Project help design an op-amp

[ericgibbs]

MikeMl said: Good reason to use the (after simulation) LTSpice schematic from which to wire up the circuit on the whiteboard. I always do. I dont bother drafting a schematic until I simulate, and then breadboard.

This is also my chosen method.
E

 

[spec]

This is also my chosen method.
E

Eric,

Would you ever produce a circuit diagram?

spec

 

[ericgibbs]

spec,
In answering an OP's question, I would try to determine the draft specification of what the OP is trying to achieve and try to determine the technical ability of the OP. [ with minimal effort on my part.]

From that I would do a quick mental analysis on the best way to way to get the result he wants.
Draw up a LTSpice sim and test my proposed solution and try to optimise the circuit.
If it feasible to do so, I would bread board the circuit.

If it looks OK, I would add text to the LTS sim, explaining to the OP the operation of the circuit.
Post a screen clip of the circuit, together with more helpful text.

Usually the OP will come back saying he cannot get the components I have used and ask can he use alternatives or some other problem, so I would retry the Sim and then reply.
This can be a reiterative procedure as you well know ...

In producing a final circuit diagram, I would normally only do that if posting to my Blog pages.

E

 

[spec]

Eric

The only fundamental difference there is sim or no sim. One thing I would say is that sim schematics are more difficult to follow than well drawn conventional schematics: the OP implied this. This is especially relevant when the wide range of people viewing ETO are considered; not all are experts like you and MrAl. In an ideal world, it would be great if sims could work from conventional schematics, as the board layout function does in ECADs. But then we would have nothing to discuss. Some ECADs do have this capability to a degree.

In industry the circuit diagram is the central document. It leads to the board layout, parts lists (BOM), mechanical design, and product documentation among other things. These areas would be covered by different people who simply would not understand sim schematics; they have enough problems following conventional schematics, especially for critical applications where speed and extreme precision may be involved. And if you are working in an integrated design environment, with many engineers and other staff working on the same system, the control would be extremely difficult and, in most cases, impossible. It is bad enough using conventional schematics.

Thus, there is the problem of errors going from the sim schematic to the conventional schematic. As I said before, errors are endemic especially as you have different disciplines working on the same job: wireman, draftsman, component engineers, reliability engineers, safety engineers, technical publications, and so on. Also, sim schematics do not include the ancillary components that make a practical circuit work, decouples, gate stoppers etc. They can be included but rarely are, certainly on ETO.

The other thing is that a sim schematic approach becomes unmanageable and error prone for large and complex circuits, where even conventional schematics can be a nightmare, especially if they are badly drawn- another subject that has already been discussed elsewhere on ETO.

Can I just say that I am not anti sim; quite the reverse.

spec

 

[Kerim]

In my case, I do not less than 50 silly mistakes when I design a new product; most of them in writing the firmware (I use doing it in assembly only to get the most of the MCU I have in term of speed, space and timing accuracy). So every time I start a new design, I prepare myself to be patient to correct one error after another. The reason I do mistakes is that I focus first on the general topology then gradually I explore and fix one detail after another which, in turn, may likely affect another detail which was already assumed being good.

To me in the least, designing (a controller, for example) is playing with trade-offs in order to let the required functions be as accurate as possible, practically and/or economically speaking. It happens that, since I was a teen, I have enjoyed playing with numbers, symbols, equations and logic. In fact, the greatest joy that my human living flesh lives is during the process of thinking... thinking out of the box, of course.

Kerim

 

[spec]

You sound like a typical engineer, with an inquiring and inventive mind. Many of us on ETO have been bitten by the electronics bug. 50 silly mistakes: is that all.

Just a couple of stories about errors:

I reviewed an aircraft transponder project; the specification said that the transmit power should be X Watts for a range of Y kilometers. So the team went designing away, but they were designing a system that guaranteed that transponder would not work at the specified range because, from the RADAR equation, the customer had defined a power at the limit of detectability. Boy did that cause the project some problems. The transponder was battery powered and getting 10 Watts radio frequency was bad enough; 100w was almost impossible (rough figures). The person writing the specification subsequently got sacked.

