1. Home
  2. Electronics Design
  3. Simple DC-DC converter, FWB, 12/24/48V to 320 or 640V, 500-1000W+

Simple DC-DC converter, FWB, 12/24/48V to 320 or 640V, 500-1000W+

First, some background.

I have designed an SCR trigger board using an IRS24541 self-oscillating FWB driver and two dual MOSFETs FDS6930B (30V, 5A, 0.055 RdsOn, 2W). I made a couple of transformers using EPCOS E187 or E16-8-5 cores with N27 ferrite having an Al of 950 nH. I used about 20 turns of #32 AWG wire on both primary and secondary, to get 12 VDC out with 12 VDC input. Running at 65 kHz I get about 11.5 VDC output into a 250 mA load (about 3W). The windings measure about 400 uH with 25 uH leakage inductance, and they seem to run very cool.

I ordered some samples from

formatting link
with target price of about $10 each, and they measure about 500 uH with only 2 uH leakage inductance. My transformers used a split bobbin for high isolation (for 480 VAC mains, so 4kV test), while his are layer wound with heavy insulation. They also work fine.

The drive is 50% duty cycle with only a couple hundred nSec dead time, so the rectified output is essentially solid DC with just a tiny droop during the crossover. I used 220 uF in parallel with 1 uF ceramic. I was somewhat concerned that there might be high current spikes on the square wave transitions, but that does not appear to be the case. Some waveforms:

formatting link

formatting link

My first prototype transformer:

formatting link

A second prototype with heavier wire #28:

formatting link
formatting link

The board with the sample transformers:

formatting link

Now for the present project. I want to generate 320 or 640 volts nominal at

500 to 1000 watts or more, from batteries of 12, 24, or 48 VDC nominal. I would not expect any more than 500 watts from the 12V source. I plan to make a transformer using an E55/28/21 core pair of N27 material, for which Al=5800 nH. At 50 kHz, I think 3 turns on each of four primary windings would be about right, with about 50 uH inductance, and about 16 ohms. The 24V P-P is about 12 volts RMS and about 1.2 amps no load. The secondary would have 38 turns each on four windings which would provide about 150 volts each, connected in series or parallel for 600 or 300 volts.

I ran a simulation that seemed to show about 95% efficiency. I will probably use a PIC to provide the 50 kHz square wave PWM, and a pair of FAN7382 hi-lo drivers. The LTSpice file is at the end of this post. I'd appreciate any opinions on this design. It just seems too simple but seems to work well in simulation and actual circuit.

Thanks,

Paul

==================================================== Version 4 SHEET 1 880 840 WIRE -160 16 -464 16 WIRE 0 16 -160 16 WIRE -560 80 -640 80 WIRE -464 80 -464 16 WIRE -464 80 -480 80 WIRE -160 112 -160 16 WIRE -224 144 -384 144 WIRE 0 144 0 16 WIRE -224 192 -224 144 WIRE -208 192 -224 192 WIRE 240 192 176 192 WIRE 320 192 304 192 WIRE 368 192 320 192 WIRE 416 192 368 192 WIRE -384 208 -384 144 WIRE -160 208 -320 208 WIRE -464 224 -464 80 WIRE -464 224 -544 224 WIRE -48 224 -256 224 WIRE -256 240 -256 224 WIRE 176 240 176 192 WIRE 224 240 176 240 WIRE 320 240 320 192 WIRE -160 256 -160 208 WIRE 48 256 -160 256 WIRE 112 256 48 256 WIRE 176 256 176 240 WIRE -640 272 -640 80 WIRE -544 272 -544 224 WIRE -464 272 -464 224 WIRE 416 272 416 192 WIRE -320 288 -320 208 WIRE -320 288 -384 288 WIRE 320 288 320 240 WIRE 0 320 0 240 WIRE 0 320 -256 320 WIRE -320 352 -384 352 WIRE -80 352 -256 352 WIRE 176 352 176 336 WIRE 256 352 256 240 WIRE 256 352 176 352 WIRE 0 368 0 320 WIRE 48 368 0 368 WIRE 112 368 112 336 WIRE 112 368 48 368 WIRE 224 368 224 240 WIRE 256 368 224 368 WIRE 320 368 320 352 WIRE -544 384 -544 336 WIRE -464 384 -464 336 WIRE -464 384 -544 384 WIRE -384 384 -384 352 WIRE -256 384 -256 352 WIRE 0 384 0 368 WIRE -160 400 -160 256 WIRE 176 432 176 352 WIRE 256 432 176 432 WIRE 320 432 320 368 WIRE 416 432 416 352 WIRE 416 432 320 432 WIRE -80 464 -80 352 WIRE -48 464 -80 464 WIRE -320 480 -320 352 WIRE -208 480 -320 480 WIRE -80 480 -80 464 WIRE 416 496 416 432 WIRE -640 528 -640 352 WIRE -464 528 -464 384 WIRE -464 528 -640 528 WIRE -384 528 -384 464 WIRE -384 528 -464 528 WIRE -320 528 -384 528 WIRE -256 528 -256 464 WIRE -256 528 -320 528 WIRE -160 528 -160 496 WIRE -160 528 -176 528 WIRE 0 528 0 480 WIRE 0 528 -160 528 WIRE -320 608 -320 528 FLAG -320 608 0 FLAG 416 496 0 FLAG 368 192 out FLAG 48 256 p1 FLAG 48 368 p2 SYMBOL ind2 96 240 R0 WINDOW 0 -21 38 Left 2 WINDOW 3 -49 83 Left 2 SYMATTR InstName L1

