Little question on opamps (vs. FETs)

[rockhorst]

Sorry, it's question time again. I'm gonna get a practical guide to electric components/circuits soon (any suggestions?), but in the mean time it would be great if someone could clarify the following:

When would one use an opamp and when a FET. I know that an opamp consists of FETs. Basically I have noticed that FETs are generally used as a substitute for tubes, mainly for use in amps and simple boosts.  Opamps tend to surface in distortion pedals.

While we're on the subject of opamps: could someone explain to me how a voltage follower acts when it's not yet in equilibrium? The equilibrium case when Vout = Vin is easy to derive mathematically, but what happens in the first split second that the opamp is activated? How does it iterate the initially large Vout down to Vin?

Also, how do I carry out calculations on opamp circuit when a diode is involved?

For pictures of the mentioned circuits, see the Wiki:
http://en.wikipedia.org/wiki/Operational_amplifier_applications

Again, I apologize for the relative noob-nes to electronics 

[brett]

Hi

When would one use an opamp and when a FET. I know that an opamp consists of FETs. Basically I have noticed that FETs are generally used as a substitute for tubes, mainly for use in amps and simple boosts.  Opamps tend to surface in distortion pedals.

Op-amps usually consist of many BJTs (bi-polar junction transistors, which are quite different to FETs), resistors, diodes and a smattering of other components.  They often use JFETs or CMOS devices at the inputs to increase input impedance.  e.g. TL07 1/2/4 s have JFETs and input impedances of many megohms.

Op-amps essentially arrange many transistors to achieve:
low noise
high input impedance
high bandwidth (not usually an issue for audio gear)
extremely high gain

Gain is controlled by using negative feedback.  In non-inverting mode, a small voltage would result in a wild upward swing in voltage unless the output is connected to the -ve input.  If there is no resistance between output and -ve input, the output voltage never exceeds the +ve input voltage.  In practice, they are virtually identical, hence the circuit is considered a voltage follower.

JFETs use electrical fields to vary the resistance of a channel of semiconductor material.  Look up "JFET channel depletion enhancement" on google to find out how they work at the atomic level.  At the circuit level, they provide slightly non-linear amplification that is slightly analogous to triode vacuum tubes.  There is a control, called a "gate" (like a tube's grid), a cathode called a "source" (that's where the electrons come from), and a sink for electrons called a "drain" (like a tube's plate).  A small change in the voltage at the grid (and a miniscule current) causes a large change in the amount of current flowing from source to drain.  Simple JFET circuits have voltage gains from 3 (guitar pre-amp) to 100 (mu-amp).

Maybe I've made some things less clear, but it's a start.  Ask more questions and read more stuff like Wikipedia.

[R.G.]

To elaborate on Brett's correct answer:

When would one use an opamp and when a FET.

That's a little like saying when would one use a horse versus using a cow. They can both pull a cart or a plow, but if what you're trying to do is get milk, there is a clear answer. There is another answer if you want to ride one. It depends on what you're trying to do.

Opamps can do many things that FETS cannot (by themselves at least) do. FETs are very high input impedance so they do not load down cicuits much. They are modest gain, and modestly accurate as a follower. They have a high variability from one unit to the next. Opamps have a moderately high to high input impedance, sometimes incorporating FETs to get that. They have insanely high gains that are controlled by feedback components, so the variation from unit to unit almost does not matter.

One uses an opamp when one wants to do tightly accurate things. One uses a FET for the high input impedance, or for the side effects of the FET, such as a variable channel resistance.

Basically I have noticed that FETs are generally used as a substitute for tubes, mainly for use in amps and simple boosts.  Opamps tend to surface in distortion pedals.

FETs are NOT a substitute for tubes. The emulations of tube amps which employ JFETs subbed for tubes generate distortion which is then shaped by the various frequency response networks from the amps. Shaped frequency responses of almost any distortion provide voicing that makes the distortion reminiscent of the original thing; this is the fundamental way that the JFET substitution emulators work. But JFETs do not distort like tubes.

Opamps are used in distortion pedals because of their insanely high open loop gains. There is enough gain there to use it to make the nonlinearities of the clipping components like diodes show through.

It all depends on what you're trying to do. You have to know that before you can make any sense of a choice.

Like the song says - you gotta know when to hold'em, know when to fold'em.

[rockhorst]

That's a little like saying when would one use a horse versus using a cow

That's actually more or less what made the question pop up, R.G.

