Love your videos! As an analog IC designer, let me make some comments: most integrated technologies are build on a p-type substrate, which means that what you said about NMOS bulk being always connected to ground is true. Not for the PMOS though, they are built on an isolated n-well and they have a separate connection that can be tied to its source (another reason why sometimes PMOS input pairs are preferred over NMOS, other than 1/f noise). If it’s an n-type technology, it’s the other way around for the bulk connection. Also lots of technologies allow for triple well, which means that even the NMOS can be isolated. This is usually more expensive in terms of area, but sometimes necessary. Especially for more recent technology nodes, where headroom is a problem, and for mixed-signal designs(basically all designs today are mixed), as the switching noise from digital may couple via the substrate to the sensitive analog parts of a system on chip. Keep on the good work, and thanks for everything
YES, I was just about to write that. I worked as a layouter on analog IC chips for automotive industry and we used several different technologies that allowed for differently polarized n-well pockets and one technology that allowed splitting the substrate into few independent sections. This is somethimes a crucial component of the design and there are two main reasons why. The first one is, as mentioned by rscreepin, noise reduction. I remember routing an input amp section for integrated parking sensor, which was powered from it's own low-noise regulator. Second reason is the ability to work with "high" voltage (in our case, it was up to about 60V). Transistors have smaller breakdown voltage than the body diod of n-well to p-type substrate, but you can cascade them and process larger signals. Also, I would like to add to the 2:12 claim, that you wouldn't want to put the voltage the "wrong way" around. Somethimes, you would. The body diod of discrete FETs is often used for "active" rectifier aplication. For example pretty much every lithium battery protection circuit has a FET transistor, which is connected the "wrong way around" and works as a diode.
@@Lantertronics You may find this of interest: A Simple Zero Debounce Delay debouncer for an SPST switch spacetimepro.blogspot.com/2023/09/switch-debouncer.html
I have been designing semis since the mid 70's with Intel, S3 and Bell Labs. Great talk professor - Almost all this design work is now in the development tools. I haven't studied this in 50 years so thanks for the refresh which was very different than I had in University. Thanks for the video and also the references.
A few notes: 1.) For NMOS, yes, it’s the bulk tied to ground. For PMOS you have a well in which the bulk sits. Of course, other technologies using isolating substrates like quartz, diamond or sapphire can have specialised bulk connections tied to either + or -, there are no absolute rules in design besides the physical characteristics of the device. 2.) BJTs are actually better described as current controlled devices because the maximum current flowing through the collector is in a direct linear relationship with the number of holes or electrons moved by the base-emitter current. Also, technically speaking BJT could theoretically be symmetrical devices, but their characteristics would be garbage compared to asymmetrical ones doped differently. 3.) There is a transistor kinda like in-between a BJT and a MOSFET called a JFET. Those have excellent linear characteristics, are indeed voltage controlled devices in normal operation and can be used in the anti-pinch region too for a few tens or hundreds of millivolts. Unfortunately, most of those are normally ON devices. Some MOSFETS are designed as normally ON too but the doping of conductive channel section is a bit more special. 4.) This is actually another thing that’s almost exclusively seen in MOSFETs. They can behave either as normally ON or normally OFF devices, which is remarkable since other technologies usually don’t allow for such a degree of fine-tuning. This versatility is why we’re seeing MOSFETs in the widest array of applications, from power switching and signal processing to digital computing.
Random thoughts: 1) The RCA Cosmac 1802 has a silicon-on-sapphire version that's in Voyager spacecraft. I only mention this because few people have heard of the 1802 and I think it's really cool and want more people to have heard of it. :) 2) I much prefer the voltage-controlled viewpoint, since you're generally driving your circuits with a voltage. When I teach our junior analog electronics class I find it much easier to explain BJTs circuits in terms of transconductance, and the base current as an unfortunate defect that you just kind of have to live with. I have some more recent videos titled "BJTs are VOLTAGE-Controlled Current Sources!" and "More Story Time with Professor Lanterman (The BJT Control Controversy) where I blather on about this more. ;) Sometimes people say the current controlled viewpoint is better than the voltage controlled viewpoint because the Ic = Ib beta equation is linear whereas the Ebers-Moll equation is exponential. But I think that just shuffles the complexity elsewhere instead of really getting rid of it. And beta is kind of a dodgy anyway, being a function of Ic (beta droop) and a bunch of other stuff. 3) I'm kind of sad that JFETs seem to have gone out of style. I see them used as approximate variable resistors in compressors/limiters and as muting switches in mixing consoles, but the only other place I see them used seems to be guitar pedals.
@@Lantertronics I see you love JFETs too. Did you know some modern JFETs are actually used in high voltage switching applications where MOSFETs tend to get their oxide layer fried? eu.mouser.com/datasheet/2/412/UJ3N120035K3S_Data_Sheet-3177188.pdf Check out these fancy silicon carbide JFETs.
A quibble: There is nothing wrong with the reverse polarity connections on a discrete MOSFET. It is done all the time with P-channel MOSFETs serving as a reverse polarity protection diode. When power is connected correctly, the MOSFET is reverse connected and biased ON so there is very little drop. If the power connections are reversed, the part is biased "normally" and is off so no destructive current flows. In some analog work a MOSFET is used to effectively short a small AC voltage to ground or vary a resistance to ground over a range of values. This is done with very small signals.
@@Lantertronics synchronous buck/rectification circuits and ideal diodes in general make use of this kind of thing quite a bit. It's the new "hotness" where hot is actually cool lol
Hello prof. Lanterman! I have been watching your videos since the COVID quarantines (pedals/effects/synths are a personal hobby of mine) and I am very happy you touched on the EKV MOS model as I am pursuing my PhD with the author of the EKV2.6 model (and mastermind behind the EKV3 model)! And I was about to comment about asymmetrical devices, but you covered it!
I respectfully disagree. Vgs is still a useful approximation even for IC design since the bulk terminal is usually not changing too much. You do have to be careful about Vth varying with bulk voltage and account for that when doing the initial design (this can be almost entirely avoided with SOI processes, which also allow more freedom to adjust circuit parameters using the back gate). Additionally, the square law approximation is very back of the envelope for modern processes anyway, short channel effects like mobility degradation and velocity saturation are generally more important in my opinion.
Indeed, I probably should have mentioned SOI (Jennifer mentioned it in her list of bullet points I read from but I skipped over it). I'm not sure how common SOI is though?
@@Lantertronics The latest open source PDK from skywater is SOI (SKY90-FD), as are a few offerings from global foundries such as 22FDSOI. It seems like it is used more for analog/RF processes. I don't know of a digital-centric process that is SOI, but I am not well versed in all the processes out there.
@@sellicott The GF processes used for AMD K8 (first x86_64 hardware) up to AMD's Piledriver chips (32nm SOI, ~2012). Also IBM processes used for the original Xbox 360, Play Station 3, and Wii processors. Unless you count "state of the art microprocessor" not as "digital-centric", but I can't think of anything more classically digital-centric.
@@namibjDerEchte Cool I didn't know that. I work with RF/mixed signal designs myself, so I don't really follow what processes are used for the high-density digital flows. I am pretty confident the latest Intel/TSMC processes are not SOI, so I had not gotten the impression that SOI had less benefit for digital design (other than the the things that make it beneficial in general).
@@sellicott Modern 14nm and sub processes are all FinFETs. They are completely different ball game. Most of outdated university professors that teach basic courses that include transistor models have no clue about them in reality. And going sub 10 is yet another totally different world where equations don't make sense but only a very rough estimations.
2:18 "There is nothing stopping you from doing something pathological with it and putting the voltage the wrong way. Of course you don't want to do that." Having the source of an N-MOS positive relative to the drain is actually desirable when using them for power gating (common source pair for low-loss bidirectional control) in a BMS or other power distribution/management situation or when using them for synchronous rectification. You would likely also use this sort of setup in a bidirectional boost/buck converter. There are plenty of MOSFETs packaged specifically for those applications. It may not make sense for logic but is plenty useful for power handling.
I really should make a follow up video talking about that. When I was making the video I was thinking about small-signal amplifiers -- analog signal processing sort of things. So I (sadly) missed this entire important application area. I'd need to do a lot of study though, power electronics isn't my area.
Or I could cheat and point to another resource... do you know of another video here on UA-cam by another content creator that talks about such topics well?
@@Lantertronics I'm not aware of anybody doing deep-dive videos into power conversion circuitry, not much of a market for that on YT. I suspect most people looking for this sort of insight will follow the datasheet and appnotes rabbit hole. Almost certain Bigclive has at least one power bank or other gadget tear-down where he explains why many battery monitoring chips use a dual-FET packages to isolate the battery. Pretty sure I have seen a multi-cell BMS review that explained the differences between contactor and MOSFET disconnect. Synchronous rectification and bridge-less power factor correction are quite well documented, resources there are plentiful. Bidirectional DC-DC converters are somewhat of a rarer breed, though I suspect interest for it will proliferate with the need for USB-PD to accommodate a broader range of anything-to-anything at the sub-200W end of the power spectrum and DC power sharing between home energy storage and BEV at the 10+kW end.
That's why you can use mosfets for muting audio for example. There an external substrate/bulk terminal is very useful because you can connect it to a negative supply so the substrate formed diodes dont short the signal out (at least until it reaches the negative rail).
Its funny how at the same time I can grasp the concepts, the equations look unaproachable. I'm a simple man who designs and builds guitar effects for myself as a hobbie, so, my understanding of mosfets is basically: "you put a zenner in it so it wont fry from stactic, then you make 3 stages of it, and you get sweet distortion."
It's okay. I literally have a degree in this stuff and the math is still something that gets me. Especially since I haven't done this type of work in a decade, and those are graduate level courses he's referencing. On the other hand, undergrad didn't really cover MOSFETs. Something interesting about that distortion. Because there's a capacitor in the MOSFET, the time it takes to turn on is based on the gade source **current** (Please correct me if wrong). That's one of the reasons MOSFET drivers exist. You may be able to get a similar amount of distortion by using a current limiting resistor in front of a single MOSFET.
I"m going to suggest to her that she create some Playlists to help organize her videos -- right now her catalogue is hard to navigate. But there is so much content there. Also definitely check out Brad Minch's videos. :)
Good video! It's worth noting that the "body effect" you describe is more important for nMOS than for pMOS. PMOS transistors sit in their own n-well. So it is easier to connect source to the body. nMOS are typically just on the substrate..
flashback to VLSI class when we were trying to design standard cells and I coupled the substrate to the source in a cascode configuration instead of doing the correct thing and coupling the substrate to v- by default it all makes sense now
Even that point isn't exactly true either -- it's only the case in circuits that are powered. In the dark silicon of an IC, moss *does* accumulate over time. It's up to the designer to decide if minimizing the moss accumulation is worth suppressing at the expense of leakage current.
this is absolutely fantastic. i am a huge fan of learning how things really work after the rules of thumb you've been given in life start to break down. there is enormous value here. thank you for your efforts. i have hit that subscribe button so freaking hard.
I never understood the difference between drain and source until one day I hooked up a mosfet backwards and realized that turning on the gate still made it bridge the body diode. Was a fun thing to learn that day :)
Dang ... as of this month it's ecactly 40 years ago when I started my internship at Philips Semiconductors. I could not stop listening to some of the chip design gurus that were too willing to explain the design of the new high speed cmos stuff they developed. It's fun being reintroduced to this at a higher level - I am never too old to learn 😁 I have never looked at SOI processes but I guess that since this includes extra steps (you need to grow a silicon layer on the insulator), this is a more expensive process that will only be used for special applications like high speed analog / RF
For me it only matters how volt threshold is. How high the switching voltage can be. How much current I can switch. What the internal resistance is like. And whether the cooling capacity is sufficient. This is all in the datasheet. If the MOSFET can still be found in the craft box, everything works.
Two notes: in (analog) integration circuit design, although in most cases (yet not always), bulk does not have to be connected to V+/V- (if for some reason you want to affect Vt). Second - bjt is actually a (base) current driven device rather than voltage device (Vbe)
Technically that's true about BJTs, but I think when designing with BJTs it's a lot more natural to think of it as a voltage-controlled device, and think of the base current as kind of a parasitic effect (although of course it's fundamental to the operation of the BJT).
BJT being driven by the base current is a useful first approximation; but the relationship is between collector current and base-emitter voltage as you can see in the Ebers-Moll equation
The base current is actually just an undesirable side effect that you are forced to accept. The actual control is carried out by v_be. However, because the latter has an effect on this parasitic side effect, a derivation can be made on the basis of the side effect. Works in practice and many things in life are - rightly - simplified for practical purposes. It is anchored in the world of electrical engineering in many places, such as the simple example of energy transport by electricity. The actual transporter is the electromagnetic field, but electron flow is used in practice.
I really enjoyed your short seminar/lecture. I have occasionally done microelectronic design since the early 80s and always thought Vgs was such a hack when source and bulk were not tied together. Thanks for the refresher. Also going to look at the other great resources you have linked in. Cheers
This puts me in the mind of Newtonian physics: it works as a way to describe the stuff we can see, more or less closely enough to get the job done, but it doesn't actually cover all possible cases and even when it does work it's not absolutely accurate then either. But generally speaking we can forget that when you throw a baseball it gets slightly heavier.
The twist here is that I think the voltage-control view actually makes it *easier* to analyze most transistor circuits, if you teach it the way Marshall did. It' like if general relativity made teaching easier than Newtonian physics. ;)
Actually the MOSFET will allow current to flow in either direction between the terminals and there are many reasons to do so. Switching AC currents is a common one as well as active rectification which utilizes the intrinsic diode of the bulk connection to create low loss diode circuit out of the transistor. In this usage current is only permitted to flow through the the transistor when the current is flow in the reverse direction. Also note in many applications the source terminal may not be or must not be consider to be connected to a power rail but instead floats offset from the rail which means that gate bias must be referenced to the source to provide reliable turn-on and turn-off. Also consider that how the gate is formed on the transistor the capacitance between the designated source and gate terminals is not the same and gate to drain capacitance changes when the device conducts so the terminals are not arbitrary but you can still use the device with proper attention to the gate bias as though the terminals don't matter. But keep in mind the parasitic diode will conduct when forward biased.
I seem to recall that the idea for a field-effect transistor preceded the invention of the BJT. Maybe not - have to look it up. Not to forget (and not shown), most "street" MOSFETs have a parasitic drain-source diode, 0ften enhanced for reverse current carrying capability in switching circuits and also sometimes doped to Zener at the Vds rating. Also, the street variety pretty much the only ones I have and have used) will happily work with the drain-source voltages reversed. The important point is to apply the gate voltage relative to the source and the gate supply can be isolated by transformer or opto-coupler. Also, the gate-source is often reverse-biased to enhance switch-off time. And not to forget the body diode in those common models is always present. I've known some to wonder why the thing still conducts with a reverse drain-source voltage applied. Properly wired these devices make a great bi-directional switch. I've used them in many apparently "strange" ways, for instance in decorative battery night lights where the MOSFET is biased off while a small solar cell charges the battery during light hours and allows LED current to flow at night.
I really need to make a follow-up video about the parasitic diode and its use of the MOSFET in current switching applications. I was primarily thinking about linear applications and just totally overlooked this when making this video.
@@Lantertronics J-FETs are truly symmetrical. They're not normally used in digital circuits because the negative threshold voltage is a pain to work with.
In a symmetric BJT the emitter and the collector are also interchangeable. Even in a non-symmetric BJT the emitter and the collector can be interchanged. The BJT is not going to work as well, but it works!
Due to the fact that pinouts for small BJTs aren't standardised, I've accidentally built entire amplifier stages with collector and emitter reversed and only noticed by chance after they've been in operation for weeks 😳
1:46 If the bulk is tied to one of the terminals, that makes the terminals different. But you repeat that the only difference between the terminals is determined by the voltage difference between the terminals ,and nothing to do with the bulk connection. That seems unclear or contradictory. You mentioned "putting the voltage the wrong way", but if there's no difference between the terminals, then there's no "wrong way".
The Ebers-Moll equations relate *all* the currents and the voltages at the terminals. BJTs are usually introduced as "current-controlled devices," but that's not very helpful from the point of view of designing or analyzing circuits with BJTs. It's much more illuminating to view them as voltage-controlled devices, and just dealing with the base current as something of a parasitic effect (although of course current must be flowing through the base for the physics to work).
I mean, it's just easier to analyze a 2-domain system than a 3-domain system all of the time. It's simply that the bulk was the least consequential to get rid of. Though I suppose a discrete bulk could make for some interesting differential amplification applications 🤔
Hi Aaron, I am working on a design for a bidirectional high current switching system (Id should be positive or negative). In a perfect world this is feasible with 2 pmos transistors and abusing the intrinsic diode of one and the channel from the other. This presents the problem of diode drops that I would rather avoid. After seeing this it seems like biasing in terms of Vgb(ulk), not Vgs, a bidirectional switch with high current capabilities seems feasible? are there off the shelf options that you know about? Any designs you would be willing to show off? Do off the shelf parts exist where the bulk is accessible to the designer exist? Am I thinking about this wrong? Thank you.
I must confess I'm at the edge of my own understanding on this -- I'm basically parroting what I've learned from my colleague Jen Hasler. So I'd be hesitant to make any specific suggestions, especially when it comes to high currents (and Jen's area of expertise is low-power electronics). I've been looking at a company called Advanced Linear Devices, they make matched duals and quads that give access to the bulk connection (same for all on a chip). Don't know if they could handle high enough current -- if you needed to add "high" to the word "current" in describing your application, probably not -- but it might be worth checking out?
Thanks for the video. Technically speaking vacuum tubes have at least 5 terminals, they need power for the black body emitter 🙂 I guess what you are saying is Technically true for any AC device and the naming convention is likely there for simplicity and consistency when using them in the field. Something I noticed at university is many of the lecturers, particularly the ones with more letters before or after their names, didn't have the real world experience to appreciate how devices are actually used in the field, so aspects of their knowledge was not always meaningful to industry.
I recall vividly tying the gate to the substrate in every schematic of MoSFET I was working on. This was 45 years ago, when the devices had just reached the market. It is somewhat sad to know, that my drawings from the time at the school, today would be marked as erroneous....
Wild coincidence, I was just looking at those! Advanced Linear Devices makes some matched duals and quads that give access to the bulk connections. They even have ones that use "floating gate" technology to let you precisely program the threshold voltage (you can get ones that you can program and you can also buy them pre-programmed). I plan to make a video about them at some point.
Nice thanks this makes sense. I have a FPAA kit, never tried it tho. I always was confused with common base / gate circuits. It does seem more intuitive to think of a 4 terminal device, the Gate to bulk voltage opening up the channel , and p vs. N is just the polarity of that gate to bulk voltage. And you're saying for analog vs cmos you have to think this way. Sounds like there should be some d/dVs and d/dVg terms instead of making the expression for Vthresh so complicated?
Yeah, in Jennifer's course she writes down the full nonlinear equations and writes like V0 + Vdelta, etc., and it turns out the math often works out so you can find things like gain without explicitly linearizing anything. It's wild, and I haven't seen things done that way in any textbooks I have.
Thanks for sharing, this video was quite insightful and it left me questioning some things, particularly CMOS gate design. Why is it that we only use PMOS as pullup and only NMOS as pulldown? I was content with the Vgs explanation (PMOS can only pull down to Vth-Vss, NMOS can only pullup to Vdd-Vth), but I didnt realize I was making an assumption that didn't apply to CMOS ICs generally, if the bulk isn't shorted to source (and vgs no longer applies), why are CMOS circuits designed the way they are?
Well, I do find it fun and interesting, to see different professors say the same thing, in different ways. Sometimes those differences in description subtend decades. I don't feel that I was ever misled. Having had the privilege of working at both the integrated level and the discreet level, I kind of knew all this, maybe because I have a tendency to remember little details like this. The happy, handy vgs formulas apply only when the bulk is tied to the source. For discreet MOSFETs, that is the case. On-chip, we can get that functionality only by giving the MOSFET its own bulk tub for source connection. For logic circuits, in which the MOSFETs of a particular flavor (N or P) share a bulk well, we just live with the fact that most sources are segregated from the bulk, making sure that the last signal to arrive at the logic gate is driving the MOSFET with the source tied to that common tub. In analogue design, the source should be tied to the bulk for best performance and predictability. But adequate models will allow predictable design without that accommodation. Thanks for mentioning the Bulk-centered, symmetrical EKV model. It continues to serve us well, and the modeling equations are a thing of beauty.
2:40 correct me if i misunderstood the sentence, but in fact there are some addons inside of the mosfets - for example relieving the parasitic capacitance, aren't they?
Your point about the Drain and Source being the same is only true for certain MOSFETS. You need to clarify that you are talking about a low voltage device constructed using a lateral structure. If you look at higher voltage vertical structures the Drain and Source are very different. Also the same applies for your description of the Bulk substrate.
Interesting. I made a post somewhere about n-channel depletion-mode MOSFETs being used as high-side switches. There is a pervasive view that a charge pump or a higher rail voltage is needed to trigger the gate by creating a vgs higher than the voltage being switched. The way this is usually explained is totally wrong. The MOSFET does not know what is going on in the rest of the circuit. Say we're building a H-bridge with identical MOSFETs. For one positive rail and one ground in the whole circuit - the so-called "charge pump" to be used to "act as a voltage doubler" does nothing of the sort. In those explanations, those arguments measure voltage at the pump relative to circuit ground - which is irrelevant to the MOSFET. What that charge pump capacitor actually does is to create a floating ground at the source terminal of the high-side MOSFET (as far as the MOSFET experiences). To that MOSFET, vgs is then larger than the threshold. It is useful to think about charge alone at the source and the gate rather than voltage and current. These are charge-controlled devices not voltage controlled as normally said.
In her analog lectures that I'm sitting in on, colleague Jennifer Hasler talks often about the capacitive aspects of MOSFETs and the benefits of sometimes thinking about charge instead of voltage.