When wordprocessors came out they were a gosend because getting large complex documents correct was a long and arduous task. The procedure was that the engineers would write much of the text and supply the supporting data and the Techical Publications department would turn that into a formal document for external distribution. Some of their documents were fantastic.

We asked them to use spell checkers to save time correcting errors. Their response was that they were professional authors and could spell- unlike engineers- and they did not need spell checkers. Subsequently, when we went through their documents and red-lined the spelling errors, they changed their minds, especially when we asked then to pay for correcting the errors out of their department funds, rather than project funds.

If you examine almost anything in detail you will find errors, some gross.

Remember the saying: The person who never made a mistake, never made anything.

spec

 

[ericgibbs]

In industry the circuit diagram is the central document. It leads to the board layout, parts lists (BOM), mechanical design, and product documentation among other things.

spec,
The vast majority of OP's on this Forum are asking for simple advice, which they want to have explained in a simple, straight forward manner.
Any OP asking for help that has obviously a good engineering background will get a proportionate response from me.
So, I try to gauge the technical level of my response, to the abilities of the OP.

I think my approach has worked well over the past 9 years.

Not wishing to sound picky, I do have a little experience in this matter.
Perhaps checking the Information under my Profile would be informative.

Eric

 

[spec]

spec,
The vast majority of OP's on this Forum are asking for simple advice, which they want to have explained in a simple, straight forward manner.
Any OP asking for help that has obviously a good engineering background will get a proportionate response from me.
So, I try to gauge the technical level of my response, to the abilities of the OP.

I think my approach has worked well over the past 9 years.

Not wishing to sound picky, I do have a little experience in this matter.
Perhaps checking the Information under my Profile would be informative.

Eric

Eric

And from me too

You don't need to explain- I have read a lot of your stuff on ETO way before I joined

spec

 

[JoeJester]

One thing I would say is that sim schematics are more difficult to follow than well drawn conventional schematics: the OP implied this.

I disagree. One can draw the simulator schematics clearly, although, I've seen some hurriedly assembled schematics that I may or may not attempt to redraw. You don't draw schematics on the simulation programs, you assemble them. The reason they are more difficult to follow is based on how the assembler chooses to present the material. How many times have we seen poorly assembled schematics and the person's breadboard circuit does NOT work, as it is as disorganized as the schematic.

 

[JoeJester]

When wordprocessors came out they were a gosend because getting large complex documents correct was a long and arduous task.

When wordprocessors came out, so did the ability to copy and paste. I've seen documents where the copy and paste injected different standards. Of course all of this could be resolved when the "editors" reviewed the material.

 

[spec]

I disagree. One can draw the simulator schematics clearly, although, I've seen some hurriedly assembled schematics that I may or may not attempt to redraw. You don't draw schematics on the simulation programs, you assemble them. The reason they are more difficult to follow is based on how the assembler chooses to present the material. How many times have we seen poorly assembled schematics and the person's breadboard circuit does NOT work, as it is as disorganized as the schematic.

I disagree with your disagreement JJ.

spec

 

[spec]

When wordprocessors came out, so did the ability to copy and paste. I've seen documents where the copy and paste injected different standards. Of course all of this could be resolved when the "editors" reviewed the material.

Hy JJ,

Yes, like any powerful tool you can get into all sorts of pickles with word processors especially in the area of formatting. The early versions of Word were fun in that respect. But that is mainly because the template and preferences had not been configured. At out Technical Publications department we had a relatively young ex RAF chap, who was not only a good author but could do anything with Word and other word processors, so it is possible. Most of the problem is that we don't learn the ins and out of Word, like we would a microprocessor for example.

All I can say is that Word transformed, beyond all recognition, the process of getting documents assembled . Not only that, but you could reuse the boiler plate, and in our documents there was tons of it, with no effort at all. In many cases you could do the same for the specific content too. Then you get automatic index generation, automatic page numbering, automatic headers and footers, automatic watermarking- the list goes on and on.

