Op-Amp Design References and Tools -- Fundamentals?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

TBailey

New member
Joined
Jan 6, 2006
Messages
3
Location
Chicago, USA
Hi there.

First off, I just wanted to thank any and all of you who've contributed to the discussions in the "Discrete Op-Amp" meta-thread. Many of them are wonderfully well-informed and inspiring; they make me excited that there are people out there who really love analog design, and study it for no other reason than that.

Now, that having been said:

I'm designing my very first discrete operational amplifier, and I have some questions.
In real life, I'm a (newish) design engineer for a toy company and mostly do C and assembly programming, but analog is a lot more fun. My first experiences with electronics were as a repair technician, and my first circuit designs were all audio related -- amplifiers and synthesizers, sensor interface circuits for piezos, passive RFID stuff -- but all of it was using monolithic parts. Until now I haven't fussed around very much with discrete transistors, other than sticking a high-current output stage onto something else, or making a monostable or the like.
Lately, I've gotten it in my head that until I really understand the intricacies of how an op-amp works internally, I won't really understand the nuts and bolts of my analog designs, and that the best way to understand op-amps is to build some.

With that out of the way:
My questions to start with are mostly about the books I should have read and the gear I should ideally have:

*1.) So to get ready for this project I went back through the transistor chapter in Horowitz and Hill, and have been plowing through the 1973 edition of the National Semiconductor Linear Applications as per the recommendations on the Meta thread. I've been using a Bob Pease book on troubleshooting and have gotten really into the Barrie Gilbert sections in an IC design book I've got. The question:

What other reference works would you all recommend?

*2.) I have a pretty normal test bench, plenty of leads, scope, function generator, etc etc. I don't have a distortion analyzer, or a spectrum analyzer. I've noticed that much of the time when the people on this list
make reference to THD and, when they can, the harmonic series of that THD. It seems like that would be a really fundamental indication of how an op-amp would sound in the real world -- and it's something I'd like to know more about. I'd love to be able to hear a tone, and know if it has a lot of 3rd harmonic, say, and be familiar with that would look like on the scope. It seems like that skill would be invaluable in audio design.

My question: How are you all making those measurements? Do all of you die-hards have audio precisions and whatnot?

(I have a Tek 500 series frame, and I could probably pick up a used AA501a and use that -- but would that help?)

*3.) Again, I can't find many good references on the subject of origins of distortion outside of audiophile cat-fights. Occasionally, Gilbert will say something about emitter area mismatch in a diff-pair leading to 2nd order harmonic distortion, or ohmic resistance leading to a dominant third harmonic, or somebody on this list will throw out something to the effect of "class A amplifiers have even order distortion", but where can I read about why this happens, and how to know how it'll happen, and how much?

*4.) I've never used a circuit simulator. I've noticed that opinion seems to be divided on that practice in this group. I'd be just as happy getting by without one -- but will I be missing out on something by not doing it?

*5.) Anything else I should know / do / have before I get down to the breadboard?

I know this is a lot to ask, and that these are pretty fundamental questions, but any pearls of wisdom you all might be able to drop about this would fall on very appreciative ears. This field seems totally fascinating, but there's a lot to know. Maybe that's why.

Again, thanks in advance, and you all really have something special here.

Yours,

Todd Bailey
 
[quote author="TBailey"]*2.) I have a pretty normal test bench, plenty of leads, scope, function generator, etc etc. I don't have a distortion analyzer, or a spectrum analyzer. I've noticed that much of the time when the people on this list make reference to THD and, when they can, the harmonic series of that THD. It seems like that would be a really fundamental indication of how an op-amp would sound in the real world -- and it's something I'd like to know more about. I'd love to be able to hear a tone, and know if it has a lot of 3rd harmonic, say, and be familiar with that would look like on the scope. It seems like that skill would be invaluable in audio design.

My question: How are you all making those measurements? Do all of you die-hards have audio precisions and whatnot?

(I have a Tek 500 series frame, and I could probably pick up a used AA501a and use that -- but would that help?)[/quote]

I dunno about anyone else, but for distortion measurements (harmonic distortion and high-frequency IM) I use the spectrum analyzer in Adobe Audition and a CardDeluxe audio card. I've developed a series of measurements (440Hz harmonic distortion, 19+19.5kHz IM distortion, 19+19.5+9.6kHz, etc.) that show up all sorts of misbehavior.