That's all true, but for practical purposes, with an enhancement device switching a this high side, the gate voltage needs to exceed the power supply. A depletion mode device would solve that problem for turning on, but it would require its gate voltage to be pulled below the negative supply, to fully switch off.
@@alunjones3860 I'll think about what you said further. However, in my lab is a H-bridge commutating a BLDC motor using my method. In further testing I will see if I can measure a lack of switching off completely and other transient effects.
@@projector7141 A depletion mode device requires the gate to be below the source voltage to fully switch off. If your circuit doesn't achieve this, the device will not turn off. It's that simple. Whether this is a problem or not depends on what you're doing.
Well, it's only valid if you let the threshold voltage itself be a function of the source-to-bulk voltage. So the *meaning* of Vgs would be constantly changing with the source voltage. It's not meaningful in the way that Vbe is for a BJT or Vgk is for a tube.
Interesting! At age 75 my recollection is that when I first encountered BJTs in the late 1950s as a prepubescent boy, most 'popular electronics' magazine descriptions of the BJT considered it to be a current-controlled device, and beta [the ratio Ic/Ib] was the most important parameter [after max Vce, of course]. This was all very strange to me who had initially learned how vacuum tubes worked as grid voltage controlled devices. In those dark ages no-one in my circle of friends measured currents, only voltages. We all lusted after a real VTVM with its near-infinite input impedance! Steep learning curves...
Aaaaah... I really wish I could send a message back in time and ask the marketers of BJTs to *not* describe them as current controlled current sources. That would make sense if you were driving the base with a current source, but you're almost always driving it with a voltage, and it's more natural to think of the BJT as voltage controlled, and the base current as being something akin to a parasitic effect form the point of view of the designer (albeit one inherent to the operation of the BJT).
@@Lantertronics It kinda makes sense to describe BJTs as current-controlled, due to the non-linear characteristics of the diode between base and emitter. This way, describing a BJT transistor as voltage-controlled becomes harder/less predictable to work with unless you always assume the resistor often added at the emitter or in series with the base. In isolation, yes the base-emitter voltage does have something to say, but the base-emitter current can vary a LOT, with very little variation in base-emitter voltage once you are out of the diode cutoff region.
@@TheMovieCreator You could describe it either way, but you're usually driving the base with a voltage source and operating in some approximately linear region around some DC bias point anyway. People often site the simpler Ic = beta Ib relationship as the reason to describe it as a current source, but it's not really helpful in designing since you're not generally thinking about driving the base with a current source. Generally you want the base-emitter current to vary a lot, since by definition that's your gain. :)
@@Lantertronics I think the difference in perspective here comes from looking at a transistor as an isolated component vs a component in circuit. While most BJT amplifiers are driven by a voltage source indeed, an amplifier has additional support-components that will affect what assumptions you can make. Say for example, if you bias for class A, you assume a constant 0.7V difference across base to emitter, regardless of the input voltage of the amplifier as a whole. The voltage swing/gain of the output is then determined as the voltage over a resistor *external* to the transistor. The transistor dictates the current through (and as such voltage over) this resistor, but the base of the transistor itself should still remain quite fixed at approximately 0.7V away from the variable voltage at the emitter. While in the overall circuit as a whole, the current through the resistor is regulated by the voltage into the amplifier, the signal driven through the transistor as seen as an isolated component is a swing in current. So while the amplifier for sure is voltage-controlled, and has voltage gain, the BJT transistor in isolation of everything else is still current controlled according to my line of thought. The fact that this implementation is voltage-controlled stem from the application the transistor is put in, and is not an inherited parameter of the component itself.
hi there, i used a mosfet once, in a relais replacement cirquite board design. it still did not work. question for ya. is a mosfet a group of SOLID compuonands wich you throw with 10 at them at every noob design. in amy random configuration. and the somehow... grow themself in it. ? and choose their function accordi gly to give the designer a good feeling without lowing to much fuses ?
I wouldn't suggest hooking up a bunch MOSFETs down and hoping they do what you want... you definitely want to think through what you want to do and what each MOSFET is doing. That said, I've seen some distortion pedal designs by Dani Ever that seem to defy any kind of conventional analysis. They really do look like she just randomly hooked up transistors until she got something that sounded interesting to her.
I think the EVK model is a good way of approaching MOSFETs. It is simple enough to be understood intuitively and gives good results in simulations. Also, it references everything to the bulk, this is important to intutively understand MOSFETs.
Wild coincidence, I just recorded a few snippets about the EKV model for an upcoming video. :) Yup, my colleagues Jennifer Hasler and Brad Minch are all about the EKV model and referencing everything to the bulk.
There are *several distinct types of MOSFET.* I am here to talk about some similarities and some differences among them, between them, and the older electronic device types. Some MOSFETs are more like Bipolar Transistors, while others are more like Vacuum Tubes, in the sense of whether the device is Normally ON or Normally OFF with respect to Current Flow. Vacuum Tubes are Normally ON, meaning a disconnected Grid will let the full current flow between Cathode and Plate, while Bipolar Transistors are Normally Off, being in the high-resistance state when Base *current* is not flowing. Among MOSFETs, usually only the Enhancement-mode MOSFETs are discussed, for instance as common switching elements in most integrated circuits and discrete MOSFETs. These devices are Normally OFF, that is, minimal current flows at zero gate-source *voltage.* _This is slightly similar but not quite like a current-driven Bipolar Transistor;_ neither is it like a Vacuum Tube. Meanwhile with a depletion-mode MOSFET the device is Normally ON at zero gate-source voltage, making them rather analogous to a Vacuum Tube, aside from being able to work at rather low voltages. The Normally ON nature of a depletion-mode MOSFET can be uniquely effective in certain low-power, low-voltage situations such as Joule Thief circuits, where the nominal 1.5-volt alkaline battery can be run down much farther than when a bipolar transistor is used as the switching element. A similar circuit may be found useful for bootstrapping from half a volt to the several volts needed to turn on Enhancement-mode MOSFETs.
That's true, but the only really difference is the threshold voltage, with a depletion mode device having a negative threshold. It's possible to use an enhancement more MOSFET in the depletion region, by biasing the gate below the source, in which case the leakage current will often be lower and a depletion mode device can also be operated in the enhancement region, which often gives it a lower drain resistance.
that pesky intrinsic silicon reverse drain-source diode that's strapped to source with most off-the-shelf parts... an application having a FET that fully switches in both directions would have to be in a situation with operating Vds swings less than a diode drop (over full temp. range). Maybe some kind of crazy AC millivoltage switch/clamp. Back to the central point of video- I can't say I'd ever been affected by the 'Vg-s' vs 'gate-bulk' issue. I've so tightly controled the Vgs conservatively anyway to prevent pesky miller switch-off things happening over all load conditions which includes squegging after a single point failure, this kind of swamps out the Vgs-th issue :-)
That's all interesting, but are there 4-terminal MOSFETs on the market at all? I can't remember if I've seen one in last 10 years of active electronics design. Maybe there's some special parts out there, but almost all of MOSFETs on the market are 3-terminal devices without separate bulk connection. The only 4-termonal device I've encountered is a double-gate one for RF application. Please, name a few 4-termnal FETs I can find on like Digikey or Mouser, it would be curious to read their datasheets.
I probably should have been clearer in the video -- this issue applies to analog VLSI design where the bulks connections are all hooked to one of the rails by construction. I pickup up this rant from Jennifer Hasler (I'm sitting in on her class this semester) who feels like she's constantly having to untrain students from habits used to discrete MOSFETs. Technically speaking most CMOS op amps or whatever you buy have these issues going on, but some other designer has thought through all that for you so can just use the chip as a black box. There are MOSFETs with bulk connections available -- matched pairs and quads from Analog Linear Devices. I've seen them here and there at various vendors, but the vendor that seems to have the biggest stock is Digikey. They all have part numbers starting with ALD, and you can find info about them on the ALD website. I'm looking at those since I'd like to experiment with analog integrated circuit design techniques on the breadboard instead of just simulation. They have some parts that use "floating gate" technology that let you *program* what threshold voltage you want. You can buy ones that you put in a programmer, or buy ones that are pre-programmed for various threshold voltages (including zero!)
1:35. "The collector is filed differently from the emitter". I don't think that's true, collector and emitter are doped either as N type material or P type material.
Wow, a most interesting video. Now, I need a video to explain which class is good for what design application? Like AI or switching power inverters, etc.? 😎 Thank you.
The Ebers-Moll equation equations relate *all* the voltages and currents at the terminals. BJTs are usually introduced as being current-in, current-out devices, but from the point of view of designing with BJTs, it's much more useful to think of them as a voltage-controlled device, and deal with the base current as something of a parasitic effect that you just have to deal with (of course, current flowing through the base is fundamental to the physics of the device).
Ic = exp (Vbe /VT) (1 + Vce / VA) = βIb . Both Ibe and vbe correlates with Ic. It's both current-controlled current source and voltage-controlled current source. You just can pick one of the two models depending on your usage, e.g. impedances you drive the circuit with. Cited from Ali Hajimiri's lectures on BJT
good points; these are not nuances as the implied conditions for the various equations and notation used are rarely explained between discrete and integrated devices ;-)
It may not be for everyone. But there's not a lot of resources here on UA-cam for those who crave deeper knowledge, and I hope to help fill that gap. :)
@@Lantertronics True, I actually just received my final marks an hour or so ago. So I just became an electronic engineer 🤣🤣🤣. Yeah, in my advanced studies, the resources are quite scarce. Then you mostly rely on academic scholarly work. Sometimes I did wish that UA-cam had some other stuff. So I appreciate what you are doing.
“Modern MOSFETs - anything in the last 30 years.” I graduated from Tech with a EE in ‘91. Hard to believe my knowledge is no longer “modern” Serves me right for thinking when I graduated that 60s engineering was old and not used.
What is the electron current and the hole current? No one can really explain it. They write so many formulas, but no one can explain it in human language.
You've hit on the reason transistors are so much more complicated to understand than tubes. I can imagine the electrons "boiling" (not a technically correct term but an evocative one) off the cathode and being attracted to the plate, and being repelled by a negative potential on the grid. But once you get into the solid state physics it's so much more complicated.
This is really focused on designs where you have multiple MOSFETs on the same die. It's not really about single MOSFETs. I should have been clearer about that from the outset. Sorry to have wasted your time.
7:18 there is a fault in the line. bjy means bipolar junction transistor right? so she should not type "approaches used for older BipolarJunctionTransistor transistors and" Man i hate when my electronic friend say "shell i use transistor or mosftet" damn such idiotos.
I really didn't like this video. Just didn't get the point of it.... it made a big fuss about some pedantic points, and then totally failed to explain why it would matter, to whom it would matter, and (most importantly for electrical engineers) when the blue smoke would be released. I frequently design with mosfets and based on this video there is nothing I would do differently...
This applies to the use of MOSFETs in an integrated circuit, where the bulk connection needs to be taken into account. I can see regular Vgs calculations not applying if there's a component between the source of the MOSFET and V-, such as another MOSFET, if the bulk is fixed to V-. A discrete MOSFET has the bulk connected to the source internally, so this problem doesn't apply.
I guess this is important when you design something that relies on linear characteristics of a fet. Most people use them just as switches and/or pwm control.
