Op-amp specs tutorial?

[xyzzy]

Dave,

First, thanks for all your great videos so far.  I've learned a great deal.

How about shooting a tutorial about how to select the right op-amp?

I suspect many of us already understand "ideal" op-amps, but would benefit greatly from a de-mystification of all their non-ideal aspects as described in their data sheets.  Ideally, we'd learn about things like offset voltage (and zeroing thereof), CMRR, bandwidth, slew rate, how a "precision" op-amp differs from a standard one, rail-to-rail output, etc.

You might even identify a few "jellybean" parts, show us how their output compares to a LM741 in an identical test circuit, and explain which specs contribute to what differences in the scope traces.

Thoughts?

[george graves]

+1

[Psi]

I think he already has like 20 other videos in his todo list.

But yeah, id be keen for a practical opamp specs video too.
So many of the opamp usage docs out there are just math and don't explain the specs practically.

[ThievingSix]

+1, i liked how he went through the voltage regulator datasheet in the PSU design video. I can't for the life of me read any IC data sheet other than the basic voltages/temps.

[Bored@Work]

He, I am sure if you take http://www.ti.com/lit/an/slod006b/slod006b.pdf , condense it into a ready-to-use script for a one-hour video [a Dave-hour, aka 1:15], and pitch it to Dave, that Dave would consider it.

In other words, what you lot are after is a feel-good video that creates the impression you know everything about op-amps after you watched the video. But the thing is, no video in the world will help you to avoid putting some work into it, and learning what you want to know. But that you could do without any video. Start with reading the above mentioned reference. And drop that silly "I can't for the life of me ...", "it is all math ...". Or as we say here on the EEVblog forum, have you considered knitting as a hobby?

[Psi]

In other words, what you lot are after is a feel-good video that creates the impression you know everything about op-amps after you watched the video.

No! and please stop trying to put people down like that.

What we're after is a practical understanding of the finer points in various opamp types and limits.
Not everyone learns the way you obviously do.
Some of use need a practical understanding before the theory/math makes any sense.

But the thing is, no video in the world will help you to avoid putting some work into it, and learning what you want to know. But that you could do without any video. Start with reading the above mentioned reference.

Incorrect, videos that explain things in a practical or real-world way very much help people to successfully use components before they fully understanding the theory.
This boosts confidence and, if/when a circuit doesn't work, you can have a look at the theory and fill in any blanks, which is when people learn things.
Trying to learn multiple pages of theory all at once is an inefficient use of time when you only need to know 25% of it to get your design working correctly.

drop that silly "I can't for the life of me ...", "it is all math ..."

No one in the thread said "I can't for the life of me" so please don't quote it.

There is nothing silly about wanting a more practical view rather than the math.
The problem with the math/theory is it's all built on top of more math/theory. It's very hard to just read a specific section because it talks about, and references, other aspects and sections.
So the math/theory is only useful if you either
- Sit down for many hours and try to get the entire 464 page document into your head so you can understand how it all fits together.
- Already have a practical understanding of how the device works and only want to understand a specific section.

[Tepe]

drop that silly "I can't for the life of me ...", "it is all math ..."

No one in the thread said "I can't for the life of me" so please don't quote it.

He wasn't quoting anybody. He was giving examples of typical statements. There is no problem in that.

There is nothing silly about wanting a more practical view rather than the math.

Unless you want a deeper understanding.

The problem with the math/theory is it's all built on top of more math/theory.

I fail to see how that is a true problem.

[Psi]

There is nothing silly about wanting a more practical view rather than the math.

Unless you want a deeper understanding.

That still doesn't make it "silly"
If you want a deep understanding of the theory then studying a practical example would be "insufficient" but definitely not "silly"

The problem with the math/theory is it's all built on top of more math/theory.

I fail to see how that is a true problem.

It's not a problem per se, it just makes it hard to read about how theory property X works without first understanding property Y,Z,K and L.

It's much easier to explain property X practically without referring to the deeper theory. Then later on you can come back to the theory and it just "clicks" with little to no effort required.

[poptones]

There is nothing silly about wanting a more practical view rather than the math.