Peace,
Paul
 
I like a book by Dennis Feucht, Handbook of Analog Circuit Design, for many things, among them his discussion of distortion mechanisms. The book itself may be hard to come by but I believe he supports a CD ROM through a website that has updated and expanded material (EDIT yes---see www dot innovatia.com---certainly the way to go, since the book doesn't turn up on bookfinder at all, and Amazon has only one independent seller willing to part with an ex-library copy for a trifling US $689.50 !!!).

For op amp design there'e the old Burr-Brown sponsored tome by Tobey, Graeme, and Huelsman. Pease IIRC likes a book by Jiri Dostal, Operational Amplifiers. I have it but haven't gotten into it much, and glancing at it now I see no mention of distortion per se.

These and others should turn up in searches here, as they have been discussed before.
 
[quote author="Dan Kennedy"]I got a lot out of Gray and Meyer's "Analysis and Design of Analog Integrated Circuits". Mine is a 2nd Edition, it's probably been revised by now...[/quote]

I was disappointed that rev. 4, adding co-authors Hurst and Lewis, has absolutely nothing about JFETs. :evil:
 
Thanks to everyone for all the recommendations -- what with all of these and the books I already ordered, I can safely say I have more to read about the subject than I will ever be able to humanly finish.

Looking through the list, I decided to order Gray and Meyer text (I've never seen such glowing reviews of an electronics textbook on Amazon since I picked up H&H as a dewey-eyed English major with his first Weller) -- even if it might not have anything on JFETs. (NB: it's always seemed to me like JFETs are the most reclusive of the transistor family tree -- I use mosfets all the time for mundane PWM and power switching, and BJTs for current sources, amplifiers, etc -- and although I once ordered some discrete JFETs for completeness' sake, I can't say I've ever quite known what to do with them. But)
I also wrote to Dennis Feucht about getting a copy of his Analog Circuit Design CD, which looks very useful, and I found a used copy of the Burr-Brown book for next to nothing. I nabbed all the PDFs you all recommended, too.

As far as Doug Self goes, I've read the stuff on his site and breadboarded some of his circuits, but he's so acerbic. He obviously has spent hundreds of hours backing up what he believes, and knows what he's talking about. But that whole reactionary absolutist my-way-or-the-highway tone is pretty hard to get through. Plus, he describes BJTs as "voltage operated devices" -- certainly not how I'd come to think of them. If you all feel strongly that book would be really useful, though, I'll stop being snotty, suck it up, and buy it.

Thanks again so much for all the pointers. I expect I'll post again soon to ask you all again about your test rigs -- I hope these questions aren't too longwinded or simple.

Yours and thanks,

Todd
 
Self is an arrogant b*st*rd egregiously lacking in civility, and I mean that in the nicest possible way. But he does give us all a good example to avoid (which I guess I'm not doing very well here :razz: ).

However, he is talented and does know his stuff pretty well.


About FETs: although great strides have been made in the processes to reduce the density of surface and interface states, MOSFETs are still a good deal noisier than good JFETs at low-medium audio frequencies. They started out just terrible.

Since FETs, at audio frequencies, can be paralleled for a square-root-of-n-device voltage noise reduction, until the capacitance starts to load things down they are able to beat about most anything at even fairly low source impedances. The gate leakage current does begin to accumulate, and doubles for about every 10 degrees K, so eventually this can start to be noticeable. For truly heroic designs you can cool them too, and the transconductance goes up as well down to nearly liquid nitrogen temperatures.

But they are a pain in the ass to use in production, since the parameter spreads are large compared to bipolars.

EDIT: asterisks inserted a la Vonnegut, in Breakfast of Champions :razz:
 
[quote author="bcarso"]
Since FETs, at audio frequencies, can be paralleled for a square-root-of-n-device voltage noise reduction, until the capacitance starts to load things down they are able to beat about most anything at even fairly low source impedances. The gate leakage current does begin to accumulate, and doubles for about every 10 degrees K, so eventually this can start to be noticeable. For truly heroic designs you can cool them too, and the transconductance goes up as well down to nearly liquid nitrogen temperatures.