@@helmutlord4335I share this sentiment, and I personally wouldn’t have phrased things quite this way, but I didn’t read this as necessarily impolite or out of bounds. I think this kind of frank honesty is entirely appropriate when it’s sincere, and (this is my intuition, not an educated or informed position) probably clearer to non-native English speakers as well as more directly useful as feedback to the author. Constructive criticism is one of the first things I learned formally as a film student, and it’s carried over a lot of valuable skills into my technical career, I wish more fields spent time teaching students how to consume media, think critically about it, and identify where the ideas being presented could be more effective and relevant to the audience. It’s been interesting to see how this has played out with UA-cam as an interactive medium. Many thanks to the Professor for sharing, and OP for expanding my own understanding of what I am watching.
I like the content, but the title and thumbnail are click bait. Most people that use something don't need to fully understand how it works if they can use a simplfied model. I feel like you really understated how this is only applicable if you are designing/manufacturing the mosfets
Fair point. I changed the "BAD!" in the thumbnail to "Misleading!" -- indeed, there's times where using Vgs is fine; I'm just wanting people to think more about what's behind it.
I wind wind up feeling compelled to use such clickbate-ish techniques, because for better or or worse -- they work. This channel isn't monitored through ad revenue, so I'm not seeking clicks for my personal finances, I'm seeking clicks because I really want to share these ideas from engineering with people. This video had over 1,000 views in less than 10 hours, and I know if I had titled it "Using Vgs is Only Fully Valid for Discrete MOSFETs" or "The Body Effect is a Hack, Here's a Better Way to View It" it wouldn't have gotten nearly that much attention. UA-cam encourages content creators this direction by reporting thinks like click-through rate. Actually, this kind of pressure from the "algorithm" and what it does to the thinking process of content creators would be an interesting video in its own right.
It is applicable when you are designing ICs really. Or when you are using the one and only CD4007. With that chip, you can actually measure and experimentally confirm that Vgs is a misleading term. IC manufacturers also use “4007-ish” setups to bond out transistors from test wafers used to characterize the process.
@@absurdengineering Advanced Linear Devices have some matched duals and quads with a shared substrate too -- they're like a refined version of the CD4007 where you have access to all the terminals (they aren't pre-wired like in the 4007).
@Lantertronics that's neat. I haven't seen the extra substrate connections on an IC before. Also, I understand the struggle to try and get more views. That's unfortunately part of being on UA-cam. Thanks for considering my opinion, though
Don't worry. It is only worth knowing if you need to know it. It is just a little device used in electronics to either amplify or be used as a switch. You only really need to know about it if you design or repair electronics. It is like you don't need to know what a camshaft is to drive a car - only if you work with repairing or designing engines etc.
See my latest videos titled "BJTs are VOLTAGE-Controlled Current Sources!" and "More Story Time with Professor Lanterman (The BJT Control Controversy)" that address this issue.
@@Lantertronics Modern CPUs have never advanced in performance and energy efficiency because of physics. My 2022 Dell G15 gaming laptop uses Intel Core i5, using Alder Lake architecture based on FinFET technology.
Aaron, the voltage between base and emitter (Vbe) does not change much after 0.66V for silicon as it is just a p-n. Bipolar transistors are driven by the current, not voltage. Are you really a "professor"?
I wish I could go back in time and find the people originally describing the BJT as a current-controlled current source and make them change. They've caused a tremendous amount of confusion. The Ebers-Moll equations relate all the currents and voltages at the terminal of the BJT. It's not that calling it a current-controlled current source is wrong, it's that it's just as valid to say it's a voltage-controlled current source, and it's much more helpful in designing and analyzing circuits to think of it that way. Most circuits apply the input as a voltage at the base, not as a current source. The traditional form of the Ebers-Moll equations give Ic as an exponential function of Vbe. It's best to flip your "does not change much after 0.66V for silicon" which is a current-focused view around -- you may bias your circuit around 0.66V, but the small signal variation gives you a large change in current for a small change of voltage at the base, which is great! That's how you get gain! And then the current flowing through the base is something more like a parasitic effect the designer just has to deal with. Remember in most transistor circuits you are using a feedback mechanism to stabilize the gain to a particular value, and in that case as beta becomes higher (more ideal), the base current goes to zero.
I would invite you to explore two resources. One is the ECE3050 Analog Electronics website created by my colleague Marshall Leach (google it, sometimes UA-cam deletes links). You will see how powerful it is to view the voltages as inputs.
The other resource I would invite you to explore is my ECE3400 Analog Electronics playlist; I teach it using Marshall Leach's approach, which is to develop Theveinin and Norton equivalent circuits looking into the various terminals of the BJT in terms of Thevenin equivalent circuits seen looking out the base and the emitter. You will see how powerful the voltage-focused viewpoint is.
Consider the resistor. Is it a voltage-controlled current source or a current-controlled voltage source? Both views are technically correct, but which is more useful depends on what's driving it.
@@Lantertronics Resistor is a passive component, so it is never a source of anything. You are too caught up with the equations and carried away from reality. Try feeding a voltage source of 3V without resistor into the base and see BJT burn. Try feeding the same 3V to a gate of JFET, MOS or a lamp and they will handle it ok with conceptually similar output. Therefore, bipolar transistors operate way differently from lamps and FETs. BJT will never open if no current flows into the base. FET works fine with zero current through the gate. It is basic physics, not even college level. So, Aaron, put away the math for a month or two, get some soldering and measuring equipment and real semiconductors. You will be surprised how things are sometimes very different from SPICE.
I don’t think the premise is very valid. The gate voltage, or charge, or whatever you’d like to call it, in relation to the source voltage controls the inversion state of the channel. Further, saying “there are more complexities than the square law” is patently obvious and mentioned in textbooks. Third, you provide no solution such as is done at Berkeley with the gm/Id methodology. Teaching kids to use their brain and the simulator is a lot harder than following some arbitrary model like EKV. How much experience do you have with FinFet design? The body effect becomes less and less relevant for modern bulk transistors, not more and more, and with decreasing supply it’s harder and harder to cascode. Also don’t agree that the bulk has to be connected to VCC/VSS, most modern CMOS processes offer triple well as a standard option.
Wow, its almost like in high school you learn about Newton's gravity and in college you learn about general relativity. They mislead us about gravity!!!! What's next? square next, square law modelling isn't true? The fact that Vs has a different coupling effect makes no big issues in designing other than body bias effect. Saying you will be marked wrong if you use Vgs is downright obnoxious and assholer-y
Since Covid, really... a lot of schools went online. Most of my colleagues taught classes live over Zoom, which I think generally doesn't work very well. A computer screen is a fundamentally different medium than a live lecture and attempts to educate via it need to be adapted to that format. I leaned into it and learned to edit video as I went.
This channel isn’t part of the UA-cam Partners program. I don’t receive ad revenue; I do that so I can remain focused on my education mission. If I am creative with my titles it’s because I want to help people to learn electronics.
@@focu625 And I am sorry if this video didn't live up to your expectations. When I do use clickbaity titles, I try to put extra work into making sure the video delivers. If you are interested in electronics (I assume that's why you clicked on it), I'd invite you to check out my ECE3400 Analog Electronics playlist. The lectures are more straightforwardly titled. ;)
None of this makes any difference to me as an engineer. Warning to Engineers: You're about to waste 9 minutes by watching this. This is why I can't stand physicists (I view this guy as more of a physicist than an engineer). Endless rambling about nothing without any practical value.
It's quite relevant to any engineers thinking about integrated circuit design, which I probably should have made clear up front. I am sorry to have wasted your time. I'd invite you to check out some my ECE3400: Analog Electronics playlist, which focuses on BJTs and may be more relevant to your interests. Or, if you are interested in audio applications, you can check out my "Analog Circuits for Music Synthesis" or "Guitar Amplification and Effects" playlists.
Love your videos! As an analog IC designer, let me make some comments: most integrated technologies are build on a p-type substrate, which means that what you said about NMOS bulk being always connected to ground is true. Not for the PMOS though, they are built on an isolated n-well and they have a separate connection that can be tied to its source (another reason why sometimes PMOS input pairs are preferred over NMOS, other than 1/f noise). If it’s an n-type technology, it’s the other way around for the bulk connection. Also lots of technologies allow for triple well, which means that even the NMOS can be isolated. This is usually more expensive in terms of area, but sometimes necessary. Especially for more recent technology nodes, where headroom is a problem, and for mixed-signal designs(basically all designs today are mixed), as the switching noise from digital may couple via the substrate to the sensitive analog parts of a system on chip. Keep on the good work, and thanks for everything
Pinning this comment. Thanks!
YES, I was just about to write that. I worked as a layouter on analog IC chips for automotive industry and we used several different technologies that allowed for differently polarized n-well pockets and one technology that allowed splitting the substrate into few independent sections. This is somethimes a crucial component of the design and there are two main reasons why. The first one is, as mentioned by rscreepin, noise reduction. I remember routing an input amp section for integrated parking sensor, which was powered from it's own low-noise regulator. Second reason is the ability to work with "high" voltage (in our case, it was up to about 60V). Transistors have smaller breakdown voltage than the body diod of n-well to p-type substrate, but you can cascade them and process larger signals.
Also, I would like to add to the 2:12 claim, that you wouldn't want to put the voltage the "wrong way" around. Somethimes, you would. The body diod of discrete FETs is often used for "active" rectifier aplication. For example pretty much every lithium battery protection circuit has a FET transistor, which is connected the "wrong way around" and works as a diode.
@@Lantertronics You may find this of interest: A Simple Zero Debounce Delay debouncer for an SPST switch
spacetimepro.blogspot.com/2023/09/switch-debouncer.html
There is SOI processes
I have been designing semis since the mid 70's with Intel, S3 and Bell Labs. Great talk professor - Almost all this design work is now in the development tools. I haven't studied this in 50 years so thanks for the refresh which was very different than I had in University. Thanks for the video and also the references.
A few notes:
1.) For NMOS, yes, it’s the bulk tied to ground. For PMOS you have a well in which the bulk sits. Of course, other technologies using isolating substrates like quartz, diamond or sapphire can have specialised bulk connections tied to either + or -, there are no absolute rules in design besides the physical characteristics of the device.
2.) BJTs are actually better described as current controlled devices because the maximum current flowing through the collector is in a direct linear relationship with the number of holes or electrons moved by the base-emitter current. Also, technically speaking BJT could theoretically be symmetrical devices, but their characteristics would be garbage compared to asymmetrical ones doped differently.
3.) There is a transistor kinda like in-between a BJT and a MOSFET called a JFET. Those have excellent linear characteristics, are indeed voltage controlled devices in normal operation and can be used in the anti-pinch region too for a few tens or hundreds of millivolts. Unfortunately, most of those are normally ON devices. Some MOSFETS are designed as normally ON too but the doping of conductive channel section is a bit more special.
4.) This is actually another thing that’s almost exclusively seen in MOSFETs. They can behave either as normally ON or normally OFF devices, which is remarkable since other technologies usually don’t allow for such a degree of fine-tuning. This versatility is why we’re seeing MOSFETs in the widest array of applications, from power switching and signal processing to digital computing.
