Talk:Integrated circuit

From Wikipedia, the free encyclopedia

WikiProject on Electronics This article is part of WikiProject Electronics, an attempt to provide a standard approach to writing articles about electronics on Wikipedia. If you would like to participate, you can choose to edit the article attached to this page, or visit the project page, where you can join the project and see a list of open tasks.
B This article has been rated as B-Class on the quality scale.
High This article has been rated as High-importance on the importance scale.
This article has been reviewed by the Version 1.0 Editorial Team.

I'm not sure I'd characterize the 4000 series as the successors to 7400 line. My understanding was that there were uses where the 7400 would be preferable and others where the 4000 would be preferable. And when CMOS really did eclipse TTL, both of these were commonly getting replaced with ASICs. Ckape 06:44, 23 Dec 2003 (UTC)

Contents

[edit] Naming conventions

Any information on why a particular IC, say the 555 timer IC or any of the 74 ICs, is named so would be appreciated. Is it purely chosen by the manufacturer or is there some convention involved?

That is a great questions. The manufacturers, having technical backgrounds, tended to assign numbers as the names of products, such as Intel 40065, 386, Texas Instrument 5408 or 7409 series, Fairchild F8, etc. Some companies only selected prime numbers such as the Teradyne J937 (a semiconductor memory tester), etc. Since the Bell Labs studies which proved that a phone number should have no more than 7 digits to be easily remembereds, the manufacturers tended to select product names with fewer than 7 digits, as well. But a successful product meant that the product name needed to be kept, so a 4-digit or 5-digit product name, such as the 6502, tended to be the end of the line. The Motorola 6800 expanded to the 68000, then 68010, 68020, etc. The Intel 4004, 8008, 8080, 8088, 8086, 80186, 80286, 80386, 80486 was going to be usurped by an AMD 586. So the 586 series was renamed to a protectable name, "Pentium (TM)", breaking the numeric series, but protecting the brand. HP's audio oscillator, the 101, used by Disney in the production of Fantasia in the 1930s, was so numbered because Hewlett and Packard wanted it to sound like there was a full product line, so they started numbering from 100 instead of 1. Ancheta Wis 17:06, 26 Mar 2005 (UTC)
So the quick answer is: Convention only, under the complete control of the manufacturer, but using the Intellectual Property rights available at the time of production or invention.
The use of standards by some consortium, say under some API or programming language standard or the like is not in use at this time. The ISO networking layers, for example, were overkill, and were ignored by the development of real protocols like TCP/IP. Ancheta Wis 17:06, 26 Mar 2005 (UTC)
someone I asked said the 555 timer IC was called so because of the internal potential divider formed by three 5K resistors. Is this correct?
I can't answer that. My coworkers liked them because they were easy to use, which would be the only reason to remember the product. Ancheta Wis 20:16, 26 Mar 2005 (UTC)
The JEDEC assigned a 4 digit number to industrial tubes starting with 5000, and going up to the 7000 by the time that tubes started to go out of favor, Transistors were numbered based on an older tube numbering system which resulted in the 2N0000 style numbers. It is likly that this may have prompted firms to use 4 digit numbers in picking names for IC products.cmacd 13:26, 16 March 2007 (UTC)

[edit] lowercase i

Since alot of devices have a name that starts with a lowercase I (e.g. iMac, iPod, ix86, etc.); does the I stand for integrated for those devices? --SuperDude 22:21, 16 Jun 2005 (UTC)

That's likely in many cases, although the 'i' in Apple products you mentioned is a reference to the fact that they are Internet oriented products. --Nbettencourt

[edit] The principle of dielectric isolation

The text says "Noyce credited Kurt Lehovec of Sprague Electric for the principle of dielectric isolation caused by the action of a p-n junction (the diode) as a key concept behind the IC". I've never heard of that principle, and a search using Google only retrieves exactly the same sentence. It seems to me to make little sense - dielectrics are insulators, and in IC fab we would expect dielectric isolation to be caused by an oxide, nitride or similar layer on top of a conducting layer. Can anyone shed light on whether this principle really exists? --Phil Holmes 10:46, 30 August 2005 (UTC)

