90V DC-DC converter IC?

That IC requires 12V at (typ) 14mA for the 12V_IN supply.

A linear drop would burn up 1W (78V * 14mA). I suggest some simple switching circuit to lower the duty cycle.

That's the easy part.

You (OP) may need to look at a dual conversion - 90V to some intermediate value with a discrete solution (doesn't have to have particularly good ripple performance) and then an IC based solution.

The easy part above is to use a high voltage transistor with a zener at the base to provide some Vreg at the emitter. Make sure that Vreg is lower than the output of the supply so that once the supply starts up, the supply output provides chip power, not the transistor.

Cheers

PeteS

I did not suggest a linear drop for the intermediate system. A linear drop *during startup* for the final system would burn minimal power. I assume you understand switched powering systems.

Cheers

PeteS

Look at Power Integrations Linkswitch-TN series. Aimed at mains, but should be adaptable. Cheap and stocked by Digikey.

If you're feeling adventurous you might want to try adapting this :

adaptable.

:

You'll find Power Integrations DPA series very easy to use in either flyback or forward modes:

That kluge has been exposed time and again for being a bad design...

TI/Unitrode has a UCC3889 for low power offline apps. I'm not sure it's guaranteed to work down to 90V. 90 to 5v is a huge step down ratio. I recommend using a transformer. Look at the LM5020 (National). Runs up to 100V and has some good flyback applications. Unfortunately the conversion ratio makes it difficult to do something simple.

There is no ready made solution other than a DC-DC converter module end product.

Other than that you have a 1:18 step-down ratio and fairly high voltage which will require more engineering than you're used to :

Anyone who brands something 'bad' or 'good' is an idiot.

There is only more or less appropriate for any given situation. All applications are different, and have different requirements.

one man's 'bad' is another's 'the only thing that makes a design financially possible'

The circuit is trash. I posted an analysis of it a few months back, the feedback is wrong among other things. It's a clueless hack.

Analysis notwithstanding, it works. May not be as 'good' as an LM2574, but you don't always need all the bells & whistles.

How about adding the modified Larkin oscillator? See

Note the L2 C2 tank at the emitter. L1 L2 are 80% mutual inductors. Don't forget R2 D2 too.

Why not throw away the first 30 volts in a zener, and then use the LTC regulator? The power dissipation would be small.

You could do a switcher with 10 or 12 parts, but it's probably not worth the effort.

John

whistles.

But fred likes bells and weasels.

There is also the LM5574 - goes up to 75V so you only need to throw away 15 :). I've used it recently.

yep! with a bit of effort it can be improved from "appalling s*it" to merely "bad".

I built one just for fun. Try starting it into a shorted load. By the time you add enough bits (in USian: components) to fix the glaring defects, you have more than enough bits to build a proper SMPS.

that being said, its a pretty good effort from someone who knows not what they do ;)

Cheers Terry

So it isn't suitable for starting with shorted loads.

As I said, there is no such thing as 'bad', just more or less appropriate for any given situation. This circuit is not appropriate for situations where the load may be shorted at startup.

There's also the LM5009, good for 100V (a bit more on the switch) but whether it can be used without the zener depends on the spec on that "90V" including transients etc.

Best regards, Spehro Pefhany