square-rooter, and so on. These function modules were contained in 40 floor-to-ceiling rack-mount panels and were interconnected using thick patch cords that represented the machine’s “program.” ENIAC was actually a physically configurable electronic equation simulator, not an electronic computer in the modern sense. Even so, it’s not hard to imagine ENIAC capturing technophile Frankel’s imagination.
In his book Eniac, The Triumph and Tragedies of the World’s First Computer, Scott McCartney describes ENIAC:
“What the army got was a thirty-ton monster that filled 1,800 square feet—the size of a three-bedroom apartment in some cities. It had forty different units, including its twenty accumulators, arranged in the shape of a U, sixteen on each side, and eight in the middle, all connected by a ganglion of heavy black cable as thick as a fire hose. It was 1,000 times faster than any numerical calculator, 500 times faster than any existing computing machine. It could perform 5,000 addition cycles per second and to the work of 50,000 people working by hand.”
ENIAC, the world’s first electronic computer, was not programed but wired up to solve a problem. It became operational just in time for Stan Frankel and Nick Metropolis to run the calculations for “the Los Alamos problem,” which sought to find out if a hydrogen bomb was feasible. Photo from Los Alamos Science
The ENIAC “programmer” developed a wiring chart that showed how ENIAC’s function units were to be connected together and switch settings that “loaded” the required constants into the machine’s function tables. A team of technicians then set up the problem-specific wiring and flipped the function-table switches to enter constants associated with the problem. Input data for the problem entered ENIAC by means of IBM punch cards and the ENIAC then performed the simulation it was wired to run. Computed results were punched onto more cards using an IBM card punch.
Parts of ENIAC’s design resembled earlier mechanical calculators in operation. For example, its counters were decimal ring counters instead of the binary counters used in modern computers. Decimal ring counters closely mirror the way mechanical calculators (the Marchant’s, Friden’s, and Monroe’s of the day) use decimal counting wheels to perform their calculations. The reason this fact is significant is that ENIAC’s design will influence the way Frankel, who studies this design very closely in 1945, thinks about computer design over the next 20 years.
The Los Alamos Problem
Frankel and Metropolis returned from the Moore School in the summer of 1945 after learning how to program ENIAC. Together with Frankel’s wife Mary, they started preparing an ENIAC program to perform a most rudimentary Super calculation. It became known as “The Los Alamos Problem.” However, ENIAC was far too small to run the full-fledged simulation required to solve the problem so Frankel and Metropolis oversimplified the simulation. The specifics and certainly the results of The Los Alamos Problem are still classified. However, the calculation grossly oversimplified the question of fusion ignition by reducing the problem to a set of three partial differential equations that modeled the behavior of one-dimensional deuterium-tritium systems with the intent of computing the ignition temperature and related characteristics for such a system. This oversimplified calculation omitted many known physical effects so the model was surely far from accurate.
In fact, Frankel and Metropolis considered the entire effort to be an “exercise” that would simply demonstrate whether or not the problem could be calculated at all. they didn’t expect to get a useful quantitative result from this particular effort.Even so, the problem that ran on ENIAC was the most complex physics calculation ever attempted at the time. The Los Alamos problem started to run on ENIAC in November or December of 1945. Even oversimplified, it flushed many bugs out of the prototype computer because it consumed and exercised more than 95% of the machine’s capacity (99% by some accounts). By January of the next year, the Los Alamos problem had successfully run to completion.
Teller, who had the Super disease, didn’t see the ENIAC results as too simple to be useful. He persuaded Oppenheimer to approve a blue-ribbon panel of current and former Los Alamos scientists to review the ENIAC results in April, 1946. Although the calculations were necessarily simplified and rudimentary, Teller assisted in writing the panel’s conclusion, which stated that a Super bomb could likely be constructed and that it would probably work. Among others, Frankel also signed this report. As with the neutron-diffusion calculations he performed in 1942, Frankel’s ENIAC work helped to further this next stage in nuclear weapons development.
Nick Metropolis
Photo from Los Alamos Science
(Note: Another version of this story, says that the H-bomb calculations proved too complex for ENIAC to handle and that Frankel subsequently ran the problem on a computer at the Eckert-Mauchly Computer Corporation in Philadelphia. However, this version of the story would push the calculations out at least a couple of years because Eckert and Mauchly founded their company on December 22, 1947. This date is later than the April, 1946 review of the ENIAC results and Frankel was no longer working at Los Alamos by 1947. Frankel’s own interviews conducted by Robina Mapstone of the Computer Oral History Project confirm that the calculations did run on ENIAC in late 1945 and early 1946. There has been some conjecture that Frankel and Metropolis ran Monte Carlo calculations (see below) simulating the hydrogen-fusion process on the ENIAC but that seems unlikely because the original ENIAC was not a stored-program machine. Metropolis has stated clearly, in print, that he and Frankel prepared a simplified mathematical model of the thermonuclear reaction that was “realistically calculable” by ENIAC after receiving their training from the Goldstines. Although some limited capability for stored-program operation was later added to ENIAC, the Frankel and Metropolis would have had enough problems just running a the oversimplified, one-dimensional fusion simulation during the ENIAC’s shakedown runs in late 1945 and early 1946. Adding the extra burden of computing random statistical variations to simulate the probabilistic nature of nuclear reactions seems unlikely. However, the ENIAC experience will spur both Frankel and Metropolis to independently start developing computerized implementations of Fermi’s Monte Carlo analysis technique before the decade ends.)
Although the calculations Frankel and Metropolis ran on the ENIAC and the results of those calculations are still classified more than half a century after the fact, the world knows that the H-bomb is feasible through practical demonstration. Both the United States and the USSR exploded thermonuclear bombs in subsequent tests. Frankel and Metropolis published some of their ENIAC results in a paper that appeared in Physical Review in 1947. This paper documents the energies needed to cause heavy atomic nuclei to split and the calculations were done on ENIAC. Significantly, the 1947 Physical Review paper ends with an acknowledgment thanking the Moore School’s ENIAC team for its help in running the calculations for this paper, and those for an unspecified prior problem. That prior problem was obviously the still-classified “Los Alamos problem.”
In February of 1946, after finishing the computations for the Los Alamos H-bomb problem, Frankel and Metropolis had told Goldstine that they planned to take positions at the University of Chicago. The war is over and the physicists are returning to academia after ably serving their country’s defense needs. Chicago is Nick Metropolis’ home town and he wants to return there. Teller, Frankel, Metropolis, and Eldred Nelson joined the University of Chicago’s newly formed Institute for Nuclear Studies (now called the Enrico Fermi Institute), which was created in 1945 after the war ended, to keep together, as much as possible, the team of Chicago-based Manhattan project scientists and engineers.
The Post-War Computer Race and Frankel’s Fall
Just one year later, Frankel is no longer at the Institute for Nuclear Studies. He moved back to his hometown: Los Angeles. Frankel hated Chicago. Eldred Nelson goes with him. Nelson’s wife also hated Chicago. By the end of 1947, Frankel has formed a consultancy with Eldred Nelson and the firm of Frankel & Nelson in Los Angeles consults on problems in applied mathematical physics. Their clients include the electronics division of Hughes Aircraft and Northrup. At Northrup, Frankel and Nelson work on solving navigational errors in missile guidance systems. However, early in 1949, the US Army cancels Frankel’s security clearance because his father had been a Communist. The consulting firm of Frankel & Nelson ceased to exist and Nelson joined Hughes (later, TRW). Nelson recalls this period as a particularly painful time.
(In a few more years, J. Robert Oppenheimer would also be stripped of his security clearance. Both Oppenheimer and Frankel had served their country well during World War II, but both had too closely brushed against the “wrong” causes. McCarthyism and the Red panic would run rampant. When Oppenheimer tried to block development of the H-bomb, he incurred the wrath of the wrong people who decided to remove him from his position of power in the atomic-weapons community. Oppenheimer was “burned at the stake” (in the words of Herbert Grosch). Frankel was treated badly and cast out of the atomic brotherhood, along with several other Los Alamos and Manhattan Project scientists, although Frankel apparently was not publicly barbequed the way that Oppenheimer and many writers and actors from Hollywood were.)