Probably a patent A Google search for "Kurt Lehovec Sprague Electric" yields US patent 3 029 366 awarded on April 10, 1962, filed April 22, 1959.. Lehovec said "I never got a dime out of it (the patent)." The concept might be peculiar to bipolar transistor technology, which Fairchild exploited, of course. Ancheta Wis 12:22, 30 August 2005 (UTC) Lehovec's innovation was to consider the operation of a reverse-biased p-n junction, not merely the electrostatics of materials forms, which appears to be where you are taking this.
Kurt Lehovec's scientific/technical autobiography: LED, Solid Electrolyte Battery, MOS Surface States, p-n junction isolation, and more!
Robert Noyce credits Lehovec here: "Microelectronics", Scientific American, September 1977 Volume 23, Number 3, pp. 63-9. I will add this to the article unless someone beats me to it. Ancheta Wis 02:39, 31 August 2005 (UTC)
Lehovec deserves recognition and monetary reward for his role in the IC revolution; Kilby got a Nobel prize. We have one founder of the IC revolution still alive. I guess that means I owe Kurt Lehovec an article in Wikipedia.

Just done the search and come up with http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=HITOFF&d=PALL&p=1&u=/netahtml/srchnum.htm&r=1&f=G&l=50&s1=3,029,366.WKU.&OS=PN/3,029,366&RS=PN/3,029,366 which is the full text of the patent. I've OCRed it and it only mentions dielectric once - in a piece on capacitors. Whilst I accept that a pn junction can function as a dielectric, I maintain that this is unusual usage. It seems to me that Noyce actually credits Lehovec with the invention of pn junction isolation in ICs.

I've tried to find the Scientific American article in full and failed - the best I can find is: "In an article entitled ""Microelectronics", published in "Scientific American", September 1977 Volume 23, Number 3, pp. 63-9, he wrote:

"The integrated circuit, as we conceived and developed it at Fairchild Semiconductor in 1959, accomplishes the separation and interconnection of transistors and other circuit elements electrically rather than physically. The separation is accomplished by introducing pn diodes, or rectifiers, which allow current to flow in only one direction. The technique was patented by Kurt Lehovec at the Sprague Electric Company"." at http://www.virtuallystrange.net/ufo/updates/1997/sep/m01-033.shtml

Again - no mention of the principle of dielectric isolation. --Phil Holmes 20:11, 31 August 2005 (UTC)

I can change the sentence to "p-n junction isolation" per your note, or you can, for making the catch. Ancheta Wis 21:22, 31 August 2005 (UTC)
Something like "Noyce credited Kurt Lehovec of Sprague Electric for the principle of p-n junction isolation caused by a biased p-n junction (the diode) as a key concept behind the IC."


To be fair, Jean Hoerni deserves equal billing. Ancheta Wis 11:28, 1 September 2005 (UTC)

[edit] The influence of silicon dioxide

One of the fundamental features of silicon as a semiconductor is that it is easy to grow SiO2 on it, and that this glass layer forms a reliable barrier to dopants, and is (relatively) easily etched with good fidelity. I'm convinced that this is one of the main reasons why silicon dominates IC fab (tho' there are others - mechanical strength and convenient band gap included). I feeel strongly tempted to write a "why silicon" section, but I'm not sure where to put it. It could go somewhere in this article (but where?) or in the fabrication article, or somewhere else. Any thoughts? --Phil Holmes 08:54, 7 September 2005 (UTC)

Why silicon would fit naturally in the fabrication article. Ancheta Wis 22:23, 7 September 2005 (UTC)

[edit] SSI/MSI/LSI

Were these the names given at the time? The article makes it sound like that, though I doubt it. The MSIs were probably considered very high-tech at the time, not "medium," just like our most advanced CPUs today will seem pathetic in a few years. Could that be clarified? Twilight Realm 02:08, 27 September 2005 (UTC)

