DC/DC powver converter, 500W

I would use a microcontroller to PWM switch multiple power MOSFETs. The idea is simple, but the devil is in the detail construction.

Look up "buck" controllers.

...Jim Thompson

-- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona Voice:(480)460-2350 | | | E-mail Address at Website Fax:(480)460-2142 | Brass Rat | |

| 1962 | It's what you learn, after you know it all, that counts.

Well, I would use a single LM339. Possibly two. Which route is more effecient?

(Ok, I lied. I'd use... maybe 40 transistors. I can probably wire it with

20, though. Even more optimal than the 339's, if not as compact as an integrated circuit.)

(I would still be lying, but only if the voltage weren't so low. If you had said 240V instead of 24V, I'd do it in about 10 tubes. ;-)

Tim

-- Deep Fryer: a very philosophical monk. Website:

Yes, voltage is low, but current is high. For 20A or so, you need to switch at very low duty cycle and spread across many MOSFETs. I am not sure how to do that with LM339.

Well, I meant the controller part, obviously...

Switching 20A or so with LM339s would need about a thousand of them, a slight excess from my estimate. ;-)

Tim

-- Deep Fryer: a very philosophical monk. Website:

nah, 20A is bugger all. just use a decent FET - at 5mR, 400A gives a measly 2W dissipation. a few SO-8 FETs would do the trick, sans heatsink ;)

now now Tim, use the worst-case spec of 6mA....3332 of them, so 833 packages :)

Cheers Terry

20A output. 24V drop across 5mR would give a lot more than 400A. The trick is to turn it off before it gets more than 880A.

That's certainly not for a SO-8.

For example, the ST250NS10 (ISOTOP power pak), can draw 220A. You can pulsed it at 880A, but less than 10% duty cycle.

I would use several of them to get a decent output power.

if its a switcher (madness otherwise) then I'm right.

I've built tens of thousands of 3.3V 14A smps using a single SO-8 dual FET for rectifier and freewheeling "diode", sealed up inside a plastic box with no heatsinking at all.

which is pantloads more than 20A. Thus proving my point.

Depends on the input, your 50W smps can certainly be SO-8. But the OP wants 500W, 48V in and 24V (20A) out.

Which is much bigger and heaver than SO-8. ISOTOP is not SO-8

so use a few more FETs. conduction losses will still dominate. the point is that a few FETs can drop IIR down to trivial levels.

switching losses usually dont dominate at 50V, and 500W is bugger all power.

I've made a 150A 24V battery switch using SO-8 FETs, again no heatsink at all. I just used lots of them.

Depending on what imprecision you can accept, you could consider an H-bridge (hip4081A & MOSFETs) running from a 50% square wave, into a 2:1 transformer and rectifier, for OUTPUT = 1/2 INPUT, less losses. It'd certainly be simple, without any controller.

--
 Thanks,
    - Win

Hi,

if you want it to be simple, Vicor and Astec do nice modules with just 4 connections, at about .5$/w, which can be worth it for small quantities (dont underestimate the complexity of high efficiency high power DC/DC). If you need one piece, watch ebay, i often see 48v->something converter modules real cheap (as i need a 12v/5v 20A module right now, if id need a 48v/5v id find 12v input modules all time...)

If you want to do it yourself, be prepared to do some serious work. I have done some smaller switching regulators and its not simple to get it to work the way it should, especially not while getting good effeciency and even worse complying with EMC limits. Layout, component seclection and so on are really critical at this power level. For a starting point, look at Linear Techs nice LT1339. Its datasheet already gives you some hints about component selection but thats not nearly all of it. You need a well designed PCB, not only being capable of the high currents (70 or better 105µ Cu) but also satisfying the necessary low inductances (2 layer with real vias is the minimum, 4-layer would be better). This is something you really cant do on a piece of prototype board. For troubleshooting, you better have some good instruments, you will need it.

I've just designed the half-bridge LM5100 into a product. Its nice, cheaper than a HIP2100, faster and with a gruntier output.

add a peak-current trip (a-la UC3842) to deal with the initial charging of output caps. I'd do it with a 74HC14, and a pair of HIP2100s.

Cheers Terry

A few mongo transistors and a couple of these, and a water-cooling arrangement, maybe a swamp cooler:

Cheers! Rich