USA 04The Birth of NC and the Machining Center

Parsons' idea and MIT's NC milling machine (1940–1955)

Helicopter blades and an IBM calculator

The story begins at Parsons Corp. in Traverse City, Michigan. During World War II this company worked out how to make the rotor blades for Sikorsky's R-5 helicopter and took on their manufacture^[2][3]. The blade cross-section is an airfoil, and checking that the shape is correct requires templates cut to the outline of the section. The coordinates of the points on a template outline had to be calculated one by one from the equation of the airfoil.

Around 1947, company president John T. Parsons and engineer Frank L. Stulen came up with a method: have an IBM punched-card calculator compute these coordinates, then move the table of a boring machine by hand to match the resulting table of numbers, drilling a close series of holes along the template outline^[3][2][4]. The idea was to position the machine from calculated numbers. In that case, why not record the numbers as holes in cards or tape, read them, and move the machine automatically? This thought became the starting point of NC.

The Air Force contract and MIT

Parsons presented this idea to the U.S. Air Force in January 1949^[5], and in June of the same year signed a contract with the Air Force's Air Materiel Command (AMC) to build a milling machine driven by numbers on cards. According to Noble, the contract was dated June 15, for $200,000 over 21 months^[4][5]. The Air Force jumped at it because the jet age had begun, and large parts with complex curved surfaces, such as wing skins and spars, had to be machined accurately and in quantity. The MIT reports give the original target as integrally machined wing skins^[6].

Parsons chose MIT's Servomechanisms Laboratory as a subcontractor, as experts in servos (control that makes a position follow a command). The laboratory had been founded in 1940 by Gordon S. Brown^[1], and numerical control research at MIT began in July 1949^[6]. Here the direction changed significantly. Parsons' idea was to position at set points one after another, but MIT moved toward a method that drove three axes simultaneously and continuously to cut curved surfaces^[6].

The continuous-motion method made the equipment larger and cost more money and time. In February 1951, when the equipment was about 30% complete, Parsons Corp. stopped funding. The Air Force's AMC took over the research through a direct contract with MIT, and this contract lasted about nine years, until 1960^[6]. In July of the same year, a control system moved the first axis from paper-tape commands alone^[6]. MIT's history of electrical engineering says that the project engineers William M. Pease and James O. McDonough coined the name “numerical control”^[17].

1952: An NC milling machine converted from a Hydro-Tel

In March 1952, MIT's NC milling machine was completed^[6]. The base machine was a government-loaned Cincinnati Hydro-Tel (a hydraulic tracer milling machine), whose original drive and control equipment was removed and replaced with equipment built in the laboratory. All three motions (table, spindle head and cross slide) were controlled numerically and synchronized so the machine could follow any path. Commands were given on punched paper tape^[6].

The controller consisted of about 270 vacuum tubes, 170 telephone relays and 300 germanium diodes on vertical panels, housed in an L-shaped frame^[10]. Equipment the size of a small room was running a single milling machine. In September 1952 the laboratory formally unveiled it to invited people from the aircraft and machine tool industries. According to Noble, 215 people from 130 companies came to see it over three days^[4]. The explanatory booklet handed out was reprinted several times, with 2,483 copies distributed, and by 1956 about 30 organizations and more than 3,000 people had seen the machine at MIT^[6]. That same September, Pease wrote “An Automatic Machine Tool” for the popular science magazine Scientific American^[9]. This article later became the starting point of NC in Japan as well (see “Links with Japan” below).

Two patents: positioning and the servo system

There are also two patents. Parsons and Stulen filed a patent on May 5, 1952 for an apparatus that positions a machine tool numerically, granted on January 14, 1958 as U.S. Patent 2,820,187^[7]. On the MIT side, four people (Jay W. Forrester, Pease, McDonough and Alfred K. Susskind) filed a patent on a numerical control servo system on August 14, 1952, granted on December 18, 1962 as U.S. Patent 3,069,608. By the time it was granted, the rights to this patent had passed to John T. Parsons^[8]. Parsons held both patents and later licensed them exclusively to Bendix (next section)^[17]. In 1985 Parsons received the National Medal of Technology together with Stulen, and in 1993 he was inducted into the National Inventors Hall of Fame^[70][27].