A practical view without math means sticking pins into a perfboard and seeing what happens. That's not science, it's alchemy.

There was an orange National Semiconductor applications book sold at Radio Shack in the 70's. I talked my dad into buying me this because it had all sorts of practical circuits in it - phono and tape preamps, radios, and so forth. It also had many articles discussing the practicalities of things liek calculating S/N ratio by measuring noise voltage and current at a few data points and mathmatically applying root mean averages in order to predict the S/N ratio of a circuit over a given bandwidth. There was A and C weighting charts, standards, discussion of noise currents and offset currents and impedance calculations for equalization.

I read that thing until it was dog eared. At first it was building the stuff in the book, but over time I wanted to understand more about what I was doing so I started reading those "math" articles - and (only) then it began to make sense.

I'm afraid there is no substitute for experience. Get in there and tackle the math until you're over your head, then step back and break it down until you get it. No one can do this for you, not even with video.

[Bored@Work]

In other words, what you lot are after is a feel-good video that creates the impression you know everything about op-amps after you watched the video.

No! and please stop trying to put people down like that.

I am just telling you the facts of life. You know, running around and begging everyone to dumb everything down so you can be an engineer, too, won't work.

I will never be an opera singer in my life. Because I lack the talent, and the motivation to put the hard work into it to become one. But I am not running around and begging everyone to dumb opera singing down to the point that it is easy for me. So I don't need to know tomes and notes, and I am not bothered the least with the finer practical points in interpreting a piece off music and its limitations.

Practical understanding? Yeah, sure. There is so much opportunity out there to learn. But you lot are just looking for an excuse to not do it. Want practical understanding? Pull up one of those million websites where they show you how to plug an op-amp onto a board, those thousands of books, or those interesting web applications, and WORK to gain some understanding, instead of waiting to get spoon feed with a video.

drop that silly "I can't for the life of me ...", "it is all math ..."

No one in the thread said "I can't for the life of me" so please don't quote it.

+1, i liked how he went through the voltage regulator datasheet in the PSU design video. I can't for the life of me read any IC data sheet other than the basic voltages/temps.

[Psi]

I am just telling you the facts of life. You know, running around and begging everyone to dumb everything down so you can be an engineer, too, won't work.

Why do you always word your posts to try to put people down?
The world isn't an evil place out to get everyone with limited knowledge.

It's about limiting the level of initial detail to whats needed to use a component and the common traps for young players.
You don't need to know everything about an opamp to start using them effectively.
It's like saying someone might as well give up on being an engineer if they don't understand how an opamp works at the atomic level.
Most engineers don't need all the theory to do there job well, it helps to know the theory sure, but only at those times when you need it. You normally pick up the theory automatically as you need it for projects.
The key is to know enough about everything to be able to use it.

+1, i liked how he went through the voltage regulator datasheet in the PSU design video. I can't for the life of me read any IC data sheet other than the basic voltages/temps.

Fair enough, i totally missed that for some reason.
I retract my statement about that.

[george graves]

Well, by board@work's logic - Dave shouldn't make any videos at all.  He must think Dave's whole approach to teaching is "spoon feeding" a bunch of babies.

I agree with Psi.  There are too many people on this board that are are down right mean spirited - it's almost as if they feel threatened that someone might learn what they know.  Gasp!  Oh the horror! 

[Stephen Hill]

I too wouldn't mind seeing an electrical engineering video about op amps. It would certainly be better than videos about consumer lighting.

[free_electron]

There is nothing silly about wanting a more practical view rather than the math.

Unless you want a deeper understanding.

Math does not guarantee or give you a deeper understanding.

Math is just a shorthand notation format that explains the relations between parameters of a device. Anything you can express in that shorthand format you can express in a multitude of different ways.

Memorizing the equations still gives you no 'deeper' understang of anything. There is nothing 'deeper' to understand to begin with.
The circuit is the circuit is the circuit, and it damn well behaves as the electrons run. An no amount of maths or other notation formats is going to change that or your understanding of it.

