From: <Michael.M<argraf@al...>  20050727 16:14:40

Hello, >> Concerning the point 2, I will try to elaborate: I need to simulate a >> coil L(I) where the value of >> the coil depends on the value of a current calculated in the circuit. >> How do I make it with the equation module? > > I just want to clarify my TODO list item. So I ask who listens: > > It is straight forward to implement a feature which enables qucsator to > take a equation variable's content and fill it into the property of > some component if: > > 1) the variable is a constant (the result) > 2) the variable depends on a sweep variable > 3) the variable depends on another components property > > But problems occur (as in the above mentioned case) when > > 4) the variable depends on another components operating point > 5) the variable depends on a simulation result (e.g. node voltage > or branch current) > > Is it necessary for these two cases to simulate, apply property, > simulate, apply property, simulate, etc. ... until no (or little) > change in simulation results occurs? We need to repeat single > simulations until results converge? Or am I on the wrong track? > > Please help me clarifying! > > Thanks in advance, > Stefan. I would like to put in my opinion on that issue: I think the abovementioned method is not very usefull! What Patrick needs is a nonlinear inductor, i.e. an inductor with hysteresis. Or do you mean the inductance depends on a current somewhere else in the circuit ? Then you would need a symbolical defined multiport. The abovementioned method leads to too much hidden faults and it is not "simulationlike" as Stefan already tried to explain. So, to sum up: Qucs should have a nonlinear inductor and/or a symbolical defined device, right ??? Regards, Michael 
From: Stefan Jahn <stefan@gr...>  20050727 11:30:15

On Di, 26.07.2005, 11:17, Patrick Cornille sagte: > Dear Stefan Hello! > Concerning the point 2, I will try to elaborate: I need to simulate a > coil L(I) where the value of > the coil depends on the value of a current calculated in the circuit. > How do I make it with the equation module? I just want to clarify my TODO list item. So I ask who listens: It is straight forward to implement a feature which enables qucsator to take a equation variable's content and fill it into the property of some component if: 1) the variable is a constant (the result) 2) the variable depends on a sweep variable 3) the variable depends on another components property But problems occur (as in the above mentioned case) when 4) the variable depends on another components operating point 5) the variable depends on a simulation result (e.g. node voltage or branch current) Is it necessary for these two cases to simulate, apply property, simulate, apply property, simulate, etc. ... until no (or little) change in simulation results occurs? We need to repeat single simulations until results converge? Or am I on the wrong track? Please help me clarifying! Thanks in advance, Stefan. 
From: <Michael.M<argraf@al...>  20050727 16:14:40

Hello, >> Concerning the point 2, I will try to elaborate: I need to simulate a >> coil L(I) where the value of >> the coil depends on the value of a current calculated in the circuit. >> How do I make it with the equation module? > > I just want to clarify my TODO list item. So I ask who listens: > > It is straight forward to implement a feature which enables qucsator to > take a equation variable's content and fill it into the property of > some component if: > > 1) the variable is a constant (the result) > 2) the variable depends on a sweep variable > 3) the variable depends on another components property > > But problems occur (as in the above mentioned case) when > > 4) the variable depends on another components operating point > 5) the variable depends on a simulation result (e.g. node voltage > or branch current) > > Is it necessary for these two cases to simulate, apply property, > simulate, apply property, simulate, etc. ... until no (or little) > change in simulation results occurs? We need to repeat single > simulations until results converge? Or am I on the wrong track? > > Please help me clarifying! > > Thanks in advance, > Stefan. I would like to put in my opinion on that issue: I think the abovementioned method is not very usefull! What Patrick needs is a nonlinear inductor, i.e. an inductor with hysteresis. Or do you mean the inductance depends on a current somewhere else in the circuit ? Then you would need a symbolical defined multiport. The abovementioned method leads to too much hidden faults and it is not "simulationlike" as Stefan already tried to explain. So, to sum up: Qucs should have a nonlinear inductor and/or a symbolical defined device, right ??? Regards, Michael 
From: Stefan Jahn <stefan@gr...>  20050728 07:04:19

