Talk:MOSFET

From Wikipedia, the free encyclopedia

splitup of article Field effect transistor in progress, please see Talk:Field effect transistor Pjacobi 20:45, 19 Jul 2004 (UTC)

• Done Pjacobi 21:19, 19 Jul 2004 (UTC)

Contents 1 Copied from pre-splitup Field-effect transistor 1.1 mosfet symbols 1.2 MOSFET section 1.3 Analog circuits 2 Big mistake 2.1 Deleted Section 2.2 Comment 3 Resistance Calculations.... 4 Graphical Representation of MOSFET operation in circuits 5 Need some help on MOSFET and Square wave high amp inverting 6 BJTs better for some digital circuits? 7 Need for tidying? 8 Terminology Error 9 Modes of Operation 10 Editted image 11 SVG symbols 12 Thermal runaway 13 Body Effect 14 Primacy of MOSFETs excessively big-worded 15 What does it do? 16 Threshold symbol?

[edit] Copied from pre-splitup Field-effect transistor

Most of what I've done here doesn't need any explanation, but there's one correction I think warrants a note. I replaced every instance of "glass" with "oxide". This is because the silicon dioxide layer under a gate is not glass. Glass would not work. A glass is an amorphous solid - irregular arrangement of atoms. The common usage of "glass" happens to be an instance of this. SiO₂ in MOSFETs is crystalline, not glass. -- Tim Starling

Small correction: amorphous oxides can work perfectly well as gate oxides. see link. At least as far as I know, these films are almost entirely amorphous. The best films are done with ECR, but you can also use PECVD and get decent results. On physical grounds, I don't see any reason why a gate dielectric has to be crystalline in "work". --User:Dgrant

In fact, crystallinity is an undesirable effect in gate dielectrics, as it creates grain boundaries, which can dramatically increase gate leakage. User: Stonecold21

You're the man Dave, I'll take your word for it. Before you came along, I didn't know amorphous oxides were used for anything other than window glass :) I notice that the resistivity of the oxide in your reference is many orders of magnitude less than for crystalline silicon. That would degrade performance somewhat, but at least the breakdown voltage is still high. -- Tim Starling

"Oxide" is more correct than "Glass". A glass is not only amorphous but it also contains lots of impurities of highly mobile ions as softener, making it very inadequate as gate oxide. Dgrat is right about amourphous oxides, although they will result in very instable transistors. In organic transistors, even polymers are used as gate dieletric. --Qdr 17:12, 17 Jul 2004 (UTC)

---

what does "whereas those to the left abstract from the body contact." mean? It doesn't make any sense to me, or at least is doesn't convey the indended meaning, in my mind. dave

It doesn't mean anything. I changed it to something which makes sense, and is probably right. I seem to remember seeing some FET-like structures with the body insulated from the backside, but I don't think they do that for MOSFETs. -- Tim Starling 00:39 May 14, 2003 (UTC)

You have that in SOI (Silicon on Insulator) FETs, but these work slightly different.--Qdr 17:12, 17 Jul 2004 (UTC)

[edit] mosfet symbols

The schematic for your MOSFET shows a solid line connecting the Source and Drain. Does this not indicate a depletion mode MOSFET?. An enhancement mode MOSFET is symbolically shown with a dashed line between the Source and Drain.

The arrows for the 'metallurgical' contacts point at the bottom of the N diffusions. The metal contact is on the top surface. Shouldn't your arrows point to the upper surface of the N regions?.

All textbooks use different notations. I can't remember right now what is right, and what is depletion, enhancement, etc... However, should use standard IEEE conventional symbols, whatever those are, and we should state that they are the IEEE standard symbols. If there are some common "misuses" of the symbols out there, then we should mention that. So first thing I think is to check IEEE and see if there are standard symbols, and secondly to check something like Art of Electronics and see what it uses. dave 22:06, 16 Oct 2003 (UTC)

IEEE: http://ewh.ieee.org/soc/es/Nov1998/14/education/

You are right about the 'metallurgical' junctions, they are on top of the diffusion region and are not the same! Modern FETs do not only use high doping, but also metal silicides at this place.--Qdr 17:12, 17 Jul 2004 (UTC)

[edit] MOSFET section

I'm planning on doing a lot of work on the MOSFET section. I hope to discuss how MOSFETs are evolving to smaller and smaller submicron dimensions, and the problems designers are encountering...obviously non-technically. I've created two subcategories I want to expand upon--why MOSFETs are so popular and the problems with scaling. Rmalloy 13:47, 14 Jul 2004 (UTC)

I think some work needs to be done in the introduction to MOSFET...most important part. for later. Rmalloy 18:38, 14 Jul 2004 (UTC)

Thanks for your efforts! I assume at some point it would make sense to put all MOSFET stuff in separate article. Agreement? Pjacobi 19:26, 14 Jul 2004 (UTC)