SYMATTR Type ind SYMATTR SpiceLine Rser=1m SYMBOL ind2 192 240 M0 WINDOW 0 -26 34 Left 2 WINDOW 3 -40 69 Left 2 SYMATTR InstName L2

SYMATTR Type ind SYMATTR SpiceLine Rser=10m SYMBOL nmos -208 400 R0 SYMATTR InstName M1 SYMATTR Value IRLH5036 SYMBOL nmos -48 384 R0 SYMATTR InstName M2 SYMATTR Value IRLH5036 SYMBOL voltage -640 256 R0 WINDOW 123 0 0 Left 2 WINDOW 39 7 147 Left 2 SYMATTR SpiceLine Rser=50m SYMATTR InstName V1 SYMATTR Value 48 SYMBOL cap 304 288 R0 WINDOW 3 24 64 Left 2

SYMATTR InstName C1 SYMATTR SpiceLine V=500 Irms=1.01 Rser=0.01359 Lser=0 SYMBOL res 400 256 R0 SYMATTR InstName R1 SYMATTR Value 100 SYMBOL nmos -208 112 R0 SYMATTR InstName M3 SYMATTR Value IRLH5036 SYMBOL nmos -48 144 R0 SYMATTR InstName M4 SYMATTR Value IRLH5036 SYMBOL voltage -384 192 R0 WINDOW 123 0 0 Left 2 WINDOW 39 -17 112 Left 2 WINDOW 3 -289 379 Left 2 SYMATTR SpiceLine Rser=10 SYMATTR Value PULSE(0 10 0 40n 40n 9.5u 20u) SYMATTR InstName V4 SYMBOL voltage -256 224 R0 WINDOW 123 0 0 Left 2 WINDOW 39 -25 113 Left 2 WINDOW 3 -433 374 Left 2 SYMATTR SpiceLine Rser=10 SYMATTR Value PULSE(0 10 10u 40n 40n 9.5u 20u) SYMATTR InstName V5 SYMBOL ind -576 96 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 5 56 VBottom 2 SYMATTR InstName L3 SYMATTR Value 10n SYMATTR SpiceLine Rser=10u SYMBOL polcap -480 272 R0 WINDOW 3 24 64 Left 2

SYMATTR InstName C2 SYMATTR Description Capacitor SYMATTR Type cap SYMATTR SpiceLine V=63 Irms=185m Rser=50m Lser=0 SYMBOL cap -560 272 R0 SYMATTR InstName C4

SYMATTR SpiceLine Rser=50m SYMBOL res -272 544 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 0 56 VBottom 2 SYMATTR InstName R3 SYMATTR Value 10m SYMBOL voltage -256 368 R0 WINDOW 123 0 0 Left 2 WINDOW 39 -95 126 Left 2 WINDOW 3 -424 321 Left 2 SYMATTR SpiceLine Rser=10 SYMATTR Value PULSE(0 10 0 40n 40n 9.5u 20u) SYMATTR InstName V2 SYMBOL voltage -384 368 R0 WINDOW 123 0 0 Left 2 WINDOW 39 -59 103 Left 2 WINDOW 3 -307 280 Left 2 SYMATTR SpiceLine Rser=10 SYMATTR Value PULSE(0 10 10u 40n 40n 9.5u 20u) SYMATTR InstName V3 SYMBOL diode 240 176 M90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName D1 SYMATTR Value RF1005TF6S SYMBOL diode 256 224 M90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName D2 SYMATTR Value RF1005TF6S SYMBOL diode 320 352 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName D3 SYMATTR Value RF1005TF6S SYMBOL diode 320 416 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName D4 SYMATTR Value RF1005TF6S TEXT 16 392 Left 2 !K1 L1 L2 0.98 TEXT -712 544 Left 2 !.tran 200m startup

Reply to
P E Schoen
Loading thread data ...