Maths is just easier to read for certain people , while another group needs a more plastic or visual explanation. and there is nothing wrong with that. either way works. Math is just one of the 'languages'. it is not the language that matters but the information stored in it.
So it's got nothing to do with 'dumbing it down'. You can convey the same information in other ways. Just because i don't know chinese doesn;t mean i should be blocked from reading the wisdoms of confucius. Somebody went through the trouble to translate them and explain it in a way i can understand them too. Saying you can't be an engineer without math is saying like you can't read conficius unless you read chinese.

besides  , there is a LOT of things going on with opamp that can't be put in equations anyway. Try to pour the concept of input operational range in an equation for example ... good luck with that. i can explain it in 3 sentences and anyone will understand it. Or Common mode rejection.

And you only truly understand something if you can explain it to someone else and make that person understand it as well. Anything else is just parrot work.

[poptones]

The gain of an op amp is the most fundamental parameter we need concern ourselves. It is usually set by the ratio of two resistors.

Ratio

How do you perform even the most basic "practical" calculation without using math?

BTW: 1.5V + VDD < Vi < VSS-1.5v

[VonKlitzing]

He, I am sure if you take http://www.ti.com/lit/an/slod006b/slod006b.pdf , condense it into a ready-to-use script for a one-hour video [a Dave-hour, aka 1:15], and pitch it to Dave, that Dave would consider it.

Here's an idea, since you are such an opinionated expert on absolutely every facet of EE why don't you do it. There's some great stuff in document you linked but it's surrounded by another 300 odd pages of stuff you will never use or need. The OP was just asking for a video about a few handy tips in their use and what traps to look out for when starting out using them. Everyone has to start somewhere and most need a helpful point in the right direction in differentiating between what's vital information and whats just nice to know.

It's not a matter of laziness, it's trying to make the learning process a little bit more efficient and quite frankly your sneering tone is extremely unhelpful to those new to these forums and the whole subject of electronics in general.   

To the OP:

You may find this video by Afrotechmods helpful...

[Psi]

The basic opamp stuff like setting the gain and inverting vs non inverting isn't what most of us are after.
Personally, I mostly do digital design but sometimes need to use opamps in products and have done so many times, so i have a good grasp on all that basic stuff and some of the advanced stuff.

As xyzzy said in the first post, we already understand ideal opamps.
The idea in this thread was to understand the non-ideal aspects and opamp variations

Offset voltage (and zeroing thereof), CMRR, bandwidth, slew rate, how a "precision" op-amp differs from a standard one, rail-to-rail output, differential opamp vs normal opamp, etc. There are many opamps available.

I remember Dave talked about chopper-stabilized opamps in a past video with a good practical explanation of them.

[free_electron]

The gain of an op amp is the most fundamental parameter we need concern ourselves.

T
he gain of an opamp is not 'set' by a ratio of resistors. The gain of an opamp is defined at design time and is a very large number. This gain is dependent on a great deal of parameters. Some of these are chemical parameters and semiconductor physics...

What you are doing is provide feedback with external circuitry and limiting this design-time gain. Besides, i can use the opamp in feed-forward as well , or i can even select to use it in open loop ( as a comparator for example )

So there you go : your mathematics notation fails you as it is incomplete and ony applicable in 1 specific little case. and it does not explain how come the real gain fo theopamp is frequency dependent and signal dependent ( opamps do have a slew rate too ) and what if the opamp only has a 5 volt supply , is wired to have a system gain of 100 and gets 1 vpp on its input ? is your 'maths equation' going to give you the right number at the output ( how are you going to provide notation for the fact the output will clip , may go over the slew rate )

You will end up with pages of equations that need analyzing and their results compared. Each of these equations will give conflicting results becasue none of them can accomodate for all parameter. The only practical way is to notate all this stuff down into a netlist and feed it through a mathematical processor called 'Spice'... and we all know how far you can trust that crap. as far a you can throw the models being fed into it. The spice engine is solid , but as with all abstract maths systems : the output is only as good as the input. And part of the input is the device model. And this one is seriously lacking in many cases.

To make sense of all this math jiberish you will need a human that has a diffeent kind of understanding and knows when to throw away what output. A computer still can't do that.