I dunno, I'm new here and don't know what the protocols are. As long as info is easily accessible it makes little difference to me. I'm not going to touch stuff like that. Rmalloy 20:06, 14 Jul 2004 (UTC)

It could make sense to have a rather generic introduction to FETS on this page and move all the details (different types, processing, materials) to other pages. --Qdr 17:12, 17 Jul 2004 (UTC)

• • • • Splitup plan, please comment under each point if necessary. If no active disagreement is seen, I'll do the splitup around 2004-07-19 21:00 UTC. Pjacobi 18:49, 18 Jul 2004 (UTC)
        • The MOSFET specific parts of Field Effect Transistor will be moved.
        • This applies to section MOSFET (currently 1.1) and DMOS (currently 1.5).
        • It will go to http://en.wikipedia.org/w/wiki.phtml?title=MOSFET&redirect=no which is currently a redirect.
        • The JFET, MESFET, and HEMT sections are not yet substantial enough to be moved to separate articles.

Sounds good, but I think also the other types of FET devices should be moved somewhere as they are way too specialized for a generic introduction. Maybe an article about "special" or "exotic" FETs? --Qdr 19:59, 18 Jul 2004 (UTC)

Lumping together JFETs and MESFETSs into Field effect transistor (exotic) (or [[Field effect transistor (bizarre)]?) will break my heart ;-). I'd vote keeping them (temporarily) in the main article, or as a second choice, make all separate articles, even when HEMT will be a short one. Pjacobi 21:34, 18 Jul 2004 (UTC)

I vote for separate articles, maybe that is also an incentive to extend the individual articles a little. And btw, there should also be a link to TFTs in the main FET article. --Qdr 22:17, 18 Jul 2004 (UTC)

I second this. With some diagrams all of these transistor types would be good separate articles. I think Field effect transistor should be a list of links. Maybe some generic discussion. Rmalloy 00:14, 19 Jul 2004 (UTC)

Rmalloy: There are some inaccuracies in your additions: The reason for using polysilicon as a gate material is the reduction of interface states and the self aligned S/D diffusion. Replacing it with metals (for example TaN, TiN) is subject of current research. Current gate oxide thicknesses are way below the 20nm you stated, I changed it to 2nm. But somebody should look up an accurate number. The problem with thin oxides is not breakdown, but leakage by quantum mechanical tunneling of electrons through the oxide. To remedy this, the industrie works on high-k dielectrics.--Qdr 17:41, 17 Jul 2004 (UTC)

QDR: First-off, I must admit that I haven't worked in this area for 2 years. But I think I am right about the polysilicon gate. The self-aligned S/D diffusion process would work equally well with a metal gate. I went back and looked at some stuff I wrote on www.everything2.com when I was a grad student in this area. Look at the article MOSFET that describes and shows diagrams of the fabrication process. The self-alignment process work the same if the gate were aluminum. I'm 99% sure that the reason you can't use aluminum is that there is a high-temperature annealing step after the gate is deposited, and this would melt aluminum. Now the reason the annealing must be done after the gate is deposited relates to the self-alignment process (S/D to be annealed created after gate), so we could be both right in a sense.

I looked up the melting points of the metals you mention, and they have very high melting points. When I was a grad student, I don't recall much effort into using these metals as gates. I don't know why, so I won't argue. But I do remember heavy emphasis on the silicides, like I wrote in this article. I know what surface states are, but I don't see how they relate.

2nm is the accurate number. I meant 20 angstroms. 2nm was considered an absolute cutoff.

Now my memory is that 2nm is still a long way for an electron to tunnel. I seem to recall that, like I said, the oxide broke down, creating states in the oxide that acted like rocks across a river that you can jump across, facilitating tunneling. So I won't argue with you here...we might both be right. And it's not worth arguing. But I'm 99% sure that 2nm was the cutoff, considered absolute by the chief technology officer of TSMC. And I agree about the high-K dielectrics, and mentioned it in the article. Rmalloy 18:54, 17 Jul 2004 (UTC)

I did some quick websearching. An Intel site claims successful operation with 1.2nm of gate oxide, so I guess I'll have to bite my tongue on 2nm. And I'm reminded of another key issue for gate materials--work function. The current setup, where the source, drain, and gate are all doped heavily at the same time gives the gates the proper work function for the transistor type--NMOS or PMOS. Successful metal gate processes would require two kinds of metals--one work function for each transistor-type. It's dawning on me that the choice of gate materials involves several issues, including all the ones we've mentioned and probably several more.

