Operational Amplifier Design

[ameya]

I have to design an operational amplifier for a analog course. Could
some one please tell me how do I start with the design ? Which
topology should I use ? How do I go about with all the design
constraints ? Which constraints need more attention while designing ?
Thank You.
Ameya

 

[Dbowey]

ameya posted:
<< I have to design an operational amplifier for a analog course. Could
some one please tell me how do I start with the design ? Which
topology should I use ? How do I go about with all the design
constraints ? Which constraints need more attention while designing ?
I suggested you explain to the person who gave you this task, that you lack the
knowledge to attempt it.

Don

 

[Ken Smith]

I have to design an operational amplifier for a analog course. Could
some one please tell me how do I start with the design ?

First you have to decide on an amplifying technology. Op-amps can be made
with bipolar transistors, JFETS, MOSFETs, tubes, magnetic-amplifiers and
amplidynes or any combination of these.

Usually you base the selection on what devices you can obtain that have a
hope of meeting the specifications you are designing to.

Which
topology should I use ?

If you download the datasheet for an LT1028, you will see a good starting
point. I'd suggest you insert 2 extra gain stages and make it work rail
to rail. The LT1028 is a fairly old design so it doesn't swing rail to
rail.

How do I go about with all the design
constraints ?

It is usually best to take them in the following order:

(1) Crossover distortion.
(2) Maximum side fumbling.
(3) Underpass distortion.
(4) Output cross conduction.
(5) Supply voltage.
(6) Supply current.
(7) Transient image distortion
(8) Spectral purity.

Which constraints need more attention while designing ?
Thank You.

The one that says "you are supposed to design an op-amp".

 

[Ban]

I have to design an operational amplifier for a analog course. Could
some one please tell me how do I start with the design ? Which
topology should I use ? How do I go about with all the design
constraints ? Which constraints need more attention while designing ?
Thank You.
Ameya

You need a differential input stage, for example a differential amplifier
with a current mirror as a load, an intermediate current amp, in this case a
darlington with a current source as a load and an output stage for high
voltage swing and low O/P resistance.

+-----------------------------+----+
| | |
/ \ / \ |
( I )25uA ( I )500uA |
\_/ \_/ |
| | |/
+---+---+ +--------|
| | | |>
Vn |< >| V diode |
o-| |-o Vp - | Vout
|\ /| | +----o
| | V diode |
+---+ +---+ Ccomp - |
| | | | | |<
\| | |/ | ||33p +--------|
|-+-| +-||---+---+ |\
<| |> | || | | |
| | | |/ | |
.-. .-. +----| | |
| | | | |> |/ |
1k| | | |1k +-| |
'-' '-' |> |
| | | |
+-------+--------------+----------+---+
(created by AACircuit v1.28 beta 10/06/04 www.tech-chat.de)

special attention has to be brought to the frequency compensation. The Ccomp
lowers the first time constant so far, that the second one kicks in only
after the gain is below 0dB. The phase margin at this frequency should be
higher than 65° to get good settling times and little overshoot.

 

[John Popelish]

I have to design an operational amplifier for a analog course. Could
some one please tell me how do I start with the design ? Which
topology should I use ? How do I go about with all the design
constraints ? Which constraints need more attention while designing ?
Thank You.
Ameya

You study how others have done it.

Start here:
http://www.national.com/an/AN/AN-A.pdf

 

[Ban]

+-----------------------------+----+
| | |
/ \ / \ |
( I )25uA ( I )500uA |
\_/ \_/ |
| | |/
+---+---+ +--------|
| | | |>
Vn |< >| V diode |
o-| |-o Vp - | Vout
|\ /| | +----o
| | V diode |
+---+ +---+ Ccomp - |
| | | | | |<
\| | |/ | ||33p +--------|
|-+-| +-||---+---+ |\
<| |> | || | | |
| | | |/ | |
.-. .-. +----| | |
| | | | |> |/ |
1k| | | |1k +-| |
'-' '-' |> |
| | | |
+-------+--------------+----------+---+
(created by AACircuit v1.28 beta 10/06/04 www.tech-chat.de)

special attention has to be brought to the frequency compensation.
The Ccomp lowers the first time constant so far, that the second one
kicks in only after the gain is below 0dB. The phase margin at this
frequency should be higher than 65° to get good settling times and
little overshoot.

BTW you can buy this circuit commercially, it is called LM741.

 

[Tim Wescott]

I have to design an operational amplifier for a analog course. Could
some one please tell me how do I start with the design ? Which
topology should I use ? How do I go about with all the design
constraints ? Which constraints need more attention while designing ?
Thank You.
Ameya

Really, this should be taught during the course of the course. I've
never actually taken that one, but I had friends who did when I was an
undergrad, and I TA'd the lab for it when I was getting my Master's
(best course I never took -- I had to know all the important stuff for
the lab, so I walked away with lots of good info for free).

