"P E Schoen" <email@example.com> wrote in message
> when I return to that one I can try the ferrite cores again. I have both
> #26 and #52 material (which I used). Maybe the #26
> is better?
Yuck, eww! #26 and 52 are powdered iron numbers, not ferrites. Both are
"high permeability", which is only 75. You really want >500 for a
transformer, and >10k for a good transformer (pulse transformers, CTs, small
signal isolation, etc.).
Common powdered iron toroids are yellow/white (#26), which is very lossy,
only good for low ripple smoothing chokes, or blue/green (#52), which is
about half the losses and same performance otherwise, which isn't really
saying much as it'll still cook off easily with much ripple. These are
common on older motherboards, where the smaller cores can tolerate some
ripple at moderate frequencies, but even then they can get quite hot.
Typical numbers for ferrite include Fair-Rite #77 and #78, Ferroxcube 3C80,
3C90 and such, and Magnetics types R and P for power (mu ~ 2200), #43 for
high frequency (mu ~ 800, good up to ~MHz), 3F3 and 3F4 for high frequency
power (up to about 1MHz), etc.; 3E6 and such, Magnetics type W for high-mu,
Obviously, you only get the best effective permeability with minimum gap.
Line filter chokes actually go so far as to use not just toroids but also
rectangle and "digital figure 8" shapes, single piece no mating faces, for
absolute maximum permeability. Such ferrites go up to 15k, even 20k
permeability: when it gets that high, it's a fragile thing, easily spoiled
by ambient bias fields and current imbalance, but when small signal levels
are all that matters (swallowing up a little RFI on the line), it does a
> I can't go much higher in frequency with my prototype design because I'm
> using a PIC18F2420 which does not have a PWM module. But when I redesign
> the board I can use a PIC that has built-in PWM, or I could use a
> separate device to drive the transformer. I'm using an L298 module
> (which is way overkill, and I'm only using half). So maybe I'll try an
> IRS2453D which is a self-oscillating full bridge driver that costs only
> about $1.35 and can drive some small MOSFETs. I don't need regulation.
> I have a 12V supply and I just want about 8-12 VDC on the secondary for
> the gate drive circuitry.
Frequency isn't a big deal; I've done the same thing with a discrete
oscillator (what could be simpler than a two transistor multivibrator?), a
little current boost (more discrete switches..), and a common mode EMI
filter choke ran "sideways". Such inductors have tons of leakage between
windings, but the price is right for the amount of isolation you get.
Independent of size, I've found most CM chokes are only good for a watt or
two: if you run at a higher frequency, you can supply more voltage, but the
higher frequency drops more voltage across the leakage inductance, making
the supply "squishier"; for some desired output stability (like 10%),
maximum load current goes down, as a result, power remains roughly constant.
I personally wouldn't use L298, because it's a slow bipolar device with lots
of voltage drop (about 2V total under typical load, IIRC). Gate drive chips
are the best choice these days; same basic operation (it pulls up to the
supply, it pulls down to ground), some even have disable pins (tristating
the output), and many are available in multiple units per package (a quad
driver with disable is equivalent to a CMOS L298, sans ground return pins).
With such inductive loads (as the powdered iron cores), you might need
schottky clamp diodes on the outputs to use CMOS chips; all that magnetizing
current might otherwise induce latchup. (Gate drivers rated for maybe 1A
output are probably good for 0.25A reverse current; design accordingly.)
Dual "complementary" packaged transistors are also handy. Sometimes too
handy: I once made the mistake of doing this,
forgetting that discrete transistors don't have the same softness of
monolithic CMOS circuits! Poor tantalum capacitor was probably sweating its
balls off, while the transistors pulled 10-20A peak currents out of it for
~50 nanoseconds each cycle. (Solution: nix the cap, soften the supply rail
with series resistance and inductance. Supply bypass is actually a BAD
thing sometimes, and it's important to see when!)
SO-8 MOSFETs and arrays of course come in many sizes, these just happened to
sink that much current under switching conditions. Protip: SO-8s are more
common than DPAKs or other packages. Bizarre when you'd rather have a
compact, heatsinkable SOT89 than a flimsy SO-8, but I guess that's just how
the market is. I've gathered a lot of SO-8 dual FETs from hard drives,
probably what's "driving" the market, among others.
For many purposes, a plain 2N7002 + BSS84 pair wired as a dumb CMOS inverter
is equivalent to most smaller monolithic drivers (0.5-1A peak current), with
the advantage that they are discrete transistors and can be operated much
Anyway, as it's late and I'm rambling again, I'd suggest a flyback converter
instead. Something with UC3843 perhaps, like so:
(UC3842 shown, because the supply is 18V; 3843 has a lower UVLO threshold,
allowing it to work on >8V supplies.)
The feedback winding can be local, in which case both output channels have
the same mediocre regulation; it can be placed on one output (via TL431 and
opto, in the common arrangement), in which case one output is perfectly
regulated and the other is mediocre. (Note that, since flyback stores
energy, but it demands large ripple, gapped ferrite is preferred over
powdered iron in the transformer core.)
Or you can buy the whole thing in an e.g. RECOM isolated converter, albeit
FYI, are you aware your message encoding is quoted-printable? It makes it
very difficult to quote as I have to enter the ">" manually. Also, the word
wrap is slightly off by my reckoning. IIRC, traditional is ~76 characters,
yours might be set 77-78, which isn't bad, just looks funny after a few
layers of quoting.
Deep Friar: a very philosophical monk.