Just wanted to know if anyone has ever had any problems with gapping EE cores for flyback converters. I wouldn't know for sure, but I think distributors can sell cores with the gap you need, and they put this gap on the center leg.

Well, I have also heard that it is possible to use paper spacers between the core halves. I have used paper spacers before for low peak primary current levels with success. But, the other day, I had some serious issues, which I believe were caused by using paper spacers. I am curious if anyone else has run into this problem.

I was building another DCM flyback, this time with a peak current of about 2 A. I was using an EE core with paper spacers placed on the outside legs, and a copper ribbon wrapped around the outside of the transformer to deal with the stray field from the outside legs. Well, in a few seconds, my FET was toast, literally. It went up in flames and the Source leg detached from the package every time. So, my suspicion was the stray field. My suspicion was confirmed after the outer winding of my transformer melted.

The problem immediately went away when I replaced the gapped core with an MPP toroidal core (wound with triple insulated wire). The converter worked as planned with the toroidal core.

So again, does anyone have similar difficulties with gapping flyback cores? From what I understand, the gap on the middle leg is shielded by the windings on the bobbin. But what is the best way to deal with the fringe field if the outer legs are also gapped? Does anyone know of any good tools to grind a gap on a ferrite core? Last weekend, I used sandpaper and water and it took me a good 3 hours to grind away 0.6mm. Even then, the gap wasn't even so I scrapped it anyways.

It is sad that I have that much time on my hands. But hey, right now I am in between ships and on vacation. There is nothing like trying to gap a ferrite core on vacation! But seriously, I am looking into some kind of tool that can grind an even gap into ferrite that does not require so much time and thumb muscle. Anyways, please let me know if you have had EMI failures from your gapped flyback core, and if you have ever tried MPP toroids.

When gapping the side legs, you have to make sure there are no conductors near the gap. Once you put the copper ribbon around, the core, you put copper right next to the HF fields and you can get a lot of heating and an unhappy inductor.

I have used sandpaper succesfully, but messy as you have observed.

The problem with MPP is the core loss is much higher than ferrite, and the toroid only works well for equal or simiary turns counts. Otherwise it is awkward to wind and hard to maintain repeatability.

With only 0.6 mm gap, your fringing fields shouldn't be that excessive I wouldn't think.

So, are saying that using the copper ribbon might be causing the problem? I did notice it got quite hot (actually, very hot)...In fact, so hot that it melted the secondary winding. I would think .6mm gap wouldn't cause too much fuss as you mentioned.

Yes, winding that toroid can be a huge pain. The good thing though is I've noticed is that leakage inductance is quite a bit lower with a toroid (about 1/3 the leakage of an EE core). This would be nice for that RCD clamp...But like you say, the core loss might not be acceptable. It would all depend I guess. Anyways, thanks for the comment.

You would need to calculate the FRINGING....I keep the fringing no greater than 10%....
When the fringing is significant, you have an increase in inductance by the same factor, thus an increase in Flux Density... so re-iteration on the calculations... To keep fringing low increase the number of gaps in the magnetic path length.. Recently I had to use 6 gaps on each leg of an E core to keep the fringing just under 10%...this was to prevent the chill plate in the core from getting heated up from the fringing H...
I once had to put several gaps in a toroid made from $$ Supermendur $$ ....the toroid was cut twice, thus 4 gaps...then re-assembled with Wakefield 152 and thin outer band made from stainless...
Provide requirements/spec of the EE core ...

Use a powder core. Nice distributed air-gap and no nasty fringing problems

Calculate fringing...Hmm...this is a new concept for me. I would think fringing would be linked to leakage inductance as you alluded to above (increased fringing equals increased leakage inductance...is that what you are saying?). Did you use a copper band with that E core you had 6 gaps on?

Ok the EE core is used in a DCM flyback. Here are the EE core specs and requirements:

Relevant converter specs:

1. Converter power: 50W
2. Peak primary current: 2.6A
3. Primary # of turns: 30
4. Secondary # of turns: 30
5. Required primary inductance: 150 uH
6. Switching freq: 100kHz

Core specs:

1. E30/15/7-3C90
2. Core material: 3C90 (Bsat = 380mT @ 100C)
3. Effective volume: 4000 cubic mm
4.Effective magnetic path length (le): 67mm
5. Effective area: 60 sqmm
6. Mass of core half: 11g
7. Required gap length on center leg: 0.236 mm (ok I got the actual gap length wrong earlier). If my math is correct, if all three legs are gapped to the same length with spacers, then the actual gap lengths would be 2/3 the above gap length. I usually just wind the bobbin and then add spacers to the outside legs until my inductance is correct, rather than measure the actual gap.