Feel free to clean up, correct, or add to anything I wrote. These issues are complex and there are many issues to discuss...maybe it would be best to avoid difficult issues like this altogether in an encyclopedia. Phew I'm glad I left this field! Rmalloy 16:35, 18 Jul 2004 (UTC)

The oxide thickness is a moving target, IMO around 2nm is a good guess. Intels 1.2 nm oxide was probably already nitrited oxide. There are many candidates for new metal gates, however the ones I quoted have been announced by the IMEC recently. And yes, there are various leakage mechanisms (poole frenkel, schottky emission into insulator valence band, direct tunneling, fowler-Nordheim) and they are enhanced by soft breakdown, however they are usually not regarded as breakdown itself.

I am pretty sure that the main incentive to use polysilicon gates was the self aligning SD process, back in the 70ies. I do not know the exact problem with Al gates and the self aligning process, but for example the spacer oxide would be pretty difficult to apply to an Al-Gate. Anyways, all of this is way too detailed for a Wikipedia article, the best way is to formulate it as generic as possible. --Qdr 19:59, 18 Jul 2004 (UTC)

Good point about the spacer. I have the melting thing stuck in my head...I must have picked it up somewhere. I agree this is all too detailed for Wikipedia, but I guess I thought some explanation for the polysilicon gate would be advisable. At first glance, polysilicon is a very strange choice for gate material. You obviously know what you're talking about, so don't hesitate to delete or change anything I wrote. Rmalloy 00:14, 19 Jul 2004 (UTC)

[edit] Analog circuits

I've added all I feel comfortable adding about MOSFETs in analog circuits. I wish someone could discuss analog stuff, since everything is so digital digital digital. Rmalloy 23:38, 14 Jul 2004 (UTC)

---

A quesiton: "The MOSFET's strengths as the workhorse transistor in most digital circuits does not translate into supremacy in analog circuits, in which the bipolar junction transistor (BJT) has traditionally been seen as the transistor of choice, due largely to its high gain." This does not seem correct, since FETs have near-infinite gain - essentially no current flows into the gate. Glengarry 21:09, 15 Jul 2004 (UTC)

Bad subject-verb agreement for one thing :(.

Ok, this is not an area I know a lot about, but let me try to explain as best I can. "Gain" invariably means "small signal voltage gain" (output signal voltage / input signal voltage). The fact that a MOSFET gate allows no DC current isn't relevant (though its true). The job of analog circuits is to handle small signals.

Suppose a transistor is being used for amplification in an analog circuit, it is "DC biased" to put it in the high gain regime. A small signal voltage is applied between gate and source (or between base and emitter), creating a small signal current from drain to source (or from collector to emitter). The ratio of this small signal current to the small signal voltage is called "transconductance." My sense is that BJTs have substantially higher transconductance than MOSFETs. For a tiny bit of support, see http://www-inst.eecs.berkeley.edu/~ee130/SP03/homework/hw13soln.pdf. This small signal current may drive a resistive load, giving an small signal output voltage of the current times the resistance of the load. Thus you end up with a higher "gain" (output voltage/input voltage).

I'm really shaky on everything analog, and that's why I made such a vague statement. I just felt like it would be inappropriate to only talk about digital stuff when analog circuits are very important. I think what I wrote is essentially correct, but if someone wants to delete it that's fine. I'd rather someone teach me though! Rmalloy 00:42, 16 Jul 2004 (UTC)

The gain issue is easily misinterpreted and is always good to start a flame war at news://sci.electronics.* or http://www.diyaudio.com. I'd write something into article but for the fear of a edit war! In essence there are four quantities which can be seen as geen dV(out)/dV(in), dI(out)/dV(in), dV(out)/dI(in), and dI(out)/dI(in). Of course MOSFET score big on dV(out)/dI(in) and dI(out)/dI(in) in NF, as no input current flows, but the practical significance is more that there is moe leeway in designing the preceeding stage.

What's more the problem with MOSFETs in discrete designs, is the variabiliy of there threshold voltage.

Pjacobi 07:27, 16 Jul 2004 (UTC)

It's not true that no *small signal* input current flows into a MOSFET. Current is constantly charging and discharging the MOSFET gate, so dI(in) != 0. In fact, capacitors, like the MOS capacitor, are short-circuits to high-frequency current. I think this topic is probably best left alone in the article unless someone is an expert on the subject, so we avoid misinformation.Rmalloy 13:04, 16 Jul 2004 (UTC)

Agreed that the transconductance of a BJT is favorable compared to a MOSFET. I'll made the change in the article. Thanks, Glengarry 14:56, 16 Jul 2004 (UTC)

I edited the analog section of this article to add a few things. I took out this section from it because I thought it sounded kinda weird and it isn't exactly true:

"Some analog circuits are designed solely using MOSFETs in a fabrication process specialized for digital circuits because it is advantageous to incorporate digital and analog circuits onto the same chip and digital fabrication processes are less expensive."

Any difference between the analog and digital fabrication process would have to be changed to incorporate the analog components. I am also not aware of any difference, although that doesn't mean there isn't one.

