pot core

I designed a laser controller in 2002, and it's licensed to a contract manufacturer. Parts are headed for end-of-life. Old Xilinx FPGA, 68K CPU, things like that.

It has a custom transformer in the isolated dc/dc converter.

formatting link

formatting link

The guy who designed it, long gone, was a former Signal Transformer employee and he went wild: litz wire, bifalar, the works. We can't get the header any more so I have to redesign it.

I hacked one onto the board. It uses #30 solid Beldsol (thermal strip) wire, which seems to be OK. No bifalar or any fancy winding tricks. At about 18 watts out, overall efficiency is 85%.

formatting link

formatting link

formatting link

In theory, one should fill the available window area with copper, but this works about half full, with the #30 that I have handy. I suppose I could let the CM build it just like as shown, but maybe I'll make a small PC board with some sort of pins, to replace the surface-mount header.

What's neat about a pot core is that you can peek in through the gaps and see how hot the wire is running.

formatting link

formatting link

I drilled the board and added a 4-40 screw to hold everything together. Luckily the only thing I hit was ground plane. It doesn't look like the mag field induces significant current/loss in the stainless screw. Most of the field must be confined to the ferrite.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin
Loading thread data ...

Most pot cores use a nylon screw there or a stainless strap around them.

Cheers

Reply to
Martin Riddle

I wound the primary, tested that, then wound one secondary and tested that. Worked OK, so I decided to add the final secondary. I took off the nut and the top core half but forgot to turn the power off. It melted the bobbin into a blob, and I had to start over. The ferrite seems to matter.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

On Sat, 07 May 2016 17:23:46 -0700, John Larkin Gave us:

formatting link

You should still spin up some quads of mag wire to approach his winding methodology. It isn't that hard and your efficiency should jump up a few points. What frequency are you running at?

Reply to
DecadentLinuxUserNumeroUno

On Sat, 07 May 2016 17:23:46 -0700, John Larkin Gave us:

You can use a nice epoxy to hold it together. "Stycast" is a bit brittle, but others have enough elasticity to be good choices.

He may also have used a gapped core or flush faced cores with a gap during build time. Transformer tape can be applied to the face of one to get a 1 mil gap or the like. Zero gap usually results in crossover point losses.

Reply to
DecadentLinuxUserNumeroUno

On Sat, 07 May 2016 20:44:15 -0400, Martin Riddle Gave us:

Doesn't matter. The transformation field is all contained within the core. His 4-40 is likely stainless and not all that magnetic either. There are also brass screws to use. A nylon screw for larger pot cores like that would never survive any kind of shake test regimen. Bad choice except for very small core sets.

Reply to
DecadentLinuxUserNumeroUno

910Dsh17.pdf

R_5.JPG

It's bifilar winding - which means two strands. You twist the two wires bef ore you wind them onto the former, so they both see exactly the same magnet ic field (to one part in a billion if you do it right).

06217.JPG
06213.JPG
06220.JPG

So you could use thicker wire, giving a lower winding resistance. Newark st ocks 28 AWG

formatting link

and you could buy a single reel for $33.18

06202.JPG
0039.jpg

The classic mistake is to put a shielding can around the core and fix it wi th central screw, creating a shorted turn.

If you replace the pot core pair with an RM core pair, you can use two spri ng clips to hold the two core halves together around the former and winding , and the spring clips tend to be a available with a projecting pin which y ou can solder into mounting holes in the board.

With an ungapped core, about 99.9% of the flux is confined to the ferrite, whose permeability tend to be a thousand or so time higher than air.

--
Bill Sloman, Sydney
Reply to
bill.sloman

It raises the inductance of the winding by a factor of 1000 or so, if the core is ungapped. That tends to be a perceptible difference.

--
Bill Sloman, Sydney
Reply to
bill.sloman

** No need for a air gap with the push-pull topology that has been used.

There is a nasty trap with using gapped pot cores, the gap is in the centre leg and when you tighten the mounting bolt - guess what happens ?

..... Phil

Reply to
Phil Allison

On Sat, 7 May 2016 18:17:03 -0700 (PDT), snipped-for-privacy@ieee.org Gave us:

Not really. It is multiple parallel lines meant to increase current capacity while keeping the ease in winding the thing at a reasonable level. A solid wire of the same circular mil area would be far harder to work with, and would exhibit work hardening issues and be subject to fracture point creation, and would have a slightly lower efficiency due to skin effect, depending on the fo. So it also provides "more skin in the game".

A poor man's litz configuration, if you will.

Reply to
DecadentLinuxUserNumeroUno

On Sat, 7 May 2016 18:35:45 -0700 (PDT), Phil Allison Gave us:

No shit. Several, in fact. It is better to make your own gap when needed.

That is what a proper thickness fiber washer is for.

We used flush polished cores and placed a single layer of 1 mil transformer tape on one face or the 1.5 mil variety. The proper tuning of such a circuit is almost an art, and it must be done such that it is repeatable across multiple transformer winders, and even contract makers' results.

The zero crossing point causes problems and efficiency hits without it. People cry that the transformer is lossy or whatever, but a PWM fired circuit will need a slight gap.

Reply to
DecadentLinuxUserNumeroUno

What is the crossover point?

Reply to
John S

300 KHz. I can't see where bifalar winding would help. Only 1/2 of the primary is driven at a time, so leakage inductance doesn't matter.

Litz might help a tiny bit but isn't worth the trouble.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

Stick with Nylon screws, their much more fun to watch melt.

Cheers

Reply to
Martin Riddle

I did super-glue the bottom core to the board, to make it easier to wind. I figure the bolt is enough to hold things together, maybe with a fancy lock nut.

I'm using exactly the same core, ungapped. For this topology, the more inductance the better.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

On Sat, 7 May 2016 21:43:18 -0500, John S Gave us:

The two points at which the transistors (or FETs) turn on and off. They can (and do) fight each other. The gap spreads that point out and circuit efficiency goes up. No gap and they bang into (fight) each other. It comes down to slew rate. The best place to mitigate it is right there between the two core halves.

Reply to
DecadentLinuxUserNumeroUno

On Sat, 07 May 2016 20:01:47 -0700, John Larkin Gave us:

He didn't use any Litz. If he claimed to, he meant of the hand configured variety as that which I described. None of those are "bifilar wound" either.

He simply placed several strands together in parallel and twisted them up to make a higher current capacity line with a flexible, easily windable nature. It simply has the added benefit of providing a bit more skin for the same circular mil area. And yes, it does work slightly more efficiently at 300kHz than a single solid strand of equivalent ampacity would.

I have wound hundreds of transformers and trained assembly personnel in the methods as well, and have spent many hours at the bench proving various functional aspects of them, some of which were related here.

Reply to
DecadentLinuxUserNumeroUno

Sure, the screw doesn't get any flux, unless the core is gapped. Even still, it should be minimal. That's what the ferrite is FOR, of course, to concentrate the field.

Jon

Reply to
Jon Elson

That UCC3808 chip has provision for dead time, to reduce losses from crossover conduction. We have it set up for about 60 ns. The fets run cold, so there's no significant simultaneous conduction.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

Skin depth at 300khz in ~5mils, so #30 is just about right. Maybe #29 or #28 would suffice, any larger would not have any advantage.

Now what kind of current is going thru it? Thats where copper losses will come into play, maybe the bifilar was used to lower losses to get the

Reply to
Martin Riddle

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.