Why is FET voltage controlled and BJT current controlled?

anonymoussome

A bit awkward question but still it confuses me......

A BJT is a current controlled device because its output characteristics are determined by the input current.
A FET is voltage controlled device because its output characteristics are determined by the Field which depends on Voltage applied.

Now the question is that current is also generated due to Voltage and still BJT is current controlled and FET voltage controlled.

Any Ideas?

waht

A transistor is a non-linear device, and so we use different models to describe its behavior. The current control model is used often because it is approximately linear. Another models like the Eber-Molls model is used to address the voltage control behavior. It is non-linear because of the base-emitter junction acts like an exponential, so different approximations are used to linearize it.

Carl Pugh

They are both voltage controlled, or if you prefer they are both current controlled.

On bipolar transistors, the voltage varies with the temperature, but the current amplification stays constant.

On FET transistors, the voltage amplification stays constant, but the current varies with temperature.

anonymoussome

So it is just our preference as to which model we choose?

dlgoff

http://en.wikipedia.org/wiki/Transistor#Types

waht

Yes, it depends on the problem you are solving. In general, whenever there is current flowing, there will be a potential difference. So if you increase current flowing in the base, you have to increase the base voltage. If you increase the base voltage in an NPN, and due to constant base-emitter diode voltage drop, a proportional voltage will increase above internal emitter resistance and thus pulling more collector current.

cabraham

Current is not generated "due to voltage". This cannot be overemphasized. Current and voltage are generated in unison. One cannot exist w/o the other. In a p-n junction, such as the b-e junction in a bjt, the change in base-emitter voltage, vbe, cannot take place until after the change in base current, ib. The reason a bjt is called a current controlled device goes back to the Ebers-Moll 1954 paper.

In the E-M model, the bjt is modeled as a pair of back to back diodes with the base being the common anode. Each p-n junction, the c-b and the b-e, is approximated by Shockley's diode equation

Id = Is*(exp(Vd/Vt) - 1).

The b-e and c-b jcns have differing doping densities, so the "Is" in a diode, scaling current, is denoted by "Ies" for the b-e jcn, and "Ics" for the c-b jcn. Hence,

Ic = alpha_n*Ies*(exp(Vbe/Vt) - 1) - Ics*(exp(Vbc/Vt) - 1).

Ie = Ies*(exp(Vbe/Vt) - 1) - alpha_i*Ics*(exp(Vbc/Vt) - 1).

For Ic, the collector current, if alpha_n, the normal mode emitter to collector current gain, is zero, we only get leakage current due to reverse biased c-b jcn. The "-Ics*(exp(Vbc/Vt) - 1), is this current, the 2nd term in the Ic equation. Since Vbc is negative and near 0, we get "Ics*(0 - 1)" or just simply "Ics".

But the Ebers-Moll model adds a current-controlled current source as follows:

Ic = alpha_n*Ie. When the emitter injects electrons into the base, rather than recombine in the base region as a diode does, the base is so thin, the electrons are drawn into the collector region due to the strong electric field in the c-b jcn.

Thus the collector current originates from the emitter current. The same electrons which make up the emitter current travel into the collector and become collector current.

The alpha factor describes "transistor action". If alpha is 1, we have a perfect device, Ic = Ie, as all electrons survive the trip through the base, and no holes are injected from the base to emitter. We approach this condition by using light doping in the base and heavy doping in the emitter.

If alpha is 0, there is no transistor action, and we just have 2 back to back diodes. The base current Ib, and base-emitter voltage, Vbe, are still present, but Ic = 0. Without high alpha values, Ic is small and transistor action is poor.

Ic is controlled by alpha and Ie, but Ib and Vbe are necessary quantities. Since the b-e jcn must be forward biased, Ib and Vbe are absolutely necessary to obtain Ie. Without Ib/Vbe, Ie would be 0, and Ic = 0, regardless of alpha.

