Making an op-amp with discrete components

[canadianpoet2012]

I've often wondered if it would be possible to construct an op-amp with discrete components! The floor plan is famous for only having one capacitor, along with 20 transistors and 11 resistors. There's nothing special to it and it isn't really that complex. Although it would require a custom circuit board.

So I have two main questions;

a) Is it actually possible, or would you come across noise problems or oscillation etc?

b) Does anyone know of the component values, or an accurate floor plan with the values on it?

Any comments would be most welcome.

 

[Styx]

the main problem that you will run into is missmatch in PNP/NPN characteristics. On a IC opamp the characteristics are pretty much the same. This makes the current mirrors used within opamps very ccurate and stable.

why do it anyway? it would be abg 20x more expensive than the IC variant and abt 1000x bigger

 

[Russlk]

Many years ago, befor the UA741, I built op amps from discrete components, I even had a vacuum tube in the front end of one. I don't have any of those schematics anymore and I would not do it in the same way today anyway. In the 1950's Heathkit put out an analog computer comprised mainly of vacuum tube op amps.

 

[bmcculla]

This sounds like a good excercise for simulation. You can build an op amp in spice and fiddle with it. I did an opamp project in school using spice - it was a great way to learn about transistors.

Brent

 

[Optikon]

the main problem that you will run into is missmatch in PNP/NPN characteristics. On a IC opamp the characteristics are pretty much the same. This makes the current mirrors used within opamps very ccurate and stable.

why do it anyway? it would be abg 20x more expensive than the IC variant and abt 1000x bigger

Where I work, some of our designs are pushing the edges of the state of the art and we find MANY occasions to do this sort of thing, mainly because nobody makes what we want/need.

There are ways to compensate for the mismatching.

With that said, why make what you can buy? I saw the other day a decent JFET opamp that was around 29 cents in small quantities.. Hard to beat that! :wink:

 

[fat-tony]

There are ways to compensate for the mismatching.

Like buying matched pairs

 

[Nigel Goodwin]

Most discrete audio amplifier designs are simply high power opamps, many high quality phono preamplifiers are also just discrete opamps. Using discrete components gives you the huge advantage that you can easily use capacitors - which is difficult in IC technology.

 

[canadianpoet2012]

Thank you for all of your comments and advice. I was aware of the mismatch problem, I guess the easiest way to fix that is to purchase matched pairs for the mirror parts of the circuit.

As for why do it; why not? It would be interesting to see how close I could get it to a real 741's characteristics. Plus it's just interesting to see if it's possible! I'm of the opinion that all this new fangled digital stuff is very boring! All this little black boxes - not humming, or glowing bits (hopefully!)

 

[Styx]

sorry if sounded like "there is no point, why do it"

As to a project sounds good. But still the mismatch will be a problem. Even from so called same-batch I had problems getting decent matching for a simple current-mirror let alone a decrete diff-amp front end that you will need.

 

[Optikon]

There are ways to compensate for the mismatching.

Like buying matched pairs

Yes, that certainly makes it easy.. but it's not the only way. For example, if I have a drifty circuit that maybe doesn't even use a transistor, I could make 2 drifty circuits and arrange my topology such that they both contribute in an equal but opposite manner thus compensating for the drift contribution. And so on and so forth, lots of clever ways..

Another good example of this is the forward biased diode in series with a zener diode. The voltage error due to zener drift is compensated for by the same kind of drift of the regular diode but it is in the opposite direction.

 

[Dean Huster]

A discrete op amp

Op amps have always been around, whether made with vacuum tubes or transistors. There were companies that made op amp modules. They were pricey and they had a lot of limitations.

Even with today's matched components, you'll have a heck of a time building a discrete op amp that has any decent bandwidth, high slew rate, good CMRR, low input offset current, good noise immunity, high input impedance and high open loop gain. They couldn't ever do all of that when the modules were being professionally built. It took the integrated op amp to began to get all the specifications to start moving toward their ideal points. Always you have compromise. You usually don't get high-gain with high bandwidth, etc. It's only been in the last 10 or 15 years (the IC op amp has been on the market for at least 35 years) that the bandwidths have gotten into the hundreds of megaHertz without that much compromise on other specs, and that's been with a lot of different companies putting a lot of bucks into research and design.

There's nothing wrong with trying this project out. Just don't expect to even approach even the specs of the lowly 741. Doing it with a simulator as suggested might be a more worthwhile investment of your time so that you can use the time saved designing a project that will be more practical.

Dean

 

[ante]

The greatest challenge as I se it is the fact that the op-amp is built on one chip. The benefits from this are enormous. And one of the biggest is temperature; all components are exactly the same temperature. This is impossible to achieve with discrete components. The performance is not likely to get anywhere near the real thing.

Ante :roll: