# Help with discrete op-amp

Hi, my name is Quan and I am new to this forum. I am trying to build a discrete operational amplifier but I am very weak with circuit designing since my concentration is in communication. Anybody here have a lot of experience building analog op-amp circuits? If you are in the area then that's great but I am sure we can still work together through the internet otherwise.

The spec for the design is as follow:
Two inputs: 1 plus and 1 negative
Discrete devices (BJT, FET, TUBES) only
Gain between 1000 – 1500 to provide a maximum +/10 output signal
Frequency response DC – 10 Mhz +/- 2db
Has to be able to drive a 50k load
Input impedance of at least 1 Megohm
Output impedance less than 50 ohms
Temperature range: 25 – 100 degrees Celcius

Please e-mail me at quan24@gmail.com if you can help. Thank you in advance!

Last edited by a moderator:

Mentor
Hi Quan, You shouldn't have to pay anybody to make this for you -- is it for a school project? Are you getting instruction in how to design opamps and other analog ICs?

A discrete opamp will have mediocre performance compared to a monolithic IC version, so this must be for some sort of learning/class setting. I'd recommend that you get the textbook "The Art of Electronics" by Horowitz and Hill, and read chapters 1-3. Chapter 3 is on opamps, and by the time you've read through chapters 1-3, you will understand how to design and build your own discrete opamp circuit, and you will be able to make one that meets the specs that you list. The book is a fast read, so you will be able to get through chapters 1-3 in less than a week, I would think.

The most important thing in building a discrete opamp is that you will need to use dual transistors that match well -- those are needed for the differential amplification stages. If the transistors are singles, their matching will not generally be good enough to be used for a real diff amp stage. The H&H textbook talks about matched dual transistors in several places, and gives example part numbers that you can use in your design.

BTW, the one spec that you list:

"Frequency response DC – 10 Mhz +/- 2db"

does not make sense for the opamp alone. The gain-phase plot will have a gain characteristic with a single pole rolloff (from the dominant pole position at low frequency, usually set with a Miller capacitance in the first diffamp's output or the 2nd amp stage's output) out to the 2nd pole near the unity gain crossover. The spec you list sounds more like a performance spec for some feedback configuration of the opamp, like when it's used as a unity gain follower.

Get the H&H book if you can. You will learn all you need for basic electronics from the book, and it's really a fun read. Good luck! -Mike-

So you are saying I should make an op-amp with 10G GBW and then use external feedback to get the correct gain and frequency response?

Can't I just the op-amp with the dominant pole at 10 Mhz from the beginning so I don't have to use any feedback circuit?

Also, are u free for more advice? :) I will check out the book as you suggested and work on this circuit. Can I send you pictures of my schematics along the way if I am stuck on something? I am having trouble finding people who are willing to give me guidance on this circuit

The circuit has to provide a gain from 1000 from dc to 10 Mhz so that's the 3db. This is not the unity gain frequency.

Mentor
Quan24 said:
So you are saying I should make an op-amp with 10G GBW and then use external feedback to get the correct gain and frequency response?

Can't I just the op-amp with the dominant pole at 10 Mhz from the beginning so I don't have to use any feedback circuit?

Also, are u free for more advice? :) I will check out the book as you suggested and work on this circuit. Can I send you pictures of my schematics along the way if I am stuck on something? I am having trouble finding people who are willing to give me guidance on this circuit

The circuit has to provide a gain from 1000 from dc to 10 Mhz so that's the 3db. This is not the unity gain frequency.

A gain of 1000 at 10MHz? That's pretty dang fast, Quan. Are you sure that's what the specs mean? Especially for a discrete opamp design, that's very fast. That puts unity gain at 10GHz. Good luck. And good luck closing feedback around a 10GHz opamp... And even an open-loop amp at 10GHz with a gain of 1000 is a very tall order. Definitely not a beginner electronics project. No, I think that you are misunderstanding that specification, or else the person who gave you that spec is mistaken.