At any rate, it was the same in two universities, 2500 miles and several
years apart. As the course unfolded the relevant bits and pieces of the
circuit were analyzed in class and built for the lab. I can't imagine
that a halfway decent instructor won't make sure that you have the
information you need when you need it.

 

[Kevin Aylward]

First you have to decide on an amplifying technology. Op-amps can be
made with bipolar transistors, JFETS, MOSFETs, tubes,
magnetic-amplifiers and amplidynes or any combination of these.

Usually you base the selection on what devices you can obtain that
have a hope of meeting the specifications you are designing to.

I would suggest a bog standard one. pnp diff input feeding a current
mirror load, feeding an npn stage with a current mirror load. e.g.
http://www.dself.dsl.pipex.com/ampins/dipa/dipa.htm, the circuit
somewhere near the bottom of the page.

However, my preference is for the second stage also to be a diff pair
with a current mirror load.

If you download the datasheet for an LT1028, you will see a good
starting point.

Oh dear. I would suggest a simplar one.

I'd suggest you insert 2 extra gain stages

Oh dear.. Again not a good start. A *total* of 2 gain stages are a
better starting point for a novice. Its can be very difficult to get 3
stage and more to be stable.

and make
it work rail to rail.

The LT1028 is a fairly old design so it

doesn't swing rail to rail.

That is not why it doesn't swing rail to rail. Your pretty new at this
amplifier design aren't you

Rail to rail op-amp have a number of compromises that are better avoided
unless you actually require rail to rail output. For example, it is
usually much better to have push pull emitter followers as the output
devices. Why do you think this is?

It is usually best to take them in the following order:

Here we go again...

(1) Crossover distortion.
(2) Maximum side fumbling.

?

I see your fumbling here.

(3) Underpass distortion.
?

(4) Output cross conduction.
(5) Supply voltage.
(6) Supply current.
(7) Transient image distortion

Oh dear...why do I bother...

(8) Spectral purity.

Overall to the above, No.No.No.

First things first. Its always a trade off between accuracy, BW and
power.

What is the op-amp to do?

Input and output volts and current required?

For a novice, a good starting point is say 50ma out with +/- 15 volt
rails.

The fundamental real design issue is *stability*. The dc design is
usually straightforward. However, one often spends 95% of the time on
getting the bastard stable.

What unity gain frequency is required?

Better accuracy requires more feedback, hence high ugf. etc...

The one that says "you are supposed to design an op-amp".

The one than says don't give advice unless you actually know what you
are talking about

And lastly, the best introductory tutorial I know of is:

http://www.national.com/an/AN/AN-A.pdf

My advice is anyone contemplating designing amplifiers understand this
paper in complete detail. Its a true goldmine.

Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

 

[Kevin Aylward]

You study how others have done it.

Start here:
http://www.national.com/an/AN/AN-A.pdf

Dash...I made my post before I read this one. Its nice to know that
there are a few who are on the ball.

Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

 

[Pooh Bear]

I have to design an operational amplifier for a analog course. Could
some one please tell me how do I start with the design ? Which
topology should I use ? How do I go about with all the design
constraints ? Which constraints need more attention while designing ?

Frankly, it sounds like a thankless task.

No less than thousands of designs of op-amps have been manufactured over
the years by many IC manufacturers in various attempts to optimise their
designs according to the semiconductor processes available to them at the
time and also to meet differing market requirements / sectors ( e.g low
power / low voltage / high voltage / load impedance / low distortion / DC
accuracy / HF bandwidth / low-high bias current - you name it ! ).


You can choose whatever constraints you choose. Since you'll be designing
a discrete op-amp no doubt - at least you won't have certain
semiconductor *processes* and their constraints to worry about.


Sounds like a poorly conceived project to me. Is the lecturer a bit
clueless ? Seen that before.


Graham

 

[Leon Heller]

Frankly, it sounds like a thankless task.

No less than thousands of designs of op-amps have been manufactured over
the years by many IC manufacturers in various attempts to optimise their
designs according to the semiconductor processes available to them at the
time and also to meet differing market requirements / sectors ( e.g low
power / low voltage / high voltage / load impedance / low distortion / DC
accuracy / HF bandwidth / low-high bias current - you name it ! ).


You can choose whatever constraints you choose. Since you'll be designing
a discrete op-amp no doubt - at least you won't have certain
semiconductor *processes* and their constraints to worry about.


Sounds like a poorly conceived project to me. Is the lecturer a bit
clueless ? Seen that before.

