Subject: Re: standard packages for 1076.1
From: siep.onneweer@philips.com
Date: Mon Oct 01 2001 - 11:41:25 PDT
LS,
I would like to give some brief counteraguments to Peter Wilsons point of
view. However, I do not consider myself an authority in the field and will leave
it up to the working group to carry the discussions further and draw its conclusions.
>
>
> Hi,
>
> Question for Discussion : Electrical Systems (Magnetic Domain) Through
> variable Flux or Flux Rate ?
>
> This is an issue that occasionally pops up (every 10 years or so) in
> modeling
> magnetic systems. There is certainly a case to made for using flux rate
> on the standpoint of direct transfer between electrical & magnetic
> domains and hence the direct duality of "power". So to understand the
> difference let's explain the basic terms, then I will outline the issues
> for discussion as I see it.
>
> Case 1 : FLUX RATE
> Across : MMF & Through : dFLUX/dt
>
> Case 2 : FLUX
> Across : MMF & Through : FLUX
>
> Dualities :
> Electrical => Magnetic
> Current & MMF
>
> Magnetic=>Electrical
> Derivative of Flux & Voltage
>
> Instantaneous "Power":
> Electrical VI
> Magnetic MMF*dFLUX/dt
>
Very clear.
> I have several points to make at this stage.
>
> The first point to note is that on the duality boundary in both cases
> there is a direct scaling between current and
> MMF. In the duality of magnetic->electrical however in the flux rate
> case scaling, but in the flux case the derivative is taken. So the
> advantage of using flux rate is that there is no derivative required.
> BUT, and this is a BIG BUT for magnetic modeling, using the rate of flux
> precludes the use of initial flux conditions. This is crucial for
> initializing the state of magnetic components and is therefore a
> significant drawback in using rate of flux as the through variable. In
> my own case, this is a potential showstopper. Ah, the response will go,
> initialize H or MMF instead, but with hysteresis in magnetics, there are
> an infinite number of states of flux for a single value of H, and so it
> is not trivial or even possible, to use this approach practically in a
> general purpose way. Applications that may have problems include
> permanent magnets (magnetic recoil & hysteresis + initial conditions),
> Magnetized materials - i.e. ran for several cycles and some remanence
> remains, PM Machines, Power Transformers....etc.
In the example 'coreja' -with hysteresis- that was included with my email to
Dr. Mantooth I have made both M and H state variables. H is always
proportional to the MMF, and M can be initialized to an arbitrary value
within the outer hysteresis loop. I believe this should solve the issue.
The real difference is that the state variable (in this case M instead of flux)
is an internal quantity of the magnetic model and not an interface quantity.
The flux density and flux can be calculated from M and H by B = MU0*(M+H)
and PHI = B*AREA.
I can see that initializing a structural model of a magnetic circuit,
e.g. a model of a transformer core with several flux paths containing
magnetic material, and also an airgap and leakage flux paths, is a
problem. A practical approach is to create a parasitic flux path that
takes care of the DC solution. One should be careful that it does not
influence the dynamic behaviour.
Admittedly this is not a very elegant approach. The problem is analogous
to a number of electrical capacitors connected in series. Here too the
DC solution is not well defined. The described solution is analogous to
adding high-valued resistors in parallel to the capacitors.
Although I recognize the difficulty of initializing models built
this way, in my opinion it is not a showstopper.
>
> The second point I would make at this stage is to move to the
> fundamental units [1] . As far as I can see, the only (main?) reason to
> make this change is for convenience in a specific instance/application.
> Does this justify not using the fundamental physical quantities
> generally? I would argue not. In the case specifically cited, of power,
> it is trivial to create the derivative of flux, so what is the big deal
> ? Does that really necessitate the change of the fundamental physical
> system in general ? I think not. This is epecially the case when counter
> examples are given (as done previously).
I do not agree that this is just a matter of convenience. You can
compare magnetic flux to its dual the electrical charge. Nobody would
argue that charge would be a better interface quantity on the electrical
side than current. Flux or magnetization serve to model energy storage
in the magnetic component. If flux is an interface quantity, the
magnetic model is not reactive any more and the energy storage is
modeled in the interface between the elecrical and magnetic domains!
This is entirely unphysical, and is the fundamental reason I oppose to
this approach.
>
> The third point I would make is that there are differences between the
> types of models generally used in the two domains that dissipate power.
> Taking a resistor component - the power is simply VI or I^2R. Notice
> that above I used the term "power" in inverted commas because in
> magnetic cores that exhibit hysteresis there is the real power
> dissipated and the apparent power dissipation (analagous to VI and Power
> in an electrical system - not the same). The real power dissipated in a
> core with hysteresis is in the area of the BH loop. To just use the
> product of MMF and dFLUX/dt will only be the actual real power at the
> end of a cycle. To properly obtain the real power requires the
> distinction in the magnetic domain of real and reactive power.
>
I was not referring to the power dissipated in the magnetic components
but to the instantaneous power transferred between them which forms the
basis for modeling dynamic physical systems.