Shortly after Frankel loses his security clearance, Professor Gilbert D. McCann recruits Frankel to become the head of a newly created digital computing group in CalTech’s Engineering Division in Pasadena, California. The job is to provide computing services to CalTech graduate students, much in the same way that Frankel provided these services at Los Alamos. At CalTech, Frankel meets PhD student Bernie Alder, who was working on simulating the interaction of atoms using statistical mechanics and liquid theory. Alder will eventually become one of the first scientists to work at the Lawrence Livermore National Laboratory when it opens in 1952. Frankel started to work with Alder on his PhD thesis problem at Cal Tech. Together, they decide to try using a statistical-analysis technique developed in the 1930s by the famous nuclear physicist Enrico Fermi.
Frankel’s experience with ENIAC in 1945 has spurred him to contemplate the use of electronic computers to explore statistical sampling techniques. Frankel is especially well suited to this line of research because he’s had a long-standing interest in random physical phenomena. He often played solitaire and poker and he contemplated the statistical nature of the card games while he’s playing. He is intensely interested in certain distributions of prime numbers, which he called “lucky numbers,” and he’ll eventually write a paper on the topic.
Further, Frankel’s first brush with computers through the programming of ENIAC alerted him to the fact that computers will alter the way people solve problems. His 1947 paper in Physical Review, written with Nick Metropolis, discusses the use of iterative summation to replace manual integration. Frankel and Metropolis wrote the following footnote in this paper:
In treating this problem with a desk calculator the use of a table of the first elliptic integral in this step would be far easier. However, in setting up the problem for the ENIAC the economy in program controls and programming labor of this procedure seemed to us adequate compensation for the increase in computing time. It seems likely that the use of high speed calculating machines will often effect changes of this kind in the economy of calculating procedures. (Emphasis added.)
Frankel clearly understood the new research vistas made possible by the invention of the computer. He also understood that calculations using computers could take fundamentally different problem-solving approaches so he was clearly primed and ready to apply computers to the non-classical study of statistical physical phenomena when he met Bernie Alder. The analysis technique the two decide to try is now called Monte-Carlo analysis and, since Alder’s and Frankel’s pioneering work, it has become essential to a wide range of scientific and technical work from nuclear weapons development to VLSI chip design.
However, the IBM tabulation equipment that Alder was attempting to use during the late 1940s proved woefully inadequate for Monte-Carlo analysis so Frankel traveled to Manchester, England in the summer of 1950 where the University of Manchester and Ferranti Ltd. were building a stored-program electronic computer (the Ferranti Mark I, a commercially re-engineered version of the prototype Manchester Mark I computer that had been built in 1948). Frankel learned how to program the Ferranti machine (also called the Manchester Mark II) and he successfully ran Alder’s Monte-Carlo analysis on it. However, Alder’s thesis advisor was a classical physicist and didn’t believe in Fermi’s “unproven” statistical-computation technique. Consequently, he discouraged Alder from publishing his work. As Alder said, if your thesis advisor doesn’t believe in your work, you don’t publish.
Meanwhile, a team of Los Alamos scientists including Nick Metropolis, Arianna Rosenbluth, Marshall Rosenbluth, Augusta Teller, and Edward Teller independently developed Fermi’s Monte-Carlo method into a computerized numerical-analysis technique. They subsequently published a paper titled “Equation of State Calculations by Fast Computing Machines” in the Journal of Chemical Physics during 1953. It is the first published paper on computerized Monte-Carlo analysis to appear. Alder notes that his Monte-Carlo work with Frankel received a footnote in Metropolis’ 1953 paper crediting the two CalTech researchers with simultaneous development of the computerized version of Fermi’s Monte-Carlo technique. Alder and Frankel eventually did publish their work in the Journal of Chemical Physics in 1955, but the earlier paper published by Metropolis and his colleagues became the classic paper that’s widely cited by other researchers today.
The Red Scare Leads to Consulting
Loss of his security clearance also barred Frankel from the inner circles of nuclear and computer research for nuclear weapons development. He stopped attending the annual main event for computer designers called the Joint Computer Conference, which was held throughout the 1950s and 1960s by the IRE (Institute of Radio Engineers), the AIEE (American Institute of Electrical Engineers), and the ACM (Association of Computing Machinery).