Ib and Vbe are necessary , whereas Ic and alpha are both necessary and sufficient to create Ic. If the base region was very wide, and all electrons recombine in said base, alpha is 0. We have Ib and Vbe, but since alpha = 0, Ic is only leakage current Ics. We have Ib and Vbe but no transistor action as alpha = 0. Ic is NOT controlled by Ib or Vbe. Ic is controlled by Ie and alpha.

We control Ic with Ie.

With FETs, the gate current is necessary as is the gate to source voltage. But controlling gate current is too hard. The gate current, Ig, needed to produce a specific Id value, drain current, varies with frequency as well as temp and bias current value. Also, a FET presents a very high g-s impedance. A current source at the g-s terminals would charge the non-linear capacitance to the point where punch-through occurs. A voltage source at the g-s terminals charges this capacitance to a safe value. Vgs is used to control Id. A FET does not lend itself to current control. Voltage control must be used for driving FEs. Both are necessary, one is controlled directly, the other is indirect.

The bjt is controlled via Ie, with Ib and Vbe being indirect. Sometimes, Ib is used as controlling current, such as saturated switch applications. The base is overdriven to fully saturate the collector-emitter. We never drive the b-e terminal voltage directly, but rther indirectly.

The FET uses direct voltage control, with Ig indirectly defined. A voltage source is connected from g-s, and the resistances and capacitances determine Ig.

Is this helpful?

Claude

anonymoussome

i think so. thanks a ton.

bitrex

Someting I don't understand is, since Ies is such a small number, how does the Ebers Moll model accurately describe the emitter current? I mean, it's possible to have transistors with many milliamperes or amps flowing through them, and I don't see how you get those numbers when you're multiplying everything by 10^-16 or some such.

Edit: Also, if you have a chance - how would the above equations be different when the transistor is not in the active region, but in saturation with both the base to emitter and base to collector diode forward biased?

cabraham

Good questions. Regarding the Ies value of 1e-16 to 1e-11, here it is. For a signal device, if Ies = 1e-15 (1.0 femtoamp), with Vbe = 0.65V, and T (temperature = 298 K (25 C), then Vt = nkT/q, where n = 1.0 in this region of operation, k = Boltzmann's constant = 1.3806e-23 joule/Kelvin, & q = 1.602e-19 C (charge on 1 electron), so that Vt = 0.0257 V, or 25.7 mV.

So if alpha = 0.99 typ, then Ic = 97.19 uA. If Vbe = 0.60V, then Ic = 13.87 uA. For a heftier device, if Ies = 1e-12 (1.0 pA), then at Vbe = 0.60V, Ic = 13.87 mA, etc.

The Ies values given in the fA & pA range are accurate in the operating region of uA and mA of collector current. The reason is as follows. The thermal voltage, Vt = nkT/q, is proportional to absolute temp, and "n", a factor which varies as current changes.

When Ic is in the uA to mA range, n is very close to 1. But at low values of current, below a uA, n increases, eventually reaching 2 in the nA region. This is explained in semiconductor physics. In this region, the E-M equations are not as accurate.

The exponential/logarithmic relation between I & V in a p-n junction, holds for about 5, or maybe 6 decades of current for a bjt, and up to around 10 decades for certain diodes. With a bjt there are 2 phenomena affecting the relation.

The first was just mentioned, the variation of "n" at low current values. Another is the c-b junction leakage current. Let's say that Ics is 1.0 nA. Even though the b-e jcn value of leakage, Ies, is 1.0 pA, the Ic minimum value is 1.0 nA. Even though the b-e junction could be at 0 V, there is always the leakage from c to b due to Vcb & Ics. This spoils the log-linear relation between Ic & Vbe, as well as Ie & Vbe. The E-M equations account for the c-b leakage Ics, with the 2nd term. This is why the log-linear function for I vs. V does not extend down into the pA or fA region.