Random thoughts:
1) The RCA Cosmac 1802 has a silicon-on-sapphire version that's in Voyager spacecraft. I only mention this because few people have heard of the 1802 and I think it's really cool and want more people to have heard of it. :)
2) I much prefer the voltage-controlled viewpoint, since you're generally driving your circuits with a voltage. When I teach our junior analog electronics class I find it much easier to explain BJTs circuits in terms of transconductance, and the base current as an unfortunate defect that you just kind of have to live with. I have some more recent videos titled "BJTs are VOLTAGE-Controlled Current Sources!" and "More Story Time with Professor Lanterman (The BJT Control Controversy) where I blather on about this more. ;)
Sometimes people say the current controlled viewpoint is better than the voltage controlled viewpoint because the Ic = Ib beta equation is linear whereas the Ebers-Moll equation is exponential. But I think that just shuffles the complexity elsewhere instead of really getting rid of it. And beta is kind of a dodgy anyway, being a function of Ic (beta droop) and a bunch of other stuff.
3) I'm kind of sad that JFETs seem to have gone out of style. I see them used as approximate variable resistors in compressors/limiters and as muting switches in mixing consoles, but the only other place I see them used seems to be guitar pedals.
@@Lantertronics
I see you love JFETs too.
Did you know some modern JFETs are actually used in high voltage switching applications where MOSFETs tend to get their oxide layer fried?
eu.mouser.com/datasheet/2/412/UJ3N120035K3S_Data_Sheet-3177188.pdf
Check out these fancy silicon carbide JFETs.
A quibble: There is nothing wrong with the reverse polarity connections on a discrete MOSFET.
It is done all the time with P-channel MOSFETs serving as a reverse polarity protection diode. When power is connected correctly, the MOSFET is reverse connected and biased ON so there is very little drop. If the power connections are reversed, the part is biased "normally" and is off so no destructive current flows.
In some analog work a MOSFET is used to effectively short a small AC voltage to ground or vary a resistance to ground over a range of values. This is done with very small signals.
Good point!
@@Lantertronics synchronous buck/rectification circuits and ideal diodes in general make use of this kind of thing quite a bit. It's the new "hotness" where hot is actually cool lol
Hello prof. Lanterman! I have been watching your videos since the COVID quarantines (pedals/effects/synths are a personal hobby of mine) and I am very happy you touched on the EKV MOS model as I am pursuing my PhD with the author of the EKV2.6 model (and mastermind behind the EKV3 model)!
And I was about to comment about asymmetrical devices, but you covered it!
That's awesome! Who is your advisor?
@@Lantertronics prof. Matthias Bucher, at the Technical University of Crete.
Greetings from Greece by the way!
I respectfully disagree. Vgs is still a useful approximation even for IC design since the bulk terminal is usually not changing too much. You do have to be careful about Vth varying with bulk voltage and account for that when doing the initial design (this can be almost entirely avoided with SOI processes, which also allow more freedom to adjust circuit parameters using the back gate). Additionally, the square law approximation is very back of the envelope for modern processes anyway, short channel effects like mobility degradation and velocity saturation are generally more important in my opinion.
Indeed, I probably should have mentioned SOI (Jennifer mentioned it in her list of bullet points I read from but I skipped over it). I'm not sure how common SOI is though?
@@Lantertronics The latest open source PDK from skywater is SOI (SKY90-FD), as are a few offerings from global foundries such as 22FDSOI. It seems like it is used more for analog/RF processes. I don't know of a digital-centric process that is SOI, but I am not well versed in all the processes out there.
@@sellicott The GF processes used for AMD K8 (first x86_64 hardware) up to AMD's Piledriver chips (32nm SOI, ~2012). Also IBM processes used for the original Xbox 360, Play Station 3, and Wii processors.
Unless you count "state of the art microprocessor" not as "digital-centric", but I can't think of anything more classically digital-centric.
@@namibjDerEchte Cool I didn't know that. I work with RF/mixed signal designs myself, so I don't really follow what processes are used for the high-density digital flows. I am pretty confident the latest Intel/TSMC processes are not SOI, so I had not gotten the impression that SOI had less benefit for digital design (other than the the things that make it beneficial in general).
@@sellicott Modern 14nm and sub processes are all FinFETs. They are completely different ball game. Most of outdated university professors that teach basic courses that include transistor models have no clue about them in reality. And going sub 10 is yet another totally different world where equations don't make sense but only a very rough estimations.
Thanks very much for covering this material and for the useful references to the works of Professor Minch and Professor Hasler.
2:18 "There is nothing stopping you from doing something pathological with it and putting the voltage the wrong way. Of course you don't want to do that." Having the source of an N-MOS positive relative to the drain is actually desirable when using them for power gating (common source pair for low-loss bidirectional control) in a BMS or other power distribution/management situation or when using them for synchronous rectification. You would likely also use this sort of setup in a bidirectional boost/buck converter. There are plenty of MOSFETs packaged specifically for those applications.
It may not make sense for logic but is plenty useful for power handling.
I really should make a follow up video talking about that. When I was making the video I was thinking about small-signal amplifiers -- analog signal processing sort of things. So I (sadly) missed this entire important application area.
I'd need to do a lot of study though, power electronics isn't my area.
Or I could cheat and point to another resource... do you know of another video here on UA-cam by another content creator that talks about such topics well?
@@Lantertronics I'm not aware of anybody doing deep-dive videos into power conversion circuitry, not much of a market for that on YT. I suspect most people looking for this sort of insight will follow the datasheet and appnotes rabbit hole. Almost certain Bigclive has at least one power bank or other gadget tear-down where he explains why many battery monitoring chips use a dual-FET packages to isolate the battery. Pretty sure I have seen a multi-cell BMS review that explained the differences between contactor and MOSFET disconnect.
Synchronous rectification and bridge-less power factor correction are quite well documented, resources there are plentiful.
Bidirectional DC-DC converters are somewhat of a rarer breed, though I suspect interest for it will proliferate with the need for USB-PD to accommodate a broader range of anything-to-anything at the sub-200W end of the power spectrum and DC power sharing between home energy storage and BEV at the 10+kW end.
That's why you can use mosfets for muting audio for example. There an external substrate/bulk terminal is very useful because you can connect it to a negative supply so the substrate formed diodes dont short the signal out (at least until it reaches the negative rail).
Its funny how at the same time I can grasp the concepts, the equations look unaproachable. I'm a simple man who designs and builds guitar effects for myself as a hobbie, so, my understanding of mosfets is basically: "you put a zenner in it so it wont fry from stactic, then you make 3 stages of it, and you get sweet distortion."
It's okay. I literally have a degree in this stuff and the math is still something that gets me. Especially since I haven't done this type of work in a decade, and those are graduate level courses he's referencing. On the other hand, undergrad didn't really cover MOSFETs.
Something interesting about that distortion. Because there's a capacitor in the MOSFET, the time it takes to turn on is based on the gade source **current** (Please correct me if wrong). That's one of the reasons MOSFET drivers exist. You may be able to get a similar amount of distortion by using a current limiting resistor in front of a single MOSFET.
Woah I didn’t know Dr Halser has her own UA-cam channel with lecture material! Will definitely take a look!
I"m going to suggest to her that she create some Playlists to help organize her videos -- right now her catalogue is hard to navigate. But there is so much content there. Also definitely check out Brad Minch's videos. :)
Good video! It's worth noting that the "body effect" you describe is more important for nMOS than for pMOS. PMOS transistors sit in their own n-well. So it is easier to connect source to the body. nMOS are typically just on the substrate..
flashback to VLSI class when we were trying to design standard cells and I coupled the substrate to the source in a cascode configuration instead of doing the correct thing and coupling the substrate to v- by default
it all makes sense now
They won’t tell you this in the data sheet, but mosfets contain absolutely no moss.
You win Best Comment on this video. :)
Even that point isn't exactly true either -- it's only the case in circuits that are powered. In the dark silicon of an IC, moss *does* accumulate over time. It's up to the designer to decide if minimizing the moss accumulation is worth suppressing at the expense of leakage current.
this is absolutely fantastic. i am a huge fan of learning how things really work after the rules of thumb you've been given in life start to break down. there is enormous value here. thank you for your efforts. i have hit that subscribe button so freaking hard.
Welcome! Check out my ECE3400: Analog Electronics playlist and my Analog Circuits for Music Synthesis playlist.
Huge thanks for a list of recommendations! I was interested in FPAAs some time ago but never really delved deeper into them.
You are very welcome! Enjoy!
I never understood the difference between drain and source until one day I hooked up a mosfet backwards and realized that turning on the gate still made it bridge the body diode. Was a fun thing to learn that day :)
Dang ... as of this month it's ecactly 40 years ago when I started my internship at Philips Semiconductors. I could not stop listening to some of the chip design gurus that were too willing to explain the design of the new high speed cmos stuff they developed. It's fun being reintroduced to this at a higher level - I am never too old to learn 😁
I have never looked at SOI processes but I guess that since this includes extra steps (you need to grow a silicon layer on the insulator), this is a more expensive process that will only be used for special applications like high speed analog / RF
For me it only matters how volt threshold is.
How high the switching voltage can be.
How much current I can switch.
What the internal resistance is like.
And whether the cooling capacity is sufficient.
This is all in the datasheet.
If the MOSFET can still be found in the craft box, everything works.
Two notes: in (analog) integration circuit design, although in most cases (yet not always), bulk does not have to be connected to V+/V- (if for some reason you want to affect Vt). Second - bjt is actually a (base) current driven device rather than voltage device (Vbe)
Technically that's true about BJTs, but I think when designing with BJTs it's a lot more natural to think of it as a voltage-controlled device, and think of the base current as kind of a parasitic effect (although of course it's fundamental to the operation of the BJT).
@@Lantertronicsbut the voltage doesn't really vary over a meaningful range between one amp and exploding, where the current is nearly linear
@@jxtq27 I'm not sure I'm following you. Check out my ECE3400: Analog Electronics lectures here on UA-cam to get a sense of what I'm talking about.
BJT being driven by the base current is a useful first approximation; but the relationship is between collector current and base-emitter voltage as you can see in the Ebers-Moll equation
The base current is actually just an undesirable side effect that you are forced to accept. The actual control is carried out by v_be. However, because the latter has an effect on this parasitic side effect, a derivation can be made on the basis of the side effect. Works in practice and many things in life are - rightly - simplified for practical purposes.
It is anchored in the world of electrical engineering in many places, such as the simple example of energy transport by electricity. The actual transporter is the electromagnetic field, but electron flow is used in practice.
I really enjoyed your short seminar/lecture. I have occasionally done microelectronic design since the early 80s and always thought Vgs was such a hack when source and bulk were not tied together. Thanks for the refresher. Also going to look at the other great resources you have linked in. Cheers
This puts me in the mind of Newtonian physics: it works as a way to describe the stuff we can see, more or less closely enough to get the job done, but it doesn't actually cover all possible cases and even when it does work it's not absolutely accurate then either.
But generally speaking we can forget that when you throw a baseball it gets slightly heavier.