Although he had been cast out of this inner circle, Frankel did not stop working with computers. By 1950, Frankel had become one of the most accomplished and experienced computer scientists in the country. While at CalTech, Frankel started designing a simple computer. He wanted to explore the lower limits of what a simple computer might be. While consulting and visiting the electronics division of Hughes Aircraft, Frankel had been encouraged to stuff his pockets with off-spec germanium diodes. Hughes had just started to make these devices and the company had wastebaskets full of functional devices that it couldn’t sell because they failed to meet their advertised specifications. However, they were ideal for digital design.
Frankel called his minimal computer design MINAC. He built a breadboard of the simple computer based on the Hughes germanium diodes and magnetic-drum memory. He’d learned about magnetic-drum memory while visiting ENIAC’s designers, Eckert and Mauchly, in Philadelphia. By that time, Eckert and Maucly had left the Moore School and were working on UNIVAC. Mauchly gave Frankel a lot of advice. Frankel also visited a computer group at the National Bureau of Standards, Raytheon’s computer group, and the Whirlwind group at MIT.
When he returned to CalTech, Frankel found a physics graduate student named James Cass who was very good in the machine shop. Cass hand built a magnetic drum for Frankel while Frankel and his secretary build magnetic heads for the device. By 1954, he’d finished a breadboard of the machine. Librascope, a southern California company located in Glendale, licensed the MINAC design from CalTech (Frankel’s employer), hired Cass to help turn the design into a production-ready product, and hired Frankel as a computer design consultant. Librascope put the MINAC into production, naming it the LGP-30.
The First Personal Computer
The LGP-30 may well be considered the first personal computer. It was an air-cooled, desk-sized, single-user machine. It ran on “only” 1500 Watts of power from a regular 110-volt room outlet and required no special air conditioning. It was also extraordinarily successful. More than 500 LGP-30s were built and sold in the late 1950s and early 1960s, at a time when the sale of one, two, or three units per design was the norm.
This ad for Stan Frankel’s LGP-30 personal computer appeared in the Proceedings of the IRE in May, 1957.
Stan Frankel的作品Frankel described the LGP-30’s design in detail in a paper published in the March, 1957 issue of the IRE Transactions on Electronic Computers but the first production LGP-30s were actually built in 1956. Frankel minimized the number of active components needed in the LGP-30 through the extensive use of solid-state diode logic and by placing all of the machine’s memory and registers on a rotating magnetic drum. Consequently, the LGP-30 has only 113 vacuum tubes (and 1450 diodes). At $27,000, it was a relatively inexpensive machine for its day, although Frankel’s goal had been to design an even cheaper machine.
The lineage of the LGP-30 is very confusing. It was first built by Librascope and “LGP” stood for “Librascope General Purpose.” Then General Precision of Glendale, California bought Librascope and “LGP” became “Librascope General Precision.” However, General Precision was just the machine’s manufacturer. The company partnered with the Royal McBee Corporation of Port Chester, New York (a division of Royal Typewriter), which marketed the computer, trained customers, and serviced the machines through its data-processing division. So the advertisements you see in magazines of the day are for Royal McBee LGP-30 computers and the product manuals for the LGP-30 are Royal McBee publications. Librascope survives as Lockheed Martin Librascope. However, the company no longer makes LGP-30s.
A BASIC Contribution
Dartmouth College in Hanover, New Hampshire bought an LGP-30 in 1959. Two Dartmouth researchers, John Kemeny and Thomas Kurtz, used the LGP-30 to develop computer-programming languages that undergraduates could understand, learn, and use. FORTRAN and ALGOL were apparently deemed too complex for the average undergraduate student of the early 1960s. Kemeny and Kurtz developed several simplified programming languages on the LGP-30 including DARSIMCO (Dartmouth Simplified Code), DART, ALGOL 30, SCALP (Self-Contained ALGOL Processor), and DOPE (Dartmouth Oversimplified Programming Experiment). None of these languages became a widespread success but they provided excellent preparation for the main event.
Stan Frankel This ad for Stan Frankel’s LGP-30 personal computer appeared in the Proceedings of the IRE in April, 1959.