They were named at the time that LSI was the state of the art. But when VLSI (very large-scale ICs) was the state of the art, the projected masks could no longer be checked by tacking their images on the walls during the verification stage, as they were simply too large. Today verification is all done by computer anyway, with CAD programs. --Ancheta Wis 16:06, 5 November 2005 (UTC)

[edit] Etymology

Something that has puzzled me for a long time: why are these slithers of silicon called dice when they don't at all look like dice? Please help to unpuzzle me! Thanks, --Maikel 16:04, 2 November 2005 (UTC)

Slice and dice is a common phrase; it simply refers to the act of cutting or dicing vegetables, like cucumbers or onions. Dice is the plural of the noun die, which need not refer to the gambling objects. Silicon boules, the long cylindrical ingots are first sliced into wafers, and then, after the fabrication process, then diced into rectangular die. If you look at an IC (out of its package), you can see the die. --Ancheta Wis 12:09, 5 November 2005 (UTC)
Wow, Wikipedia DOES provide useful knowledge after all! ;-) Thanks! --Maikel 13:34, 5 November 2005 (UTC)

[edit] Artwork

The term artwork has been in use on engineering drawings for electronic circuits, including printed circuits, since before the invention of the IC. I object to the use of the phrase 'silicon artwork' solely for those constructs which might be construed as subject to artistic license. Hence I reverted. Real estate on a chip is valuable and subject to signoff. A chip such as a microprocessor is constructed by teams of hundreds of people. It would be difficult to get by a review. It is true that custom chips might be built by only a few minds, and that it is cheaper to build chips which might share a wafer in a custom run, and that it would then be possible to permit artistic license, but the occurrence would definitely not be on an industrial scale. --Ancheta Wis 22:03, 18 February 2006 (UTC)

A far better thing to do would have been to reword the bit to be acceptable to you rather than reverting it. I don't care for it much myself, but I moved it here from CPU where it really didn't belong. On WP, in general you should reword rather than revert. -- Gnetwerker 23:08, 18 February 2006 (UTC)

An already existing page Chip art, covers this topic more thoroughly than Silicon Doodling. Hence I replaced Silicon Doodling with a re-direct, and changed the link. LouScheffer 02:18, 19 February 2006 (UTC)

Completely appropriate -- thank you. -- Gnetwerker 03:05, 19 February 2006 (UTC)

[edit] Die or Dice

Due to the line "Each device is tested before packaging. The wafer is then cut into small rectangles called 'die'. Each die is then connected into a package using aluminium (or occasionally gold) wires which are welded to pads, usually found around the edge of the die". Actually it is die or dice?

Dice is singular, die is plural. Die is singular, dice is plural. This was always honored at Intel during 80s and 90s, and doubt it has changed. -- Gnetwerker 15:59, 28 February 2006 (UTC)
Wrong way round : see [1] -- Alf Boggis [[User_talk:Alf_Boggis|(talk)]] 12:11, 1 March 2006 (UTC)
Duh, yes of course, I fat-fingered it when typing. You are right. I fixed it above. -- Gnetwerker 18:15, 1 March 2006 (UTC)
Both wrong: "dice" is plural, "douse" is singular :-). But "die" is standard terminology, corrupted in semicon by non-native English speakers, so that you may quite often read "dice" as singular. mikka (t) 18:23, 1 March 2006 (UTC)

The definition die=singular; dice=plural is correct. However, the original sentence wasn't good English - imagine if it had said "cut into small rectangles called square" - it should be "cut into small rectangles called squares". So I've made it "dice" in that sentence and put a small parenthetic comment about die/dice. --Phil Holmes 13:07, 5 March 2006 (UTC)

Dice is clearly the plural form of die when you're playing Craps or Yahtzee, but not when you're talking about integrated circuits cut from a wafer. According to several dictionary entries [2] and even other wikipedia articles [3] [4], the correct plural form of die is "dies" in this context. My best explanation for this is that the term "die" was probably borrowed from the metalworking industry, where repeated stamps or punches are commonly called "dies". ~ Fanblade 20:52, 24 April 2006 (UTC)