The twist here is that I think the voltage-control view actually makes it *easier* to analyze most transistor circuits, if you teach it the way Marshall did. It' like if general relativity made teaching easier than Newtonian physics. ;)
I always worked with discrete mosfet components (from miliamps to hundred of amps), is the first time I see what means all the lines in the mosfet
Actually the MOSFET will allow current to flow in either direction between the terminals and there are many reasons to do so. Switching AC currents is a common one as well as active rectification which utilizes the intrinsic diode of the bulk connection to create low loss diode circuit out of the transistor. In this usage current is only permitted to flow through the the transistor when the current is flow in the reverse direction. Also note in many applications the source terminal may not be or must not be consider to be connected to a power rail but instead floats offset from the rail which means that gate bias must be referenced to the source to provide reliable turn-on and turn-off.
Also consider that how the gate is formed on the transistor the capacitance between the designated source and gate terminals is not the same and gate to drain capacitance changes when the device conducts so the terminals are not arbitrary but you can still use the device with proper attention to the gate bias as though the terminals don't matter. But keep in mind the parasitic diode will conduct when forward biased.
Eurorack? Engineering stuff I don't really understand but want to learn? Yeah that's an instant subscribe for me.
I seem to recall that the idea for a field-effect transistor preceded the invention of the BJT. Maybe not - have to look it up. Not to forget (and not shown), most "street" MOSFETs have a parasitic drain-source diode, 0ften enhanced for reverse current carrying capability in switching circuits and also sometimes doped to Zener at the Vds rating. Also, the street variety pretty much the only ones I have and have used) will happily work with the drain-source voltages reversed. The important point is to apply the gate voltage relative to the source and the gate supply can be isolated by transformer or opto-coupler. Also, the gate-source is often reverse-biased to enhance switch-off time. And not to forget the body diode in those common models is always present. I've known some to wonder why the thing still conducts with a reverse drain-source voltage applied. Properly wired these devices make a great bi-directional switch. I've used them in many apparently "strange" ways, for instance in decorative battery night lights where the MOSFET is biased off while a small solar cell charges the battery during light hours and allows LED current to flow at night.
You are right, the FET came first, but the folks working on it had trouble getting it working, which led to the BJT.
I really need to make a follow-up video about the parasitic diode and its use of the MOSFET in current switching applications. I was primarily thinking about linear applications and just totally overlooked this when making this video.
@@Lantertronics J-FETs are truly symmetrical. They're not normally used in digital circuits because the negative threshold voltage is a pain to work with.
In a symmetric BJT the emitter and the collector are also interchangeable. Even in a non-symmetric BJT the emitter and the collector can be interchanged. The BJT is not going to work as well, but it works!
Due to the fact that pinouts for small BJTs aren't standardised, I've accidentally built entire amplifier stages with collector and emitter reversed and only noticed by chance after they've been in operation for weeks 😳
BJTs are often used in reverse, for example as a bidirectional digital voltage translator.
This is restricted only to small signal or rf mosfets, power mosfets have allways the bulk connected to either drain or source terminals.
What I’m curious about is what solid state component is analogous to a pentode? It would seem there isn’t one.
To my knowledge there isn't one.
1:46 If the bulk is tied to one of the terminals, that makes the terminals different. But you repeat that the only difference between the terminals is determined by the voltage difference between the terminals ,and nothing to do with the bulk connection. That seems unclear or contradictory. You mentioned "putting the voltage the wrong way", but if there's no difference between the terminals, then there's no "wrong way".
The tube and FET are voltage controlled, the BJT is current controlled, not voltage controlled.
The Ebers-Moll equations relate *all* the currents and the voltages at the terminals. BJTs are usually introduced as "current-controlled devices," but that's not very helpful from the point of view of designing or analyzing circuits with BJTs. It's much more illuminating to view them as voltage-controlled devices, and just dealing with the base current as something of a parasitic effect (although of course current must be flowing through the base for the physics to work).
Lectures 6 and 7 from my ECE3400 video series here on UA-cam dig into this more.
GaN FET HEMT devices are on the other hand mostly symmetrical and their reverse current characteristics are usually documented in datasheets
so the "really" is analogous to "akshually" in software...very informational video.
Thanks!
I mean, it's just easier to analyze a 2-domain system than a 3-domain system all of the time. It's simply that the bulk was the least consequential to get rid of. Though I suppose a discrete bulk could make for some interesting differential amplification applications 🤔
Thank you Sir!
You are welcome!
I've been told that mosfets only block in one direction. That suggests that mosfets are asymmetrical.
Hi Aaron, I am working on a design for a bidirectional high current switching system (Id should be positive or negative). In a perfect world this is feasible with 2 pmos transistors and abusing the intrinsic diode of one and the channel from the other. This presents the problem of diode drops that I would rather avoid. After seeing this it seems like biasing in terms of Vgb(ulk), not Vgs, a bidirectional switch with high current capabilities seems feasible? are there off the shelf options that you know about? Any designs you would be willing to show off? Do off the shelf parts exist where the bulk is accessible to the designer exist? Am I thinking about this wrong? Thank you.
I must confess I'm at the edge of my own understanding on this -- I'm basically parroting what I've learned from my colleague Jen Hasler. So I'd be hesitant to make any specific suggestions, especially when it comes to high currents (and Jen's area of expertise is low-power electronics).
I've been looking at a company called Advanced Linear Devices, they make matched duals and quads that give access to the bulk connection (same for all on a chip). Don't know if they could handle high enough current -- if you needed to add "high" to the word "current" in describing your application, probably not -- but it might be worth checking out?
@@Lantertronics Thanks for the detailed response! I'll look into this! For me, high current is like 5A lol.
@@TheSlowerMonkey Heh, yeah that would turn the ADL devices into a pile of melted plastic in about a nanosecond. ;)
Thanks for the video. Technically speaking vacuum tubes have at least 5 terminals, they need power for the black body emitter 🙂
I guess what you are saying is Technically true for any AC device and the naming convention is likely there for simplicity and consistency when using them in the field.
Something I noticed at university is many of the lecturers, particularly the ones with more letters before or after their names, didn't have the real world experience to appreciate how devices are actually used in the field, so aspects of their knowledge was not always meaningful to industry.
I recall vividly tying the gate to the substrate in every schematic of MoSFET I was working on. This was 45 years ago, when the devices had just reached the market.
It is somewhat sad to know, that my drawings from the time at the school, today would be marked as erroneous....
You really put us on game with those recommendations sir
Awesome. Thank you for this.
This is a really clear video thanks
You are welcome! :)
Pretty sure you can get discrete 4-terminal FETs for those adventurous designers! Interesting video, thanks.
Wild coincidence, I was just looking at those! Advanced Linear Devices makes some matched duals and quads that give access to the bulk connections. They even have ones that use "floating gate" technology to let you precisely program the threshold voltage (you can get ones that you can program and you can also buy them pre-programmed). I plan to make a video about them at some point.
Nice thanks this makes sense. I have a FPAA kit, never tried it tho.
I always was confused with common base / gate circuits. It does seem more intuitive to think of a 4 terminal device, the Gate to bulk voltage opening up the channel , and p vs. N is just the polarity of that gate to bulk voltage.
And you're saying for analog vs cmos you have to think this way.
Sounds like there should be some d/dVs and d/dVg terms instead of making the expression for Vthresh so complicated?
Yeah, in Jennifer's course she writes down the full nonlinear equations and writes like V0 + Vdelta, etc., and it turns out the math often works out so you can find things like gain without explicitly linearizing anything. It's wild, and I haven't seen things done that way in any textbooks I have.
Thanks for sharing, this video was quite insightful and it left me questioning some things, particularly CMOS gate design. Why is it that we only use PMOS as pullup and only NMOS as pulldown? I was content with the Vgs explanation (PMOS can only pull down to Vth-Vss, NMOS can only pullup to Vdd-Vth), but I didnt realize I was making an assumption that didn't apply to CMOS ICs generally, if the bulk isn't shorted to source (and vgs no longer applies), why are CMOS circuits designed the way they are?
Well, I do find it fun and interesting, to see different professors say the same thing, in different ways. Sometimes those differences in description subtend decades. I don't feel that I was ever misled. Having had the privilege of working at both the integrated level and the discreet level, I kind of knew all this, maybe because I have a tendency to remember little details like this. The happy, handy vgs formulas apply only when the bulk is tied to the source. For discreet MOSFETs, that is the case. On-chip, we can get that functionality only by giving the MOSFET its own bulk tub for source connection. For logic circuits, in which the MOSFETs of a particular flavor (N or P) share a bulk well, we just live with the fact that most sources are segregated from the bulk, making sure that the last signal to arrive at the logic gate is driving the MOSFET with the source tied to that common tub. In analogue design, the source should be tied to the bulk for best performance and predictability. But adequate models will allow predictable design without that accommodation. Thanks for mentioning the Bulk-centered, symmetrical EKV model. It continues to serve us well, and the modeling equations are a thing of beauty.
2:40 correct me if i misunderstood the sentence, but in fact there are some addons inside of the mosfets - for example relieving the parasitic capacitance, aren't they?
but for power mosfets does we have the bulk? And in power mosfets drain is in the back (vertical mosfet).
Great video!
Thanks!
Your point about the Drain and Source being the same is only true for certain MOSFETS. You need to clarify that you are talking about a low voltage device constructed using a lateral structure. If you look at higher voltage vertical structures the Drain and Source are very different. Also the same applies for your description of the Bulk substrate.
I mention it at the very end, but you're right, I should have made that clear up front.
Have to scroll tooo far to find this comment among all the fanboys cheering author who actually mislead them
Interesting. I made a post somewhere about n-channel depletion-mode MOSFETs being used as high-side switches. There is a pervasive view that a charge pump or a higher rail voltage is needed to trigger the gate by creating a vgs higher than the voltage being switched. The way this is usually explained is totally wrong. The MOSFET does not know what is going on in the rest of the circuit. Say we're building a H-bridge with identical MOSFETs. For one positive rail and one ground in the whole circuit - the so-called "charge pump" to be used to "act as a voltage doubler" does nothing of the sort. In those explanations, those arguments measure voltage at the pump relative to circuit ground - which is irrelevant to the MOSFET. What that charge pump capacitor actually does is to create a floating ground at the source terminal of the high-side MOSFET (as far as the MOSFET experiences). To that MOSFET, vgs is then larger than the threshold. It is useful to think about charge alone at the source and the gate rather than voltage and current. These are charge-controlled devices not voltage controlled as normally said.
In her analog lectures that I'm sitting in on, colleague Jennifer Hasler talks often about the capacitive aspects of MOSFETs and the benefits of sometimes thinking about charge instead of voltage.
That's all true, but for practical purposes, with an enhancement device switching a this high side, the gate voltage needs to exceed the power supply. A depletion mode device would solve that problem for turning on, but it would require its gate voltage to be pulled below the negative supply, to fully switch off.
@@alunjones3860 I'll think about what you said further. However, in my lab is a H-bridge commutating a BLDC motor using my method. In further testing I will see if I can measure a lack of switching off completely and other transient effects.
@@projector7141 A depletion mode device requires the gate to be below the source voltage to fully switch off. If your circuit doesn't achieve this, the device will not turn off. It's that simple. Whether this is a problem or not depends on what you're doing.
really insightful and thanks by the way😅
Thank you for your kind words!
So, in that model VBS changes the depletion capacitance, and thus the kappa?
It’s mentioned that Vgs =Vg-Vs is never valid for bulk devices.
I missed something; how can this not be correct?
Well, it's only valid if you let the threshold voltage itself be a function of the source-to-bulk voltage. So the *meaning* of Vgs would be constantly changing with the source voltage. It's not meaningful in the way that Vbe is for a BJT or Vgk is for a tube.
@@Lantertronics there’s something wrong here. By definition Vab = Va- Vb .
So I think the idea you are trying to get across could be worded better.
@@yjweaver5108 Indeed, perhaps a better way to say it would be it's "not useful."
Theoretically i know this in case of discrete mosfets. Integrated circuit info is somewhat new but logical from the stand point of technology
I'm hoping over time more IC design will become available to the masses (maybe through FPAAs).
Interesting! At age 75 my recollection is that when I first encountered BJTs in the late 1950s as a prepubescent boy, most 'popular electronics' magazine descriptions of the BJT considered it to be a current-controlled device, and beta [the ratio Ic/Ib] was the most important parameter [after max Vce, of course]. This was all very strange to me who had initially learned how vacuum tubes worked as grid voltage controlled devices. In those dark ages no-one in my circle of friends measured currents, only voltages. We all lusted after a real VTVM with its near-infinite input impedance! Steep learning curves...
Aaaaah... I really wish I could send a message back in time and ask the marketers of BJTs to *not* describe them as current controlled current sources. That would make sense if you were driving the base with a current source, but you're almost always driving it with a voltage, and it's more natural to think of the BJT as voltage controlled, and the base current as being something akin to a parasitic effect form the point of view of the designer (albeit one inherent to the operation of the BJT).
@@Lantertronics It kinda makes sense to describe BJTs as current-controlled, due to the non-linear characteristics of the diode between base and emitter. This way, describing a BJT transistor as voltage-controlled becomes harder/less predictable to work with unless you always assume the resistor often added at the emitter or in series with the base. In isolation, yes the base-emitter voltage does have something to say, but the base-emitter current can vary a LOT, with very little variation in base-emitter voltage once you are out of the diode cutoff region.
@@TheMovieCreator You could describe it either way, but you're usually driving the base with a voltage source and operating in some approximately linear region around some DC bias point anyway. People often site the simpler Ic = beta Ib relationship as the reason to describe it as a current source, but it's not really helpful in designing since you're not generally thinking about driving the base with a current source. Generally you want the base-emitter current to vary a lot, since by definition that's your gain. :)
@@TheMovieCreator I should make another video about this.
@@Lantertronics I think the difference in perspective here comes from looking at a transistor as an isolated component vs a component in circuit. While most BJT amplifiers are driven by a voltage source indeed, an amplifier has additional support-components that will affect what assumptions you can make.
Say for example, if you bias for class A, you assume a constant 0.7V difference across base to emitter, regardless of the input voltage of the amplifier as a whole. The voltage swing/gain of the output is then determined as the voltage over a resistor *external* to the transistor. The transistor dictates the current through (and as such voltage over) this resistor, but the base of the transistor itself should still remain quite fixed at approximately 0.7V away from the variable voltage at the emitter. While in the overall circuit as a whole, the current through the resistor is regulated by the voltage into the amplifier, the signal driven through the transistor as seen as an isolated component is a swing in current. So while the amplifier for sure is voltage-controlled, and has voltage gain, the BJT transistor in isolation of everything else is still current controlled according to my line of thought. The fact that this implementation is voltage-controlled stem from the application the transistor is put in, and is not an inherited parameter of the component itself.
hi there, i used a mosfet once, in a relais replacement cirquite board design. it still did not work. question for ya. is a mosfet a group of SOLID compuonands wich you throw with 10 at them at every noob design. in amy random configuration. and the somehow... grow themself in it. ? and choose their function accordi gly to give the designer a good feeling without lowing to much fuses ?
I wouldn't suggest hooking up a bunch MOSFETs down and hoping they do what you want... you definitely want to think through what you want to do and what each MOSFET is doing. That said, I've seen some distortion pedal designs by Dani Ever that seem to defy any kind of conventional analysis. They really do look like she just randomly hooked up transistors until she got something that sounded interesting to her.
I think the EVK model is a good way of approaching MOSFETs. It is simple enough to be understood intuitively and gives good results in simulations. Also, it references everything to the bulk, this is important to intutively understand MOSFETs.
Wild coincidence, I just recorded a few snippets about the EKV model for an upcoming video. :)
Yup, my colleagues Jennifer Hasler and Brad Minch are all about the EKV model and referencing everything to the bulk.
There are *several distinct types of MOSFET.* I am here to talk about some similarities and some differences among them, between them, and the older electronic device types. Some MOSFETs are more like Bipolar Transistors, while others are more like Vacuum Tubes, in the sense of whether the device is Normally ON or Normally OFF with respect to Current Flow. Vacuum Tubes are Normally ON, meaning a disconnected Grid will let the full current flow between Cathode and Plate, while Bipolar Transistors are Normally Off, being in the high-resistance state when Base *current* is not flowing.
Among MOSFETs, usually only the Enhancement-mode MOSFETs are discussed, for instance as common switching elements in most integrated circuits and discrete MOSFETs. These devices are Normally OFF, that is, minimal current flows at zero gate-source *voltage.* _This is slightly similar but not quite like a current-driven Bipolar Transistor;_ neither is it like a Vacuum Tube.
Meanwhile with a depletion-mode MOSFET the device is Normally ON at zero gate-source voltage, making them rather analogous to a Vacuum Tube, aside from being able to work at rather low voltages.
The Normally ON nature of a depletion-mode MOSFET can be uniquely effective in certain low-power, low-voltage situations such as Joule Thief circuits, where the nominal 1.5-volt alkaline battery can be run down much farther than when a bipolar transistor is used as the switching element. A similar circuit may be found useful for bootstrapping from half a volt to the several volts needed to turn on Enhancement-mode MOSFETs.
That's true, but the only really difference is the threshold voltage, with a depletion mode device having a negative threshold. It's possible to use an enhancement more MOSFET in the depletion region, by biasing the gate below the source, in which case the leakage current will often be lower and a depletion mode device can also be operated in the enhancement region, which often gives it a lower drain resistance.
Interesting!@@alunjones3860
that pesky intrinsic silicon reverse drain-source diode that's strapped to source with most off-the-shelf parts... an application having a FET that fully switches in both directions would have to be in a situation with operating Vds swings less than a diode drop (over full temp. range). Maybe some kind of crazy AC millivoltage switch/clamp.
Back to the central point of video- I can't say I'd ever been affected by the 'Vg-s' vs 'gate-bulk' issue. I've so tightly controled the Vgs conservatively anyway to prevent pesky miller switch-off things happening over all load conditions which includes squegging after a single point failure, this kind of swamps out the Vgs-th issue :-)
I think "squeezing" may be my favorite engineering term.
That's all interesting, but are there 4-terminal MOSFETs on the market at all? I can't remember if I've seen one in last 10 years of active electronics design. Maybe there's some special parts out there, but almost all of MOSFETs on the market are 3-terminal devices without separate bulk connection. The only 4-termonal device I've encountered is a double-gate one for RF application. Please, name a few 4-termnal FETs I can find on like Digikey or Mouser, it would be curious to read their datasheets.
I probably should have been clearer in the video -- this issue applies to analog VLSI design where the bulks connections are all hooked to one of the rails by construction. I pickup up this rant from Jennifer Hasler (I'm sitting in on her class this semester) who feels like she's constantly having to untrain students from habits used to discrete MOSFETs. Technically speaking most CMOS op amps or whatever you buy have these issues going on, but some other designer has thought through all that for you so can just use the chip as a black box.
There are MOSFETs with bulk connections available -- matched pairs and quads from Analog Linear Devices. I've seen them here and there at various vendors, but the vendor that seems to have the biggest stock is Digikey. They all have part numbers starting with ALD, and you can find info about them on the ALD website. I'm looking at those since I'd like to experiment with analog integrated circuit design techniques on the breadboard instead of just simulation.
They have some parts that use "floating gate" technology that let you *program* what threshold voltage you want. You can buy ones that you put in a programmer, or buy ones that are pre-programmed for various threshold voltages (including zero!)
Can you share the document link?
Good idea! I just added it to the description.
1:35. "The collector is filed differently from the emitter". I don't think that's true, collector and emitter are doped either as N type material or P type material.
Reversed collector and emitter behaves terribly, because the doping densities are different
Thanks a lot for this (again) great explanation!
You are welcome!
If drain Source doesn't matter, then just connect them in reverse
Big FET doesn't want you to know - wake up, people!!!
Wow, a most interesting video. Now, I need a video to explain which class is good for what design application? Like AI or switching power inverters, etc.? 😎 Thank you.
Good video, but I take note that a bipolar is current driven and thus characterized by the base current I (be) not V (be).
The Ebers-Moll equation equations relate *all* the voltages and currents at the terminals. BJTs are usually introduced as being current-in, current-out devices, but from the point of view of designing with BJTs, it's much more useful to think of them as a voltage-controlled device, and deal with the base current as something of a parasitic effect that you just have to deal with (of course, current flowing through the base is fundamental to the physics of the device).
Lecture 6 and 7 from my ECE3400 lectures series here on UA-cam goes into this in gory detail. ;)
Ic = exp (Vbe /VT) (1 + Vce / VA) = βIb . Both Ibe and vbe correlates with Ic. It's both current-controlled current source and voltage-controlled current source. You just can pick one of the two models depending on your usage, e.g. impedances you drive the circuit with. Cited from Ali Hajimiri's lectures on BJT
i have been enlightned
good points; these are not nuances as the implied conditions for the various equations and notation used are rarely explained between discrete and integrated devices ;-)
I am an electronic engineer, I think this knowledge is not for everyone. Most people are just hobbyists.
It may not be for everyone. But there's not a lot of resources here on UA-cam for those who crave deeper knowledge, and I hope to help fill that gap. :)
@@Lantertronics True, I actually just received my final marks an hour or so ago. So I just became an electronic engineer 🤣🤣🤣. Yeah, in my advanced studies, the resources are quite scarce. Then you mostly rely on academic scholarly work. Sometimes I did wish that UA-cam had some other stuff. So I appreciate what you are doing.
I did not know this.
we were MOSled?
hah!
@@Lantertronics It's kind of dissapointing on how few bad jokes are in the comments section.
Do they still teach Murphy's Law in Engineering?
“Modern MOSFETs - anything in the last 30 years.” I graduated from Tech with a EE in ‘91. Hard to believe my knowledge is no longer “modern” Serves me right for thinking when I graduated that 60s engineering was old and not used.
Next semester I'm teaching my "Guitar Amplification and Effects" class so I can stay in the 1960s.
Ah! Did you have any classes with Marshall Leach or Phil Allen?
@@LantertronicsMarshall Leach - a great teacher!
What is the electron current and the hole current? No one can really explain it. They write so many formulas, but no one can explain it in human language.