On Mi, 27.07.2005, 16:41, Michael.Margraf@... sagte: > Hello, Hi! >>> Concerning the point 2, I will try to elaborate: I need to simulate a >>> coil L(I) where the value of >>> the coil depends on the value of a current calculated in the circuit. >>> How do I make it with the equation module? >> >> I just want to clarify my TODO list item. So I ask who listens: >> >> It is straight forward to implement a feature which enables qucsator to >> take a equation variable's content and fill it into the property of >> some component if: >> >> 1) the variable is a constant (the result) >> 2) the variable depends on a sweep variable >> 3) the variable depends on another components property >> >> But problems occur (as in the above mentioned case) when >> >> 4) the variable depends on another components operating point >> 5) the variable depends on a simulation result (e.g. node voltage >> or branch current) >> >> Is it necessary for these two cases to simulate, apply property, >> simulate, apply property, simulate, etc. ... until no (or little) >> change in simulation results occurs? We need to repeat single >> simulations until results converge? Or am I on the wrong track? >> >> Please help me clarifying! >> >> Thanks in advance, >> Stefan. > > I would like to put in my opinion on that issue: > I think the abovementioned method is not very usefull! But points 1) through 3) may be useful? > What Patrick needs is a nonlinear inductor, i.e. an inductor > with hysteresis. Or do you mean the inductance depends > on a current somewhere else in the circuit ? > Then you would need a symbolical defined multiport. > The abovementioned method leads to too much hidden faults > and it is not "simulationlike" as Stefan already tried to > explain. > So, to sum up: Qucs should have a nonlinear inductor and/or > a symbolical defined device, right ??? I agree for points 4) and 5). Cheers, Stefan. 
From: Stefan Jahn <stefan@gr...>  20050728 09:04:21

On Do, 28.07.2005, 10:22, Patrick Cornille sagte: > Dear Stefan and Michael > 2) I still have problem with the printing command which scrambled > everything in the version 007. Ah. So it worked with version 0.0.6 or previous ones? So you may ask Markus Bongard <markus.bongard@...> who made the package for MacOS if it has possibly something to do with the used Qt version. Otherwise... we'll see. > 3) I did not succeed to open your file which was written on the version > 008 but I did succeed to open it as a text and compared > it with the version of my file and see my errors. I understand that we > must add .v or. i to the variables. The simulation did work perfectly. To use the (0.0.8) schematic I sent to you you could also have change the version number in the .sch file by hand... You must add .v or .i if you use the AC simulation. For transient it's .Vt and .It and for DC it's .V and .I, you can use every variable you find in the dataset as displayed by the diagram dialogs. > 4) I need to calculate some kind of paratransformer (see the paper > below) where the inductance of the coil > depends not only on the current flowing through the primary coil but > also on the current flowing through the secondary coil. > > I need to simulate a law of the form: > > L(Ip,Is)=[10.5*tanh((IpI0)/2))]*[10.5*tanh((IsI0)/2))] > > where Ip and Is are currents flowing through the coils and I0 is a DC > polarization current. > As for all parametric study, the values or R(I), L(I), C(V) depend on > the values of the currents or voltages somewhere > else in the circuit. I do not know the best way to do it in your > software but I need to do this simulation rapidly. Possibly another kind of component may help you?: I have a current controlled voltage source or a voltage controlled current source in mind. It is possible to have such a component with a (nonlinear) symbolic equation at hand. Is it possible to break your specific problem down to an equivalent circuit using these kind of components? > I am ready to make a financial contribution to the programmer who will > do it for me because this will help me a lot. Is this for your work, I assume? Cheers, Stefan. 
From: <Michael.M<argraf@al...>  20050729 06:44:22

Hello to all, >> 2) I still have problem with the printing command which scrambled >> everything in the version 007. > > Ah. So it worked with version 0.0.6 or previous ones? So you may ask > Markus Bongard <markus.bongard@...> who made the package for MacOS > if it has possibly something to do with the used Qt version. > Otherwise... we'll see. The printing is done by Qt, so I don't have much influence on it. Furthermore, I don't have a Mac to test it. Can you print it into a *.ps file and sent it to me ? >> 3) I did not succeed to open your file which was written on the >> version 008 but I did succeed to open it as a text and compared >> it with the version of my file and see my errors. I understand that >> we must add .v or. i to the variables. The simulation did work >> perfectly. > > To use the (0.0.8) schematic I sent to you you could also have change > the version number in the .sch file by hand... > > You must add .v or .i if you use the AC simulation. For transient it's > .Vt and .It and for DC it's .V and .I, you can use every variable you > find in the dataset as displayed by the diagram dialogs. I will add this topic in the Qucs help. >> 4) I need to calculate some kind of paratransformer (see the paper >> below) where the inductance of the coil >> depends not only on the current flowing through the primary coil but >> also on the current flowing through the secondary coil. >> >> I need to simulate a law of the form: >> >> L(Ip,Is)=[10.5*tanh((IpI0)/2))]*[10.5*tanh((IsI0)/2))] >> >> where Ip and Is are currents flowing through the coils and I0 is a DC >> polarization current. >> As for all parametric study, the values or R(I), L(I), C(V) depend on >> the values of the currents or voltages somewhere >> else in the circuit. I do not know the best way to do it in your >> software but I need to do this simulation rapidly. > > Possibly another kind of component may help you?: I have a current > controlled voltage source or a voltage controlled current source in > mind. It is possible to have such a component with a (nonlinear) > symbolic equation at hand. Is it possible to break your specific > problem down to an equivalent circuit using these kind of components? Ahh, so you need a transformer with a strange kind of hysteresis. (Indeed it is not really a hysteresis). A symbolical defined 4port would perfectly fit your need. But at the moment there is no component in Qucs that can handle your problem. If time is critical for you, the best thing you can do is to create a new component in Qucs. I don't know how complex your transformer is. But I will have a look into the publication you sent. Is this the only component you need ? Which simulation types do you need to perform ? Regards, Michael 
From: <Michael.M<argraf@al...>  20050729 06:24:15

