DC-DC conversion from 1 Kilovolt DC down to 12VDC (isolated)

Got a link to one?

Given that it's not isolated, that's a huge amount of parts to replace a simple buck switcher.

John

This was just an illustration for what I meant with a tapped transformer architecture. I wouldn't build it that way, one reason being that a DC imbalance can be bad. And mine never had this many parts :-)

Doing a direct buck from 1000V down to 12V is possible but you'd have to deal with a rather extreme duty cycle. Whether you have an inductor or a transformer isn't much of a cost difference anymore these days if there is a stock part that fits.

I use that architecture in several of my products, where the customer supplies 24VDC or some such but wants me to isolate it. I usually run open-loop, which is pretty simple.

This is pretty simple, but the transformer is custom.

ftp://jjlarkin.lmi.net/28S910D.pdf

But the fet drain voltage peaks at twice Vin, which would make 1000 volts into 2000 volts.

Yeah, I guess a high-ratio buck has the same problems as a high-ratio boost. Capacitances get'cha. So... two-stage buck?

John

gate/

avoidance.

Those are very suited for open loop. The UCC3808 is a bit pricey for me though.

Just watch out that it never stops or there'll be smoke.

Yeah, that's the other downside. Tying CS to ground is like putting a brick on the accelerator as "cruise control".

I'd just push-pull capacitively into a transformer. That could also run open loop if the accuracy doesn't have to be great. It would be nice to find a huge CCFL transformer that has enough oomph to be able to do this job (in reverse configuration). Paralleling gets old if it's too many and you never know whether they count the HV turns accurately.

the

gate/

isn't

avoidance.

he also wanted isolation rated for 3kV, so why not a buck from 1000V to 300V in front of a standard ~320V input 3KV isolation rated 12V supply

-Lasse

the

gate/

isn't

my

avoidance.

That's cute. We're buying 150 watt MeanWell PFC-corrected universal-input switching supplies for $30. Front that with a simple buck, or maybe squirt 400 VDC in midway, after the PFC boost stage.

John

heap

at

en the

g gate/

isn't

. *BOOM*

ly my

o do

. At

its

idance.

or

I thought it was isolated. It should be if it isn't. I dont think you buck switcher will be that simple either.

You can easily isolated this one. The only connection from output to input is the voltage feedback line. Can be isolate by modulating plus ferrite, linearized optocoupler (the ones with two photodiodes), etc.

the

gate/

isn't

avoidance.

There are lots of isolated switchers that input 300 to 400 volts DC, like the tail end of the MeanWell. A very crude buck could take the

John

cheap

that

between the

gate/

isn't

probably my

do

At

its

avoidance.

buying the right standard supply will have the added bonus of a piece of paper that says rated for 3KV

since that was asked for it is probably important

-Lasse

o cheap

that

tween the

ding gate/

sh" isn't

... *BOOM*

bably my

g to do

cap. At

of its

avoidance.

n for

d it

Well as has already been discussed at length there are few suitable power devices available. Then you need to protect against a shorted device from destroying the "meanwell supply". Then you need an auxilary supply to power the buck regulator, that also has to run from the 1kv input.

cheap

between the

gate/

isn't

probably my

do

At

its

avoidance.

Hey, if electronic design was easy, anybody could do it.

John

Bingo!

Especially when things go high voltage the number of engineers willing or able to do the job often shrinks substantially. In this case one could squeak by using a "reservoir" trick. Have the HV slowly charge a cap to, say, 15V and then zener it off right there. Doesn't take much energy if you can allow several seconds or even tens of seconds. The zener tugs at the base of a BJT and when base current flows that BJT pulls the PWM chip of this pre-buck out of UVLO ... vroooom. Like kick-starting an old Harley. After x attempts you have to pause and catch your breath but it's supposed to start after the first attempt.

After start-up the output of the MeanWell switcher takes over and feeds the PWM chip of the pre-buck.

Regarding shorts the MeanWell (hopefully) has overload protection already inside.

Forgot to mention: There must be a hysteretic latching so the PWM chip keeps operating until this cap is depleted to some reasonable lower values. UVLO funtions often have that built-in so there's very little or nothing in extra parts needed. If for some reason the MeanWell won't start or is in overload shutdown or whetever the pre-buck just keeps starting and stopping at a low duty cycle.

This still needs isolation, so you can't easily use the MeanWell output to take over from the startup supply. But a few turns on the buck inductor could make the sustaining power.

MeanWell or not, it might make sense to first pull the 1000 volts down to something easier to handle, like 200 or so, with a simple buck.

John

You could use a cheap telco transformer with sufficient dielectric rating and an oscillating gate driver to bring 12V across, unregulated. Or use a cheap DC/DC converter, maybe like this, costs around $5:

Yes, if the cost difference between a real 1kV switcher and the MeanWell is huge. Which it probably will be. But one has to keep in mind how the dielectric withstanding is defined. Often it is not rated 3kV for any serious duration.

too cheap

ite that

ess I

between the

winding gate/

"ish" isn't

UT* ... *BOOM*

probably my

hing to do

ns cap. At

ome of its

on avoidance.

ason for

ould it

ok.

ace

First you say its a "simple" buck switcher, and now...?

You have to know your limitations, otherwise you make a mountain out of a molehill.

Nice big fat hot resistor maybee.

Doesn't take much

Thata a long time for power up.

The

Isolation

What they usually have is a fuse, filter, mov, rectifier, cap, on the input. 1kv in, mov explodes, fuse blows if your lucky, otherwise 1kv goes to main caps.....