Hi
Why do we always bias a mosfet in Saturation mode? Is it becaus gm is more in saturation or is it because gm is more linear in saturation?

Thanks
Salil

 

We don't always bias mosfet's in saturation mode. However, the term saturation describes the active mode for mosfets, whereas saturation describes the linear mode for BJT's (called triode mode on mosfets)

 

Hi
Why do we always bias a mosfet in Saturation mode? Is it becaus gm is more in saturation or is it because gm is more linear in saturation?

Thanks
Salil

In the saturation mode, a MOSFET's source-drain current is (ideally) independent of the source-drain voltage and, instead, a function of the gate-source voltage. It thus acts as a voltage-controlled current source.

In the triode region, a MOSFET's source-drain current, for a given gate-source voltage, is a (roughly) linear function of the source-drain voltage. Thus it acts as a voltage-controlled resistor.

Both operating modes find extensive use in circuits.

 

We don't always bias mosfet's in saturation mode. However, the term saturation describes the active mode for mosfets, whereas saturation describes the linear mode for BJT's (called triode mode on mosfets)

I've always understood (unless my brain has been outgassing again) that the linear mode in a BJT was another name for the forward-active region because the collector current was linear with respect to the base current. I've also seen this referred to quite often as the constant-current region. So the three primary regions are (1) cutoff, (2) saturation, (3) active, linear, or constant-current.

On the other hand, the term "linear" would not be a suitable descriptor for the active region in a MOSFET because, here, the source-drain current goes as the square of the gate-source voltage. Since the source-drain current is mostly independent of source-drain voltage, I have also seen this region referred to frequently as the constant-current region. I have (but not often) seen the triode region of a MOSFET referred to as the linear region because, for a given gate-source voltage, the source-drain current is roughly linear with regards to the source-drain voltage.

So the corresponding regions in the MOSFET are (1) cutoff, (2) triode or linear, (3) saturation or constant-current.

 

"linear" to describe the roughly linear relationship of collector current to collector voltage for a BJT in saturation. I use the term descriptively, not formally. It wouldn't have made much sense to say "..saturation describes the saturation mode..."

I've never heard the active region described as linear.

 

Hi
Why do we always bias a mosfet in Saturation mode? Is it becaus gm is more in saturation or is it because gm is more linear in saturation?

Thanks
Salil

Outside saturation mode; the MOSFET is a voltage controlled current limiter - can be quite useful if you know how to design a self compensating bias generator.

 

I just had this pounded into me on ETO last month. The linear mode is when a mosfet is turned on hard, as in a switch. The saturation mode is when its in the resistive region, like in an amplifier.

 

That sounds so backwards!

However...you have to go with what causes proper communication.

 

That sounds so backwards!

However...you have to go with what causes proper communication.

It's actually correct. The definition of saturation is reversed from BJT's. This causes much confusion

 

"linear" to describe the roughly linear relationship of collector current to collector voltage for a BJT in saturation. I use the term descriptively, not formally. It wouldn't have made much sense to say "..saturation describes the saturation mode..."

I've never heard the active region described as linear.

I did a quick Google search and found several sites, including textbooks, that use the term linear region and active region synonomously and others that specifically call it either the active linear or the linear active region in order to cover their bases. I didn't fine any that called the saturation region the linear region. If I look hard enough, I probably would, though.

I understand the motivation for describing the saturation region of a BJT as "linear", but I don't know how good a description that is. Datasheets will often have a curve of Vcesat versus collector current and this is quite non-linear. But this is usually at a fixed value of β (typically 10 for small signal transistors) and so this doesn't really address how linear the Ic vs Vce curves would be at fixed base current.

On a somewhat related note, I have never once seen in a datashee the classic characteristic curves for a transistor -- the family of curves for Ic as a function of Vce for different base currents that is almost always how transistors are introduced. I've conclused that no matter how useful this might be from a teaching standpoint, it must not have enough practical utility to warrant inclusion.

 

I just had this pounded into me on ETO last month. The linear mode is when a mosfet is turned on hard, as in a switch. The saturation mode is when its in the resistive region, like in an amplifier.

But it is when it is turned on hard that it is in the resistive mode -- namely that the drain-source voltage is roughly proportional to the drain-source current. In the saturation region, it is like a voltage-controlled current source.

 

It's actually correct. The definition of saturation is reversed from BJT's. This causes much confusion

Yep. Which why I tend to use the term "active" in both cases and "triode" for the MOSFET. I haven't found a better term than "saturation" for the BJT because I don't think "linear" is a good one since, to the best of my knowledge and experience, it is more likely to be used for the active region than the saturation region but, more importantly, the fact that it is used for both by different camps is just asking for confusion.