It's quite a popular project on degree courses; a former colleague of mine
had to do something similar. IIRC, his SPICE simulation worked OK, but the
actual circuit built from BJTs etc. didn't.

Leon

 

[Paul Burridge]

I have to design an operational amplifier for a analog course. Could
some one please tell me how do I start with the design ? Which
topology should I use ? How do I go about with all the design
constraints ? Which constraints need more attention while designing ?
Thank You.
Ameya

Oh hell, just copy a 714; no one will notice. ;-)

 

[Winfield Hill]

Paul Burridge wrote...

Oh hell, just copy a 714; no one will notice. ;-)

If you meant the 741, with its unusual and distinctive input stage,
it certainly would be noticed. If you meant the uA714 precision
opamp, that's an awkard and overly-complicated starting design.

 

[Ken Smith]

Ken Smith wrote: [..op-amp design ..]

made with bipolar transistors, JFETS, MOSFETs, tubes,
magnetic-amplifiers and amplidynes or any combination of these.

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[....]

If you download the datasheet for an LT1028, you will see a good
starting point.

Oh dear. I would suggest a simplar one.

Oh dear, K.A. thinks the posting was serious even with the suggestions I
made and the fact that I was answering an obvious homework question and
the more suggestions below.

[... snippage ...]

?

Go search on "reto-encabulator" or "turbo-encabulator". There was a
thread mere month ago where we all giggled about that one.

 

[Ken Smith]

You need a differential input stage, for example a differential amplifier

You are answering someone's homework question for them. This person
either (a) skipped a lot of classes (b) has the worst instructor in town
or (c) is unable to absorb the information from class.

If it is (a) you are encouraging him to continue to skip classes and not
learn. This may result in him having the paper but not the knowledge that
goes with it. It is no service to him.

If the situation is (b) I think we would be much better off explaining why
things should be done rather than what should be done. We could offer him
a complete solution but I don't think he would learn much from it.

If the answer is (c), it is best that the person fail now before he has
invested too much time in the effort. He may have abilities in other
areas. If he fails his op-amp design question but does well in optics or
something, he is likely to switch to a better option.

 

[Ken Smith]

Paul Burridge wrote...

If you meant the 741, with its unusual and distinctive input stage,
it certainly would be noticed. If you meant the uA714 precision
opamp, that's an awkard and overly-complicated starting design.

The LM324 would be a much better example to start with. IIRC, it has a
darlington in the input that could be left out.

 

[PaulCsouls]

I have to design an operational amplifier for a analog course. Could
some one please tell me how do I start with the design ? Which
topology should I use ? How do I go about with all the design
constraints ? Which constraints need more attention while designing ?
Thank You.
Ameya

If you are serious about learning how to do it but are having trouble
understanding your teacher and textbook, I suggest getting Analysis
and Design of Analog Integrated Circuits by Paul A Gray and Rober G.
Meyer. I bought a used copy for 7 dollars.

Paul C

 

[PaulCsouls]

It's quite a popular project on degree courses; a former colleague of mine
had to do something similar. IIRC, his SPICE simulation worked OK, but the
actual circuit built from BJTs etc. didn't.

Leon

Op Amps require matched components. You'll have a lot trouble trying
to build it with discretes.

PaulC

 

[Ban]

You are answering someone's homework question for them. This person
either (a) skipped a lot of classes (b) has the worst instructor in
town
or (c) is unable to absorb the information from class.

If it is (a) you are encouraging him to continue to skip classes and
not learn. This may result in him having the paper but not the
knowledge that goes with it. It is no service to him.

If the situation is (b) I think we would be much better off
explaining why things should be done rather than what should be done.
We could offer him
a complete solution but I don't think he would learn much from it.

If the answer is (c), it is best that the person fail now before he
has invested too much time in the effort. He may have abilities in
other
areas. If he fails his op-amp design question but does well in
optics or something, he is likely to switch to a better option.

--

Ken, this information is available all over the net, and in every book. I
can see, you never visited a course like this. The principal design
equations include deriving the Bode-diagramm by calculating the open loop
gain and phase, the bias conditions and quiescent current in all stages, the
output drive capability, input and output impedances of every single stage.
I added a few typical currents. Of course it gives the disciple a kick
start, but maybe that is what is needed.
It is also for guys like you to get an estimate, compare it with what you
wrote for example which was pretty much off. At least he can derive some
useful hints I hope.

 

[Pooh Bear]

It's quite a popular project on degree courses; a former colleague of mine
had to do something similar. IIRC, his SPICE simulation worked OK, but the
actual circuit built from BJTs etc. didn't.

There's a lesson there. Learn how to design practical circuits that work before
using simulation. Too many graduates know loads of theory and precious little
practice about making things ( that work ).


Graham