[edit] Big mistake

[edit] Deleted Section

The channel in a MOSFET is connected on each end to source and drain terminals which are oppositely doped in relation to the channel, and highly doped so that they form low-resistance "ohmic contacts" with metal wires. It is well-known among electrical engineers that a p-n junction allows current to flow only in one direction, from p-type semiconductor to n-type. Since the structure of the MOSFET consists of back-to-back, but oppositely directed, p-n junctions, the MOSFET allows no current to pass in the "off" state, in which no voltage is applied to the gate.

[edit] Comment

I've overlooked this when doint the split up. The first sentence is somewhat O.K. But there a no pn-junction in the MOSFET conduction path. It's: ohmic contact - n+ - n-channel - n+ - ohmic contact. Some hard work to do here, perhaps a new picture is needed, adding the n-channel zone under the gate. Pjacobi 21:31, 23 Aug 2004 (UTC)

The picture is perfect. It illustrates the MOSFET's *construction* and was never meant to show the various depletion, inversion and accumulation layers that form, widen, compress and vanish during its *operation*. It would take four pictures just to explain the MOSFETs most important operating conditions (thermal equilibrium, cut-off, strong inversion, saturation), and more such conditions exist. It is my current belief that Wikipedia entries should be edited in the style of a vivid and concise encyclopedia, but not as exhaustively as if for a comprehensive engineering or physics textbook.

Also, two pn-junctions definitely exist. It is precisely the reverse-biased drain junction that prevents current flow in cut-off condition, for instance. The "ohmic contact - n+ - n-channel - n+ - ohmic contact" conduction path suggested above exists in strong inversion exclusively and gets disrupted in saturation although current continues to flow! Kaeslin 17 Sept 2004.

Sorry, if I got this wrong, feel free to revert. But isn't the MOSFET typically depicted as a unipolar device, especially in popular presentations, stresseing the difference to the BJT? So that the picture, even when perfectly illustrating the construction may give a false idea of operation to the non-technical reader? Pjacobi 09:47, 17 Sep 2004 (UTC)

Yes, the label "unipolar" is sometimes being used since the charge carriers are electrons or holes exclusively (in n- and p-channel devices respectively) as opposed to BJTs where both types of carriers are involved in the same transistor. I might indeed reinclude the figure if I find leisure to do so. Kaeslin 16:00, 20 Sep 2004 (UTC).

One thing that is definitly wrong is the remark about the "metallurgical junctions". If they are to be mentioned, they should be properly marked on top of the diffusion regions.--Qdr 16:28, 17 Sep 2004 (UTC)

No, the term "metallurgical junction" refers to the borderline between n- and p-doped semiconductor regions. What you want between the diffusion regions and the metal plugs that connect to them are ohmic contacs, not junctions. Kaeslin 16:30, 20 Sep 2004 (UTC).

[edit] Resistance Calculations....

The wiki says, "Conceptually, MOSFETs are like resistors in the on-state, and shorter resistors have less resistance." From the basic R = (pL) / A surely reducing the length would not change the resistance?

EG: for this example lets make rho=5; let the width of the channel=2 and we will vary the length. If we start with the length at 10, the resistance will be: (5 * 10) / (10 * 2) = 2.5 If we reduce the length to 5, the resistance will be: (5 * 5) / (5 * 2) = 2.5

(NB: this is my first wiki edit and I've probably broken loads of conventions. Could someone please edit/delete this, whatever is appropriate)

The A in your formula is the area formed by the other two directions, so to speak width and height of the channel, not length.

But the article chapter you are referring to is nevertheless a but fishy .

Your edit was just fine, you only missed to use the "auto signing" feature of the software, putting four tildes at the end of your discussion statement, would give user and date/time info.

Pjacobi 22:50, 5 Dec 2004 (UTC)

[edit] Graphical Representation of MOSFET operation in circuits

Would it not be usefull to include Vin/Vout graphs? i.e. what the 'resistance' across the MOSFET does as the gain-source voltage increases? Yossarian 10:19, 11 Jun 2005 (UTC)

[edit] Need some help on MOSFET and Square wave high amp inverting

Guys and gals, can you help me a little bit. I have this idea of buildng a square wave AC source from a constant current 300A DC. I can, hopefully, use an H bridge and MOSFETs. My question is, if I use (stepped down) AC to control the gate of the MOSFET, the resulting voltage would not be square wave, or would it be? Any thoughts? thanks. I will be watching this page.

[edit] BJTs better for some digital circuits?