But I find that most people have no problem with understanding what you mean when you talk about a BJT being in saturation, so that doesn't seem to be a point worth stressing about. But people DO get confused and think that when you talk about a MOSFET being in saturation that you are talking about the triode region (and I used to make that mistake a lot, too). So I think it IS worth avoiding that term used that way and call it the "triode" region. Of course, you still have to be aware of how other people can and do use the terms so that you can properly interpret what THEY say.

 

Yep. Which why I tend to use the term "active" in both cases and "triode" for the MOSFET. I haven't found a better term than "saturation" for the BJT because I don't think "linear" is a good one since, to the best of my knowledge and experience, it is more likely to be used for the active region than the saturation region but, more importantly, the fact that it is used for both by different camps is just asking for confusion.

But I find that most people have no problem with understanding what you mean when you talk about a BJT being in saturation, so that doesn't seem to be a point worth stressing about. But people DO get confused and think that when you talk about a MOSFET being in saturation that you are talking about the triode region (and I used to make that mistake a lot, too). So I think it IS worth avoiding that term used that way and call it the "triode" region. Of course, you still have to be aware of how other people can and do use the terms so that you can properly interpret what THEY say.

You prefer the term triode to linear? Triode is referring to a mostly obsolete device (three terminal vacuum tube) but calling it the linear region is highly descriptive because acts as a resistor which is obviously a linear device.

I agree with everyone about the unfortunate confusion about the term saturation in BJTs and MOSFETs.

There is another region that is used often in practice called subthreshold. The device is technically in triode but it is still operated as a voltage amplifier. In that regime gm is low but gm/I is excellent so very low-power circuits can be constructed.

 

You prefer the term triode to linear? Triode is referring to a mostly obsolete device (three terminal vacuum tube) but calling it the linear region is highly descriptive because acts as a resistor which is obviously a linear device.

I agree with everyone about the unfortunate confusion about the term saturation in BJTs and MOSFETs.

The term "triode" comes from the vacuum tube days. IIRC, the terminology that grew around FETs was motivated by the terminology of the vacuum tube era while the terminology that grew around BJTs was more motivated by the device physics viewpoint.

It's not that I prefer the term "triode" as much as I have seen the term "linear" cause confusion because people that are used to having "saturation" and "linear" regions in the BJT don't do well when both of those terms are used for MOSFETS but the meanings are essentially reversed.

So I prefer to use "active" for both to describe the constant-current region of operation and would love to use a single term to describe the other region that couldn't be easily confused with the active region of either. But I'm not aware of any such term that, at least that has any following. So I settle for "triode" (because it is unique and unambiguous, not because of the term "triode" itself or how one would think of a vacuum tube triode but just because it is a term that is not used, in relation to FETs or BJTs, for any other purpose). I would be satisfied if the use of the term "triode" also became synonomous with the saturation region in a BJT -- again, not because the definition of the original meaning of the term applies, but because it would be a consistent and unambiguous terminology.

There is another region that is used often in practice called subthreshold. The device is technically in triode but it is still operated as a voltage amplifier. In that regime gm is low but gm/I is excellent so very low-power circuits can be constructed.

Yep. Many of the IC circuits I worked on operated the amplifiers in the deep subthreshold region usually for just that reason, although there were sometimes other reasons that were directly related to being a good match for the lunatic fringe designs we often got involved with.

 

The term "triode" comes from the vacuum tube days. IIRC, the terminology that grew around FETs was motivated by the terminology of the vacuum tube era while the terminology that grew around BJTs was more motivated by the device physics viewpoint.

I think that's true. There are a lot of old vacuum tube terms floating around. For example, the term "cascode" came about as a combination of the words "cascade" and "cathode", which doesn't make much sense when you're using MOSFETs or BJTs.

It's not that I prefer the term "triode" as much as I have seen the term "linear" cause confusion because people that are used to having "saturation" and "linear" regions in the BJT don't do well when both of those terms are used for MOSFETS but the meanings are essentially reversed.

So I prefer to use "active" for both to describe the constant-current region of operation and would love to use a single term to describe the other region that couldn't be easily confused with the active region of either. But I'm not aware of any such term that, at least that has any following. So I settle for "triode" (because it is unique and unambiguous, not because of the term "triode" itself or how one would think of a vacuum tube triode but just because it is a term that is not used, in relation to FETs or BJTs, for any other purpose). I would be satisfied if the use of the term "triode" also became synonomous with the saturation region in a BJT -- again, not because the definition of the original meaning of the term applies, but because it would be a consistent and unambiguous terminology.