By 1963, the LGP-30 has become outdated and Dartmouth replaced it with General Electric GE-225 and Datanet-30 computers. Kurtz supervised the development of a timesharing system for the GE computers. At the same time, Kemeny developed a compiler for the next experimental Dartmouth programming language, the Beginner’s All-purpose Symbolic Instruction Code—or BASIC.
Over the next 20 years, Dartmouth BASIC serves as a foundation for the personal computing world. HP developed its own dialect of BASIC for its minicomputers during the late 1960s and built BASIC interpreters into all of its top-end desktop calculators and computers during the 1970s and 1980s. Indirectly, Frankel’s LGP-30 enabled HP’s rise to the top of the 1970’s personal computer market by creating an early machine environment that encouraged the development of a personal style of computer use. Thanks in part to Frankel’s dream of and work towards the creation of a personal computer, the right companion programming language already existed when HP developed suitable desktop-computer hardware.
A code-compatible, cost-reduced, transistorized version of the LGP-30 called the LGP-21 appeared a few years after the LGP-30 is introduced. According to an interview with Cass, the LGP-21 is a transistorized copy of the LGP-30. The machine’s design uses 460 transistors and about 300 diodes. It runs three times slower than the LGP-30 (an artificial handicap placed on the machine to allow it to be sold for less than the LGP-30, which Librascope continues to offer). The LGP-21 cost much less than the LGP-30 ($16,200 in 1963).
CONAC the Computer and a Microwave Interlude
A little more than two years after the introduction of the LGP-30 (but only one year after he published a paper on the LGP-30’s design), Frankel published a really unusual paper on computer design. It appeared in the September, 1959 issue of IRE Transactions on Electronic Computers. The paper, titled “A Logic Design for a Microwave Computer,” details Frankel’s consulting efforts on behalf of General Electric to develop a computer using microwave devices—traveling-wave tubes and microwave diodes—as digital components. In his paper, Frankel admitted that flip-flops are difficult to create with microwave components. He proposed using delay lines as an alternative for building register storage in a microwave computer. It’s not a new idea. Many early stored-program computers used delay-line memory. Frankel would soon recycle this idea of delay-line storage in his transistorized calculator design.
If he had succeeded, Frankel would have helped GE build a GHz computer thus leapfrogging the industry by 30 to 40 years. However, no such computers appeared and GE eventually sold off its computer hardware business so it’s reasonable to assume that Frankel ran into design problems that were either too expensive to solve or that could not be solved at all. However, the biography that appeared with Frankel’s 1959 article claims that he at least did finish the logic design of the microwave computer. It doesn’t say if the machine was ever built.
Stan, ERMA, and Frankelstan’s Revenge
In 1957, General Electric and IBM competed for a contract from the Bank of America to develop the ERMA (Electronic Recording Method of Accounting) system. GE won. Stanley Frankel did not work on ERMA’s mainframe design but he did develop a small computer for the ERMA system that was to be used as a “paper processor.” This auxiliary processor captured data from newly developed MICR (magnetic ink character recognition) check readers and stored the data on magnetic tape for subsequent processing by ERMA’s mainframe processor, a GE-100 (later renamed the GE-210). Frankel’s auxiliary processor (we would call it an “embedded processor” these days) was considered elegant but it suffered from the lack of a user interface. Programs had to be debugged with an oscilloscope. Consequently, ERMA programmers dubbed the small auxiliary processor “Frankelstan’s revenge.” Ultimately, Frankel’s design was not used for ERMA. GE added a second GE-100 mainframe to ERMA’s design for other reasons and thus eliminated the need for an auxiliary processor.
The 1959 biography in IRE Transactions on Electronic Computers also provided a list of other computer designs that Frankel either developed or helped to develop during the 1950s including the CONAC (the Continental Automatic Computer, nicknamed “Connie,” a more advanced, drum-based machine developed by the Continental Oil Company during the period of 1954-1957). Frankel’s Monte-Carlo paper published with Alder in the Journal of Chemical Physics says he’s at the Theoretical Research Group of the Continental Oil Company in Los Angeles by March, 1955 (he’s actually consulting there, helping to develop a computer called CONAC). Oil companies will eventually become the second big users of supercomputers for analysis of seismic data gathered in the search for new oil reserves, following the explosive growth in the use of supercomputers for nuclear weapons design at the US national laboratories.