Your explanation is plausible, but I stand by my reportage of actual usage in the chip industry, above. This is, of course "original research", so I won't edit based upon it. :-) -- Gnetwerker 06:20, 25 April 2006 (UTC)
I used to work in a clean room making ICs and packaging them. In our work we always referred to the diced components as dice, not dies. I believe dies is incorrect, and the theory that is comes from the device used to cut metal components (as in "tap and die") is also incorrect. Citing other articles that make the same mistake doesn't justify the error. However, I guess this is unlikely to be resolved, and may be an instance of local usage (I'm from the UK). It seems to me the best is to allow both, so I'll edit to that effect. --Phil Holmes 12:05, 30 April 2006 (UTC)

[edit] Definitions: IC, hybrid circuit, multi-chip module

It is important not to spread meanings beyond usefulness. An IC, per history of the term, consists of a monolithic circuit. The current dual-core CPUs still qualify because they still use the same silicon; the multi-die flash memories do not qualify because they are separate chips in a hybrid package. Same with the flash memories packaged a separate USB interface chip -- its hybrid packaging.

When matched transistors are an important consideration, the processing of an IC can provide intrinsically matched components without preselection. There is no point in calling tube circuit an IC. This just makes the term a curiosity.

The historical use of a separate glass assembly to hold multiple filaments calls to mind the use of chimneys to vent multiple flues, in Europe, thus avoiding the chimney tax. Interesting, but of small economic importance, compared to the huge stream of improvements on the IC.

When you examine the history of the hybrid circuit, look at the packaging. There are any number of important technologies, like the magnetic bubble memory or the multi-chip module etc which echo the concept of the blivet; interesting, but not in the main stream.

If you like, we can work on the prose in hybrid circuit to include history like the magnetic bubble memory or the multi-chip module which were tried and discarded by major manufacturers like IBM. But a triode/tetrode/vacuum tube etc just isn't part of an article about the IC. --Ancheta Wis 21:45, 3 July 2006 (UTC)

I disagree with the whole concept of this. The article has been written about one very specific type of integrated circuit - what is often described today as a monolithic integrated circuit. The term 'Integrated Circuit' existed before these devices came into being. Indeed the circuit shown on the illustration in the hybrid circuit article was (and is still entitled to be) called an integrated circuit. When the newer types appeared, the older types were called 'thick film integrated circuits' and the newer types were known as 'thin film integrated circuits', since they used a process similar to that used for producing thin film transistors (an early type of the variety used in LCD displays). Indeed, surface mount components, popular today were also around at the time when thick film integrated circuits were popular.
I have seen the Loewe multi valve device in several museums around Europe and in at least one (It may have been the Deutse Museum, Munich), it was described as the world's first integrated circuit. It is precisely that - in integration of parts to make a functional modular circuit. To be pedantic, a printed wiring board with components on is a 'integrated circuit' but that is stretching the point too far.
The contributor who added the Loewe device to the article is quite correct. In my view he (or she) has not gone far enough, in that the article now seems to describe the beginning and end of the integrated circuit story, but completely omits the middle. For example, there were several integrated circuits produced in the heyday of the valve computer era for use in analogue computers - plug in operational amplifiers, though they were spoilt by having a removeable (and replaceable) valve. It is fair to say that the term 'integrated circuit' itself only seems to have been coined for the thick film devices (the manufacturers' literature described them as exactly that), but the concept was far from new, though the term 'module' was often used until something better came along.20.133.0.14 08:14, 4 July 2006 (UTC)
I agree that the Loewe device is a module. What I am objecting to is the broadening of the usage of the term IC. An IC is a unitary component. It came about after a thorough understanding of the physics stemming from Dirac's sea of states whereas the tetrodes stem from the electron gas. The electron gas had to be continuously generated by thermionic emission (heating of an electrode) but the sea of states are quantum mechanical. The term IC evokes the use of semiconductor materials, whereas the tetrode came about from a different physical mechanism.
I agree that William Shockley is indebted to the triode/tetrode/vacuum tube. He explicitly writes that he was attempting to create a solid state vacuum tube. This discussion really belongs to the history of electronics and the encyclopedia needs such an article. --Ancheta Wis 17:54, 4 July 2006 (UTC)
After visting Wikipedia:WikiProject Electronics I see that the vacuum tube is neglected. Semiconductor device#History of semiconductor device development has some of the history but the triode/tetrode/vacuum tube are not discussed. Thus there appears to be room for a history article. Anyone interested? --Ancheta Wis 01:31, 5 July 2006 (UTC)
History of electrical engineering treats the triode / tetrode / vacuum tube technology in only a few sentences. More discussion of analog circuits, for example in its use in radio engineering, sound reproduction, cinema, and television could be expanded, for example.
In seeking a home for the anon. contribution, several of us have added it to vacuum tube.

