Talk:MOSFET

Totally Useful Article

As a board and ASIC level designer engineer for the past three decades, I think this article provides necessary information and is a good reference for new and experienced designers. It is definitely not for semiconductor physicists, because they will be way above this article, however it likely will be above non-technical persons who likely won't visit this page anyway or who won't need to read more than the introductory paragraphs. MOSFETs are complicated devices and if you don't know at least what is given in this article then you very likely won't be able to read the data sheets of the various types of discrete MOSFETs or of ICs whose interface outputs/inputs are made of such MOSFETs in order to design proper circuits. This article is a good and easily accessible resource for new designers or old designers - who may only do one or two designs every year or two. Of course, it's also a similarly useful resource for engineering, technology and technologist students to use in addition to their textbooks, which may provide more or less information. So, please don't dumb this article down. Indeed, it should be improved to present the information needed in the best way possible to allow these types of people to do real work. — Preceding unsigned comment added by 207.107.66.194 (talk) 22:58, 19 December 2016 (UTC)

• While I agree with what you are saying, the argument can (and perhaps should) be made that wikipedia is not for people who are designers, but a general resource. Designers should perhaps consult other sources (textbooks on the matter, for example). Hence, it might make sense to simplify the article at least somewhat.

TheUnnamedNewbie (talk) 19:12, 4 January 2017 (UTC)

As an uninitiated seeker of knowledge about MOSFETs, I have to agree with TheUnnamedNewbie. The "design engineer" has been indoctrinated (or perhaps innoculated is a better term) with the terminology and concepts needed to understand this article. I find articles that use terms not previously defined to be near useless. If the author decides not to define their terms, this is the most judicious time to use links to articles where those terms are defined.

I hope my reasoning is not found to be out of line. Dwightfowler (talk) 18:26, 1 May 2024 (UTC)

analog

The article says that analog electronics is mostly MOSFET. Not quite sure what analog is mostly now, I wonder about stereo amplifiers. Since mine isn't so new, it might not be representative. Are stereo amplifiers now mostly MOSFET? Gah4 (talk) 12:25, 19 May 2020 (UTC)

In 2020, probably so. I think most are class-D switching amplifiers these days (in every phone, for example), and even high-end class-AB amps use mosfets (see this video). Dicklyon (talk) 17:53, 4 July 2022 (UTC)

bipolar

The article seems to describe how much better MOS is than bipolar. But, other than main memory, computer technology was bipolar for about 10 years (a long time in computer years) after MOS ICs became available. Looking back, we can see the path technology took, but it wasn't so easy to see at that time. Dennard scaling is nice, but the technology to do it took some time, and even then wasn't always used. The operating voltage of a bipolar transistor depends on the energy gap, not on its size. In the early years, MOS, and especially CMOS, was much slower than TTL and ECL. MOS took over main memory, replacing magnetic cores, many years before it took over large system CPUs. Gah4 (talk) 18:09, 19 May 2020 (UTC)

Yes, I was thinking we needed more about the transition. Initially it was far from obvious MOS would win, there were big problems with stability that needed to be solved, and it was many times slower than bipolar. Bell Labs wasn't very interested. Moore (as in Moore's Law) took time to be convinced. One of the sources I was reading was saying how IBM actually tried to cancel their MOS program on multiple occasions, but the individual engineers and researchers resisted, and the management never quite followed through. It was the advantages of MOS in mass production and integrated circuits which eventually seems to have won the battle, but it took years. At the moment the article rather implies that it took over the world following Atalla and Kahng's report, and it wasn't really like that. It took people by surpise when NASA went with MOS on integrated circuits because it was very new technology. This appears to be the important point: "MOS technology was much slower than the standard transistors found in integrated circuits, but its simplicity of fabrication allowed manufacturers to incorporate more transistors into integrated circuits. The new MOS technology promised to solve one of spacecraft designers’ growing predicaments, the need for greater electronic capability on board for communications and other functions." [1] (p.240) IMP-D is 1966. The Sharp QT-8D seems to have been one of the first really successful commercial applications, the cheapest electronic calculator to date, and its low power usage meant the first battery powered calculator, the QT-8B, could be launched in 1970. And then yes, sure, the applications were endless. But there's quite a long way from 1960 (MOS invented) to 1966 (IMP-D launched) and then 1969 (mass produced calculators). Merlinme (talk) 21:26, 19 May 2020 (UTC)