But they are a pain in the ass to use in production, since the parameter spreads are large compared to bipolars.[/quote]

W/r/t paralleling: does that mean that when you buy a discrete low-noise JFET it is actually many paralleled devices on one die (like a power mosfet) or does that mean that you would hand match individual monolithic JFETs and use them in parallel? It seems like that would require a lot of tinkering.
Further, am I reading you correctly, that when you say: "they are able to beat about most anything at even fairly low source impedances" that noise figures are worse somehow when the impedance driving a circuit is low? I always though that it was a rule of thumb to have a low impedance source driving a higher impedance load (and again, my experience with FETs in motor drives and the like is that they are driven more quickly from a low impedance) -- but I imagine I must be missing something. Do you think you could say a little bit more about that statement?

Thanks again,

TB
 
FETs are characterized by, among other things, an effective channel length (~distance between drain and source). Making the width larger is the equivalent of paralleling devices.

A few very low-capacitance low-transconductance devices look something like the simple structure of the basic JFET diagram shown in texts when you pop the hood, although you get a top view when manufacturers show the chip (or when you peer at it out of the case) and thus you have to infer the 3D picture. But most JFETs, including garden-variety general-purpose types, have several fingers of drain and source contacts, so can be seen as the equivalent of paralleled devices.

When you try to make a really massive paralleled device in one pass the yield suffers, unless you are more conservative in how far you push the process. Due to the clever technology that's evolved for memory and other VLSI chips, the power MOSFETs now can be massively parallel internally with acceptable yields; there has been no economic motivation to do the same with JFETs. So, it is easier to parallel devices in separate packages, knowing that each one is tested before shipment.

To your second question: if "current" noise is small, the "voltage" noise of an amplifying device determines what the lowest noise figure is with a given resistive source. Since the resistive source's noise goes as the square root of the resistance, lower source impedances require lower voltage noise devices for the same amplifier noise contribution.

Bipolars have lots of current noise, so (a) paralleling them to reduce voltage noise, or (b) running them at higher currents until the base sort-of-contact resistance's thermal noise becomes dominant, or both (a) and (b), eventually results in the current noise flowing in the source getting comparable to the voltage noise. JFETs have such low low-frequency current noise (essentially the shot noise and excess noise in the gate leakage current) that they can, by comparison, be paralleled until the cows come home. However, you may not be able to live with the input capacitance after a while.

Bipolars are wonderful and indispensable, as are insulated-gate devices. I was just disappointed that all trace of junction FETs disappeared in the latest Gray et al. tome.
 
[quote author="bcarso"]I like a book by Dennis Feucht, Handbook of Analog Circuit Design, for many things, among them his discussion of distortion mechanisms. The book itself may be hard to come by but I believe he supports a CD ROM through a website that has updated and expanded material (EDIT yes---see www dot innovatia.com---certainly the way to go, since the book doesn't turn up on bookfinder at all, and Amazon has only one independent seller willing to part with an ex-library copy for a trifling US $689.50 !!!)..[/quote]

And another seller on Amazon UK has one - might be the same copy, might not - for £952.76!!!

Which leaves me wondering what to do now. I have an absolutely mint copy on my shelf - I bought it, opened it, shut it again and it hasn't been touched since. (Don't get me wrong, it is a SUPERB book if you're into op-amp design, but not if you like to just buy your op-amps and use them.)

If it's going to be useful to someone here I'd rather they made me an offer for it than it carried on gathering dust on my shelf.
 
I looked at my copy in a whole new light when I saw that Amazon price Andy! I guess that it must have been a fairly small initial printing.

Actually the book is good for amplification in general---Feucht clearly learned well when he worked for Tektronix. The material includes a great deal about predominantly open loop designs, as required by scope vertical amplifiers. The material on wideband compensation I had most of in some course notes from another ex-Tek guy, albeit in a quite terse form. The bits on thermal distortion are most helpful, extending the material by Addis (Tek great, again) in one of Williams's books. There is also lots of good stuff on nonlinear circuits.

I shudder at how close I came to never pulling the book off the shelf when I saw it at Opamp Technical Books in LA (what, me needing yet another book on circuit design?). Opamp always gets top dollar too, in this case $65 US in 1990.
 
[quote author="bcarso"]Actually the book is good for amplification in general---Feucht clearly learned well when he worked for Tektronix. The material includes a great deal about predominantly open loop designs, as required by scope vertical amplifiers. The material on wideband compensation I had most of in some course notes from another ex-Tek guy, albeit in a quite terse form. The bits on thermal distortion are most helpful, extending the material by Addis (Tek great, again) in one of Williams's books. There is also lots of good stuff on nonlinear circuits.[/quote]

Having looked again - it is literally 13 years since I last looked at it - you're dead right, there is excellent stuff here if you want to 'push the boundaries' of your chips.
 