Whirlwind and the question of “who makes the tape”

Even with the machine running, there was another wall. To cut a curved surface, the path of the cutter center has to be divided into fine points and the coordinates of every point punched into the tape. Calculated by hand, a single part would take weeks. MIT used a computer then running at the same university, Whirlwind I, to calculate the tool paths and produce tapes that could go straight to machining^[10]. In 1952–54 John H. Runyon organized the calculation routines used on Whirlwind, and in 1955 Arnold Siegel built a prototype of two-dimensional automatic programming^[6]. From the problem of “moving the machine” to the problem of “making the commands”: here the history of NC begins to overlap with the history of computers and software.

Numericord and the Air Force's 100 machines

Machine tool makers moved too. In July 1953 Giddings & Lewis of Fond du Lac, Wisconsin asked MIT to build a prototype commercial NC unit, which was delivered in May 1955^[6]. In June 1955 the company demonstrated its magnetic-tape NC “Numericord”, developed with GE, running an aircraft spar and skin milling machine for the Air Force and aircraft makers, and announced that it would sell it commercially^[13][4]. Magnetic tape was used instead of paper tape to send the fine, computer-generated commands quickly and in large volumes.

Then in 1955 the Air Force began issuing contracts to build 100 NC milling machines. A 1974 report by the General Accounting Office (GAO) puts the total at $35 million^[12]. While private factories were still waiting to see, the military created the NC market with its own money. The historian David F. Noble later questioned this sharply (see “Noble's question” below).

Three ways NC moves a machine. NC comes in three kinds: positioning, which moves from one hole position to the next (suited to drilling and boring machines); straight-line cutting, which cuts straight along one axis at a time; and contouring, which moves two or more axes at once to cut angles and arcs. MIT's machine was 3-axis contouring from the start, built to cut aircraft curves. Low-cost positioning NC spread to small shops somewhat later. Today's CNC codes G00 (rapid traverse), G01 (linear interpolation) and G02/G03 (circular interpolation) are a legacy of this distinction (code table).

APT and the first machining center (1956–1959)

APT: describing part shapes in words

In June 1956 the Air Force restructured its contract with MIT into research on automatic programming: having a computer generate the commands for 3-axis and 5-axis contour cutting^[6]. It was led by Douglas T. Ross, head of the laboratory's Computer Applications Group, and it was here that APT (Automatically Programmed Tools) was born^[6][20]. In APT, the programmer writes the part shape and the cutter motion in near-natural language (“this line”, “this circle”, “this plane”), and the computer calculates the tool path and produces the tape. It is exactly the idea behind today's CAM.

APT was a program too big for a single laboratory to complete. The Aircraft Industries Association (AIA), a group of aircraft companies, took on the joint effort, and in May 1957, 21 programmers from 14 companies gathered at MIT to divide up the writing of APT II, which ran on the IBM 704. MIT acted as technical coordinator^[6][5]. It is an early example of what we would now call open collaborative development: building one piece of software across company boundaries.

On February 25, 1959, MIT, the Air Force and the AIA held a press conference on APT at MIT. They demonstrated APT running on the IBM 704 and gave reporters aluminum ashtrays programmed with APT and cut on MIT's NC milling machine^[6]. An ashtray with the raised lettering “APT II MIT AMC Feb 59” is now in the collection of the Computer History Museum^[37].

The Air Force's 105 machines go to the factories

The NC machines bought by the Air Force were distributed to aircraft makers' plants in 1957–58. A July 1957 special report in American Machinist counted 105 large machine tools with continuous-path NC. The machines were built by four companies (Cincinnati Milling Machine, Giddings & Lewis, Kearney & Trecker and Morey) and the NC units by five (Bendix, GE, Giddings & Lewis with Numericord, Electronic Control Systems and Cincinnati)^[15][22]. According to Noble, the 105 consisted of 5 five-axis universal contour milling machines, 24 three-axis skin mills (large machines for cutting wing skins) and 76 three-axis profile mills^[4].

Bendix and the Parsons patents

On the NC unit side, the aircraft electrical equipment company Bendix Aviation bought exclusive rights to the Parsons patents around 1956 and was supplying NC units for Kearney & Trecker machines under Air Force contracts^[15][16]. Bendix also sold a tape-preparation system using its own G-15 computer for about $70,000 (it could also be rented for $1,900 a month)^[15]. This shows that “tools for making tapes” had become products just as much as the machines themselves.

Burgmaster: a drilling machine that changes tools with a turret

Apart from the big aircraft machines, NC for drilling spread too. Burgmaster of Gardena, California built drilling machines that changed tools by rotating a turret (a rotating tool holder) carrying six or eight spindles. In an August 1958 advertisement it promoted a version fitted with GE numerical control, which performed drilling, tapping and boring in sequence from tape commands^[29]. According to Holland's study, the first Burgmaster with GE NC was sold to GE itself^[30]. The combination of positioning NC and automatic tool changing was just one step short of the machining center.

1958: The Milwaukee-Matic

The first to give that machining center a real form was Kearney & Trecker of Milwaukee (a milling machine maker founded in 1898)^[69]. Its president, Francis J. Trecker, brought in engineer Wallace E. Brainard from the aircraft company Hughes. Hughes had asked K&T to develop a line of three machines with Hughes-built NC units for machining parts, plus a tool changer, and Brainard moved to K&T with this contract in 1956. The Hughes NC units did not work well and the project was canceled, but Brainard led a K&T team that designed a machine combining the tool changer with GE's NC^[68][35].

On June 27, 1958, Brainard and four others filed a patent for a machine tool that changes tools automatically (granted on September 4, 1962 as U.S. Patent 3,052,011)^[28]. In the later lawsuit against Giddings & Lewis over this patent, the appeals court in October 1971 ruled against K&T, holding it improper that the former examiner had been used in the reissue proceedings^[47]. Then in mid-October 1958, at the AIEE (American Institute of Electrical Engineers) machine tool conference in Hartford, Connecticut, Brainard presented the Milwaukee-Matic. Its newest feature was a device that held about 30 tools in a circular tool magazine and automatically changed the needed tool into the spindle^[31][32].

It went on sale in December 1958^[34]. The Model II described in American Machinist ran on GE paper-tape NC; the spindle moved front-to-back, side-to-side and up-and-down (X 24, Y 21, Z 16 in), and an indexing table turned the workpiece to present the next face to the spindle^[33]. According to the same article, one tool change, in which the changer arm gripped the tool in the spindle and the next tool at once and swapped them, took 8.5 seconds^[33]. On a workpiece clamped once, milling, drilling, reaming, tapping and boring could be done in sequence while changing tools: the basic form of today's machining center was complete. It sold well, too, reaching 55 orders and 30 deliveries just over a year after launch^[34]. The Japanese book Nihon no Gijutsu 100-nen (100 Years of Japanese Technology) calls the Milwaukee-Matic the world's first machining center and gives the year as 1957^[24] (in trade journals of the time, it was unveiled in October 1958 and the patent was filed in June of the same year).

The name “machining center”

In January 1959, K&T called the Milwaukee-Matic “a flexible automated machining center” in a trade journal advertisement^[36]. From here “machining center” spread as the name of a type of machine. However, the phrase itself had been used since the early 1950s to mean “a machining area in a factory”, so we cannot say K&T was the first to use it.

What changed on the shop floor. Until then, drilling, milling and tapping a single part usually meant carrying the workpiece around to separate machines. Every move meant another setup, and every re-clamping shifted the position. The machining center cut both the number of setups and clamping errors by “changing the tools without moving the workpiece”. Today's ATC, tool numbers (T codes) and tool length offsets on an MC are refinements of this idea (our machining center guide).

Standards and DNC (1959–1972)

Building-block NC: Mark Century

NC units of the 1950s were expensive and custom-built for each machine. In 1960 GE advertised its NC unit Mark Century, promoting that contouring programs could be made without a computer and that functions such as positioning and contouring could be chosen and combined like building blocks^[38]. NC units became “something you could buy as a component”, and more machine tool makers began fitting them to their own machines and selling them. MIT's Servomechanisms Laboratory also changed its name to the Electronic Systems Laboratory in 1959^[1].

EIA RS-274: the origin of G-code

As NC spread, tape formats that differed from machine to machine and unit to unit became a problem. In January 1963 the Electronic Industries Association (EIA) adopted RS-273 for positioning and RS-274 for contouring^[39][40]. Both use the “word address” format, in which a block (one chunk of commands) is made by lining up address letters plus numbers, such as “G01” and “X12.5”, and the block length can vary freely. This is the origin of today's G-codes and M-codes. RS-274 was revised as RS-274-A in 1965 and RS-274-B in 1967, and RS-274-B became a U.S. national standard (ANSI X8.2) in 1968^[39].