This is interesting, from Britannica (article co-written by Sze, no less): [2] "The most important device for very-large-scale integrated circuits (those that contain more than 100,000 semiconductor devices such as diodes and transistors) is the metal-oxide-semiconductor field-effect transistor (MOSFET)... The main reasons why the MOSFET has surpassed the bipolar transistor and become the dominant device for very-large-scale integrated circuits are: (1) the MOSFET can be easily scaled down to smaller dimensions, (2) it consumes much less power, and (3) it has relatively simple processing steps, and this results in a high manufacturing yield (i.e., the ratio of good devices to the total)." Merlinme (talk) 21:57, 19 May 2020 (UTC)

It seems to me that the Self-aligned gate is the important invention. Before that, MOS transistors were not easier to make than bipolars. But now, it seems so obvious. I had not thought about that it couldn't be done with aluminum at the processing temperatures. This was also the start of Intel and the 1101 MOS RAM. But by 1970, there was 50MHz or so TTL, much faster ECL, and 1MHz SRAM. The Intel 1101, as well as I know the first MOS RAM, is 256 bits, PMOS, with +5, -7, -10 volt power supplies. Not so much later, the 1103 1K bit PMOS DRAM. These, and similar devices, were computer memory for much of the 1970's, when CPUs were TTL or ECL. The 1970's was also the time of 1MHz microprocessors. As well as I know, it was not unusual to shrink a die, but not change the oxide thickness, so not actual Dennard scaling. Changing the oxide thickness changes the supply voltage, which means not a direct replacement. The 8080 has +12, +5, -5 volt power supplies. (NMOS is slightly more convenient using positive power supplies, instead of negative, and still be TTL compatible.) Scaling of whole ICs isn't so easy to describe. Individual transistors are faster, but still have to drive the wires across the chip. Gah4 (talk) 00:29, 20 May 2020 (UTC)

AFAIK, early MOSFETs were slower than BJTs, but were smaller and consumed less power. So the earliest applications were things that require a high level of integration (like spacecraft) or low power consumption (like calculators). It wasn't until, as you mentioned, the silicon-gate came along that MOSFETs started being used extensively in computers for things like memory (replacing magnetic cores) and CPUs (with microprocessors). One of the inventors of the silicon-gate MOS IC at Fairchild, Federico Faggin, later went on to lead the development of the first single-chip microprocessor, the Intel 4004, which was fabricated on a silicon-gate PMOS IC. Intel's success with their MOS memory and microprocessors may have paved the way for MOSFETs to take over the computer industry. Maestro2016 (talk) 10:20, 21 May 2020 (UTC)

Early MOSFETs were discrete devices, so size wasn't obvious. They were mostly used where the high input impedance was needed. JFETs were also used early on, where speed was needed. I am still suspicious about the use in analog circuits, where as well as I know, bipolar are still pretty popular. That is, both discrete and IC circuits. The reference given doesn't seem to say that there are more analog MOSFETs than analog bipolars, which is what the statement seems to say. Gah4 (talk) 12:09, 21 May 2020 (UTC)

The earliest MOSFETs were around 10-20 micrometers in the '60s. Do you know what was the size of BJTs at the time? As noted by Merlinme above, MOS ICs were used in NASA spacecraft in the '60s because they allowed a higher level of integration. As for analog tech, the source says that MOSFETs dominate analog integrated circuits. Are you maybe referring to a different type of analog electronics, like discrete analog transistors or power transistors? I think I'll re-word it so that it says analog integrated circuits more specifically. Maestro2016 (talk) 16:25, 21 May 2020 (UTC)

I don't know about others, but there was much discussion last year, the 50th anniversary of Apollo 11, about the Apollo computer, which uses bipolar based RTL. As for dominate, do you count each type of IC, or each one produced? Some are very popular, produced in large quantities. I am not sure by now about op amps. Bipolars were popular for many years, or bipolar with FET inputs. Gah4 (talk) 21:53, 21 May 2020 (UTC)

I meant other NASA spacecraft, not the Apollo 11. They used MOS for their IMP/Explorers spacecraft. As for analog, the source simply says dominated. The quote: "A recent textbook on the subject of analog integrated circuits (Jorns and Martin, 1997) takes the approach that such circuits are now totally dominated by MOSFETs but includes some BJT applications." I can't exactly quantify it, but that's what it says. Maestro2016 (talk) 23:07, 21 May 2020 (UTC)