But, if we hold the Ic value in the uA to mA region, or even amps for a power device, using an Ies value in the fA to pA range, produces good results. The slope of the I-V curve is linear from let's say 100 nA to 10 mA of Ic, or 5 decades of current. If Ies is 1.0 fA, that is way below the straight line region of the I-V curve. Below 100 nA of Ic, the curve flattens out. It is not log-linear way down to 1.0 fA at all. So why can we use the 1.0 fA value if the Ic-Vbe graph does not linearly extend down that low?

The answer is that a line drawn from Ies = 1.0 fA upward extends into the nA, uA, & mA region with good accuracy. From Ic = 100 nA until 10 mA, the lines nearly coincide. Below 100 nA, the actual Ic curve is greater than that predicted by E-M based on Ies = 1.0 fA.

But we seldom operate a bjt w/ fA or pA of Ic. So the error is not a big deal. Spice models for the bjt utilize Ies values in the fA to pA range with good results.

The Ies value of 1.0 fA or pA does not represent an actual physical current associated with the device. It is a scaling factor derived from doping density, junction geometry, and other intrinsic material properties. If the I-V curve was log-linear all the way down to the Ies value, then Ies is the forward current when the (exp(Vbe/Vt) - 1) factor is equal to 1. That occurs when exp(Vbe/Vt) = 2, so subtracting 1 leaves 1. Multiply this factor by Ies, and we get Ies. That is the physical meaning of Ies.

Ies is the current which would flow at Vbe = Vt*ln2 = 17.8 mV, if the device was true log-linear down to this value of I/V.

As far as operation in the saturated region goes, yes, E-M covers that.

Let's say that we use Vbe = 0.60V again, w/ Ies = 1.0 pA, but we are in the active region. Since Vbc < 0, the 2nd term in E-M can be ignored. The exp(Vbc/Vt) is nearly 0. For example, if the collector is at 1.0V above the base, Vbc = -1.0V, and exp(Vbc/Vt) = 1.23e-17! This is virtually 0, and exp() - 1 = -1. So, the 2nd term is simply -(-Ics), or Ics, the c-b jcn leakage current.

Ic is then alpha_n*Ies*(exp(Vbe/Vt) - 1), which equals 13.87 mA. But if we increase the collector resistor so that saturation takes place, we now must deal with that 2nd term.

Let Vce = 0.10V, the device is fully on hard. Vbc = Vbe - Vce = 0.60V - 0.10V = 0.50V. Let's say that the Ics value is 20 pA. The 2nd term of E-M gives

Ics*(exp(Vbc/Vt) - 1) = 5.71 mA.

But the E-M eqn has the 2nd term subtracted from the 1st term. Hence Ic = 13.87 - 5.71 = 8.16 mA.

The actual Ic value is less in saturation, and E-M predicts the same. So E-M covers both active and saturated modes of operation. In active regions, the 2nd term is not needed, as it is small. But in saturation, we must use both terms.

Does this make sense?

Claude

bitrex

Thank you! That clarifies things a lot for me. Here's a related follow up question - when the transistor is in saturation and the base to collector junction is forward biased, if the collector is connected to a load resistor will that junction behave in a similar way to how the emitter voltage follows the base voltage in an emitter follower? I had been reading about the problem of "latchup" in op amps, and the reason given was that excessive common-mode voltage could cause one of the transistors in a differential pair to enter saturation, and that the global feedback network could switch from being negative to positive. The only way I can picture this happening is if in saturation the collector follows along with the base, causing the output to become non-inverting rather than inverting.

sokrates

There's so much techno-babble in this thread, and none of it really addresses the OP's quesiton.

Simply put, a FET is a "voltage-controlled" current source, because the output current is modulated by a the GATE VOLTAGE which changes the electro-static FIELDs inside the channel, hence the name FIELD-EFFECT transistor.

On the other hand, a BJT is a "current-controlled" source, because the output current is modulated by the amount of BASE CURRENT that is injected from the base terminal. The characteristic "amplification" of current in a BJT is directly related to the amount of electrons that enter from the base...