Hello, >>>> Concerning the point 2, I will try to elaborate: I need to simulate >>>> a coil L(I) where the value of >>>> the coil depends on the value of a current calculated in the >>>> circuit. How do I make it with the equation module? >>> >>> I just want to clarify my TODO list item. So I ask who listens: >>> >>> It is straight forward to implement a feature which enables qucsator >>> to >>> take a equation variable's content and fill it into the property of >>> some component if: >>> >>> 1) the variable is a constant (the result) >>> 2) the variable depends on a sweep variable >>> 3) the variable depends on another components property >>> >>> But problems occur (as in the above mentioned case) when >>> >>> 4) the variable depends on another components operating point >>> 5) the variable depends on a simulation result (e.g. node voltage >>> or branch current) >>> >>> Is it necessary for these two cases to simulate, apply property, >>> simulate, apply property, simulate, etc. ... until no (or little) >>> change in simulation results occurs? We need to repeat single >>> simulations until results converge? Or am I on the wrong track? >>> >>> Please help me clarifying! >>> >>> Thanks in advance, >>> Stefan. >> >> I would like to put in my opinion on that issue: >> I think the abovementioned method is not very usefull! > > But points 1) through 3) may be useful? Yes, indeed. But could lead to infinite recursion. >> What Patrick needs is a nonlinear inductor, i.e. an inductor >> with hysteresis. Or do you mean the inductance depends >> on a current somewhere else in the circuit ? >> Then you would need a symbolical defined multiport. >> The abovementioned method leads to too much hidden faults >> and it is not "simulationlike" as Stefan already tried to >> explain. >> So, to sum up: Qucs should have a nonlinear inductor and/or >> a symbolical defined device, right ??? > > I agree for points 4) and 5). Fine. Regards, Michael 
From: Stefan Jahn <stefan@gr...>  20050729 06:42:43

On Fr, 29.07.2005, 08:23, Michael.Margraf@... wrote: > Hello, Hi! >>>>> Concerning the point 2, I will try to elaborate: I need to simulate >>>>> a coil L(I) where the value of >>>>> the coil depends on the value of a current calculated in the >>>>> circuit. How do I make it with the equation module? >>>> >>>> I just want to clarify my TODO list item. So I ask who listens: >>>> >>>> It is straight forward to implement a feature which enables qucsator >>>> to >>>> take a equation variable's content and fill it into the property of >>>> some component if: >>>> >>>> 1) the variable is a constant (the result) >>>> 2) the variable depends on a sweep variable >>>> 3) the variable depends on another components property >>>> >>>> But problems occur (as in the above mentioned case) when >>>> >>>> 4) the variable depends on another components operating point >>>> 5) the variable depends on a simulation result (e.g. node voltage >>>> or branch current) >>>> >>>> Is it necessary for these two cases to simulate, apply property, >>>> simulate, apply property, simulate, etc. ... until no (or little) >>>> change in simulation results occurs? We need to repeat single >>>> simulations until results converge? Or am I on the wrong track? >>>> >>>> Please help me clarifying! >>>> >>>> Thanks in advance, >>>> Stefan. >>> >>> I would like to put in my opinion on that issue: >>> I think the abovementioned method is not very usefull! >> >> But points 1) through 3) may be useful? > > Yes, indeed. But could lead to infinite recursion. Can you give an example? In fact the equation solver finds out about recursions, e.g. a=1 b=2+a c=3d d=c is detected, and an appropriate error message thrown... Stefan. 
From: <Michael.M<argraf@al...>  20050729 06:48:46

Hello, >>>> I would like to put in my opinion on that issue: >>>> I think the abovementioned method is not very usefull! >>> >>> But points 1) through 3) may be useful? >> >> Yes, indeed. But could lead to infinite recursion. > > Can you give an example? > > In fact the equation solver finds out about recursions, e.g. > > a=1 > b=2+a > c=3d > d=c > > is detected, and an appropriate error message thrown... That's fine. Then it is alright!!! Bye, Michael 