Integrated injection logic and emitter-coupled logic bipolar technologies were used in some LSI and VLSI high-performance microprocessors, surprisingly as late as 1993. But they were rare compared to all the nMOS and CMOS microprocessors surrounding them. Dicklyon (talk) 18:28, 4 July 2022 (UTC)

They also have faster switching speed

The article says: They also have faster switching speed. It seems to me that this is a side effect of the different scaling laws. For many years, it was TTL and especially ECL that were fast, and MOS was slow. Processors were built with TTL logic and MOS memory. Also, an early problem with CMOS was latch-up where parasitic bipolar transistors would cause devices to fail. Avoiding latch-up, and shrinking to make devices faster, eventually allowed CMOS to pass bipolar in switching speed. Gah4 (talk) 21:18, 10 August 2020 (UTC)

New Reference and Links

New sentence added in MOS sensors section to specifically address some of the applications of MOS and MOSFET sensors for sensing gases. One reference added. Two links added. Nanomaterials21 (talk) 19:41, 3 March 2021 (UTC)Nanomaterials 21

Maestro2016 and Jagged 85 cleanup

Hey. Just to let everybody know @Maestro2016 the main contributor to this article and MOSFET applications, as well as hundreds of others has been banned for being sock puppet of banned user @Jagged 85, a well known vandal [[3]]. You can see Wikipedia:Requests for comment/Jagged 85 and Wikipedia_talk:Requests_for_comment/Jagged_85 for more info about him. This article and MOSFET applications contains most of the errors and exaggerations regarding Atalla/MOSFET, introduced into dozens of other articles all across Wikipedia. For instance it exaggerate Attalla and Kahng contribution suggesting that they single-handedly invented MOSFET and all of it's modern application is due to them. I'll copy what I wrote on Transistor talkpage:

"I am currently reading To the Digital Age: Research Labs, Start-up Companies, and the Rise of MOS Technology by Ross Knox Bassett. Bassett is professional historian of science, so his work is as reliable as it gets. I have huge problems with the way this article and other present history of MOS transistor. First of all as Bassett show there was very little new in Attalah and Kahng invention, as Bassett puts it:

"Atalla appears to have conceived it, but it was an invention in a different sense than the transistors of Bardeen and Brattain and Shockley. The invention of both the pointcontact transistor and the junction transistor involved novel effects. The principles that Atalla’s device used were well known; veterans in the field would have recognized them as ones that had been tried without success by Bardeen, Brattain, and Shockley. Atalla recycled these principles using the advanced fabrication techniques that Bell Labs had developed to make diffused bipolar junction transistors. In some sense Atalla’s biggest breakthrough was an intellectual one, thinking that such a device was worth making at all"(page 24). Attalah and Kahng doe not even give this device a name, again from Bassett:

"Atalla and Kahng’s writings provide evidence that even they had ambivalence about what they had done. A name is obviously one of the first steps in the serious consideration of any kind of invention, and Atalla and Kahng’s failure to name their device implies that they saw it as stillborn. They did not even identify their device as a transistor, suggesting a reluctance to even put their work into the same family line as the work of Bardeen, Brattain, and Shockley. Atalla and Kahng’s paper at the 1960 SSDRC did not establish their device as a promising subject for research or even as something recognized by the semiconductor community at large. The conference chairman made no mention of Atalla and Kahng’s work in his brief report on the technical highlights of the conference, although he did mention Bell’s epitaxial transistor. No further work on a device like Atalla and Kahng’s was presented at either the SSDRC or the Electron Device Conference over the next two years. Two articles reviewing the state of the semiconductor field in 1962 made no mention of Atalla and Kahng’s device. Their work seemed to be a dead end".

The reason MOS transistor even received attention was due two factors:first passivation of silicon surfaces by silicon dioxide gave hope the problems of semiconductor surfaces could be resolved, and second invention of integrated circuit change the way transistor are judged, making MOS simplicity attractive to some(page 13). It will take many years and many people working on it to make MOS practical. Again Basset write that, for example in IBM even in 1967 the future of MOS technology was far from clear(page 106). Contribution from people like Wanlass was just as important as Atalla and Kahng work."