Reading old programs. On tapes of the RS-274 era, the block number (N), preparatory function (G), coordinates (X, Y, Z), feed (F), spindle speed (S), tool (T) and miscellaneous function (M) appear in a fixed order. Writing in the same order is still easiest to read on today's machines because of this legacy. Puzzling over coordinates without a decimal point (is “X125” 12.5 mm or 0.125 in?) is also a leftover from the tape era (Decimal point input).

DNC: one computer runs many machines

In the late 1960s came DNC (Direct Numerical Control), which sent commands from a computer directly to NC machines without tape. The background was the arrival of cheap small computers (minicomputers). At the Chicago machine tool show that closed on October 2, 1970, DNC was the main attraction, with more than 70 DNC systems demonstrated^[46].

• GE's CommanDir: two computers running 10 machines (up to 30 possible). Built by GE's manufacturing automation operation in Charlottesville, Virginia^[46]
• Sundstrand's Omnicontrol: used an IBM 360/30 or a DEC PDP-11 and ran six machines at the show (said to handle up to 256 in all)^[46]
• Cincinnati Milacron's DNC: a configuration that could start small and expand, up to 64 machines^[46]
• Kearney & Trecker's System Gemini: up to eight machines on one Hewlett-Packard 2114B minicomputer; an IBM 1800 could link several minicomputers^[46]

The same article quotes a GE man saying that, compared with the NC machines that debuted at the 1960 show and kept breaking down, the 1970 DNC systems “worked well”^[46]. Vacuum tubes had given way to transistors and integrated circuits, and NC reliability had improved. Cincinnati Milling Machine had changed its name to Cincinnati Milacron in January of that year^[45].

The seeds of FMS

The idea of linking machines by computer moved on to the FMS (flexible manufacturing system), which also includes workpiece transport. A 1974 RAND report cites K&T's System Gemini and Sundstrand's Omni-Control as examples of a computer directly running several NC machines, and Molins' System 24 in Britain as an example that linked even transport by computer^[53] (System 24 is covered in the British chapter “NC, FMS, Industrial Decline and Revival”). At K&T, Brainard worked on computer-based production systems such as Gemini and one for Allis-Chalmers before retiring in 1969^[35]. K&T also showed the TURN-12, a multitasking machine that could also turn, at the 1970 Chicago show, but only three were built and it never became a product^[43] (Japanese Archive Chapter 7).

CNC, robots and the import wave (1973–1991)

NC that was slow to spread

NC began with Air Force backing, but it spread slowly in private factories. In 1972 NC machine shipments were worth $167 million, 15% of metalworking machinery shipments. In 1973 an estimated 21,000 NC machines were in use in U.S. factories, about 2,400 of them owned by the government^[12]. A 1983 report by the U.S. International Trade Commission (USITC) says that after commercial NC machines appeared in the mid-1950s it took nearly 20 years for them to be accepted, and that even in 1976 only 1% of the machine tools in use in the U.S. were NC. What held back their spread was the price and size of NC units^[56].

CNC and RS-274-D

Integrated circuits made NC units smaller and cheaper, and the next step was to put a computer inside the NC unit: CNC (Computer Numerical Control). In the U.S., Cincinnati Milacron's controls division brought out its first CNC in 1973, bringing a CRT screen onto the shop floor for the first time^[19]. A USITC report describes CNC as one in which "a small computer stores the program and supplements or replaces the NC unit," and lists among its advantages that it adapts easily to a variety of machines and makes programs easy to create and recall^[56]. The tape format too was unified in April 1974 by RS-274-C, which merged the separate standards for positioning, straight-line cutting and contouring into one^[39], and RS-274-D, completed in 1979 (the document is dated February 8, 1980), is still the basis of programs on many CNCs today^[59][58]. In 1982 NC was still only about 2.2% of the machine tools in use in the U.S., but NC machines accounted for 36% of that year's machine tool consumption by value^[56]. Though still few in number, NC had already become the norm for expensive machines. In American Machinist's equipment survey the next year, 1983, NC machines numbered about 103,000 (about 4.7%) of the roughly 2.19 million machine tools in metalworking, double the approximately 52,900 of the previous survey in 1976–78^[62][63].