As for the dominant pole, no, it will generally be in the 10-100Hz range. Let me point you at some opamp datasheets that should hopefully put you on the right track (in combination with the H&H textbook that I mentioned). Go to National Semiconductor's website, and check out the datasheet for the LF353, for example. It's an older FET input opamp (to get the higher Zin in your specs), but its architecture and specs are pretty typical of simpler opamps that you might try to build (a reduced functionality version of) using discrete components:

http://www.national.com/ds/LF/LF353.pdf [Broken]

There is a simplified schematic early in the datasheet, and a more complete schematic later. It would be good practice for you to calculate the performance of the simplified schematic, using parameters for discrete components like a matched dual FET that you would use as the input diffamp pair. Note also the open-loop gain plot in the datasheet, and where the dominant pole is being placed.

Also, if you are "having trouble finding people who are willing to give me guidance on this circuit", what is the assignment for? Is there no teacher? Can you describe the circumstances where you have been given this task, and what resources are available to you?

Last edited by a moderator:
I am e-mailing the professor again to verify if this is indeed true. Because I am getting repeated response of people that the design is too tough if it's indeed the case. This is a "senior design project". I do remember him multiplying the gain 1000 with the number he gave for the frequency response 10 Mhz and got a 10 Ghz gain bandwidth product. So if the GBW is 10 Ghz, isn't it true? Am I misunderstanding the definition of gain bandwidth product? And he made a comment that this blows LM741 out of the water. I would think so when LM741 is only 0.9 Mhz

I really think this senior design project is nuts. I have only taken 2 classes analyzing simple circuits like CE and etc. and I have to design this circuit. I try to ask professors but they do not make themself available to me (literally by not answering my e-mail or telling me they are too busy when I go see them in their office) and all the students are useless since it's beyond their ability. All the students who have the same projects are in the same boat as me. And if I don't finish the project, I won't graduate until I do...it sucks :(

Gold Member
I don't think it is that difficult. Just keep the open loop gain low.

es
I had a project just like this in undergrad. I think most EEs do. However we only had to have a gain bandwidth product of 100k and believe me with cheap, slow, high-tolerance, mismatched components, lousy package characteristics, and the stray capacitance and intrinsic resistance of bread boards that was tough. Perhaps only 1/3 of us actually got it to make spec.

What materials/components were you planning on using when building your opamp? Are they providing material or do you have to buy it? Do you actually have to build it, or is simulation sufficent?

I recommend doing searches online to find other discrete opamp design projects. Often you can find reports which include designs, and you can compare their requirements to your own. Also, many opamp datasheets have circuit diagrams in them learn from those. H&H is good advice.

But most importantly, find out if your understanding of the requirements is correct. If it is, spice a standard opamp design using only the best case parameters of your discrete transistors. I seriously doubt the open loop gain will be greater than 1 in the GHz range.

es
I just happened to think, how are they going to verify your design meets spec? Do you have a description of the test setup? I have a feeling you professors are playing marketing type tricks with your parameters.

I have to build it and show it to him. This is driving me nuts. He stopped responding to my e-mails. I think I have to go and bother him in his office. I don't know what is wrong with him. This is the toughest project out of all the ones that he gives out. Students who are assigned this project get their graduation delayed at least a couple of semester and it keeps on happening this way. I have not heard of one student who have actually met the spec that he gave out. I think it's a torture test given to students he dislike.

Mentor
Quan, When you go talk with the professor, take a copy of the AD8009 datasheet with you. It's an "ultrafast opamp" from Analog Devices, and it has a small signal BW of 1GHz.

This is near the state of the art for monolithic IC opamps in terms of speed, so if your prof is asking you for an order of magnitude better performance from a discrete design...

Hopefully with a datasheet in your hand, you two can talk about what he really wants from you in the project. Maybe he really means an open-loop amp, instead of an opamp that you would use feedback around. Although, doing an open-loop amp with a gain of 1000 at 10MHz is still very non-trivial.

Ok I e-mailed him and he e-mailed me back with the following response

The 10MHz is the 2db response. Therefore, the
3db bandwidth is much higher probably around
17Mhz or higher. Yes the unity gain bw product
is very high , therefore the transistors used must
be higher than around 1GHz apiece per stage
or higher based on your gain per stage.
It can be done others have already accomplished