I'm 100% with you. It's also a cultural difference. The places where I've worked the only time you see a BJT is in a bandgap reference so calling the active region of a MOSFET "saturation" isn't that ambiguous. I imagine it is different from organization to organization. You make a very clear argument for calling it "active" but as Microsoft can attest, sometimes being the best idea doesn't find you traction in the marketplace.

Yep. Many of the IC circuits I worked on operated the amplifiers in the deep subthreshold region usually for just that reason, although there were sometimes other reasons that were directly related to being a good match for the lunatic fringe designs we often got involved with.

I haven't had the pleasure of designing subthreshold circuits but I understand it's a somewhat different world! Although, in modern CMOS processes the VDD/Vt ratio is going down fast so we have to push our devices pretty hard just to get enough swing. We call it "weak inversion" because it's not actually subthreshold.

I don't know what processes you've used but nowadays we are asked to design analog circuits using processes with a gm/gds that peaks at 6 or 7! Totally changes how you approach the task.

 

But it is when it is turned on hard that it is in the resistive mode -- namely that the drain-source voltage is roughly proportional to the drain-source current. In the saturation region, it is like a voltage-controlled current source.

But that is where I had my problem understanding it. The resistive/linear/fully turned on mode is actually the place that has the least resistance and thus makes the lowest heat in the mosfet. Which just seems wrong, to me.

 

But that is where I had my problem understanding it. The resistive/linear/fully turned on mode is actually the place that has the least resistance and thus makes the lowest heat in the mosfet. Which just seems wrong, to me.

The key things are to note that it isn't the "resistive" mode because the resistance is higher than in the other mode, but because the resistance is more-or-less constant and is usable in a similar fashion as a resistor in order to get a voltage that is proportional to a current (or vice-versa). In the active mode, the resistance is much, much higher (ideally infinite) since it is behaving more like a current source and thus we don't see much change in current as the voltage changes. We effectively have a big resistor in parallel with a current source and so we can usually ignore it.

 

I think that's true. There are a lot of old vacuum tube terms floating around. For example, the term "cascode" came about as a combination of the words "cascade" and "cathode", which doesn't make much sense when you're using MOSFETs or BJTs.

That's one that cause me problems because for the longest time I insisted on trying to understand the cascode configuration in terms of being some rational variation of cascade. I finally just accepted that it's just a name.

You make a very clear argument for calling it "active" but as Microsoft can attest, sometimes being the best idea doesn't find you traction in the marketplace.

Agreed, which is why I restrain myself to using terms that are a subset of the ones in use and that others are unlikely to understand what I mean, while accepting that I have to be vigilent and be sure that I don't misunderstand what they mean when they use the rest of the terms. I don't try to invent a new term and get people to adopt it -- ain't gonna happen no matter how good the term is.

I don't know what processes you've used but nowadays we are asked to design analog circuits using processes with a gm/gds that peaks at 6 or 7! Totally changes how you approach the task.

The smallest design rules I ever used were 0.13μm. We tended to use logic processes in order to make analog chips (mixed-signal, actually, but the logic was there just to support the analog functionality). We did this for a few reasons. First, we did a lot of design of physically large chips -- like four-chips-to-a-wafer large -- so the cost savings associated with using a logic process were important. But also, types of designs we did didn't allow for the kind of Vss/bulk management that you would normally use in an analog process, so we leveraged the lower substrate sheet resistivities of the logic processes but had to design and layout our own logic cells to keep the Vss currents in the logic portion from getting injected into the substrate. We also had to contend with the fact that, being a logic process, the analog parameters weren't as well controlled or even characterized and available, so we had to design the critical analog circuits without that information or without being able to rely on the kind of matching that is usually available in an analog process.

 

But that is where I had my problem understanding it. The resistive/linear/fully turned on mode is actually the place that has the least resistance and thus makes the lowest heat in the mosfet. Which just seems wrong, to me.

In saturated region its either on or off, so there's either lots of current and very little volt drop, or no current and all the volt drop - either way P=I*V dictates that very few watts are dissipated,

In linear mode the device can simultaneously be dropping significant voltage and carrying significant current I*V gives you the resulting watts.

 

In saturated region its either on or off, so there's either lots of current and very little volt drop, or no current and all the volt drop - either way P=I*V dictates that very few watts are dissipated,

In linear mode the device can simultaneously be dropping significant voltage and carrying significant current I*V gives you the resulting watts.

First time hearing this, that very few watts are dissipated in the saturation region. Everything I've been told is to put them in the linear region, when you can, to keep them from heating/dissipating watts. Can you give a reference to this concept? I'm constantly in learning mode.