T3: a robot run by a minicomputer

Computer control of machines spread to robots too. Cincinnati Milacron entered industrial robots in 1974 with the T3, controlled by a minicomputer. Until 1982, Unimation, the earlier entrant, and Cincinnati Milacron together held more than half of the U.S. robot market^[49][50] (the T3 was developed in 1973 and launched in 1974^[51][52]). Machine tool makers had entered an age in which they built even robots to load and unload workpieces.

Mergers and the import wave

From the end of the 1970s, U.S. machine tool makers faced a wave of mergers and imports. Kearney & Trecker, which had created the machining center, merged with the transfer-machine maker Cross Company in 1979 to become Cross & Trecker^[57], and in 1991 it was bought by Giddings & Lewis^[66]. In NC units, Allen-Bradley agreed in 1982 on a joint venture combining Olivetti's NC division in Italy (OSAI) with its European business^[56].

According to the USITC report, imports' share of U.S. machine tool consumption rose from 17% in 1977 to nearly 30% in 1982, and imports from Japan grew from $105 million to $535 million^[56]. On March 10, 1983, the National Machine Tool Builders' Association (NMTBA) petitioned for a Section 232 investigation seeking import restrictions on national security grounds^[64], and in December 1986 Japan and Taiwan agreed to voluntarily restrain exports of machining centers, NC lathes and other machines to the U.S. for five years from 1987^[65][60]. The country that created NC had become a country troubled by imported NC machines. This story is traced in detail in the next chapter, “The Import Wave and Revival”.

Record-playback or numbers: Noble's question

There is a well-known critique in the history of NC. In a 1978 paper and his 1984 book Forces of Production, the historian David F. Noble argued that there had been another path to automating machine tools besides NC^[71][4].

That path was record-playback (also called teach-and-playback). The movements of a skilled machinist cutting the first part are recorded on magnetic tape, and from the second part on, the tape is played back to repeat the same movements. GE researched this method from the late 1940s and made it a product in 1954. GE later called this line the Mark series^[11][4]. With record-playback, the skilled worker on the shop floor makes the program himself, and neither computers nor mathematicians are needed.

NC, by contrast, represents the part shape as equations and numbers, and programs are made in the engineering department (and later by computer). Noble argued that both methods were technically viable, and that NC was chosen because managers and the Air Force preferred a method that moved programming from the shop floor to the engineering department, and control of the factory from skilled workers and unions to management. The Air Force's funding of MIT's development work and of NC machine purchases, he said, set the direction of the technology^[71][4]. Noble also describes in detail an attempt at a GE aircraft engine plant in which shop-floor workers tried to manage the operation of NC machines themselves (the Pilot Program)^[4].

There is also an account from the insiders' side. J. Francis Reintjes, who directed the MIT laboratory from 1953^[1], wrote a book on the course of NC's development in 1991^[72]. Read alongside Noble's book, the same events look different. This chapter does not decide which reading is correct. But NC was born as a technology to “replace the craftsman's skill with numbers”, and how to make that replacement was debated from the start, a point that still speaks to today's shop floor. Whether to use the tool paths that CAM produces or a short program written by hand by a veteran: the roots of that question lie in the 1950s.

Links with Japan

Japanese NC started with one magazine

NC in Japan began with reading MIT's reports. According to a Tohoku University paper on FANUC's growth, in October 1956, at an automatic control conference held at Waseda University, Yasundo Takahashi of the University of California showed Seiuemon Inaba of Fuji Tsushinki Seizo (the forerunner of Fujitsu) a copy of Scientific American that introduced the NC milling machine. Inaba recalled that the MIT reports became the “bible” of their research for some time^[21]. The paper does not give the article's title, but from the timing and content it appears to be Pease's article in the September 1952 issue^[9]. In 1956 Fuji Tsushinki Seizo became the first private company to succeed in developing NC and servos, and in 1958 it delivered its first commercial unit to Makino Milling Machine, creating Japan's first NC milling machine^[18][26] (Japanese Archive, Chapter 2).