Hey thanks for the tips and pointers!

Cost is the driver for using E & I cored gapped ferrite as this enables use of bobbin windings. Powder cores are normally toroids and therefore expensive to wind.
Good for aerospace as they are very stable with temperature. Until Ferroxcube came out with 3C95 material ferrite really fell over at -40C
DC flux does not cause core loss only the ac component, therefore in CCM mode the powder core could be favourable with lower ac loss in the windings and lower leakage inductance due to the better winding geometry of a toroid

Maybe you were saturating core, MOSFET failed short, current melted wire and eventually MOSFET leg? The powered iron core has a softer saturation.

As for the fringing field with all three legs gapped - the H-field is concentrated in the gap so any copper near the high H-field will have higher current induced leading to localized heating. You should try to keep some distance from a gap, a quick way to see this is put your inductor on an LRC meter to measure Q and see the Q drop when you bring the copper near the gap. For outer flux bands keep in mind they carry current to buck the fields so a thin flux band can also over heat. You need to make a flux band with thicker copper, not the 1oz copper tape, if you have high AC current ripple and a gap in the outer legs.

Good luck!

Thanks for the tips Andrew. Maybe you are right about the core saturating. It could be that I missed something there. I guess that would explain the Source leg falling off. The flux band I used was pretty thin. I will get some thicker copper and see what happens. Anyways, thanks for the post!

Jay, have you measured the peak primary current? Is it actually 2.6A? According to my calculations ,with the core you used and 30 turns ,at 150 uH your design can handle 4.0 A with only a 10% drop in inductance. This transformer design should not be saturating unless the peak primary current is actually higher than you think and higher than 4A. What is your MOSFET rated for?

I did not have the time to measure anything...I blew chunks immediately. I was running the converter open loop at a reduced bus voltage at a fixed duty of about 40% and a DC bus voltage of only 60 V. Peak primary current should have been about only 1.6A. The design input voltage is 100V (hence the 2.6 or 2.67 A).

The MOSFETs I used were actually some pretty rugged SMPS rated IGBTs (yes, much more muscle than needed, but it was what I had at the moment). I used the now obselete HGTG7n60Aa4D 600V 34A SMPS IGBT.

I didn't think that it should saturate either. I did notice the copper band was super hot, and that the secondary completely melted (pretty impressive actually). The primary winding (under the secondary) was fine. That led me to believe the stray fields from the outer gaps melted my secondary. Maybe my copper band was too thin? Maybe the copper band got so hot that the secondary melted, which then caused the primary to saturate, destroying the FET.


Im still investigating this. I do know that once I replaced the EE core with the MPP toroid, the converter worked beautifully (nice VDS DCM waveforms).

I'll post an update on my findings. Thanks for the post and taking the time to look at my converter stuff.

When I say open loop, the output filter capacitor was already charged before power was applied, so the converter never entered CCM (at least I am 99.99% sure). What I was doing was powering the converter from a 60V battery bank, and then discharging back to the same battery bank. So, over-current at start-up should not have happened, so that could not have been the problem either.

Jay,
Like Nelson, I design and build transformers for a living. I did some calculations based on the information you provided.

1. Peak flux density is only about 2200 gauss, no saturation issue for the 3C90 material.
2. The gap length you use for this core is not enough to create any huge issues with fringing flux.
3. A torched secondary can only be due to high secondary current. Any chance you had your secondary wired backward? The ouput diode?

For a hobbyist, to gap a center leg in a ferrite it is easiest to use a small diameter diamond coated drum, maybe 1/8 or 1/4" in diameter, rotated slowly in an electric screwdriver. Don't rush it or you will send chunks flying. A cheaper alternative is a strip of drywall sanding screen taped or glued to a popsicle stick.

The problem is solved and the converter now runs successfully using a gapped ferrite with the gap on the center leg only, and no flux band around the outside.

So now the converter operates well with either an MPP toroid, or a gapped ferrite (center leg, no flux band). I noticed that the transformer is quite hotter with the MPP (30F temp difference between the two). Dr. Ridley mentioned that MPP has pretty high core loss, which is probably why it gets hotter than the ferrite core. I also have 46 turns on the MPP, as opposed to the 30 on the ferrite (another reason why its hotter).

I could just leave it there...But I think my next step is to run the thing with paper spacers for gapping and no flux band and see what happens. I'd like to get to the absolute bottom of this.

Thank you all for the comments, tips, and advice! I really appreciate your helpful suggestions.