At the end of "The Primacy of MOSFETs", the article currently states, "Ironically, the BJT has some advantages over the MOSFET in certain digital circuits; digital circuit designs can incorporate BJTs to speed signals in critical locations." Could somebody clarify this? Exactly what advantages do BJTs have in what sort of digital circuits? Does the wording here indicate that BJTs are faster than MOSFETs for digital switching? (if so, shouldn't this actually be mentioned explicitly?) -- Foogod 12:16, 28 November 2005 (UTC)

• My guess is that the author didn't know what bipolar circuit it was exactly. The author was probably referring to emitter coupled logic Dunno if an article exists for it Snafflekid 19:12, 28 November 2005 (UTC)

The way you define digital circuits is where the nuance comes in. In many chip-to-chip digital communication protocals BJTs are used, but they operate on low-swing voltage signals. Is that digital, well, because they are working on '1's and '0's it can be considered digital, but because the representation of a '1' and '0' is in an unconventional low-swing voltage it's really an analog signal. So it's really a tossup. These serial communication links have gone up to 40GB/s, and then are parallelized in the digital chip with demultiplexors, and are used in the chip. Both the demultiplexors and the rest of the chip use MOSFETs. -- Jeff3000 19:58, 28 November 2005 (UTC)

Sounds like TTL. - mako 06:44, 29 November 2005 (UTC)

BJT's are currently better for at least 3 jobs: The first is in high speed switching because they don't have the "larger" capacitance from the gate, which times the resistance of the channel gives the intrinsic time constant of the process. Widening the channel reduces the resistance of the channel, but increases the capacitance by the exact same amount. Reducing the width of the channel increases the resistance, but reduces the capacitance by the same amount. R*C=k, 0.5R*2C=k, 2R*0.5C=k. You can skip most of this problem with a BJT. The second job stems from the first: When driving many other gates the resistance of the MOSFET is in series with the gate capacitance's of the other FETs, creating another time constant. This is how delay circuits work. BJT's are better to drive the other gates because they output much more current per unit area, which causes the gates to charge faster and decreases the time constant, increasing the speed. The third job is high current amplification, such as in audio and motor control applications. MOSFET arrays where possibly thousands of MOSFETs are placed in parallel( I haven't seen any chips with millions in parallel) to provide the current, although these circuits suffer from the timing issues discussed above and usually use at least 1 BJT to drive the MOSFET array.--Jeffrobins 17:28, 2 May 2006 (UTC)

I'm 99% certain that the conclusion in Depletion mode MOSFETs section is the opposite of the information given in the beginning

You're right. I've corrected it. -- Jeff3000

[edit] Need for tidying?

Hi,

I plan to contribute to both the "power MOSFET" and "MOSFET operation" sections. The problem is that the article is relatively long now, and adding content to these sections will increase the "bloated" aspect of the article.

Furhtermore, there is a mix between power qnd microelectronics MOSFET in some section which make the whole not exactly consistent (for example, the termal runaway of power MOSFET is addressed in the "scaling" section, which is specific to microelectronic MOSFETs).

I think it is time to split the article, with a main root about the generic principle of a MOSFET and its history, and then articles about integration, power, maybe manufacturing. What do you think? CyrilB 12:18, 2 April 2006 (UTC)

Well, I'll will say that nobody is against some transformations on the article. I will start by creating a different article MOSFET model to put all the equations (and demonstrate them by the way), as they add little to the understanding of the principle and might frighten people... I think the MOSFET scaling should also go in its own article, but I'm not really an expert about this issue. -- CyrilB 18:50, 6 April 2006 (UTC)

I didn't notice your first comment because the comment edit had to do with bjt operation in digital circuits. I think you're on the right step to divide up this article, but I don't think moving out the equations is the right step. There's not much that can be added to that, so that article would remain small. We have to move out a section that has room for growth. So I would suggest moving out the Power Mosfet (as it's quite different) out into it's own artcile with a summary section in this page. That would shorten the article considerably. -- Jeff3000 19:09, 6 April 2006 (UTC)

I take back my comment a little, the Power MOSFET section by itself is not big enough, so I would move all the other MOSFET types out and use just the microelectronics MOSFET on this page. We would put a disambiguation link at the top of the page, with a short section in the bottom of the page, that other types of MOSFETs also exist. -- Jeff3000 19:13, 6 April 2006 (UTC)

Actually, the equation part and the power MOSFET section are the two parts on which I plan to work, so they should grow in the near future. And both the microelectronics MOSFET and the power MOSFET share the same principle. So I think the MOSFET article should be more focussed on the history and basic principle of the MOSFET, and link to specific types. -- CyrilB 19:50, 6 April 2006 (UTC)

How bout you first work on moving out the Power Mosfet stuff, and if need be (through your additions in the equations) we'll move out the equations. In my mind the equations need to be on this page. -- Jeff3000 20:39, 6 April 2006 (UTC)

Done. But the page is still very long. Anyway, lets see how things go. -- CyrilB 21:34, 6 April 2006 (UTC)

[edit] Terminology Error

Hey, this article has a terminology error in it. I tried to correct it but it was changed back without thought and I don't want to get into an argument.