Throughout the 1950s, Frankel continued to publish scientific and technical papers that list a variety of southern California addresses as a consultant with no company or university affiliation. He also designs a small computer intended for use as an auxiliary document processor for the General Electric 100 and 210 mainframe computers, the M’AC (an “academic” design developed for a Long Beach company named Logical Design), and the NIC-NAC, a desktop calculator.
Frankel’s Last Design
In 1958, as part of a plan to expand from typewriters into a full line of office equipment, the Smith Corona Company (a leading typewriter manufacturer) acquired Marchant (a leading calculator company). Smith Corona marketed Marchant mechanical calculators under the brand name “Smith-Corona Marchant” until 1962 when the entire company adopted that brand name (SCM for short).
Electronic calculators started to appear from companies such as Friden, Sharp, Wang, Mathatronics, IME, and Olympia during the early 1960s. These early machines emulated mechanical calculators—they performed addition, subtraction, multiplication, and division but not transcendental functions. However, even though they performed the same functions and no more, electronic calculators were both faster and quieter than the mechanical versions so the marketplace rapidly adopted the new machines. SCM realized that it needed an electronic calculator in its product line, or it would shortly cede the calculator market to its competitors.
By early 1964, SCM had licensed a Frankel calculator design. Frankel’s calculator design would become SCM’s Cogito 240SR desktop calculator. Prior to engaging Frankel, SCM had hired a very bright electrical engineering graduate from Berkeley named Tom Osborne to help the company enter the electronic era. Osborne studied Frankel’s design and SCM’s production plans for a while and decided that the entire situation didn’t look very good. First, Frankel’s calculator design stored its registers in a recirculating acoustic delay line, similar in function to the rotating magnetic drum he used for the LGP-30, but much slower. The delay line needed milliseconds to recirculate the registers, which was sure to result in slow operation. Further, the design required a large number of crystal diodes that sold for 25 cents each (again like the LGP-30’s design). However, to meet cost targets, SCM planned to use reject diodes that didn’t meet spec (like Frankel’s MINAC, the prototype LGP-30). Osborne felt that plan was a recipe for disaster.
Osborne decided that he didn’t want to work on the Cogito 240SR project any longer. In fact, he believed he could develop a superior calculator design and he proposed this plan to SCM’s management. He even proposed working for free, taking only lab space from SCM until his calculator design was finished. SCM’s management refused the offer, saying it didn’t run the company that way.
SCM weighed the credentials of the freshly minted MSEE from Berkeley against the seasoned computer designer from the Manhattan Project and unsurprisingly decided to bet on Frankel, the veteran. In fact, SCM had already bet on Frankel and had not expected to hear an objection from Osborne. The decision had been made before Osborne even started his analysis. Osborne left SCM and went off on his own, created a brilliant calculator design, got another refusal from SCM, hit the pavement to find a buyer, eventually hooked up with HP, and helped to put HP into the desktop-calculator business.
The Cogito 240SR electronic calculator.
Photo courtesy of Rick Bensene. www.oldcalculatormuseum.com
SCM developed Frankel’s design without Osborne’s help and put the Cogito 240SR into production by early 1966. The Cogito 240SR was a fully transistorized calculator with a tiny CRT display that showed the contents of three registers. It employed a truly baroque mechanical keyboard that harkened back to the machine’s electromechanical calculator roots and SCM’s typewriter heritage.
Like the LGP-30, the Cogito 240SR used what seems today to be an unusual memory to implement the calculator’s register storage. Instead of a magnetic drum like the LGP-30, the Cogito 240SR used an acoustic delay line to store the 480 bits of memory the calculator requires. The acoustic delay line is a bit-serial storage device and the machine was bit-serial throughout to reduce its cost. Delay-line memories are quaint today, but they were used extensively in early computer designs because core memory was new and expensive, flip-flops required either two tubes or two transistors per bit and were therefore prohibitively expensive, and integrated-circuit memories had yet to be invented.
