Simone Mannori-3 |
Dear Shriram,
I would apologize for the late answer (we are in a constant overload situation for the next Scilab 5.0 release). I would say a big "thanks" to Masoud for his indispensable and timely help (Masoud is the main contributor of Modelica integration in Scicos). Some notes about the diagram: - before start a simulation is better use reasonable values (close to the real one) for components parameter's. Avoid this rule at your risk when are playing with POWER circuits ! I choose L=0.1Henry and R=0.1Ohm. The relative time constant is L/R = 1s. Probably the real circuits is faster: you are more expert than me. - I left the MOSFET parameter's as default. It is not a very good choice because I'm breaking the rule #1. The simulation woks quite good, but for useful results you need to tune MOSFET parameter's in order to "clone" the performances of the real one. I would suggest a "Italian Made" STB150NF55 by ST . Masoud still have MY book on MOSFET modeling, so ask him for more information about MOSFET parameter's tuning ;). - I added a small value resistor (100Ohm) in series with the gate of the MOSFET. Drive a gate with an ideal voltage source is not a very good idea. The MOSFET equivalent input circuits is a capacitance and drive a capacitor with a ideal voltage source can create issues in the calculation of the input gate current : i = C d vg / dt . If "vg" is a voltage step the input current is infinite and you have a convergence problem. In the "real world" the input equivalent circuits is a non-linear, time varying capacitance with a parallel non linear, time varying, conductance. This conductance can be NEGATIVE. If you add the inevitable inductance of the wiring, you can obtain a nice 10MHz oscillator. Normally these oscillation destroy the real device. Usually the minimum gate resistor is in the order of 10 Ohms. An alternative, safer, solution is drive the gate with a "voltage clamped" current source. - Push the KP gain is not dangerous (in the simulation) because the system is - basically - first order... BUT look at the Gate voltage: more than 30 volt ! You will destroy the real device. I introduced a saturation to limit the gate voltage (look at the second attached file). In the real circuits a Zener diode will clamp the voltage. - Using linear drive you need to add a big heath sink to the MOSFET. Use a PWM circuits is a better solution. Masoud has some prepared some ready made examples inside "demos/Electrical" folder. Please don't hesitate to contact us for further help for your application: Laurent Vaylet is a mechanical engineer with some first-hands experiences in the automotive world. Best Regards Simone Mannori - Scilab/INRIA Rocquencourt P.S. Thanks to Sylvestre for his "postman" job ;). e_test_r1.cos (195K) Download Attachment e_test_r2.cos (200K) Download Attachment stp150nf55.pdf (747K) Download Attachment |
Shriram Sampat |
Hello Simone,
Thanks a bunch for your vey detailed reply and thanks a lot to Masoud for his suggestion to use the controlled voltage source. We have a similar working setup with heatsink and ventillators(as you guessed ;-) ), bcos we need to be driving around 150A over a considerable period of time. This setup and control for it, is totally with electronics. Now we want to increase the current driving capability(new MOSFET) and a higher control frequency (around 50KHZ PID loop), and high configurability via software. The initial scicos electrical model was just a simple experiment, whether I could do a simulation of this, to aid me write the control part in software and thanks to you guys i have the full set of tools to describe the system now. One problem i have as of now, is defining the parameters for the MOSFET (attached datasheet). It is a high current (200A) MOSFET. I have asked the manufacturer for the spice parameters for his device and so far have got no updates. Can any of you guys help me parameterise this for Scicos ? I am not much of a power electronics guy, as my expertise lies more in software and small electronics. I will go through your suggestions today and complete my simulation. I will also suggest it to my circuit design colleagues while we do the new design. Thanks a lot guys for your effort and time. I really appreciate all of your help. Ram Simone Mannori wrote: > Dear Shriram, > > I would apologize for the late answer (we are in a constant overload > situation for the next > Scilab 5.0 release). > > I would say a big "thanks" to Masoud for his indispensable and timely > help (Masoud is the > main contributor of Modelica integration in Scicos). > > Some notes about the diagram: > > - before start a simulation is better use reasonable values (close to > the real one) for components > parameter's. Avoid this rule at your risk when are playing with > POWER circuits ! > I choose L=0.1Henry and R=0.1Ohm. The relative time constant is > L/R = 1s. Probably the real circuits > is faster: you are more expert than me. > > - I left the MOSFET parameter's as default. It is not a very good > choice because I'm breaking the rule #1. > The simulation woks quite good, but for useful results you need to > tune MOSFET parameter's in order to > "clone" the performances of the real one. I would suggest a "Italian > Made" STB150NF55 by ST . > Masoud still have MY book on MOSFET modeling, so ask him for more > information about MOSFET > parameter's tuning ;). > > - I added a small value resistor (100Ohm) in series with the gate of > the MOSFET. Drive a gate with an ideal > voltage source is not a very good idea. The MOSFET equivalent input > circuits is a capacitance and drive > a capacitor with a ideal voltage source can create issues in the > calculation of the input gate current : > i = C d vg / dt . If "vg" is a voltage step the input current is > infinite and you have a convergence problem. > In the "real world" the input equivalent circuits is a non-linear, > time varying capacitance with a parallel non linear, > time varying, conductance. This conductance can be NEGATIVE. If > you add the inevitable inductance of the > wiring, you can obtain a nice 10MHz oscillator. Normally these > oscillation destroy the real device. > Usually the minimum gate resistor is in the order of 10 Ohms. An > alternative, safer, solution is drive the gate > with a "voltage clamped" current source. > > - Push the KP gain is not dangerous (in the simulation) because the > system is - basically - first order... BUT > look at the Gate voltage: more than 30 volt ! You will destroy the > real device. I introduced a saturation to > limit the gate voltage (look at the second attached file). In the > real circuits a Zener diode will clamp the voltage. > > - Using linear drive you need to add a big heath sink to the MOSFET. > Use a PWM circuits is a better solution. > Masoud has some prepared some ready made examples inside > "demos/Electrical" folder. > > Please don't hesitate to contact us for further help for your > application: Laurent Vaylet is a mechanical engineer with > some first-hands experiences in the automotive world. > > Best Regards > > Simone Mannori - Scilab/INRIA Rocquencourt > > > P.S. Thanks to Sylvestre for his "postman" job ;). > > ------------------------------------------------------------ Shriram R Sampat Software Engineer - Embedded Systems EK Tronic Ing. R. Nitz Tel.: + 49/(0)7154/8311-12 FAX : + 49/(0)7154/70449 email: [hidden email] www.ektronic.de ------------------------------------------------------------ EK Tronic Ing. R.Nitz Enzstr. 9 , 70806 Kornwestheim Ust ID: DE 146160631 Geschäftsführung: Ing. Richard Nitz semikron111AR.pdf (677K) Download Attachment |
Simone Mannori |
Shriram:
> One problem i have as of now, is defining the parameters for the MOSFET > (attached data sheet). It is a high current (200A) MOSFET. I have asked > the manufacturer for the spice parameters for his device and so far have > got no updates. Can any of you guys help me parameterize this for Scicos ? Masoud has already proposed me to write a Scilab/Scicos application capable to compute Spice parameters from data sheet. I have three conferences in the following weeks, so it is better is you ask Masoud directly. A good starting point is to replicate inside Scicos the test circuit used by the SEMIKRON to plot the curve and play with the Spice parameters in order to obtain a reasonable match. >From Feb. 15th I will find time also for this development. Simone |
Free forum by Nabble | Edit this page |