The gain equation of an opamp as non-inverting amplifier is only true provided the input signal stays withing the input operational range , and the output stays within the output operational range. And if the signal falls withing the given gainbandwidth plot and does not contain edges that surpass the maximum slew rate.

The maths notation falls extremely short in this case ... as it cannot accomodate for all those extra factors.
And this is only one the most basic circuit possible with an opamp...

Yes, you need the simple basic equation so you can calculate the resistors to set the system gain.. but you also need the surrounding verbal conditions or you still will not understand why the danr thing doesnt work right ! having the equation is one thing, you also need to know in what circumstances the equation is valid.. and there are more exceptions than there are cases where the equation is right.

Besides , that standard inverting and non inverting equations  ( A = R1/R2  and A = 1+(R1/R2) R1 being output-input and R2 being input -signal or input-reference) are crap. They are oversimplifeid so they are only valid in one specific subvariant. The one where the other terminal is ground reference. apply a level there and it all goes to snot. The equation is invalid at that point.
So far for the 'maths notation' .. completely useless without knowing the scope in which the equation is applicable. Like i said : math is a shorthand notation form. but it is very limited as it does no concept of limiting factors. One of the basic things you learn in maths is that a given function can have two outcomes... up to you to decide which of the two is valid. the equation cannot decide it for you.

Besides. the circuitry around an opamp only requires ohms law to solve. if you know that the sum of all currents in a node is 0 ( every electron entering a node must leave somewhere ) and you apply ohm's law you can solve an opamp circuit for any possible case.

If AC is involved the resistor simply become frequency driven impedances and phases come into play as well. just an extention of ohms law ( current / voltage / impedance relation )

[free_electron]

Offset voltage (and zeroing thereof), CMRR, bandwidth, slew rate, how a "precision" op-amp differs from a standard one, rail-to-rail output, differential opamp vs normal opamp, etc. There are many opamps available.

'precision opamp' is a marketing buzz.

There are whole bunch of factors , like you said
- input oparational range
- input common mode
- output common mode
- output operational range
- slew rate
- input bias current
- common mode rejecten  (CMRR)
- power supply rejection (PSRR)
- input offset
- gain/bandwidth
- input capacitance

just open a datasheet for a quality opamp like an OP27 and you will be dazzled by the sheer amount of parameters they throw at you.
You will need to know what is important for your design and which ones you don't care about.

And then there are internal techniques , like chopper stabilizing.

By the way, even in switchin regulators things are starting to be very hairy ... that feedback pin you think is used to monitor the output voltage ? it's doing way more than that.... they can measure the current throught the coil , see how far it is from saturating and much more.
depending on what the mosfets in the power stage are doing they sample the output pin and the switch node and derive information from that..

anyway... if i get 5 minutes i will wirte up some stuff. but yeah. maybe dave should do a 'demisitifcation' of opamp terminology

[poptones]

So there you go : your mathematics notation fails you...

Uh huh. If I even acept your nit picking it seems more like the English language failed me. Forgive me, I wasn't writing a book.

[deephaven]

A small hint.
Practical Op-Amps have a parameter called bias current. This is the small current that the Op Amp draws from it's input pins. If you feed the inputs from very different impedances, you will get an unwanted offset voltage. So, try to keep the input impedance of the connected circuitry the same on both inputs and the offset voltages will tend to cancel out.

[free_electron]

A small hint.
Practical Op-Amps have a parameter called bias current. This is the small current that the Op Amp draws from it's input pins.

There are opamps that actually GIVE current out of their input pins. Jfet opamps like the tl082 are notorious for that.
this current can , if you use large resistance values, create enough offset to mess things up.

In precision circuitry you will often see an extra resistor between the feedback node and the actual input.

Code: [Select]


       ------r2----
      |            |
--r1--+--r3--|-\   |
             |  >--+---
        -r4--|+/
       |
      GND

R3 and R4 serve no purpose in an 'ideal opamp'. Input impedance is infinite...
In a real opamp there Is an input current. By making R3 = R4 the current drawn , or delivered, by the inputs creates a voltage drop across R3 and R4. Since both are the same value this drop is equal and thus becomes a common-mode signal, which is rejected by the opamp.
An opamp is only interested in differentials , anything common is rejected. So that's how you solve the problem.