Osborne was right about the the Cogito 240SR. It was slow. Really slow. In the words of HP’s Director of R&D Barney Oliver, the Cogito 240SR was “a miserable machine, it took forever to do anything.” It was the Cogito 240SR’s use of the slow acoustic delay line for register storage that Osborne had objected to (that and the below-spec diodes). Osborne wanted to use random-access core memory to provide high-speed register storage so that the calculator would complete it’s computations quickly.
The SCM Cogito 240SR may well have been Frankel’s last computer design. If he designed any subsequent machines, the documentation on those later computers is deeply buried. Much of Frankel’s life is poorly documented. I could find only the one photograph of him although most of the other early male members of the Los Alamos team (and their wives) seem amply photographed. The research needed to compose just the few paragraphs about Frankel on this Web page was substantial and represents the assembly of many bits and pieces from more than thirty documents over a month’s time. Research on Frankel’s life would have been nearly impossible before the creation of the World Wide Web and Google’s search engine.
It appears that the Red scare of the early 1950s may have driven Frankel almost completely underground as it did many Hollywood screenplay writers. After 1954, Frankel preferred to work as a paid consultant and avoided joining companies. He published a few papers and a book with Karol Mysels on the topic of soap films in the late 1950s and the 1960s. Frankel’s last published scientific paper (on the thickness measurement of soap films) appears in the September, 1966 issue of the Journal of Applied Physics. His work on soap films appears to be well respected and useful even today but Frankel disappeared from the history of computer development after the introduction of SCM’s Cogito 240SR in 1966. His testimony does appear in the records of the Honeywell vs. Sperry Rand trial that took place during the 1970s.
Frankel’s last published papers give a street address of 411 N. Martel, Los Angeles, California. The house is still standing. Here’s a shot of the house taken in February, 2009.
Frankel’s work may be scattered and largely unknown but he had a significant and profound effect on many fields in science and engineering, including computer design, and his dream of a personal computer did come true with the help of his technical legacy. Frankel died in May, 1978. He lived to see the introduction of early microcomputers such as the Altair, Imsai, Radio Shack TRS-80, and Commodore Pet but before the introduction of the IBM PC. Thus Frankel did get to see the early realizations of his dream of truly personal computers.
Special thanks to Rick Bensene for bringing Stan Frankel to my attention and for his efforts in documenting Frankel’s life and his work. Frankel and his story have haunted me ever since.
Information for this Web page came strictly from unclassified, publicly available information sources including personal telephone interviews with Tom Osborne, Eldred Nelson, and:
Alder, B J, Frankel, S P, and Lewinson, V A, Radial Distribution Function Calculated by the Monte-Carlo Method for a Hard Sphere Fluid, The Journal of Chemical Physics, Volume 23, Number 3, March 1955, p 417-419.
Allan, Roy A, A History of the Personal Computer, The People and the Technology, Allan Publishing, London, Ontario, Canada, 2001.
Anderson, Herbert L, “Metropolis, Monte Carlo, and the MANIAC,” Los Alamos Science, Fall 1986, p 96-107.
Bensene, Rick, “SCM Marchant Cogito 240SR Electronic Desktop Calculator,” www.oldcalculatormuseum.com/scm240sr.html.
Bethe, Hans A, “Observations on the Development of the H-Bomb,” published as Appendix II in the 1989 version of The Advisors, Oppenheimer, Teller, and the Superbomb by Herbert F York, Stanford University Press, 1976, 1989.
Bethe, Hans A, “Coments on the History of the H-Bomb,” Los Alamos Science, Fall, 1982, p 43-53.
Burks, Alice Rowe, Who Invented the Computer? The Legal Battle That Changed Computing History, Prometheus Books, Amherst, NY, 2003.
Conant, Jennet, Tuxedo Park, Simon and Schuster, New York, NY, 2002.
Feynman, Richard, Surely You’re Joking, Mister Feynman!: Adventures of a Curious Character, W W Norton & Company, New York, NY, 1984.
Fitzpatrick, Anne, “Igniting the Light Elements: The Los Alamos Thermonuclear Weapons Project, 1942-1952" PhD dissertation, Virginia Polytechnic Institute and State University, 1999.