You've hit on the reason transistors are so much more complicated to understand than tubes. I can imagine the electrons "boiling" (not a technically correct term but an evocative one) off the cathode and being attracted to the plate, and being repelled by a negative potential on the grid. But once you get into the solid state physics it's so much more complicated.
@@Lantertronics What I realized is that I didn't understand anything 😁
Who is the "who" of which you speak?
"The Secrets Big Transistor Doesn't Want You to Know" ;)
@@Lantertronics its either that or you have no idea, I prefer it's the you have no idea option
And the power mosfet with the usual reverse diode ...?
Your video does not help me one single LSB understanding and using mosfets in daily terms.
This is really focused on designs where you have multiple MOSFETs on the same die. It's not really about single MOSFETs. I should have been clearer about that from the outset. Sorry to have wasted your time.
7:18 there is a fault in the line. bjy means bipolar junction transistor right? so she should not type "approaches used for older BipolarJunctionTransistor transistors and"
Man i hate when my electronic friend say "shell i use transistor or mosftet" damn such idiotos.
MOSFETs are more powerful than the bipolar comments... There I said it.
I really didn't like this video. Just didn't get the point of it.... it made a big fuss about some pedantic points, and then totally failed to explain why it would matter, to whom it would matter, and (most importantly for electrical engineers) when the blue smoke would be released. I frequently design with mosfets and based on this video there is nothing I would do differently...
This applies to the use of MOSFETs in an integrated circuit, where the bulk connection needs to be taken into account. I can see regular Vgs calculations not applying if there's a component between the source of the MOSFET and V-, such as another MOSFET, if the bulk is fixed to V-. A discrete MOSFET has the bulk connected to the source internally, so this problem doesn't apply.
Why respond in such a critical tone?
This is really addressed towards cases when you have more than one MOSFET in a device. I probably could have done a better job of explaining that.
I guess this is important when you design something that relies on linear characteristics of a fet. Most people use them just as switches and/or pwm control.
@@helmutlord4335I share this sentiment, and I personally wouldn’t have phrased things quite this way, but I didn’t read this as necessarily impolite or out of bounds. I think this kind of frank honesty is entirely appropriate when it’s sincere, and (this is my intuition, not an educated or informed position) probably clearer to non-native English speakers as well as more directly useful as feedback to the author.
Constructive criticism is one of the first things I learned formally as a film student, and it’s carried over a lot of valuable skills into my technical career, I wish more fields spent time teaching students how to consume media, think critically about it, and identify where the ideas being presented could be more effective and relevant to the audience. It’s been interesting to see how this has played out with UA-cam as an interactive medium. Many thanks to the Professor for sharing, and OP for expanding my own understanding of what I am watching.
News flash: a triode has FIVE leads, not three.
True, but only three are interesting in terms of signals. Two of them just make it a light bulb. I'm not saying those two aren't important, though. ;)
@@Lantertronics Da devvil's in da details. 8)
I like the content, but the title and thumbnail are click bait. Most people that use something don't need to fully understand how it works if they can use a simplfied model. I feel like you really understated how this is only applicable if you are designing/manufacturing the mosfets
Fair point. I changed the "BAD!" in the thumbnail to "Misleading!" -- indeed, there's times where using Vgs is fine; I'm just wanting people to think more about what's behind it.
I wind wind up feeling compelled to use such clickbate-ish techniques, because for better or or worse -- they work. This channel isn't monitored through ad revenue, so I'm not seeking clicks for my personal finances, I'm seeking clicks because I really want to share these ideas from engineering with people. This video had over 1,000 views in less than 10 hours, and I know if I had titled it "Using Vgs is Only Fully Valid for Discrete MOSFETs" or "The Body Effect is a Hack, Here's a Better Way to View It" it wouldn't have gotten nearly that much attention. UA-cam encourages content creators this direction by reporting thinks like click-through rate. Actually, this kind of pressure from the "algorithm" and what it does to the thinking process of content creators would be an interesting video in its own right.
It is applicable when you are designing ICs really. Or when you are using the one and only CD4007. With that chip, you can actually measure and experimentally confirm that Vgs is a misleading term. IC manufacturers also use “4007-ish” setups to bond out transistors from test wafers used to characterize the process.
@@absurdengineering Advanced Linear Devices have some matched duals and quads with a shared substrate too -- they're like a refined version of the CD4007 where you have access to all the terminals (they aren't pre-wired like in the 4007).
@Lantertronics that's neat. I haven't seen the extra substrate connections on an IC before. Also, I understand the struggle to try and get more views. That's unfortunately part of being on UA-cam. Thanks for considering my opinion, though
I dont even know what a MOSFET is. Oh the depths of my lack of knowledge arrrgh!!! 😔
I highly recommend the free "Introduction to Electronics" course my colleague Bonnie Ferri has on Coursera.
www.coursera.org/learn/electronics
Don't worry. It is only worth knowing if you need to know it. It is just a little device used in electronics to either amplify or be used as a switch.
You only really need to know about it if you design or repair electronics. It is like you don't need to know what a camshaft is to drive a car - only if you work with repairing or designing engines etc.
Not to scare you, but billions if not trillions of MOSFETs know who _you_ are.
@@ericbwertzYea but they are only tiny little bits of slightly polluted silicon. They have to gain up to be capable of any semi intelligent danger.
BJTs are rather current controlled
See my latest videos titled "BJTs are VOLTAGE-Controlled Current Sources!" and "More Story Time with Professor Lanterman (The BJT Control Controversy)" that address this issue.
Quantum tunneling in planar MOSFETs
Alas quantum physics is largely magic to me. I've been trying to learn some because I'm interested in quantum computing but it's pretty overwhelming.
@@Lantertronics Modern CPUs have never advanced in performance and energy efficiency because of physics. My 2022 Dell G15 gaming laptop uses Intel Core i5, using Alder Lake architecture based on FinFET technology.
Cool
🙌
I know I'm starting to get into the proper nerd side of UA-cam when I see clickbait titles about transistors.
Aaron, the voltage between base and emitter (Vbe) does not change much after 0.66V for silicon as it is just a p-n. Bipolar transistors are driven by the current, not voltage. Are you really a "professor"?
I wish I could go back in time and find the people originally describing the BJT as a current-controlled current source and make them change. They've caused a tremendous amount of confusion.
The Ebers-Moll equations relate all the currents and voltages at the terminal of the BJT. It's not that calling it a current-controlled current source is wrong, it's that it's just as valid to say it's a voltage-controlled current source, and it's much more helpful in designing and analyzing circuits to think of it that way. Most circuits apply the input as a voltage at the base, not as a current source.
The traditional form of the Ebers-Moll equations give Ic as an exponential function of Vbe. It's best to flip your "does not change much after 0.66V for silicon" which is a current-focused view around -- you may bias your circuit around 0.66V, but the small signal variation gives you a large change in current for a small change of voltage at the base, which is great! That's how you get gain! And then the current flowing through the base is something more like a parasitic effect the designer just has to deal with.
Remember in most transistor circuits you are using a feedback mechanism to stabilize the gain to a particular value, and in that case as beta becomes higher (more ideal), the base current goes to zero.
I would invite you to explore two resources. One is the ECE3050 Analog Electronics website created by my colleague Marshall Leach (google it, sometimes UA-cam deletes links). You will see how powerful it is to view the voltages as inputs.
The other resource I would invite you to explore is my ECE3400 Analog Electronics playlist; I teach it using Marshall Leach's approach, which is to develop Theveinin and Norton equivalent circuits looking into the various terminals of the BJT in terms of Thevenin equivalent circuits seen looking out the base and the emitter. You will see how powerful the voltage-focused viewpoint is.
Consider the resistor. Is it a voltage-controlled current source or a current-controlled voltage source? Both views are technically correct, but which is more useful depends on what's driving it.
@@Lantertronics Resistor is a passive component, so it is never a source of anything. You are too caught up with the equations and carried away from reality. Try feeding a voltage source of 3V without resistor into the base and see BJT burn. Try feeding the same 3V to a gate of JFET, MOS or a lamp and they will handle it ok with conceptually similar output. Therefore, bipolar transistors operate way differently from lamps and FETs. BJT will never open if no current flows into the base. FET works fine with zero current through the gate. It is basic physics, not even college level. So, Aaron, put away the math for a month or two, get some soldering and measuring equipment and real semiconductors. You will be surprised how things are sometimes very different from SPICE.
I don’t think the premise is very valid. The gate voltage, or charge, or whatever you’d like to call it, in relation to the source voltage controls the inversion state of the channel. Further, saying “there are more complexities than the square law” is patently obvious and mentioned in textbooks. Third, you provide no solution such as is done at Berkeley with the gm/Id methodology. Teaching kids to use their brain and the simulator is a lot harder than following some arbitrary model like EKV.
How much experience do you have with FinFet design? The body effect becomes less and less relevant for modern bulk transistors, not more and more, and with decreasing supply it’s harder and harder to cascode.
Also don’t agree that the bulk has to be connected to VCC/VSS, most modern CMOS processes offer triple well as a standard option.
Damn... Watched this thinking it was about Prophets...
I do have a video where I talk about one of those. :)
ua-cam.com/video/cX_yUtzHzQU/v-deo.html
Wow, its almost like in high school you learn about Newton's gravity and in college you learn about general relativity. They mislead us about gravity!!!! What's next? square next, square law modelling isn't true? The fact that Vs has a different coupling effect makes no big issues in designing other than body bias effect.
Saying you will be marked wrong if you use Vgs is downright obnoxious and assholer-y
So how long has there been this secret cult of professors on UA-cam? Back in my day you had to pay for your education!
Since Covid, really... a lot of schools went online. Most of my colleagues taught classes live over Zoom, which I think generally doesn't work very well. A computer screen is a fundamentally different medium than a live lecture and attempts to educate via it need to be adapted to that format. I leaned into it and learned to edit video as I went.
you mislead us with your title so that you can advertise.....cheap tricks....
This channel isn’t part of the UA-cam Partners program. I don’t receive ad revenue; I do that so I can remain focused on my education mission. If I am creative with my titles it’s because I want to help people to learn electronics.
@@Lantertronics ok i am sorry....Wish you best of luck then....
@@focu625 And I am sorry if this video didn't live up to your expectations. When I do use clickbaity titles, I try to put extra work into making sure the video delivers.
If you are interested in electronics (I assume that's why you clicked on it), I'd invite you to check out my ECE3400 Analog Electronics playlist. The lectures are more straightforwardly titled. ;)
None of this makes any difference to me as an engineer. Warning to Engineers: You're about to waste 9 minutes by watching this. This is why I can't stand physicists (I view this guy as more of a physicist than an engineer). Endless rambling about nothing without any practical value.
It's quite relevant to any engineers thinking about integrated circuit design, which I probably should have made clear up front.
I am sorry to have wasted your time. I'd invite you to check out some my ECE3400: Analog Electronics playlist, which focuses on BJTs and may be more relevant to your interests. Or, if you are interested in audio applications, you can check out my "Analog Circuits for Music Synthesis" or "Guitar Amplification and Effects" playlists.
WHAT???