[edit] Stable version now

Let's begin the discussion per the protocol. --Ancheta Wis 05:37, 11 July 2006 (UTC)

  • Oppose, and not only due to my opposition to the process. The article has a reference section, but no context as to what it's sourcing. How can it be stable if someone comes up with new information or sources their additions differently, for instance? --badlydrawnjeff talk 18:18, 11 July 2006 (UTC)
  • Oppose -- how can we even start to discuss stabilizing an article if we aren't confident enough to call it good, let alone featured? Why not continue pursuing good and featured articles, and then make those "certified" versions easy to link to? I feel like, especially with articles like this one which could use attention from the Wiki community, making a "stable" version seems to imply, "no more work necessary here!" I just don't think that's accurate with all but the very best of our articles. JDoorjam Talk 21:37, 11 July 2006 (UTC)
  • Oppose. This proposal is just about dead, it would need to achieve consensus before any action could be taken under it - it has achieved the opposite. WP:CONSENSUS, WP:PROT, WP:5P all individually (never mind collectively) trump a still-under-discussion-but-largely-rejected proposal. Cynical 20:29, 30 July 2006 (UTC)
  • Oppose. A "stable Wikipedia article" is a contradiction in terms. If it was a collection of stable (aka "dead") articles, it wouldn't be Wikipedia any more. And this particular article should at least make it to FAC status first. --Wtshymanski 01:24, 1 August 2006 (UTC)

[edit] The cleanup notice

I mostly added it because the sections "Advances in integrated circuits" and "Classification and complexity" in length describes the same things. →AzaToth 02:44, 12 October 2006 (UTC)

Thank you for your note. I moved the most similar paragraphs under the same header. Everyone, please help by removing redundant sentences. --Ancheta Wis 02:57, 12 October 2006 (UTC)
The cleanup notice can probably be removed now. --Ancheta Wis 09:06, 30 December 2006 (UTC)

[edit] Advances in integrated circuits

The section Advances in integrated circuits should take into account emerging information represented by this news article, but no doubt related in greater detail and authority elsewhere: Markoff, John. "Intel Says Chips Will Run Faster, Using Less Power", The New York Times, 27 January 2007. Retrieved on 2007-01-27.  --User:Ceyockey (talk to me) 14:04, 27 January 2007 (UTC)

Another advance, currently underway, is the creation of 3D ICs. This has been added to the ITRS roadmap and is the subject of numerous articles and conferences. A whole collection of references is available at http://3d-ic.org/literature.html . The difficulty in creating a wiki on this is that the definition of 3D IC is still fluid. -- Gretchenpatti (talk) 15:53, 4 January 2008 (UTC)

Not having received any opposition to teh above, I added 3D-IC to the ULSI/WSI/SOC section. Gretchenpatti (talk) 17:53, 8 February 2008 (UTC)

[edit] Inventors and Patents

I note that the US patent 3138747, mentioned in the article, was not actually grated to Kilby, but to another TI engineer named Richard F. Steward. Is this a typo?