Frankel, S Phillips, “Elementary Derivation of Thermal Diffusion,” Physical Review, Volume 57, Number 7, April 1, 1940, p 661.
Frankel, S and N Metropolis, “Calculations in the Liquid-Drop Model of Fission,” Physical Review, Volume 72, Number 10, November 15, 1947, p 914-925.
Frankel, Stanley P, “Convergence Rates of Iterative Treatments of Partial Differential Equations,” Mathematical Tables and Other Aids to Computation, Volume 4, 1950, p 65-75.
Frankel, S P, “The Logical Design of a Simple General Purpose Computer,” IRE Transactions on Electronic Computers, March 1957, p 5-14.
Frankel, S P, “On the Minimum Logical Complexity Required for a General Purpose Computer,” IRE Transactions on Electronic Computers, December 1958, p 282-284.
Frankel, Stanley P, “A Logic Design for a Microwave Computer,” IRE Transactions on Electronic Computers, September 1959, p 271-276.
Frankel, Stanley P and Karol J Mysels, “On the ‘Dimpling’ During the Approach of Two Surfaces,” Journal of Physical Chemistry, Volume 66, January 1962, p 190-191.
Frankel, Stanley P and Karol J Mysels, “Simplified Theory of Reflectometric Thickness Measurement of Structured Soap and Related Films,” Journal of Applied Physics, Volume 37, Number 10, September 1966, p 3725-3728.
Fritz, W Barkley, “The Women of ENIAC,” IEEE Annals of the History of Computing, Volume 18, Number 3, Fall 1996, p 13-28.
Goldstine, Herman H, The Computer from Pascal to von Neumann, Princeton University Press, Princeton, New Jersey, 1972.
Grosch, Herbert RJ, Computer: Bit Slices from a Life, Third Millennium Books, Novato, CA, 1991. http://www.columbia.edu/acis/history/computer.html.
Groueff, Stephane, Manhattan Project, The Untold Story of the Making of the Atomic Bomb, Little, Brown, and Company, Boston, 1967.
Head, Robert V, ERMA’s Lost Battalion, IEEE Annals of the History of Computing, July-September 2001, Volume 23, Number 3, pages 64-72.
Howes, Ruth H and Caroline L Herzenberg, Their Day in the Sun, Women of the Mahnattan Project, Temple University Press, Philadelphia, PA, 1999.
Jennings, Tom, “Librascope/General Precision LGP-21 Computer,” www.wps.com/projects/LGP21.
Kernan, Donal Mac and Michel Mareschal, “Berni J. Alder, Interview,” SIMU Challenges in Molecular Simulations, Issue 4, Chapter II, Lawrence Livermore Laboratory.
Lee, J A N, “May in Computing History,” Computer, Volume 29, Number 5, May 1996.
Mapstone, Robina, Interview with Dr. Stanley Frankel, October 5, 1972, Computer Oral History Project, Smithsonian Institution.
Mapstone, Robina, Interview with Dr. Stanley Frankel, October 26, 1972, Computer Oral History Project, Smithsonian Institution.
Mapstone, Robina, Interview with James Cass, December 8, 1972, Computer Oral History Project, Smithsonian Institution.
McCartney, Scott, Eniac, The Triumphs and Tragedies of the World’s First Computer, Penguin Putnam, Inc, New York, 1999.
Metropolis, N, J. Howett, and Gian-Carlo Rota, A History of Computing in the 20th Century, Academic Press, New York, NY, 1980.
Metropolis, N, “The Beginning of the Monte Carlo Method,” Los Alamos Science, Special Issue, 1987, p 125-130.
Michael, George, “An Interview with Bernie Alder,” www.nersc.gov/~deboni/Computer.history/Alder.html.
Thelen, E, “LGP-30,” http://ed-thelen.org/comp-hist/lgp-30.html.
“Evolving from Calculators to Computers,” www.lanl.gov/worldview/welcome/history/22_computers.html.
“Oversight Committee Formed as Lab Begins Research,” www.lanl.gov/worldview/welcome/history/16_oversight.html.
“The Berkeley Summer Study,” www.lanl.gov/worldview/welcome/history/02_berkeley-summer.html.
