Cheap pulse 1:1 transformer

fredag den 10. november 2023 kl. 00.14.58 UTC+1 skrev Klaus Kragelund:

why a transformer instead of an isolated driver?

seems to me DK has plenty of 1:1 pulse transformers much cheaper than that double inductor

Yeah, I was looking at those. But except for LAN transformers, they approach 1USD/pcs, which is awful high

I was even thinking doing a planar transformer, but production assembly of ferrites are not straight forward

Speaking of LCSC, it's quite amazing that one can get a transformer like this for less than 10cents:

Dual inductors (like DRQ family) have super coupling but high winding-winding capacitance and insulation is just the minimal enamel so useless for anything HV.

I have even used low cost Murata and Panasonic common mode chokes as HV separated gate driver transformers!

piglet

The classic examples are bifilar wound - with twisted pair - and can be 1:1 to one part per billion. They can still have appreciable inductance, but at least it is exactly the same for each side. It takes an effort to get 10:1 ratio transfomers as good as one part in 10 million

Sounds credible.

If you stick to core shapes that are used in volume, you can get volume pricing on any drawing and spec from off-shore mfrs.

CCFL or LED <8W ballasts, or USB wallwarts are a good place to get an idea of what's used in this market.

That's what north american suppliers do, with ~ 500% mark-up.

. . . or you can just buy off the shelf from off-shore vendors with one-time-buy volumes. 10K at $0.10 . . . that's a few day's tech labor in NA . . less than you will probably spend to process the paperwork.

Sigh.

RL

Suggest you nail down your requirements first: Lpmin Vtmin, Rdcmax, Isolation class, prod test level.

Many techniques avoid the requirement of high Lp and Vt.

Be smart or pay.

RL

There's not a lot of smart transformer design discussed here. John Larkin seems to think that you ought to ignore the possibility - anything that you can buy off the shelf trumps anything that you might design, but then again he claims to get away with ignoring the parallel capacitance of transformer windings, which is bizarre,

I suppose it doesn't really increase volt-seconds asmuch as allow a lower v-s rated transformer to be used.

Between the initial turn-on and the turn-off the main transistor gate is held up by its own (or the added) capacitance so it doesn't matter if the transformer induced secondary voltage collapses.

piglet

May as well buy an SSR.

Yeah, the idea is to make the available voltseconds go further—like putting cornflakes in the meatloaf.

Cheers

Phil Hobbs

I have done ac coupled gate drive before. Does not work well with 100% dutu cycle, and has some issues with transients on the rails, which coupled through drain-gate capacitance and into the driving circuits. But, great link anyway :-)

It needs to switch fast, since transient loss might be very large, and exceed transient thermal resistance constraints

Why do you need to switch a heater fast? Those are fancy fets just for heater switching, triacs make better use of silicon?

piglet

Due to the switching loss of the FET. I am looking into heavy snubber to avoid that.

Surprised you find it worthwhile, I estimate you'll see 2.2W in the two FETs versus 2.9W in a triac - which costs less and frees up PCB area you could use instead for heatsinking?

piglet

Yes, it's only a backup solution if the current solution with the triac misbehaves. Triac costs 1 USD, FETs 5 USD combined. At lower power levels it switches, so FETs are cheaper than triac. Advantage of FETs are also no spurious turn on and other funny 4th quadrant switching problems.

This one?