[Eb7+9]

there's two aspects to fidelity that apply to audio, spectral fidelity and transient fidelity - the first one is easily covered through small-signal analysis (Bode analysis) while the later is the domain of large-signal or TRANsient analysis per Spice simulators and real time sample-scope analysis which nobody does (expensive and no standardized testing formulas) - or by ear ... this is where a significant difference CAN lie between single-ended topology and high gain NFB circuitry (op-amp/servo) responses ...

how you design with op-amps will affect the transient fidelity outcome of your circuits - reason being, op-amps (servo blocks) loose lock during fast/large enough transient waveforms - how much of the sub-sonic ringing that occurs before the op-amp resumes lock (to a varying degree) will be sub-modulated into the audio realm depending on how the op-amps are biased and interfaced with each other (often through non-linear caps - tsk tsk ...) ... this gives a hint as to how to better that transient fidelity when designing with op-amps - and also why some audio-philes totally avoid them ...

on paper op-amps look like little darlings - with all the convenient academic-making math that comes along with them ... but down on the audio farm things are different - the TS808's midrange is a good example of op-amp related transient mush, or rather why some op-amp OD's get better defined clipping when running diodes outside the NFB loop ... there's lots of examples to relate to this, if you go through schematics with this aspect in mind you will begin to see why some op-amp designs sound beter than others - and how to improve existing ones ... you may not notice it with one stage buffers, but in multi-stage designs you'll likely start noticing a difference if the op-amp blocks are not interconnecetd properly ... consider how some phasor circuits have DC-coupled op-amp blocks - this is the key ...

another thing some people overlook is the fact that the ultra low Zout of an op-amp follower only applies when the loop is in lock - a theoretical figure that applies with no signal applied - once considerable signal is applied this figure is no longer relevant ... the Japanese hi-fi market made good use of this fact in the 70's - which lead to a re-appraisal of NFB in the VT revival of the 90's ... that's why some designers advise using single-ended followers in place of op-amp ones to help maintain transient fidelity - there's not a huge advantage in using up all that gain and killing it with NFB unless you need to drive high loads ... this may explain why some vintage (pre op-amp) gear sounds so good ...

hope this sheds some light on matter

~JC

[R.G.]

That's actually more or less what made the question pop up

Yep. Let me be more clear.

You have to know (a) what you want to achieve and (b) what the characteristics of horses and cows are before you can make an intelligent choice between them. I then went ahead to describe some but not all of that. Describing all of it would require the equivalent of many books.

You use a JFET when you want some special characteristic of the JFET, such as high Zin, variable resistance region, triode-like biasing, specific clipping characteristics, etc. You use an opamp when you want DC and low-audio accuracy, utterly simple design, repeatable performance with normal parts variation, low Zout, etc. And you have to know what of those characteristics matter in the application to make a good choice.

there's two aspects to fidelity that apply to audio, spectral fidelity and transient fidelity -
...
hope this sheds some light on matter

Actually, it darkens the issue a bit by swirling some transient intermodulation distortion tweako hifi foofoo dust around it.

I'm guessing that the thrust of that post was, loosely translated, "In addition to what's already been said, one might want to use discrete JFETs because they are not high feedback parts and so do not have transient intermodulation distortion, long loop transit times, and so on like opamps do."

And that is true - if you happen to have an application that (a) has large signals that force the amplifying device out of small signal bounds (b) have fast enough signals to force loop transit time or slew rates to be exceeded *and* (c) have a reproduction system and ears that can reproduce and sense the results. But it's not a good reason to choose a JFET over an opamp as a signal booster if you don't already know what ought to go in there.

op-amps (servo blocks) loose lock during fast/large enough transient waveforms

Yes, they do. That is, there is a slew rate that the fastest the output of the opamp can move, in volts per second. If you try to force it to go faster than that, the output slews at a constant volts/sec. So quick - if an opamp slews at 13V/uS like the TL072 does, what is the fastest signal it can reproduce accurately in the 6V range of a circuit powered by a 9V battery? It can slew through that 6V in less than half a microsecond, so it can accurately reproduce a 6Vone megahertz signal, yes?

No, it can't. It doesn't have enough high frequency gain out there. That's just the fastest the output can move.

But it does illustrate the kinds of signals you have to send a modern opamp to cause it to "lose lock" as you say. A guitar pickup cuts off rapidly at 7kHz. Not much problem tracking that one. The point is, "losing lock" is not a simple thing to pin down. And it's quite difficult to hear - unless you read audiophile magazines - 

But since I'm always challenging people to put it in numbers, here are some numbers. From "Audio IC OpAmp Applications", Walter Jung, third edition:

"With an output swing of 10V (peak) and a full-power frequency response of 20kHz, the required slew rate is ... 1.256V/uS".

Jung goes on to qualify that with why and how some opamps can't do that. It makes for good reading. I highly recommend it.

TI puts it another way in its datasheet for the TL072 (not a particularly high performance opamp, but better than 741's): With +/-5V supplies, a TL07x opamp will deliver +/-3V signals to over 300kHz. So you have to work at it to make signals that will cause it to "lose lock".

how much of the sub-sonic ringing that occurs before the op-amp resumes lock

Since sub sonic ringing is easy to see on an oscilloscope, it should be easy to pin down, right? Can you post an example of sub-sonic ringing caused by an opamp which has experienced  an instance of transient distortion? This just can't be hard to find right? Because while searching for a sub-nanoscond gremlin in the signal flow is hard, sub-sonic ringing is easy to see... isn't it? But I've never seen this. So can you point me to a picture of it?

why some op-amp OD's get better defined clipping when running diodes outside the NFB loop

So some opamp ODs get better defined clipping when running diodes outside the NFB and others get better defined clipping when running diodes *inside* the NFB loop? Or did you mean that only opamp ODs with diodes outside the NFB get well defined clipping?

