ImprovedOpAmpLimited

[AHaumer]

This PR replaces #4485 which will be closed but we still can look up the discussions.
The following explanation could be used for future release notes:
Modelica.Electrical.Analog.Ideal.IdealizedOpAmpLimited has been declared as deprecated since some of the results (Modelica.Electrical.Analog.Examples.OpAmps) could not be compared between different tools, because the usage of noEvent() and smooth() for saturation prevented exact events in switching applications.
As a replacement Modelica.Electrical.Analog.Ideal.ImprovedOpAmpLimited has been implemented, thoroughly tested with different tools and used in the above mentioned examples. Saturation has been designed simpler, events are triggered by two Boolean variables that indicate saturation. Additionally a firstOrder models rise time of the OpAmp. Initialization is now without homopotopy acting on the result of the firstOrder (before saturation acts on the output).

[HansOlsson]

I'll look into it.
Just a quick style-question - is there a particular reason you switched from SI.Frequency (etc) to Modelica.Units.SI.Frequency?

[AHaumer]

@HansOlsson this was by accident.
I copied all the OpAmp stuff out of MSL, then developed and tested and copied back in.
You're right, looks clumsy. Therefore replaced all Modelica.Units.SI. by SI.

[maltelenz]

Without looking at the actual changes, I ran the full MSL tests, and got result differences in Adder, ControlCircuit and Subtracter.

Blue is the reference data, orange is the new result with this PR.

Adder:

In ControlCircuit it looks like an event became a softer transition at 0.5? This is zoomed in:

Subtracter:

[AHaumer]

Sorry @maltelenz I forgot to mention:

• In some examples I flipped the output voltage sensor vertically. That explains the drastic differences.
• Of course, in detail some results might be different (due to the firstOrder!)
  Important: The SignalGenerator runs stable!

@GallLeo please generate new reference results for Modelica.Electrical.Analog.Examples.OpAmps.*
after this PR has been accepted and merged.

[AHaumer]

After merging this PR, I'll prepare another PR containing a remake of Modelica.Electrical.Examples.ChuaCircuit using realistic parameters (mH and pF insteaf of H and F), showing the path from 2 periodic attractors to chaos and an additonal example demonstrating how the characteristic of Chua's diode could be implemented with 2 OpAmp-circuits called NIC (negative impedance converter) - that example relies on the OpAmp-model of this PR.

[HansOlsson]

In general good, but I would need some feedback on the filter-part.

[AHaumer]

In general good, but I would need some feedback on the filter-part.

You mean the firstOrder? Give me some time over the weekend to design initialization.

[AHaumer]

@HansOlsson I hope I addressed all your concerns, please re-review.
Now the firstOrder is optional (and therefore the initialization for it). It is needed only in 2 examples: MultiVibrator and SignalGenerator. The results seem to be the same in Dymola, OpenModelica and SystemModeler.
@casella @maltelenz @henrikt-ma may I kindly ask for your review?

[henrikt-ma]

I am not a domain expert, so this set of changes is too big for me to comprehend. However, I am pleased to see that @maltelenz is reporting any consequences this will have in our test suite, and that @HansOlsson is paying attention to possibly unintentional stylistic changes.

[AHaumer]

Ok there a some cosmetic changes (my fault I know) in the examples.
The main work is a nearly complete redesign of the OpAmp model, some more explanations:

• The differential input voltage is either just amplified by V0 or (optional) additionally delayed with a firstOrder, this gives the fictional intermediate voltage
• If the optional firstOrder is activated this intermediate voltage can be initialized as Linear, UpperLimit, LowerLimit.
• This intermediate voltage gets limited between positive and negative supply without triggering events to get output voltage.
• Events are triggered by 2 Boolean indicators of saturation, so saturation should be documented exactly.

Should I include a test model (meaybe in ModelicaTest?) measuring the rise time?
To meet the RiseTime stated in data sheet (e.g. LM741 0.3 µs), the time constant has to be set to V0/log(9)*RiseTime.
RiseTime is measured in a voltageFollower circuit and defined by Tr = t90 - t10 where
0.1 = 1 - exp(-t10/Tau) and 0.9 = 1 - exp(-t90/Tau), using a step input.
The differential equation for the voltageFollower taking the firstOrder into account but avoiding saturation is:
Tau*der(vOut) = V0*(vStep - vOut) - vOut.
Many applications are nearly not affected by this firstOrder, so it is deactivated.
For some applications the user would have to carefully compare time constant and transient behaviour of his input.
For some applications (MultiVibrator and SignalGenerator) the firstOrder is essential to eases numerical solution.

[AHaumer]

wait wait wait ... silly me: I forgot to check / adapt the comparisonSignals.

Then I want another change of the order of the examples. After that we get:

• 10 examples using the ready-to-use models from OpAmpCircuits
• 4 non-autonomous circuits and 1 autonomous (up to now uncritical) switching circuit
• 2 autonomous critical circuits

And I'll implement the correct measurement of rise time in ModelicaTest.

All done after the following commit (latest this evening).

[christiankral]

I wonder if ImprovedOpAmpLimited is a good choice for a component name. One day IdealizedOpAmpLimited will be removed and we might not remember exactly what was improved ...

[christiankral]

Sorry @maltelenz I forgot to mention:

• In some examples I flipped the output voltage sensor vertically. That explains the drastic differences.

We might want to add a comment in Modelica.Electrical.Analog.UsersGuide.ReleaseNotes in case someone is wondering why the results appear differently in MSL 4.1.0.

[AHaumer]

@christiankral I accepted all your valuable suggestions with 9209342.
Regarding the naming, in Modelica.Electrical.Analog.Ideal we have;

• IdealOpAmp (infinite gain, no saturation)
• IdealOpAmp3Pin (infinite gain, no saturation)
• IdealOpAmpLimited (infinite gain, saturation, compulsory supply)
• deprecated IdealizedOpAmpLimited (finite gain, saturation, optional supply)
• new ImprovedOpAmpLimited (finite gain, saturation, optional supply, optional firstOrder rise of output, stability)

The new one is not only improved comparing wih the deprecated one, but also with the older three.

[christiankral]

OK then, so we stick with the name ImprovedOpAmpLimited.

If I everybody else agrees with the name, it's fine with me, too...

[AHaumer]

Text for release notes:
ImprovedOpAmpLimited replaces IdealizedOpAmpLimited (which is deprecated), with finite gain, saturation, optional supply and optional firstOrder rise of output which improves stability.
The examples in OpAmps are adapted to use the new ImprovedOpAmpLimited and harmonized polarity of output voltage sensor.
@GallLeo @MatthiasBSchaefer please provide new reference results for all examples in Modelica.Electrical.Analog.Examples.OpAmps after this PR has been merged.

[casella]

Unfortunately the discussion on the naming was too quick for me to react. I don't really think that "Improved" is a good attribute for any Modelica Standard Library model. Same as "New". It is new or improved now, and then it will be soon obsolete, as all things human.

The name should describe what the model represents, not how its development history went.

Why don't we just call it OpAmpLimited? It's not idealized, because it has a finite gain. So, it's just an op amp. @AHaumer do you agree?