Furthermore, there is a German patent number 833366 which was granted to a Mr. Werner Jacobi on May 15th, 1952, describing the idea of manufacturing several amplifier stages on a single semiconductor by means of applying multiple electrodes to the surface of the semiconductor. It is a rather cursory patent, but nevertheless contains the basic idea of an integrated circuit, at about the same time when Dummer described the idea. Maybe it would be appropriate to include this in the list of inventors. --Stefan heinzmann 11:33, 9 February 2007 (UTC)

[edit] Design Rules and Mead and Conway

The text states: This step was largely made possible by the codification of "design rules" for the CMOS technology used in VLSI chips, which made production of working devices much more of a systematic endeavour. (See the 1980 landmark text by Carver Mead and Lynn Conway referenced below.)

This seems (to me) to imply that Mead and Conway introduced the idea of design rules.

I did not remove the reference to Mead & Conway and design rules, (later restored) but I think the removal was correct. The idea of design rules pre-dates Mead & Conway. We certainly used design rules in 1975, and Mead and Conway was published in 1980. What Mead and Conway did in this area was introduce simplified "lambda based" rules. This made it easier to get started, but was ultimately a dead end. All modern processes have rules specified in absolute terms, and include values that are weird multiples of any underlying lambda.

Does anyone else have a different view of this? LouScheffer 06:46, 19 June 2007 (UTC)

This whole section is bogus. Microprocessors do not need millions of transistors and were introduced in 1971. Mead and Conway is only relevant in terms of introducing more students to the field. Design rules were codified much earlier. Better tools and manufacturing were the keys. LouScheffer 06:56, 19 June 2007 (UTC)

I agree the whole section is bogus, since the early VLSI chips were all nMOS, and having design rules had little to do with them. I don't agree, however, that simplified design rules were a "dead end"; they are still serving their purpose, which is to simplify, to give students and others a way to do simplified portable design not tied to the detailed design rules of a fab. But, this polarization of opinion on their value is over 25 years old, and not really relevant here; I just thought I'd mention my POV on it. Dicklyon 07:08, 19 June 2007 (UTC)
Agreed that simplified rules have value, and maybe there should be an article on Lambda rules or something similar. But I don't think they were an enabler for the advances described in this section, except maybe indirectly by making it possible to educate enough designers to do the job. LouScheffer 08:05, 19 June 2007 (UTC)
I agree that "microprocessors do not need millions of transistors", since CPUs have been built entirely out of 4,100 (3-input) NOR gates.
I also agree that "lambda rules" should be mentioned in this article because of their historical and educational use, even if no one today still uses them.
If "lambda based" rules are a "dead end", why does the Magic (software) claim that Magic is still "widely used"? (Magic only uses lambda-based design rules. It can't handle "weird multiples of ... lambda"). --68.0.124.33 (talk) 19:11, 22 February 2008 (UTC)

[edit] Resistive structures

"Resistive structures, meandering stripes of varying lengths, form the loads on the circuit." This sentence is a bit ambiguous to me. Exactly what does this mean. Does this mean that these structures are the only structures that causes electric load in the IC? --Renier Maritz (talk) 00:16, 13 March 2008 (UTC)

[edit] Gates v. Transistors

How many logic gates are implemented in a typical VLSI chip? Is it half the number of transistors? Anwar (talk) 18:35, 10 May 2008 (UTC)

According to an an EDN article "Gate count as a measure of logic capacity has its roots in ASICs, in which the technique is reasonably straightforward. A gate-array ASIC's logic structure comprises a sea of fine-grained, two-input NAND gates." It takes four transistors to make a two-input CMOS NAND gate.
The problem is that VLSI chips contain many logic elements that are not two-input NANDs, so someone must decide how many two-input NANDs are equivalent to each logic element, how many transistors there are, and what the average number of transtistors per equivalent two-input NAND is. Various companies can come up with various counts. --Gerry Ashton (talk) 19:25, 10 May 2008 (UTC)