Just mucking around with a LT1533 low noise smps simulation, and I realised how little I know (what's new?) about spice, and it's too damn windy to walk up the hill for a beer (spanish, yuck)
Each inductor has a value called Parallel Resistance, and I really don't have a clue how to give it a nominal value/guess
Any rules of thumb? Nominal L's are from 470u to 8mH
" The magnetic loss can be modelled reasonably well as a parallel resistor (Rp) across the existing model. The value can again be calculated from data sheet parameters, using the quality factor (Q). In a parallel RLC3 circuit the relationship between the quality factor and inductance is given by; Q = Rp/2piFoLo "
So...I suppose.. Since V^2/Rp = power then Hot core from losses ..low Rp Cool core from gentle use...high Rp
I think Rp generally drops in value with increasing B swing or increasing frequency. It's gonna be core material dependent too..
Come to think of it...I think Rp is fake in spice.. It's not really there.. I think it's just included to explain where the power is disappearing.
But I'm just learning and thinking about this now..
great stuff, LTspice is excellent, but when doing a transformer,there are too many thingummies to make sense of, unfortuantely LTspice on yahoo groups is a bit like snowboarding through mud
Won't this cause a problem in some applications? A series resistor will have a DC drop, whereas a parallel will not. If you know the Q, you can make the parallel resistance 2*PI*f*L*Q, or a series resistor of 2*PI*f*L/Q; or, split the R between series and parallel. For an air core, you can make the Rp=10E10. A stab in the dark guess for inductors wound with decent size wire might be Q=100 for the 470u and Q=30 for the 8mH.
I've had good results simulating my transformers in spice. Most of of time I design them to avoid core saturation, or even excessive core losses, if I can. I also like to see the important transformer parameters right on my circuit, so I use a perfect transformer with its turns ratio, add a parallel input inductor for the core magnetizing inductance Lm = AL N^2, where you get AL from the core's datasheet. Getting the right value for Lm usually isn't very important.
Then, very important, I add input and output resistors for the two copper winding resistances (these do a better job of dealing with the high peak currents from rectifiers driving storage caps, etc., than a general parallel loss parameter), an ohmmeter is handy for this.
Series leakage inductance is important. This can be measured (while shorting the other side), or it can be calculated (see posts by Tony Williams) or you can try Lell = 1% or 2% of Lm as a starting estimate. You only add the measured leakage inductance once, on either the primary or secondary side - its value includes both sides. If you have multiple secondaries, measure Lell between them and place half that on each one, then measure the primary and subtract the turns-ratio-squared- adjusted value from the secondary side you had shorted.
This overall scheme gives spice results that are pretty close to the bench results, unless you're at high frequencies and high ac voltages, where core loss is an issue. You can try estimating the power loss from core-material curves, and add a parallel resistor to implement it. But generally I'm less concerned with an accurate primary power consumption estimate than I am with knowing an accurate secondary waveform result.
Let me hasten to add, my statement was for forward converters, or for push-pull, etc, as used with the LT1533. OTOH, flyback converters, etc., depend rigorously upon the Lm value, with the core's air gap closely determining AL.
For someone who makes sweeping generalizations like RC-timers and discrete logic is for newbies, you sure don't know a heck of whole lot about electronics...
Winfield snipped-for-privacy@yahoo.com posted to sci.electronics.design:
If you do your measurements correctly, you can derive the parallel resistance like components of core losses this way. And by extension for any specific core with similar windings by scaling as an approximation.
Pleased tell us more. The core loss goes as something like f^1.8 and B^2.5, etc., so you'd have to test at the right frequency and amplitude, or be confidant of the scaling factors. Is there an easy trick?
Winfield Hill snipped-for-privacy@rowland.org posted to sci.electronics.design:
Not in the least, i specified the same core with similar windings. I clarify, not more than 50 percent difference in turns, peak and average ampere turns, and volt-seconds, and maximum MMF (assuming that the difference does not approach core saturation and fairly linear cores). The idea is that it is measurable and that the measurements can be used in SPICE simulations.
Almost sorry to have to back off so much, but it is useful for single frequency of operation designs.
I don't see how it would cause problems. Real inductors really do have series resistances, and they have an honest d.c. drop if there's a d.c. flow. Putting in a realistic series resistance, then, makes the simulation more realistic (and grounded more closely on actual physical parameters).
Real inductors do not have parallel resistances carrying any meaningful current. Rp, as used in Q and filter calculations, is a mathematical convenience to model certain losses. It *is* convenient, especially for low-level signals where the core isn't being pushed, but it's not a real resistor.
That's handy stuff for RF filter work. For my occasional switch-mode power supply work, cores are loss so variable with load and frequency-- and small to begin with, by careful choice of core and windings--that they aren't that critical. I usually just estimate, and don't bother simulating them. If I'm really interested, I measure (in real life).
I don't have any proximity-effect papers from that era, although I know a lot of work was done on inductance and it's subtleties at the NBS, as they developed the theory for and created various inductance standards, etc.,. and I have copies of some mid-30's stuff.
I wonder if you could scan that gentle beast for us?
It's about 300 pages long, and now quite fragile because it's printed on toilet paper and bound with staples like raffle tickets--from the front cover through the whole thickness to the back cover. They decently covered this offence with tape. I'll see if I can make a couple of readable digital photos of the curves. IIRC they talk in terms of frequency dependent resistance for various winding styles, but don't specify how it arises.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.