HV dc/dc

If we exclude buck/boost inductors, the resulting statistics starts to deviate from the given 98% value. But these standard inductors do indeed have awfully good volumetric properties. For example, these:

I wouldn't be able to make a 47uH/9.5A in the 13.5x12.5mm form factor. The smallest would be ~15mm OD toroid. Starting at ~22mm (diameter/edge) outperforming a standard part becomes quite simple, though.

Multiple windings, high isolation voltage transformers, resonant converters with integrated magnetic components all require a custom part. It is good to have a big box of various ferrite and alloy powder cores and another with all kinds of magnet wires and copper tapes. You can have even a pretty complex part prototyped within 15-60 minutes. Once proven to work, you send the specs to your favourite coil maker.

And this all is about linear magnetics only. I can see no way to start prototyping a non-linear inductor or transformer without these magic boxes. >

I like #2. What's the bobbin? #1 could be wound with flat wire, it is much easier, and the cross-section area is larger for the same prototyping effort.

Best regards, Piotr

I'd breadboard any switcher that uses these tiny inductors. They might get really hot from AC losses. Sometimes a bunch of PCB copper pour will keep them from frying.

Given how cheap some stock parts are, and how fast a pick-and-place slams them down, there's no reason to not put some in series or parallel.

Here's some magnetics.

The flyback transformer is a Coilcraft stock part.

Those were experiments for high-frequency sinewave oscillators. I don't recall what the bobbins were; probably samples from a Lodestone kit. I wound up using a stock part in my oscillators.

Some have no specs. I've tested some to breakdown. One DRQ127 failed at 2250.

Here's a 1400 volt opamp supply:

A buck would happily accept any oversized inductor. Need 8uH, slam 47uH. DCR losses would go up by a factor of 5.9 (for a constant coil volume), AC and core losses would go down 34 times. Find an optimum somewhere in between. This option allows for high loss magnetic materials like the NiFe alloy powders (e.g. Hi-Flux):

With delta_IL~3% the inductor is basically DCR-limited and there are nice flat wires to fill the available window. Swinging chokes might be an option too, designing them is fun. These 20uH beauties are for a 50A buck-boost. DCR is below 2mOhms.

Probably would be good up to 100A, but I don't need that much current.

The inductance of uncoupled inductors connected in parallel would only be proportional to the length of wire instead of the square of it. This is good for one-offs or low-power applications, but otherwise you are wasting too much power on the excessive DCR.

Best regards, Piotr

series...

pageSize=25

flyback

es.

My first

but the

Better

y small.

nly a

t.

s care

d lab

us,

few

a few

mes in

e got the circuit from England without getting the literature reference tha t should have come with it - Baxandall, P.J, Proc I.E.E 106, B, 748 (1959.

of application notes for Linear Technology, on high frequency inverters fo r driving cold cathode back-lights used in laptop computers (application no tes AN45, AN49, AN51, AN55, AN61, AN65). Jim Williams describes the inverte r as a current driven Royer inverter, referring back to the non-resonant in verter described by Bright, Pittman and George H. Royer in 1954 in a paper ?Transistors as on-off switches in saturable core circuits? in Electrical Manufacturing."

transformers which tend end up with rather low self-resonant frequencies.

s and find a shop that would wind them for you - it isn't all that difficul t.

tting something close enough off the shelf isn't easy, even if you get down right sloppy about "close enough".

ng after high-voltage inverters and got the circuit diagram for a Baxandall class-D oscillator based device.

he 1954 Royer converter?

ade publication - it doesn't show up in the literature cited in his paper.

It was a public available patent:

So, yes, seems like he ripped it off. Nothing wrong with that, except if Ba ndaxall didn't credit Royer and Bright

r, and produces very different behavior.

Yes, that was a clever addition

Cheers

Klaus

the 1954 Royer converter?

trade publication - it doesn't show up in the literature cited in his paper .

That didn't mean that Peter Baxandall would have known about it.

Searching US patents in the late 1950's meant paying a patent lawyer. If I remember rightly, that sort of search cost about $100,000 dollars (in late

Peter Baxandall worked at the UK Royal Radar Research Establishment at Malv ern at the time

It was a pretty high-powered group, with lots of impressive academics who w ent off to become professors all mover the place.

Peter would have conformed to academic publication standards, but the Royer paper wouldn't have shown up anywhere where an academic would have run int o it at that time. I ran into it when I worked at the Plessey Pacific Centr al Research Lab in Melbourne in 1970-71 - which had fairly close links with academic electronic engineers - but it isn't a particularly impressive pap er.

Royer saturated his transformer to kill the base drive, which isn't an eleg ant approach. I just relied on the transistors running out of current gain, which is what everybody did at that time.

