SPICE - leakage inductance

Thread Starter

Ian0

Joined Aug 7, 2020
1,098
Could someone please confirm what is the correct formula for K.
I've seen:
K=1-(Lleak/Lmag)
K=1-(Lleak/Lmag)^2
and
K= SQR(1-(Lleak/Lmag))
the last one seems most accurate.

Is there any way of modelling the core loss (except the obvious one of putting a resistor across the primary which causes the real part of the magnetisation current to flow)?
 

Thread Starter

Ian0

Joined Aug 7, 2020
1,098
I've watched the university lecture, and come to the conclusion that it is best to set K to 1 and add series R and L.
 

fourtytwo

Joined May 2, 2017
80
I've watched the university lecture, and come to the conclusion that it is best to set K to 1 and add series R and L.
I use series L for leakage as for me that is the most important effect to model (snubbers etc) and doing it that way makes the schematic and values very clear. I dont bother to much with core losses in LTspice, instead using the transformer design spreadsheet to tot them up along with everything else to estimate efficiency. Hope that helps :)
 

Thread Starter

Ian0

Joined Aug 7, 2020
1,098
Actually, I'm modelling a 6kVA transformer, half-controlled bridge rectifier and inductor. What interests me is how much switching noise finds its way back on to the supply, which is from a generator.
 

fourtytwo

Joined May 2, 2017
80
Actually, I'm modelling a 6kVA transformer, half-controlled bridge rectifier and inductor. What interests me is how much switching noise finds its way back on to the supply, which is from a generator.
Ripple I can do but noise I find such a dark art! Very dependent upon component parasitics, pcb layout, wiring looms etc and very difficult to model in my experience. In my case using MOSFETS & IGBT's most noise is switching and very dependent on edge speed, sometimes I reduce speed simply to reduce noise despite the increased power dissipation. You said controlled rectifier so perhaps you are using thyristors at 50/100hZ plenty of opportunity for noise, I am guessing the inductor is in the load circuit and fairly big hence it will have plenty of capacitance and so some EMC may be generated at thyristor turn on, maybe a few turns around a ferrite in the thyristor anodes would help, TBH I nearly always try to fix EMC in the prototype with it running, apart from trying to put some basics into the circuit design (usually common mode chokes etc). As I say dark art IMOP.

Maybe in your case the generator won't bother with a bit of EMC but your neighbors might, perhaps the generator will be more bothered by power factor and nasty harmonics that I would classify as ripple rather then EMC.
 

Thread Starter

Ian0

Joined Aug 7, 2020
1,098
"switching noise" might not have been the right choice of phrase. It's not really EMC, nor is it ripple: It's the way the DC choke interacts the the leakage inductance and generator source inductance, at the point of thyristor switch-on. There is a short but severe dip on the transformer secondary voltage, which I thought was diode reverse recovery, but isn't. It is caused by the DC choke forcing its current to remain unchanged, but the thyristor current having to increase from zero to the DC current via the transformer leakage inductance.
The voltage dip then upsets the zero-crossing detector from which the thyristor trigger point is derived.
Anyway, thanks to advice received, my SPICE model agrees exactly with what I observed on the scope!
 

Thread Starter

Ian0

Joined Aug 7, 2020
1,098
It was just about modelling the transformer. (But I appreciate the offer).
The plan is to try a split-bobbin transformer which increases the leakage inductance from 800uH to 6mH. It will keep the voltage dip on the secondary side where the generator can't see it!
 

fourtytwo

Joined May 2, 2017
80
Sorry it was a stupid sort of reply I was just intrigued but I know many don't always want to disclose all they are doing :)
I am kinda thinking the "output (may not be)" inductor would appear in series with any leakage inductance & hence limit the instantaneous current BUT because I dont really know much about your application other than what I have surmised that's about as far as I can go. Please excuse my natural inquisitiveness :)
 

Thread Starter

Ian0

Joined Aug 7, 2020
1,098
It's a good old-fashioned lead acid battery charger with half-controlled bridge and output inductor.
Spot on about the leakage inductance limiting the instantaneous current.
A well-coupled transformer couples the voltage dip back to the primary, which the generator doesn't like.
I'm thinking that a split-bobbin transformer will give a bigger dip on the secondary side, and nothing much on the primary, so the generator will be happier.
 

fourtytwo

Joined May 2, 2017
80
Well you lost me a bit as in why does the output inductor perform the same function but then I cannot see the circuit diagram.
Have you experienced trouble with the generator or are you surmising there may be ?
What sort of generator is it, an inverter type or traditional 50hZ alternator ?
 

Thread Starter

Ian0

Joined Aug 7, 2020
1,098
A Perkins 2 litre diesel engine + 15kVA alternator. Problems with stability started to occur when the generator manufacturer changed alternators from Stamford to MeccAlte.
 

fourtytwo

Joined May 2, 2017
80
I don't know why I am thinking this but DC BIAS!! Is there something wrong in one half of the power or control circuit such that a DCbias is being placed on the secondary ?
This would distort the primary current waveform possibly enough to cause a similar bias in the Alternator, not good!
Of course if you ran it on the regular street supply all would seem well.
Do you have a scope for checking all this out or are you having to do it blind so to speak ?
I used to run a diesel powered lighting rig alongside sound re-enforcement a long time ago and DC bias in the lighting controllers could cause nasty problems, inexplicable heating mostly. Different Alternators behave differently in this situation.
 

Thread Starter

Ian0

Joined Aug 7, 2020
1,098
Christmas over, so I've tidied up the simulations.
On the graph, top is primary, green is the (real) concentric transformer, blue is the (theoretical) split bobbin.
The green transformer specs are measured, the blue is assumed to have the same Lmag, Rpri and Rsec; its leakage inductance is extrapolated from similar transformers.
The value for Zgenerator is a guess.
The simulated thyristors and diodes are rather too small but SPICE diodes don't fail in the same way real ones do. The real thyristors/diode is rated at 150A.
The green waveforms exactly match what I see on the scope. Hypothetically the split bobbin transformer will fix the problem, by eliminating the pulse on the primary that seems to be upsetting the generator. Last day before Christmas holidays the transformer manufacturer was telling me that two-section bobbins were not available in that size. Probably because he was thinking more about finishing for the year at lunchtime and going to the pub rather than looking in a catalogue other than Miles Platts.
 

Attachments

Top