The first word processor I ever used in anger was on the BBC Computer that I took to work from home. I can't remember the word processor name but it was not WSWIG and was very elementary in all respects. I could only scrounge a 9" green screen monitor too. But it made writing this particularly awkward document I had to write child's play by comparison to using a chisel and stone tablet.

Then spreadsheets came out- another earth shattering event. And then CAD- heaven. Then ECAD- nirvana

Having said all that, I still like writing with a fountain pen on good quality paper.

spec

 

[JoeJester]

spec

We can disagree, but in the latter part of the 20th century, technical manuals go from full schematics to just words, tidbits of circuit explanations, simple block diagrams and running wire lists. So, teaching that equipment was easy. I once developed a set of functional diagrams covering the whole system to which that equipment was connected. There was a problem, as indicated on the equipment, and my technician looked at the diagram I created while a visiting engineer grabbed that "technical manual." He hadn't found the area in the running wire list that he was seeking before my technician identified the source of the problem and fixed it.

I've seen other documentation that had nice schematics and the block diagrams were "inaccurate."

Of course these were 25 years ago, but I doubt things got better. On my side of the pond, fifty years ago there was a company that produced "Sam's photofacts", technical documentation for consumer products. I don't think they are in business anymore, as there are more disposable consumer electronics.

on edit ... they still exist.

https://www.samswebsite.com/

 

[spec]

spec

We can disagree, but in the latter part of the 20th century, technical manuals go from full schematics to just words, tidbits of circuit explanations, simple block diagrams and running wire lists. So, teaching that equipment was easy. I once developed a set of functional diagrams covering the whole system to which that equipment was connected. There was a problem, as indicated on the equipment, and my technician looked at the diagram I created while a visiting engineer grabbed that "technical manual." He hadn't found the area in the running wire list that he was seeking before my technician identified the source of the problem and fixed it.

I've seen other documentation that had nice schematics and the block diagrams were "inaccurate."

Of course these were 25 years ago, but I doubt things got better. On my side of the pond, fifty years ago there was a company that produced "Sam's photofacts", technical documentation for consumer products. I don't think they are in business anymore, as there are more disposable consumer electronics.

on edit ... they still exist.

https://www.samswebsite.com/

The manuals are the same here too. Block diagrams have always been used in handbooks and also descriptive text. The customer also had a data pack with all supporting design information.

With the advent of LSI circuit diagrams become little more than interconnections diagrams. And there was higher level interconnections diagrams to describe the overall system. Sometimes the documentation set was massive and filled a small room. It was also expensive and was good business for the company.

spec

 

[MikeMl]

Mike, your simulation of circuit #27 is incorrect. For example, at the zero point the simulator shows an output error of 9mV ...
If a perfect opamp were assumed, there would be absolutely no offset error. How a simulator can say anything else throws doubt on the simulator or the implementation. And how you can't see this by a cursory inspection of the circuit is surprising.

You are right. Your R6 and R7 form a voltage divider such that the voltage at V(n) (the inverting input) is exactly zero V. Since I am using LTSpice's "ideal" opamp (notice it has no part number designator, and no power supply connections), the opamp forces V(i) to be zero as well. Since in an ideal opamp, the bias currents and input offsets are zero, then V(spec) obviously has to be zero, as well. Notice that this does not validate the "gain" of the circuit; only the "offset".

Here is the LTSpice .DC solution confirming this. Be aware I am posting only a "thumbnail"; you have to click on the image below to make it open in its own browser tab at a scale where you can read it.



So why does this sim show a perfect zero V output, while the previous one showed an "error". Well, the answer is that in the previous sim, I plotted the difference of two node voltages. V(spec) was zero, but the 9mV offset was actually in the expression where I computed V(des) = 7.63*V(sensor) - 12.85. I allowed some round-off error in the transposed equation, and that accounts for the 9mV offset. Here is the sim again, this time I wrote the transposed equation so that LTSpice computes the coefficients to its full floating point precision (no round off), instead of me doing it on a calculator. The simulator did exactly what I asked it to...