This is essentially what he needs to know. What does it even have to do with the Ebers-Moll model?? It's just a way of modelling the device from a circuit perspective...

sokrates

No it is NOT up to our choice... The name FET refers to a specific class of devices where the current modulation is done via a VOLTAGE terminal to affect the FIELD inside the device (FIELD-EFFECT)

Bipolar Junction Transistors are not FIELD-EFFECT devices, they control the the transistor current by the AMOUNT OF CURRENT injected from the base terminal...

These are FUNDAMENTALLY different in the device level and HAS NOTHING to do with the particular circuit model that's being used to solve the "circuit" problem...

The names are given to accentuate the PHYSICAL principles of the device, hence in this context, Ebers-Moll and such "terminology" is useless and irrelevant.

sokrates

As a device physicist, most naturally I have strong doubts on what you are saying here.
Could you please point out that paper you are mentioning showing the point where "current-controlled device" is defined for the first time?

This is most probably incorrect, because the reason BJT is called a current controlled device has nothing to do with a "back-to-back" diode model created for circuit applications at a time where COMPUTERs were not as powerful as they are today to do the EXACT simulations.

Nowadays, commercial tools like MEDICI or TAURUS know NOTHING about Ebers-Moll, because it is much stronger (and accurate) to solve the problem from a charge-control point of view... Where drift-difussion equations are solved self-consistently with poisson equation.

So Ebers-Moll is NOT AT ALL fundamental and it was a mere convenience at the time.

The real reason why BJT is called a current controlled device is related to the DEVICE PHYSICS as I explained in my previous posts.

cabraham

I had the paper on my old pc. I'll find it this week and post it. It does describe the bjt as back to back diodes, with a current controlled current source. With 2 diodes, 1 forward, 1 reverse biased, E-M describes each diode in terms of Shockley's diode eqn. But since electrons emitted from the emitter, pass through the base region and are collected by the electric field of the reverse biased c-b jcn, E-M described the collector current with

Ic = alpha_n*Ie.

The emitter current controls the collector current with alpha being the factor of proportionality. Hence, at the mAcro level, a 1st order approximation, a bjt is a CCCS. This is not the most fundamental model.

I don't dispute your charge control assertion. I also agree that E-M is a "black box" mAcroscopic view of the bjt. If you wish to get down to thew mIcroscopic view, or "fundamental" if you prefer, then I agree that "charge control" is the best model. Beyond charge control, is the quantum mechanics modeling. QM is over my head. The late Richard Feynman, one of the 20th century's best physicists, said that nobody, including himself, really understands QM. I fully agree. I studied QM in modern physics class in the late 70's, and semiconductor physics 2 class in 2008. I know that I don't understand QM. I accept the published findings of the science community. I don't know enough to dispute them.

Finally, although the charge control model works best, when we externally drive a bjt, we cannot inject a specific amount of charge into a device. We control the device terminal currents and voltages. The bjt is definitely a charge controlled device, but we use a current source to control the bjt, or a voltage source plus a series resistor. This is current control, or "current drive". Internally, again I agree with your charge control model as the best description of bjt behavior.

We agree, so there's no need to argue. Thanks and as soon as I find that paper, I'll post it.

Claude

cabraham

Techno-babble? That was rude! He asked why the bjt & FET are classified as CCCS & VCCS resp. So, I gave him an historical account of the math and modeling of these devices. I did go deep into details, but one cannot explain device physics in 1 paragraph. My treatise exactly addresses the OP question, how can you suggest otherwise?

It has everything to do with E-M! E-M was the 1st model attempt at the bjt, and still stands today. It has been modified and improved per Gummel-Poon. Also, for large signal operation and/or switching mode, the charge control model is extensively used. Also. the collector current is modulated by the emitter current, NOT BASE current. Base current is necessary for bjt operation, as is base-emitter voltage, but Ic is controlled by Ie. Every semi maker affirms the same. "Current-controlled" refers to emitter, not base current. You don't know semi phy.