Universities moved early too. In February 1957 the Precision Engineering Research Laboratory of the Tokyo Institute of Technology succeeded in building a prototype NC lathe, the first NC lathe in Japan^[24][25] (Japanese Archive, Chapter 1). For machining centers, a Hitachi group built an NC universal machine in 1960 with an ATC (8 tools) that swapped tools with two arms; Nihon no Gijutsu 100-nen counts it as Japan's first domestic MC and says Japanese engineers were thinking about MCs at almost the same time as the Americans^[24]. Niigata Engineering's MC 50DIU of 1968 was developed in-house after also studying information on Sundstrand's “Omnimill” MC from the U.S., which was then attracting attention^[42].

Open-loop pulse motors

Fujitsu also chose a path different from the Americans'. In 1959 it completed an electro-hydraulic pulse motor^[18], and pursued open-loop NC without a position detector. Whereas most American NC was closed-loop, using DC motors and detectors, Fujitsu made the opposite choice, which made the equipment cheap and simple^[21]. In 1972 Fujitsu's NC division became independent as Fujitsu FANUC and announced a CNC. Around this time GE and Fujitsu are each said to have been making about 2,000 NC units a year^[18][19]. In 1974, under a contract with the American company Gettys, it began making and selling DC servo motors, and president Inaba received an award from the Numerical Control Society in the U.S.^[18].

Exports, and GE Fanuc

In 1970 the MTC-2500R NC lathe from Yamazaki Iron Works (now Yamazaki Mazak) became the first Japanese-built NC lathe to be exported to the United States^[44]. In 1975 Matsuura Machinery began exporting its MC-750V vertical MC to the United States^[54][55]. FANUC set up FANUC America in the U.S. in 1977, and in 1982 formed GMFanuc Robotics, a joint venture with General Motors^[18]. In the same year, 1982, Japan's machine tool production overtook that of the United States to become the world's largest^[60][61].

Then in 1986 GE and FANUC formed GE Fanuc Automation, owned 50/50. Its headquarters were in Charlottesville, Virginia^[21]. That was the town where GE's manufacturing automation operation, which had built the CommanDir DNC in 1970, was located^[46]. The Japanese company that had started NC by reading an MIT article had, 30 years later, formed an equal joint venture with GE, one of the original homes of NC. The joint venture was dissolved in 2009^[18].

Japan's “first” and America's “first” in NC. From America's first NC milling machine (MIT, March 1952) to Japan's first NC lathe (Tokyo Institute of Technology, February 1957) and NC milling machine (Makino and Fujitsu, 1958) took about 5 to 6 years, and from the Milwaukee-Matic (October 1958) to the Hitachi group's NC machine with ATC (1960) about 2 years. However, the year of the “first” changes depending on whether you count prototypes, products or volume production. See also the notes in the Japanese Archive's Chapter 1 and Chapter 2.

FAQ

Who invented NC (numerical control)?

The idea of moving machine tools by numbers came from John T. Parsons and Frank L. Stulen (Parsons Corp.), who signed a development contract with the U.S. Air Force in 1949 and filed a patent for a positioning apparatus in 1952 (U.S. Patent 2,820,187 in 1958). The first NC milling machine that moved three axes simultaneously and continuously was built by MIT's Servomechanisms Laboratory and completed in March 1952. The sources fit best if you think of the idea as Parsons' and the actual contouring machine as MIT's.

What was the world's first machining center?

The first known NC machine that changed tools automatically from a tool magazine is the Milwaukee-Matic of Kearney & Trecker (Milwaukee). It was unveiled at the AIEE machine tool conference in October 1958 and went on sale in December. It held about 30 tools and, with an indexing table turning the workpiece, could do milling, drilling, reaming, tapping and boring on one machine.

When was G-code established?

Today's format of lining up address letters and numbers goes back to the paper-tape standards RS-274 (for contouring) and RS-273 (for positioning), adopted by the Electronic Industries Association (EIA) in January 1963. They were later revised as RS-274-A (1965), RS-274-B (1967) and RS-274-C (1974), and RS-274-D, completed in 1979, is still the basis of programs on many CNCs today.

What is APT? How is it related to today's CAM?

APT (Automatically Programmed Tools) is an NC programming language developed from 1956 by Douglas T. Ross and others at MIT. You write the part shape and the cutter motion in near-natural language, and the computer calculates the tool path and produces the tape. Aircraft companies developed it jointly from 1957, and it was unveiled at MIT in February 1959. The idea of having a computer calculate tool paths from geometry lives on directly in today's CAM.