So back to the "gain" validation. Here the sim again, this time I asking the simulator to evaluate the "better" transposed equation, as well as asking it to compute V(spec) as a function of the parameter "Pressure". In the upper pane, I plot V(sensor), V(des), and V(spec). Within the screen pixel resolution, V(spec) falls on top of V(des), so it appears ok. In the lower plot pane, I plot the difference between V(spec) and V(des), and now we see a slight difference.

... Simulations are not the real world and the models are not verified as Texas Instruments and National Semiconductor point out. Also there is the old maxim: garbage in, garbage out.

Simulators, like computers, do exactly what you ask them to do. You can validate lots of circuits without using any models, as I have showed here.

I am not even going to respond to your points about my post 7&9. As I have already said, that was an intermediate circuit to lead the the student to a solution, rather than just solving the entire problem and just handing the solution to the student without any explanation, which is what you did initially in post 10 and 13. It was only after I nagged you about the homework issue , that you wrote the explanation of your circuit in post 17...

 

[JoeJester]

Here's Spec's circuit using the OPA model from TI.

Of course, if I wanted to get the resistance to match closer to Spec's using series resistance to get the closest I can .... Here's that circuit

 

[spec]

You are right. Your R6 and R7 form a voltage divider such that the voltage at V(n) (the inverting input) is exactly zero V. Since I am using LTSpice's "ideal" opamp (notice it has no part number designator, and no power supply connections), the opamp forces V(i) to be zero as well. Since in an ideal opamp, the bias currents and input offsets are zero, then V(spec) obviously has to be zero, as well. Notice that this does not validate the "gain" of the circuit; only the "offset".

Here is the LTSpice .DC solution confirming this. Be aware I am posting only a "thumbnail"; you have to click on the image below to make it open in its own browser tab at a scale where you can read it.

View attachment 99566

So why does this sim show a perfect zero V output, while the previous one showed an "error". Well, the answer is that in the previous sim, I plotted the difference of two node voltages. V(spec) was zero, but the 9mV offset was actually in the expression where I computed V(des) = 7.63*V(sensor) - 12.85. I allowed some round-off error in the transposed equation, and that accounts for the 9mV offset. Here is the sim again, this time I wrote the transposed equation so that LTSpice computes the coefficients to its full floating point precision (no round off), instead of me doing it on a calculator. The simulator did exactly what I asked it to...

View attachment 99568

So back to the "gain" validation. Here the sim again, this time I asking the simulator to evaluate the "better" transposed equation, as well as asking it to compute V(spec) as a function of the parameter "Pressure". In the upper pane, I plot V(sensor), V(des), and V(spec). Within the screen pixel resolution, V(spec) falls on top of V(des), so it appears ok. In the lower plot pane, I plot the difference between V(spec) and V(des), and now we see a slight difference.

View attachment 99569



Simulators, like computers, do exactly what you ask them to do. You can validate lots of circuits without using any models, as I have showed here.

I am not even going to respond to your points about my post 7&9. As I have already said, that was an intermediate circuit to lead the the student to a solution, rather than just solving the entire problem and just handing the solution to the student without any explanation, which is what you did initially in post 10 and 13. It was only after I nagged you about the homework issue , that you wrote the explanation of your circuit in post 17...

Hy Mike,

Thanks for you well put reply- much appreciated.

spec

 

[spec]

Here's Spec's circuit using the OPA model from TI.

View attachment 99571

Of course, if I wanted to get the resistance to match closer to Spec's using series resistance to get the closest I can .... Here's that circuit
View attachment 99572

Hell JJ,

What a beautifully laid out and presented post. It is so clear that even I could understand it. Your post is also much appreciated.

That would certainly look good in any project documentation.

Could I impose on you and, if you have some available time, do a run on circuit #13.

spec

 

[MikeMl]

Here is a comparison of Spec's circuit in post 13 with the "better" Transposed equation:



(click on the thumbnail)

 

[spec]

Here is a comparison of Spec's circuit in post 13 with the "better" Transposed equation:

View attachment 99576

(click on the thumbnail)

Nice job Mike.

spec