And is it possible that running the diodes inside the NFB are running the diodes in constant current drive, while the ones outside the NFB are run in current limited voltage drive?

Here's a curiousity: if an opamp has feedback diodes, its output only needs to move one diode drop up or down before the current from the diode satisfies the inputs' need to be equalized. How come that's worse than than it having to slew the full power supply to drive diodes-to-ground? Especially since we just tarred and feathered opamps for poor slew rates, high signal transit time, and subsonic ringing from transient overloads.

in multi-stage designs you'll likely start noticing a difference if the op-amp blocks are not interconnecetd properly

Can you tell us the correct way?

consider how some phasor circuits have DC-coupled op-amp blocks - this is the key

I'm having trouble translating that one.
Are you saying that you have to DC couple stages to get good transient response? Or what?

Or are you referring to the build up of amplfied offset voltages in opamp strings?

the TS808's midrange is a good example of op-amp related transient mush,

Actually, no, the TS808's midrange is a good example of designing for the most distortion to be in the midrange. See "The technology of the tube screamer" for how that happens.

another thing some people overlook is the fact that the ultra low Zout of an op-amp follower only applies when the loop is in lock - a theoretical figure that applies with no signal applied - once considerable signal is applied this figure is no longer relevant

How much is "considerable"? Can I say that I'm only going to apply 0.5 considerables to a circuit and be safe?

The "loop being in lock" as we've seen is true unless you're (a) trying to exceed the slew rate of the circuit, which is substantial for modern opamps, or (b) trying to force the circuit to deliver more current than it can at ANY speed. The loop is in lock any time the opamp can deliver enough current to the output and the output is not swinging too fast. And "too fast" is a number you can read off the datasheet, not an audiophile's note that "this amplifier has a well defined midrange with a hint of parsnips and a clove-cinnamon finish". What you write almost gives one the impression that opamps slip "out of lock" at the first hint of rain.

That being said, let's look at the OPEN LOOP output impedance. This is something one can measure, instead of only referring vaguely to. The open loop output impedance of an opamp is often in the range of tens of ohms. While that's not 0.001ohm like it might get to with feedback, it's still quite low compared to a number of other circuits with discretes. It's just not what you expected if you only looked as shallowly as the feedback output impedance numbers.

There is no alternative to knowing the numbers. Guidelines are just that, guidelines, and this gets back to what I was saying. You have to know, or at least where to find out. And audiophile magazines are NOT where to find out. They're still back on being able to hear it when sound has passed through copper wires with a fraction of a percent of oxygen in it.

[brett]

Hi.

swirling some transient intermodulation distortion tweako hifi foofoo dust around it.

Larfing till my guts hurt...
(No disrespect intended Eb7+9).

There's an interesting point to this discussion:
You guys both know a lot.  How, then do I, or an even less experienced builder, work out who to listen to and what to do?
As a scientist (sometimes) I am always impressed by the facts and figures, which is what I got from RG.  But there was some fascinating information comming from Eb7+9.  (I think it was Plato who said that if you tell a story with enough pathos, you can beat the man who uses logic).

I don't have an answer to this issue, I'm just raising it because we have many areas where there's mixes of evidence, logic, semi-truths and myth. 

This goes quite off-topic:
-----------------------------------------------
Here's an electronics analogy concerning a myth.  It contains a couple of half-truths, but hopefully it is amusing and maybe educational concerning JFETs:
There's a turtle and a hare and they decide to race.  The turtle is called JFET and the hare is called OP-AMP.  This is the story of JFET.

JFET the turtle has one leg that is really important and controls how she goes.  She calls it Gate because it swings back and forward like a Gate.  If she doesn't keep Gate happy, she's not going anywhere.  For the audio race, JFET wants to keep Volts off her Gate.  If she gets too many Volts on her Gate, it stays permanently "on", and doesn't swing back and forth like a leg should.  Even with the right amount of Volts, Gate never works 100% ok.  It is slightly bent (check out her datasheet), which means that she makes a little bit of distortion most of the time.  JFET also has two other main parts: her mouth (Sauce) and her bum (Drain).  By eating electrons from the negative rail (through Sauce) and emitting them (out Drain) she can keep running all day.  In some circuits we examine the powerful stuff that comes out of Drain, but other times we use the far less potent electrons hanging around Sauce.

If you want to make something powerful, you'll have to harness JFET and several of her mates to eat a lot of electrons.  You'll need them to help each other (a mu-amp) or stand in line and eat the powerful output of the previous JFET (a cascade). 

Some people think that making Mega Distortion wins the race, but others like just a little distortion.  The people who like just a little distortion really like JFET (because she has that bent leg). 

JFET can also do some special tricks by adjusting the volts on her Gate, but she does that when she's on the sideline, not actually in the Audio Race.

OP-AMP is a turbo-charged racing hare.  He can do a lot of things, including Mega Distortion.  Maybe his story can be told another day.

[R.G.]

Well said! I like the parable of the bent-leg JFET.

You forgot that being a JFET, she often could not tell the difference between two of her legs, that being why she uses the Gate - the others are just too confusing.