I was just giving him perspective. I studied EE at the BE & MS level in the 70's. In 2007 I returned to grad school for the Ph.D. I have this semester and next to complete my course work. I've been a practicing EE for 32 yrs. in R & D. I have developed many many pieces of equipment using discrete bjt, diodes, FETs, SCR, triac, IGBT, LED, photodiode, etc. I took semiconductor physics from the physics dept. in the 70's as a senior. As a grad student (MSEE) I took semi phy from the EE dept in '79. As a Ph.D. student, I took semi phy 2 in 2008, fabrication in 2008, and advanced sensors in 2007.

I was mentored in the 80's by a boss who became an EE just as the bjt hit the market in the 50's. In his drawer of sample parts, he had germanium diodes & bjt parts from the 50's. The part nos. were 1N3, 1N4, 2N3, etc.! He knew transistor physics quite well and I learned much from him.

If you don't see how my historical treatise based on years of post-graduate studies and mentoring from sages who have been there done that is relevant, and comes across as "techno babble", then maybe you should examine your own background. Maybe you are the one lacking in knowledge. What credentials have you got? Where do you get off rudely rebuking me? How many semi phy courses have you taken and passed? Do you have work experience in the semi industry.

The problem is that electronics is a field where everybody sees themself as an expert. Anybody that's heard of Ohm's law thinks they are the equal to a professor. What folly.

Claude

sokrates

I am not questioning anybody's expertise, nor am I making any brave assumptions on people I encounter on internet by the limited knowledge they share with other people. There's just no way to see whether you are debating with an expert or a high school student.

I don't need to show off by listing the courses I have taken or bragging about my own "credentials" and so on... This is not the right venue for that. Why don't we focus on the content instead of belaboring about the number of courses we took? I took a few SC courses, but apparently even a few of them was enough for me to set the basics right.

Back to our discussion> You still owe the community the Ebers-Moll paper where THE CURRENT CONTROLLED DEVICE concept is being used FOR THE FIRST TIME in history to distinguish BJT operation... Why don't you post that, instead of your work xp?

If you LOOK BACK and see what the OP INITIALLY asked (before being bombarded on Ebers-Moll, and history of BJTs etc...) it REALLy had an easy answer, and I am sorry, even if you are Schokley, your answer to the OP was unnecessarily complicated and off the track... Anybody who knows a decent amount of BJT and FET theory would agree. And he was drifted off to other parts of the problem without really appreciating the basic difference between a FET and a Junction Transistor.

Now before we start another major debate on the operation principles of BJTs, let us agree on one thing: The collectro current on a BJT is mainly modulated BY THE BASE current, the sole purpose of the existence of a base terminal... You might want to check an elementary SC book (say, Pierret) before insisting on this one. Because it's really a VERY, VERY rudimentary piece. Of course, the base junction is intricately connected to the emitter junction by the common bias they share, but still what really changes the output current DIRECTLY depends on the number of carriers injected from the base... Without the base current there wouldn't even be any TRANSISTOR operation, because "base" terminal is analogous to the GATE terminal of a FET in BJT... And you are saying that the collector (drain) is modulating the ON-current of a BJT (FET) ? That's obviously wrong...The emitter is needed to EMIT (supply) a large amount of carriers to the collector and BASE current determines HOW MUCH of the emitted current will make it to the collector to be collected... That's basically it. And you reach to the conclusion that I don't know semiconductor physics from this wrong premise?... But I am not going to assume the same thing, don't worry. Because I don't know you well enough.

The question was related to voltage control (Field-Effect) and current control ( Junction transistors, current injection), people don't care about what we know, they are coming here because they want intuitive and common sense answers.

Years of "sagely" advising people must have taught you how to be concise and relevant. There are hundreds of ways to address a basic question from a beginner, you chose the worst one and you blame me for that.