Load exceeding supply current

You'll need more like 32,000 uF to hold droop to 1 volt in 1 millisecond at 32 amps.

The average current for the above pattern is 10 amperes, so you need a supply with greater capacity than that, or it is going to be in current limit all the time.

A 30mF cap will sag 1 volt in 1 millisecond while delivering the whole

32 amperes. but if the supply can deliver 10 of those amperes, then 22 mF will deliver the remaining 22 amps during that millisecond while the voltage sags by a volt. It will recharge back to full voltage with the excess current capacity during the rest of the cycle. Current limit is not a problem for the regulator as long as it is designed to be in current limit, continuously. Current limiting is just regulation with a different condition controlling the pass device, except that the current is high. But any supplies are not designed to handle the temperature rise of continuous current limit operation.

Coupling the big capacitor to the supply through a big inductor will let the cap carry more of the peak current (requiring a larger capacitor) but will smooth out the current drawn from the supply (and also waste some voltage with its resistance). A bigger supply is probably cheaper than this solution.

If your power supplies are switchmode types, you have to consider :

- will they like starting into a large capacitance ?

- "current limit" can be a vague term for switchmodes. Most do not have the classic laboratory power supply current limit where voltage drops to keep current at the limit. Often a protective cycle by cycle limit cuts in, but way above the rated current, and the effect on output current depends on mains voltage. Some supplies shut down when excessive current is drawn. Most switchmodes are intended to be wired up to a load which does not have massive peaks above the average value.

- A little bit of resistance works wonders - say 20 or 50 milliohms between power supply and capacitor will help lower peak current and assist supply control loop stability. An inductor is good too.

You can get switchmodes with an external control terminal or feedback terminals, and it is possible to control the switchmode for your own purposes. We incorporated this in a 34A output supply to give current limit and precise voltage control. We had to consider control loop stability. Might be too complex for your needs.

With series power supplies, I have seen a reversed diode wired across each supply output, so one can never drive current thru the other.

Roger Lascelles

I'm thinking about one that let's you slow down the start-up.

Thank you so much for that! Exactly what I was worried about...

Yes.

This is a valid concern. It's generally bad to put supplies in series. If you expect them to behave in current limit, you're dreaming. They will never be exactly the same. You may be in the situation that only one ever current limits. You won't be able to use foldback limiting. There's probably a glitch filter in the limit that can put both supplies in a frenzy. If you're only building one system, you only have one customer to worry about. If you build a lot of them, you're asking for trouble. It's unlikely that the current limit behavior you need is in the spec. Using any component outside it's specification is a recipe for disaster.

Some bean counter in your organization is gonna save a buck by changing vendors. Or some bean counter at the supplier is gonna change one of his vendors. And all of your careful evaluation goes out the window and your replacement is on a plane to do an emergency retrofit.

Put a filter between the supply and the load that guarantees you never put the supplies into current limit. Make sure your caps are rated for the repetitive peak current you are gonna put thru them. Once you've met the specs on the supply, you can remote sense the output of the filter. Yes, there will be phase issues and potential for oscillation.

Pay attention to your resistance budget. 1V 32A gives you 1/32 ohms total resistance for all the caps, wire, connectors, etc. And that's before you get to any capacitive droop.

Buy a supply that's rated for the peak and average currents you need. The design problem is the same, it just goes inside the regulation loop and is guaranteed to work. mike

--
Wanted, Serial cable for Dell Axim X5 PDA.
Return address is VALID but some sites block emails
with links.  Delete this sig when replying.
FS 500MHz Tek DSOscilloscope TDS540 Make Offer
Bunch of stuff For Sale and Wanted at the link below.
MAKE THE OBVIOUS CHANGES TO THE LINK
http://www.geocities.com/SiliconValley/Monitor/4710/

First thank you so much for comments! Resistance budget: yes, very important, especially after I noticed (duh!) how the load will take more current from the low-series-resistance component, and that may be the power supply, not the caps as in some rosy dreams... (The caps won't provide the bulk of current.)

issues and potential for oscillation.

Yep. It's tempting to use remote sense to compensate droop. I'm thinking of... filtering the filter output... Something like 1s delay. (The 32 amps -- and droop -- happen every 8ms.) Won't help fast-changing firing patterns (I'm firing solenoids), but would catch up to a constant one.

Solenoid drivers for fuel injectors. The box must be able to drive 8 simultaneously. One runs up to 4A, then settles at 1A. Max freq 125Hz.

This needs to work at 12V as well.

I was thinking of using one unreg in series with one reg with sense divided by two. But I found a small enough variable supply (yes, size matters...) and I'm going with that.

Many go up to something like maybe 200%, but surge and periodical seem to be disjunctive concepts...

That's basically were I am heading too.

The imaginary device in the homework assignment.

Cheers! Rich

probably what's the surge rating on those supplies? If it worries you use a bigger cap

I'd say to use unregulated power supplies, but you've got that 1V limit maybe use two unregulated power supplies and then regulate their series output. (or just get a 24V powersupply.)

What sort of device has nice sharp current fluctuaions like that?

Bye. Jasen

And you're planning of firing all 8 of them at once?

125Hz on a 4 cylinder engine would be 3750 Rpm - kind of slow for an upper limit. and on an 8 half that.

It sounds more like you want do so sequential multipoint injection (or possibly direct injection) with one injector feeding each cylinder.

this way only one injector is fired at a time, and 125Hz (on each injector, with the injectors phased by 45 degree) would translate to 15000 RPM, the sort of RPM race cars, (and motorcycles) run at. and basically you have an 8 amp load with a 3A ripple. not so scary at all.

Bye. Jasen

It's used for developing a type of jet engine.