> Self is an arrogant b*st*rd egregiously lacking in civility

That's probably why I like him.

He gives you a firm foundation for disagreement.

Of course you can't get away with that chit unless you are pretty-right: Doug is.

And unlike those 'scope-designer books, Doug is all about audio. (Well, and steam engines...)

> seemed to me like JFETs are the most reclusive of the transistor family tree

Anything you can do with a JFET, you can do with a BJT. Almost. There are a few niches where JFETs rule: these were mostly old 30 years ago so they don't get into the spanking new textbooks.

> I once ordered some discrete JFETs for completeness' sake, I can't say I've ever quite known what to do with them.

Use them like vacuum tubes, except change 250V to 25V and adjust the plate resistor accordingly. About the only difference is that they have a "built-in screen grid", so if you have more than a few volts across the channel, they act like pentodes (but without screen grid supply complication).

> he describes BJTs as "voltage operated devices" -- certainly not how I'd come to think of them.

Yeah, they suck current. How much? We can't ever know. Hfe is very variable, from part to part and with current. Static Vbe is also variable from batch to batch and with current. BUT: the dynamic change of Vbe DOES follow Shockley's Law with exactitude (stay away from the device's maximum rated current where parasitic ohms lurk).

So for audio precision, we usually look at the voltage across the Base-Emitter junction. Not the voltage we apply: that may be different because there are no zero-ohm sources and that Base is sucking significant and loosely-known current. And Shockley's Law is not a nice thing for audio; too bad, but that's the best we can do.

FETs (including vacuum tubes) also aim at Shockley's Law, but miss. You can come pretty close with a generic JFET at super-low current, or quite close at audio-useful currents with the big JFETs. In a vacuum tube, the poor leverage of a wound-wire grid against a cloud of electrons masks Shockley; If we could wind the grid fine enough, we have to account for the temperature term in Shockley.

> when you say: "they are able to beat about most anything at even fairly low source impedances" that noise figures are worse somehow when the impedance driving a circuit is low? I always though that it was a rule of thumb to have a low impedance source driving a higher impedance load

Lo-Z drive, hi-Z load is more a convention than a necessity.

> noise figures are worse somehow when the impedance driving a circuit is low?

When the source is a passive resistor (a good model for strain gauges, dynamic mikes, and many other things), it has a Noise Power which can be re-interpreted as a Noise Voltage and a Noise Current, their ratio being the source resistance.

A 10 ohm resistor has a lower noise voltage (and higher noise current) than a 10K resistor.

Amplifiers have noise current and noise voltage too. Choice of device and operating condition affects these parameters. For any device, there will be an operating condition that gives lowest Vn*In and lowest noise power. In BJT, the optimum is broad: we can pick a ratio of Vn/In that approximates our source impedance. In FETs, In is nearly invariant (but very low), and Vn can only be pushed down so far.

If we can use a transformer, we can make any source work with any low-noise amplifier, even if the impedances are different. For FETs at low frequencies, In is nearly zero, the optimum source impedance is awful close to infinity, we use a transformer with a hi-Z secondary. But hi-Z windings have real audio problems. We ask how low we can go before noise gets noticeable. Historically, below 10K audio impedance, an FET was not your first choice for low noise. Recent fat JFETs can be run with very low Vn, comparable to a hundreds-ohm source, which used to be BJT turf. A BJT can reach even lower Vn, but will then have high In: you have to find a balance.

Low-noise design may conflict with BJT bias current goals, or the offset voltage of JFETs may upset a DC error budget.
 
[quote author="PRR"]> Self is an arrogant b*st*rd egregiously lacking in civility

That's probably why I like him.

He gives you a firm foundation for disagreement.

Of course you can't get away with that chit unless you are pretty-right: Doug is.

[/quote]

:green:

It's not so much his book writing and articles as his letters to Editors lambasting other writers. He did a hatchet job on Norm Thagard for example that was simply uncalled-for. It's that air of "How dare you write about amplifiers! I have already written the last word on them, until of course I vouchsafe you peons another bit of my largesse..."
 
Back
Top