Digital MOSFET circuits operate in the cut-off and linear region, while the saturation region is used mostly for analog amplifiers. Just think about it from the IV curve: when the Vds is small we are in the linear region. If using the the FET as a switch in the saturation mode you'd have a larger Vds, not good for a switch. I noticed the same problem on the switching amplifer site and corrected it, although that's probably been changed back by now too.

Am I missing something here? Did they decide to clean up the terminology and not tell me?

Sorry about all the edits, I am a newbie here (but not to engineering!). -- Dkomisar

No, you're right. It's been re-corrected. - mako 21:29, 2 May 2006 (UTC)

[edit] Modes of Operation

I deleted this statement:

In digital circuits MOSFETs are operated in cut-off and linear mode. The saturation mode is mainly used in analog circuit applications.

Its true that in steady state, the mosfets in a digital circuit are either in cut-off or linear mode. But any time a cmos stage is switching, there is a transistor in saturation mode. Rather than complicating the statement to make it accurate, I just removed it; I thought it seems like an extraneous point. —Kymacpherson 04:42, 13 June 2006 (UTC)

Thanks for fixing that. I was just reading the previous talk where it went back and forth. Some of us are old enough to remember metal-gate pMOS with saturated loads, and silicon-gate nMOS with sometimes-saturated depletion loads, too. But as you say, in CMOS the interesting parts, namely the switching events, are done by transistors in saturation. Dicklyon 05:18, 30 August 2006 (UTC)

[edit] Editted image

I made an update of that image to make it more clear. If anyone likes it, please install it. The new name is simply MOSFETs.jpg. Dicklyon 04:40, 30 August 2006 (UTC)

Nobody home here, so I'll go ahead. Revert if you disagree. Dicklyon 16:53, 1 September 2006 (UTC)

Does anyone happen to know if the pair of FETs pictured form a functional device? It looks like they could work as a digital inverter, but I can't identify the doping regions just from looking at the photo... -- mattb @ 2006-09-11T04:10Z

That layout is not compatible with the idea that one is an nFET and other a pFET, since they share a source/drain contact of one type. Nonetheless, you could bias one device to act as a saturated load, by setting its gate near threshold, and use the other gate as input, and make a functional inverter that way. But, no, it's not designed to be an inverter, I would say by looking at it. I don't know where the picture came from. Dicklyon 04:15, 11 September 2006 (UTC)

[edit] SVG symbols

Omegatron, good job on the SVG schematic symbols. I was considering doing something myself, but I'm a novice at SVG editing and not very comfortable with Inkscape yet (you know anything better on Mac OS X?). Just a couple of things I recommend changing:

• P-channel JFET: the gate arrow should come out of the source side, like the N-channel, not the drain side; just swap the labels S and D. Alternatively, JFETs are usually drawn with the arrow centered, but then there's no defined choice of S and D.
• The simplified MOSFET symbols are usually drawn with the bar extended to the same length as the triple bar in the complex symbol. For example, more like this one: [1]

I'll work on it if you prefer. Dicklyon 01:54, 10 September 2006 (UTC)

I didn't create them, except for the simplified ones. See Commons:User_talk:Jjbeard#FET_symbols. The simplified ones can be drawn either way: [2] [3] [4] I was just copying the bitmap diagram that was in the article previously. — Omegatron 02:55, 10 September 2006 (UTC)

Yeah, I didn't like it before, either. They CAN be drawn either way, but the lazy way is ugly, and less like the official way. I'll work on them at some point if you don't want to. Dicklyon 03:16, 10 September 2006 (UTC)

One more thing: a more complex issue. The complex MOSFET symbols, both depletion- and enhancement-mode, show the bulk or body terminal connected to the source. This is typical in single packaged transistors, but by no means the only widely used configuration. In typical ICs, most of the transistors have their bulk, body, or "back-gate" terminal connected to a power supply (VSS or VDD) rather than to the source; about half the time it's the same node, and of the ones with source not at VSS or VDD, the bulk is usually, probably 99% of the time, not connected to the source. So, ideally they are drawn as four-terminal devices, since that's how they are always used in IC designs. Dicklyon 02:46, 10 September 2006 (UTC)

I know. I was going to draw every permutation of transistor symbol, in the same style, size, and line width and everything, but I really don't have time. I saw these were already drawn and decided to go with them for now. I agree that both types of symbols should be shown. — Omegatron 02:55, 10 September 2006 (UTC)

[edit] Thermal runaway

You're both wrong. Or both right. A quick GBS search shows verifiable sources for both positions (that thermal runaway is "impossible"[5] or "widely known"[6]). So we need to represent both in the article, instead of a revert war. Here's one that says "less prone to"[7] Dicklyon 00:55, 9 October 2006 (UTC)

I got this on my user page from an unregistered user, and answered it, but nobody is likely to see it there so I'm copying it here:

A few days I corrected your info on the MOSFET page, relating to thermal runaway. You then changed back that section, saying in comment that it didn't make physical sense. So let me try to explain it to you... Assume that a MOSFET device turned ON works as a resistance (which is true for DC current). Now imagine a circuit that connects a N-channel MOSFET's source to the negative of a 9V power supply and it's drain to the positive of the same supply. Assume also that the gate is biased so as to turn the MOSFET fully ON and that it's ON resistance is 1 ohm. The current following in this circuit will be 9V / 1 ohm = 9 Amps. Now let the MOSFET heat up and increase it's resistance to 3 ohm (typical increase at 125º C). The current flowing will now be 9V / 3 Ohm = 3 Amps. Now for the punchline : the power dissipated on a resistance is 'Power dissipated = Resistance x Current squared'. So for a cold MOSFET you'd get a dissipation of 1 Ohm x 9 Amps x 9 Amps = 81 Watts. And for a hot one 3 Ohm x 3 Amp x 3 Amps = 27 Watts. This means that thermal runaway is impossible. If you want to you can use the more known formula 'Power = Voltage x Current'. So with MOSFET cold Power = 9V x 9 Amps = 81 Watts, with MOSFET hot Power = 9V x 3 Amps = 27 Watts. Your info would only be correct if the current flowing through the MOSFET was constant. But then the voltage applied to source-drain would have to increase, which normally doesn't happen. Hope this clears up your misconception, and if you would be so kind as to correct the page... ;-)

Carlos Azevedo

Carlos, yes, that analysis is correct for a MOSFET switch connected across a power supply. However, a more typical situation is to have the MOSFET driving a load, with a load resistance that's higher than the MOSFET's on resistance, usually by a large margin, so that the load is receiving more power than the MOSFET is dissipating during the on state. In this case, increasing the on resistance makes the power dissipated in the MOSFET go up, not down. That's why thermal runaway is still possible. It is closer to the constant-current case, and the drain voltages does increase in this typical configuration. Please do a couple of calculations for such a configuration, and let me know if you accept this analysis or not. It's probably still a good idea to revise the article to state approximate conditions under which thermal runaway is possible or not, with some references. Dicklyon 17:15, 13 October 2006 (UTC)

[edit] Body Effect

I removed this section because it seems out of place. Sure, it's included in Level 1 SPICE, but it's not required for a basic explanation of MOSFET operation. I'd like to see such material on a MOSFET modeling page. - mako 22:14, 24 October 2006 (UTC)

I agree that it could be removed from this page, but it deserves much more that MOSFET modelling. There is a real physical effect when the body is reverse or forward biased. Regards, -- Jeff3000 22:18, 24 October 2006 (UTC)

I disagree that it can be removed. A MOSFET is inherently a four-terminal device, and a qualitative understanding of it in any other than ground-source configuration depends on being aware of body effect. Dicklyon 22:26, 24 October 2006 (UTC)

As the person who added it back, I'm with Dick on this one. Just as he said, MOSFETs are four terminal devices. The body affect is only neglected for the purposes of circuit modeling if the body is shorted to the source or if the designer is doing a quick 'n dirty approximation. As I said in my edit comment, you need only look at any small or large signal model for the MOSFET and you'll find factors accounting for the body effect. -- mattb @ 2006-10-24T22:47Z

Very well, I concede. Things should be made as simple as possible, but not simpler, eh? - mako 23:51, 24 October 2006 (UTC)

[edit] Primacy of MOSFETs excessively big-worded

I think the primacy of MOSFETs section could be reworded, since it uses a wide vocabulary that isnt necessarily useful. It sounds somewhat like an advertisement does (although I dont think it was added as advertising, it just reads like one. Examples: "primacy", "possess such technical attractions", "serendipitously", "Buoyed by this stroke of good fortune" and, "electronic hegemony". Kaldosh 10:51, 26 October 2006 (UTC)

[edit] What does it do?

OK, the MOSFET, transistors and stuff, I get it.

But what does it do? What do you get out of it that you might want? What do you need to put into it?

I'd suggest reading transistor. A full discussion of "what a MOSFET does" is rather out of the scope of this article and more suited for a semiconductor physics course. In a nutshell, a MOSFET is a type of transistor that uses electric fields at an oxide-semiconductor interface to control the conductivity between two terminals. What you "get out of it" depends on how you use it. Transistors in general can be used for switching, amplification, buffering and impedence matching (followers), etc. -- mattb @ 2006-11-27T16:14Z

Or think of it by analogy with a neuron; the brain is made of billions of neurons, but what do they do? What do you put into them and what do you get out of them? Answer: all kinds of signals, depending on what role they are playing. In one simple configuration, you put in a voltage signal and get out a current signal; in another, you put in a voltage signal and get out an amplified voltage signal. I agree with Matt: start with the transistor article. Dicklyon 16:26, 27 November 2006 (UTC)