Bandaxall didn't credit Royer and Bright.

He probably didn't know about their work. Literature searching was much mor e tedious and expensive before the world wide web got going.

It meant going to a university library - as I did from time to time - and d igging through abstracts journals. You could find stuff, but it wasn't all that easy, and it was easy to miss things.

nverter, and produces very different behavior.

It has had commercial applications at medium power levels, where multiple outputs or isolators are needed. I've never heard it seriously referred to as a Baxandall circuit - it's just a current-fed inverter.

RL

If you go 20 years back, almost all the ATX PC supplies used a Royer followed by several buck outputs

Cheers

Klaus

"Baxandall" is a much tonier name though. ;)

Cheers

Phil Hobbs

Just keep in mind that there can be half an order of magnitude difference between working voltage and rated breakdown, or easily a whole order of magnitude to where it actually breaks down.

This is why we can't do hipot tests for very long, because it's gradually destructive to the parts.

With the Cockroft-Walton on the secondary it could be ok regarding breakdown. If it goes into a product I'd vet it with the coil manufacturer though.

It's in a product and seems OK so far. 350 volts peak is pretty far from the 2250 breakdown.

It's impressive how much voltage tiny magnet wire can stand. A twisted pair of #40 arced at about 1200 volts.

The more interesting part of that project was the 1400 volt opamps.

Use TIW and enjoy arcing at about 7kV.

Best regards, Piotr

IME that can bee too close. Gradual degradation sets in early.

Sure, but even at 100V manufacturers often avoid direct layer-on-layer.

Oh yeah :-)

I am working on a project with 400V right now and yesterday the almost inevitable happened. I unhooked a scope probe clip from the source of a big FET and it briefly touched the drain ... tsst *BAM* ... and now I am down one scope hook. The hook part in it had evaporated instantly. The scope set at 50mV/div came through unscathed.

We have one of the fully isolated Tek scopes. You can clip the probe "ground" thing anywhere. That's fabulous.

Boschert had an (almost) current fed inverter, using a 723 to control the buck section, nonlinearly, synced to the self-oscillating Jensen inverter.

That's one step up from the Royer - using a second small saturable (proportional-base-drive) transformer to set the operating frequency.

The inverter stage might be push-pull or half-bridge (Harada) without major changes to component count or drive transformer, depending upon the power level (130 - 400W).

Anyways, must have been hundreds of thousands of these in the market, at one time - being custom built into IBM, Burroughs and NCR hardware. So, mainframes, not PCs.

At 50W or less (ATX), it would have been a single-switch self-oscillating two-transformer flyback, using television semiconductors, until optocouplers bumped the second transformer out. There were were always IC-control-switched types of all conventional topologies, with 60Hz housekeeping control power, though some novel magnetically regulated versions showed up in Europe for a while.

G.H.Royer; "A Switching Transistor DC to AC Converter having an Output Frequency Proportional to the dc Input Voltage", AIEE Transactions on Communications and Electronics, 74, pp 322-326, 1955

J.L.Jensen; "An Improved Square-wave Oscillator Circuit", IRE Transactions on Circuit Theory, CT-4 pp 276-279, September 1957.

That would have been really helpful in this case. I am measuring 100mV signals across a shunt in the presence of huge magnetic spikes. It barely works with CH1 minus Ch2 and very careful cable routing. But, almost done now.

I learned from investigating epidemic HIPOT failures on a supplier that moved production to Mexico I qualified them in SD. It turned out that that grounding the isolated 48V secondary added more stress to the primary HIPOT test with a half dozen marginal gap locations with the slow ramp 1 second DC test that is 10% higher than 1.414x Vac level.

I modified the Sawyer HIPOT tester to reduce the probe impedance from

Then from my 50kV HIPOT tests on 5MVA OFAM transformers I discovered that the variation in oil breakdown voltage was due to nano-sized contaminants that created Partial Discharge (PD and thus) affected the initial condition to random dielectric failure. The greater the variation between BDIV or breakdown inception voltage was due to the pre-breakdown with internal PD. The ideal insulation has BDIV-PDIV=0 and the contaminated insulation has a large difference which is proportional to the BDIV standard deviation.

Increasing the HIPOT probe impedance or reducing the fault current to

Tony Stewart

Yes, current fed push-pull can be a useful topology. I am quite fond of the related Weinberg configuration.

Trying to make the Baxandall into a low distortion sine wave generator is non-trivial and fraught because it relies so critically on exact class B switching of the two transistors: if there is any non-conducting deadtime the feed inductor makes the centre tap fly high voltage, conversely if there is any conduction overlap the resonant LC tank is damped and purity suffers.

piglet