It took an age just to find a simple equation that gave me operating flux vs number of turns given all the core data. I eventually found out about the key 'volt seconds' parameter.
Provided you're not screwing the core for the tiniest 'optimum losses' the design then becomes pretty trivial, especially with the aid of something like Epcos's Ferrite Magnetic Designer program to point you in the right direction.
At the end of the day it seems you just have to accept the non-linearity and suck it and see !
Warms the cockles of me heart to see all those huggins of transistors and opamps and resistors and stuff :-).
Never having touched one, I had same trouble finding any 'key' SMPS transformer design info. (thought that's why SMPS experts offer shelves full of expensive SMPS text books?). Having to wade through reams of maths that with hindsight revealed itself as trivialities. Seems the authors are ashamed that core materials are best just described as naff and non-linear and hence make great effort to describe in painful detail any other aspect they have control over. 'Never mind the quality, feel the width'. Time was pressing, so just said bollocks to it, started out with V=L.di/dt and Mr Joule and had a 200W PSU running couple of days later. Sure, there's interesting side issues but that's just electronics. It's known as Black art only because the maths guys aren't comfortable with non-linear systems. regards john
I'm pretty confident they've done that and I plan to as well.
From inspection I expect their transformer to have been wound for low leakage inductance. There's no half width bobbins for sure.
Ah yes. I've read about that technique being used to avoid a separate inductor.
The truth is, at this stage, I'm unsure how the leakage inductance is exactly going to influence the design. I've assumed that I want good tight coupling but maybe it's not that simple.
Yup, I have diodes there too. 1N4148 fits the above requirement ( not 200 V though ) . I'd wondered about the switching time and considered shottkies. I have some
I once replaced the FET in a gatedrive smps that ran from 15V, supplying perhaps 0.3W (actually a similar topology to Grahams - forward mode transformer, no output inductor, pulse-by-pulse peak primary current limit, resonant reset). The "designer" used a 3A 1000V FET to do the job, drastically and detrimentally affecting circuit operation.
A couple other things to note:
- fet Rg is spec'd for the SPW20N60C3 at 0.54 Ohms
- IR2113 output resistance could be as high as 7.5 Ohms - 2A into a short with +15V supplies. Measure it, its sure as hell more than 1-2 Ohms. Ultimately the reverse diode cant do any better than Rdriver + Rfet. Unless you make the "diode" the B-E junction of a common-collector PNP transistor.
IME the granularity caused by available core sizes (I have never designed a custom core, but went to one of Ed Blooms fascinating planar magnetics courses where he did just that - custom pot core design for max power handling capability) greatly restricts the number of design choices.
If you can come up with a curve-fit expression for core loss (3 data points required on a core loss graph, two at fixed frequency, two at fixed flux density, or better yet read the philips app note) then the calculations become real easy as you can directly calculate losses.
I have a mathcad worksheet with arrays of parameters (Ae, le, Ve, m, MLT) for every core in the MA02 databook. I sometimes do designs with
*every* core, and pick the best result. other times I fiddle with a couple of possible cores. I very rarely use McLymans Kg method, simply because core availability is a prime concern. Many designs start with a call to my magnetics vendor - what have you got thats cheap....
for a forward-mode transformer, Lmag (and its associated non-linearity) are parasitics and can safely be ignored during 1st-order calculations. Just ensure that Bmax < Bsat at Tmax, and that you are happy with the core loss. Vin & Vout (and perhaps D) set the turns ratio, then B is chosen to suit, fixing Np and hence Ns. More B = more Pfe & fewer turns = less Pcu & Lleak & Lmag = more Imag.
For a flyback-mode transformer, there is an air gap, and the core gets a
*whole* lot more linear. for gaps above about 0.2mm you can actually ignore the core, and pretend all the energy is stored in the gap, so L = u0*Np^2*Ae/lgap
just think about the size repetitive thump you want to give the diode, and whether or not it will like it very much. Unless Vth is really low (its not) you dont care about schottky vs PN, just the series impedance of the diode - IOW how much current it can pass. A quick squiz at a PMLL4148 data sheet shows its around 2 Ohms or so, and Ifsm = 4A at 1us, so its probably OK thermally, but Rd is not that low.
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.