Now, there is more. This assumption is only true IF the node formed by R1 and R2 is low impedant ! otherwise it is possible to still pull that one out of whack since that current going into or out of the inut may still influence that node.
If r1 and r2 are in the megaohm range you don't stand a chance of getting the bias cancelled, unless you use active cancelling ( by using an additional opamp to sense the drop across R4 and feed it back into the node.  Or you add on a matched pair of fets or mosfets at the input to drop this bias current. That is for example youw you build an electrometer that can measure femtoamperes. you can't do that with a normal opamp as it may be faulting the measurement by injecting or drawing simply too much current itself...

Most opamps have bias current in the microamp range , some go all the way down to femto ampere.
Where is this important ? well, let's say you have a sensor with an output impedance of 10 megaohm. The sensor delivers a voltage between 1 and 100 millivolts. The opamp is supplied with a 5 volts power supply.

If your opamp has a input bias current of 1 uampere ... 1 microampere through 10 megaohm is ... 10 volt ! You can kiss your signal goodbye... the opamp is already well outside its operating range ( it only has 5 volts to work with ) so the input is clipping at 5 volts ( assuming it's a rail to rail input opamp and then still it would be like 4.7 volt . if it's a classic, non rail to rail input, opamp and the pump happens to be topside it may clip at 3 volts ... ).

So that circuit will never work. moreover , you are overloading the output of the sensor with a high voltage and you may destroy it !

Think this is exagerated example ? the carbon monoxide sensor in your house's fire alarms are exactly that. They are an electrochemical cell that delivers a few millivolts across a 1 megaohm resistor. Hook up the wrong kind of opamp and your system doesn't stand a snowball's chance in hell to work. As an extra:  apply more than 100 millivolts externally and you kill the sensor electrode.

So yes, input bias current is an extremely important factor in certain designs. in others ? meh , who cares.

And there is no amount of maths that is going to teach you that. You need to know what to look for and what you are working with. Sure, you will use math to see if you fall within range. But it's easier if you know what to look for without having to calculate everything everytime.

[Excavatoree]

Math does not guarantee or give you a deeper understanding.

Memorizing the equations still gives you no 'deeper' understang of anything. There is nothing 'deeper' to understand to begin with.
The circuit is the circuit is the circuit, and it damn well behaves as the electrons run. An no amount of maths or other notation formats is going to change that or your understanding of it.

Exactly.  I knew a guy in college (University) that was brilliant at mathematics.   He decided that one of the resistors in the "classic 555" circuit should be zero ohms, and dutifully plugged zero into the equations.  He couldn't figure out why it didn't work.

Simple, I said.  There has to be resistance between those two nodes.   The circuit won't work with a direct short.

But the equation says it will work - this is BS.

I never did get him to understand.

Sorry for being a bit off topic and rambley.

[GK]

A small hint.
Practical Op-Amps have a parameter called bias current. This is the small current that the Op Amp draws from it's input pins. If you feed the inputs from very different impedances, you will get an unwanted offset voltage. So, try to keep the input impedance of the connected circuitry the same on both inputs and the offset voltages will tend to cancel out.

This was basically true for all cases back in the good old days, but now a lot of op-amps are designed with input bias current cancellation. For these parts the maximum input bias current offset error often specs worse that the typical input bias current!

For these op-amps, deliberately equalising the DC impedances beween the inputs can actually make the total input offset error worse.

 

[urbanwriter]

Just a couple of short observations. Certainly happy to read that I'm not the only person who finds one or two of the regular posters to be as obnoxious as humanly possible; perhaps they missed 'hobbyists, hackers and makers' following 'engineers' in the header.

I'd also like to thank a couple posters in this thread for information that was topical and/or an effort on their part; to name a few, free_electron for what could be the script for Dave's video, to VonKlitzing for their comment and the Afrotechmods video, Psi and Excavatoree for constructive comments.

Now I actually have a very quick-to-read bit on op-amps that will help as I inadvertently blow a couple up. But that won't happen until this multi-servo, multi-555, doo-dad works the way I 'calculated' it should.

Now, back to having fun, more or less on my terms.