To give other example,in MOSFET applications it is said that "MOSFETs are the basis for modern electric road vehicles". The source for that is an obscure 30 years old abstract, which says that "Recent developments in the technology of permanent magnetic materials, power MOSFETs and microcontrollers have opened the way for significant advances in electric vehicle drive systems". Nowhere does it says MOSFET is the basis of modern electric vehicles.This is one example, but it's just a tip of the iceberg. You can also check my edits, where I did cleanup after Maestro2016/Jagged 85 to find many more examples related to source abuse relating to MOSFET and Atalla. In my opinion this article and MOSFET applications should be stabbed, or at very least massively trimmed, because it would simply take too much time to clean them up.

We can add good and neutral pieces of information later, once we verify them. Also according to Wikipedia:Banning policy:Anyone is free to revert any edits made in violation of a ban, without giving any further reason and without regard to the three-revert rule. I think other articles edited by Maestro2016/Jagged 85 can be cleaned up, as they are smaller. In general if you see exceptional claim that is sourced by hard-to-check source or you simply can't immediately verify it with high quality source, it should be deleted as this was one of Jagged 85 favorite tactics to introduce disinformation.

Thank you.

DMKR2005 (talk) 18:49, 2 July 2022 (UTC)

Yes, please revert much or all of the overstatement of the Atalla and Kahng contributions that were added by Maestro2016/Jagged 85. I'm sure their contribution was valuable and worth mentioning, but the current imbalanced over-emphasis is pretty awful. I could perhaps help, but am pretty busy with work... Dicklyon (talk) 02:27, 3 July 2022 (UTC)

I don't quite recall whether Lilienfeld was able to show how to make a working FET. Shockley also "invented" the MOSFET, I think. But Atalla and Kahng convincingly made one work, iirc. So that's important. But to state simply that they "invented" the MOSFET contributes more bias than understanding. So we should work on that. Not by spamming the names Atalla and Kahng in every possible place, but by discussing the early invention of the FET and MOSFET some place and linking it. Dicklyon (talk) 02:52, 3 July 2022 (UTC)

Maestro2016's significant edits to this article were from 27 June 2019 through 12 July 2020. In total, they made nearly 1000 edits. The article grew by 48% during this time. I don't think we need to stub the whole thing. Worst case, we restore the article to its state before this started. Not all of Maestro2016's contributions were problematic but I appreciate that removing them is arguably a safer way to deal with them than reviewing. If we are able to summon the wherewithal to review, we could end up with something better than what we had before Maestro2016 and better than the current revision. ~Kvng (talk) 14:48, 3 July 2022 (UTC)

Thanks Dicklyon and Kvng for suggestions. When I said stubbed I mean return to version before Maestro2016/Jagged85 edits. I think that's what we should do. After that we can add more information about Atalla and Kahng contribution. I have Ross Bassett To the Digital Age book with me, which is very good history book about MOSFET. After we return to pre Jagged-85 state, we can add paragraph about Atalla role in MOSFET. There is also MOSFET applications article which is spinoff of this article, and is entirely written by Maestro2016/Jagged 85. What do you think we should do with it? DMKR2005 (talk) 15:17, 3 July 2022 (UTC)

A quick look at MOSFET applications indicates the whole History section should go. I would like to review the rest. I would also like an opportunity to review Maestro2016's changes to this article. I haven't figured out how to do the diff over 1 year and 1000 edits. Any wiki tips would be appreciated. ~Kvng (talk) 15:33, 3 July 2022 (UTC)

If there's anything good in MOSFET applications, it's probably best to merge it into MOSFET and redirect to a section. I have no objection to rolling back MOSFET to a pre-Maestro state, being careful to put back constructive contributions of others if any are there. Dicklyon (talk) 17:43, 3 July 2022 (UTC)

Thanks. I think returning to pre-Maestro2016 state and adding Atalla and Kahng contributions is the way to go. DMKR2005 (talk) 21:24, 3 July 2022 (UTC)

I struck my comment about merging applications back to the main article. Maestro did the split, and it was probably a good idea. Just need to clean up his revisionist emphasis on Atalla everywhere. Dicklyon (talk) 18:32, 4 July 2022 (UTC)

Thank you. Speaking on MOSFET applications we would also need to check this article for errors and exaggeration. See for instance above about MOSFET being the basis for modern electric road vehicles DMKR2005 (talk) 20:55, 4 July 2022 (UTC)

I thought most power applications had gone to Si MOSFETs by now, but this article suggests that SiC MOSFETs and Si IGBTs are pretty hot in electric cars. Dicklyon (talk) 00:31, 5 July 2022 (UTC)