Here's a radio schematic in which each transistor is labeled by what it does functionally, in terms of RF (radio-frequency), IF (intermediate-frequency), and AF (audio-frequency) signals. Dicklyon 16:31, 27 November 2006 (UTC)

Though it should be noted that the transistors in that receiver are BJTs rather than MOSFETs. The small signal operation of BJTs and MOSFETs are modeled much the same, but the actual operation and utility of each is significantly different from the other (e.g. BJTs exhibit much higher transconductance and are thus better small signal amplifiers while MOSFETs have very high gate input impedance and are good for buffers, etc). As Dick pointed out, it's impossible to summarize the uses of any transistor in one neat statement. One quick generalization I can make is that FETs (MOSFETs primarily, perhaps HFET/HEMTs) tend to be better switches (digital circuitry) and bipolar transistors (BJTs, HBTs, etc) tend to be better amplifiers (analog circuitry), though they can both be used in either capacity. -- mattb @ 2006-11-27T19:17Z

That's a conventional generalization, but it's less true with modern FETs than it was in the old days. FETs can easily deliver any transconductance value you need, any drive level, any speed, etc. Dicklyon 19:21, 27 November 2006 (UTC)

Sure, but a bipolar transistor of similar physical dimensions will still deliver higher transconductance; just as a good heterojunction system can produce a bipolar transistor with extremely high common-emitter input impedance. BiCMOS exists for the very purpose of leveraging the advantages of both types of transistor. I still think that the generalization is valid. -- mattb @ 2006-11-27T21:09Z

I remain unconvinced. For both types, transconductance is approximately proportional to the output current DC level, with a small numerical advantage to the bipolars. The FETs, however, can often carry more current in a given area, at least in modern CMOS IC and power device technologies. My impression is that BiCMOS is no longer much used; it offers extra circuit flexibility, but there's not much net advantage. Dicklyon 22:48, 27 November 2006 (UTC)

We could go on like this for awhile, but I'd rather not seeing how it's non-critical to this article and the original question. I think we'll just have to agree to disagree for the time being.

Do consider, however, that bipolar transistors I-V relationships are based on exponential law while MOSFETs are (more or less) modeled with a square law. That is part of the reason that BJTs usually have a transconductance advantage. I'm also a bit skeptical about your current claims since you can compensate for current crowding pretty easily in the BJT if high current power electronics are your aim (I've helped design HBTs for high current applications). In general it's the bipolar transistor design that can drive more current because it is limited more by lower extrinsic series resistance than the pinch-off regime that exists in the MOSFET channel.

The IGBT comes to mind, which combines the voltage control and fast switching transient characteristics of a MOSFET with a BJT. This is specifically for high voltage/high current applications where it would be more difficult to create a MOSFET that could handle the required power. On a similar note, last I talked to my contacts at a RF power amplifier company, they use GaP/InGaP HBTs for their high power radio amps. Notice some of Skyworks' high current PAs; they use HBTs. You can be sure they'd be using cheaper transistors (like Si MOSFETs) if they could easily be fabricated to deliver the same kind of gain and power.

Anyway, as I said, I don't really want to continue this discussion since it's not very important to this article at this time. I just wanted to point out that in my humble experience, bipolar designs are inherently superior over MOSFETs for amplification and current driving. Take that for whatever it's worth to you. :) -- mattb @ 2006-11-28T00:26Z

Matt, I do appreciate your experience in this. I just like to be contrarian sometimes. Since a good chuck of my FET experience is in using them in the subthreshold (weak-inversion) region, where their transconductance per output current is almost as high as bipolars, and since that transconductance degrades only gradually as you go above threshold into the quadratic region, I allowed for a "small numerical factor" of BJT superiority there. And I certainly agree that you can get big advantages from exotic (non-Silicon) semiconductors, which are generally easier to make BJTs with than FETs. Anyway, no problem living with our differences. Thanks. Dicklyon 00:33, 28 November 2006 (UTC)

Thanks everyone, that's clear. I had looked it up being told that it was making HVDC transmission more viable, and somehow got the impression that it was a more complex device rather than just a kind of transistor.

[edit] Threshold symbol?

I changed the threshold voltage subscript in the triode/saturation/cutoff region because it was labeled incorrectly. For an n-channel MOSTFET, the threshold voltage is written as V_TN, not V_th.

Here are 76 books that disagree with you. Dicklyon 06:48, 30 November 2006 (UTC)

I've seen it written both ways and don't particularly care which convention is used as long as its consistant. -- mattb @ 2006-11-30T15:39Z

In book search, I find TH about twice as common as TN (is that because it's used for both types, as opposed to TN which is for n-type only?). Capitalization varies a lot. Dicklyon 15:46, 30 November 2006 (UTC)