Therefore I agree completely with this point. In addition for high
frequency operation such as in switched mode power supplies the
dissipated power in the inductor/transformer core is mainly due to
relaxation and eddy current losses and not to static hysteresis. Also
the losses in the windings are frequency dependent. It turns out that
these phenomena can be modeled very elegantly with the choice of
variables that I propose.
> So summarise therefore, I believe that FLUX should be the through
> variable and the FLUX RATE and POWER remain as secondary derived
> variables.
I certainly appreciate this point of view but I remain convinced of the
opposite. The fundamental approach to physical modeling of dynamic
systems, based on the exchange of power between idealized lumped circuit
elements, is applicable in many areas of physics. There is no reason to
make an exception for magnetics.
>
> To provide some background to help those not familiar with the general
> area, I
> have listed some useful papers/material as follows
>
> [1] : Fundamentals !
> [2]-[5] : Duality and Inter-domain Modeling
> [6]-[15] : Magnetic Modeling and Power Loss
>
> References:
> [1] J.C.MAXWELL, "A Treatise on Electricity and Magnetism", Clarendon
> Press
> Oxford, 1873, 3rd Edition.
> [2] E. C. CHERRY, "The duality between inter-linked electric and
> magnetic
> circuits", Proceedings of the
> Physics Society, Volume 62, 1949 pp101-111
> [3] E. R. LAITHWAITE, "Magnetic Equivalent Circuits for electrical
> machines",
> PROC. IEE, Vol. 114,
> November 1967, 1805-1809
> [4] C.J. CARPENTER, "Magnetic Equivalent Circuits", PROC. IEE.,Vol
> 115,No 10,
> October 1968,
> 1503-1511
> [5] AD BROWN, JN ROSS, KG NICHOLS AND MD PENNY, "Simulation of
> Magneto-Electronic
> Systems using Kirchoffian Networks", European Conference on Magnetic
> Sensors
> and Actuators, Sheffield,
> July 1998
> [6] W. ROSHEN, Ferrite Core Loss for Power Magnetic Component Design,
> IEEE
> Transactions on
> Magnetics, Vol. 27, No 6, Nov 1991, pp4407-4415
> [7] J.G ZHU, S.Y.R. HUI & V.S. RAMSDEN, A Generalized dynamic circuit
> model of
> magnetic cores for
> low- and high-frequency applications-Part I: Theoretical Calculation of
> the
> equivalent core loss resistance, IEEE
> Transactions on Power Electronics, Vol. 11, No 2, 1996, pp246-250
> [8] J.G ZHU, S.Y.R. HUI & V.S. RAMSDEN, A Generalized dynamic circuit
> model of
> magnetic cores for
> low- and high-frequency applications-Part 2: Circuit Model Formulation
> and
> Implementation, IEEE
> Transactions on Power Electronics, Vol. 11, No 2, 1996, pp251-259
> [9] S MULDER, Power Ferrite Loss Formulas for transformer design, PCIM
> Magazine, July 1995, pp22-31
> [10] P. R. WILSON, J. NEIL ROSS & ANDREW D. BROWN, "Dynamic
> Electric-Magnetic-Thermal
> Simulation of Magnetic Components", IEEE Conference on Computers in
> Power
> Electronics (COMPEL
> 2000), July 2000
> [11] C.P. STEINMETZ, "On the Law of Hysteresis", American Institute of
> Electrical Engineers Transactions,
> Vol. 9, 1892, pp3-64
> [12] P. TENANT, J.J. ROUSSEAU, L. ZEGADI, "Hysteresis modeling taking
> into
> account the temperature",
> European Power Electronics Conference proceedings, 1995, Volume 1,
> pp1.001-1.006.
> [13] MAXIM A., ANDREU D., BOUCHER J., "A Novel Behavioural method of
> SPICE
> Macromodeling of
> Magnetic Components Including the Temperature and Frequency
> Dependencies",
> Conference Proceedings -
> IEEE Applied Power Electronics Conference and Exposition - APEC, Vol. 1,
> 1998,
> pp 393-399
> [14] ODENDAAL W.G., FERREIRA J.A., "A thermal model for high frequency
> magnetic components", IEEE
> IAS Meeting, New Orleans, La., October 1997, pp1115-1122
> [15] MAXIM A., ANDREU D., BOUCHER J., "A New Behavioural Macromodeling
> method
> of Magnetic
> Components Including the Self-Heating Process", PESC '99, pp735-740
>
> Peter
>
> --
> -- Peter R. Wilson
> --
> -- mailto: prw99r@ecs.soton.ac.uk
> -- Homepage: http://www.ecs.soton.ac.uk/~prw99r
> -- Tel/Fax : +44 (0)23 8059 6665/2901
> --
> -- Room 53/3053, Dept of Electronics & Computer Science,
> -- University of Southampton, Highfield, Southampton,
> -- SO17 1BJ,United Kingdom
> --
Best Regards,
Siep Onneweer
Philips Semiconductors, DTG
Phone: +1 408-991-2562
Fax: +1 408-991-3355
Notes: Siep Onneweer/SVL/SC/PHILIPS
mailto: Siep.Onneweer@philips.com
This archive was generated by hypermail 2b28 : Mon Oct 01 2001 - 12:04:49 PDT