Most circuits (even battery-powered ones) require a housekeeping supply. This can affect efficiency expectations - drive power is not part of this simulation's efficiency calculation - neither are losses in the coupling magnetics.

An unregulated circuit will generally have issues at start-up, overload and shutdown, due to unlimited voltage-to-voltage energy transfer. You are depending on leakage inductances to limit high frequency current spikes which may prove difficult to define in physical construction and test of the magnetics.

Keep in mind that 12V, 24V and 48V describe battery bus voltages that have a wide tolerance range as the state of charge/discharge rate varies. This will be reflected in the output voltage. When a downstream regulation requirement results, the presence of the unregulated DC-DC section tends to become unjustified - though one level of isolation IS provided.

Low voltage high current inverters also suffer from source, bypass and hookup impedance issues, which are not easily simulated.

Without some simple form of DC current prevention in the primary, the most minor drive delay imbalances, switch impedance mismatches, or control logic glitches, can force the coupling magnetics into saturation, faster than you can say kaboom - a control mechanism to do this, with intentionally simulated imbalances are worth the simulation effort.

RL

Reply to
legg

Hey, I like that bobbin, where'd you get it, p/n?

--
 Thanks, 
    - Win
Reply to
Winfield Hill

It was a sample from Lodestone Pacific. It was one of the following part numbers:

2-02248-0-414-06-0

2-04143-1-220-04-0

Couldn't find them, but the following are pretty close:

formatting link

formatting link

formatting link

formatting link

I had to use a file to enlarge the hole in the E187 bobbins enough to accommodate the E16 core.

Paul

Reply to
P E Schoen

Thanks!

--
 Thanks, 
    - Win
Reply to
Winfield Hill

Thanks for the observations. I changed the simulation to a transformer with an 800 uH primary and 48 mH secondary, reflecting the expected inductance of

4x3 turns and 4x38 turns. I also set the coupling factor N=1 and represented the leakage inductance (or a separate added inductor in the primary) of 2 uH. This would be the same as N=0.9975, or 0.25% impedance. With that, I get an output of 292 volts and 852 watts, with input of 886 watts, or 96% efficiency. This results in a supply current of about 85 amps peak at 130 uSec and then peaks of 40 amps when output is stable at 1.6 mSec.

With 1 uH, the output rises to 326 volts and 1.07 kW with input of 1.1 kW or

97% efficiency. The supply current has a peak of 130 amps at 70 uSec and 42 amps when stable at 1.6 mSec.

With 4 uH the output drops to 235 volts and 549 watts with input of 578 watts, or 95% efficiency. The peak current slowly rises to about 48 amps at

250 uSec.

To see the effects of a variation of VdsOn, I added a 10 mOhm resistor to the source of one of the low side MOSFETs, which have a VdsOn of 3.7 mOhm. The output power is 545 watts and input of 579 watts, so that is not very significant. This causes about 77 mVDC imbalance and 291 mA DC into the primary.

I can see where primary inductance is pretty important, and it has a lot more effect on the output than I thought. Perhaps I can use an air core inductor of 4 uH so it will not saturate. I used an on-line calculator to find that a coil of 20 turns, 1" diameter and 2" long, has an inductance of

4.08 uH:
formatting link

I will also have to build a prototype transformer and test it to see what sort of leakage inductance it has. If I can use the measurements of the small transformer as a guide, it has about 0.4% leakage, which is just about what I need for the 4 uH. I may want to use a higher value, and add more turns to the output to get the voltage I want.

The output voltage is intended to be used on a VFD, which has a range of

200-400 VDC for 240 VAC output. So I really don't need regulation. But the inductance greatly affects the open circuit output voltage. With a 5k load, the output becomes 371 volts, which is close to the maximum of the VFD bus link voltage.

Paul

Reply to
P E Schoen

For unregulated DC-DC, core loss is independent of load ~ continuous. Worst case occurs at high (battery) line. Without copper loss, the

4V/turn (65mTpk at 50KHz) should generate a 20degC surface rise in an N27 E55 structure. There is some fiddling room, and iterations to be done to get a transformer that does (mostly) what you want. Air-cored magnetics are probably a waste of time and a source of radiation - the parasitic ones that are unavoidable will be tough enough to deal with.

Your simulation will show the amount of time it takes, simply to reverse current flow in the full bridge primary, within the switching cycle - a serious load regulation effect. This can cause increased HF harmonic ripple current in output filter caps.

Start-up, limit/inhibit cycles, load discontinuity and shutdown can also provoke transient overshoot on either input or output, depending on natural decoupling/filter component resonances. This will call for more care in the controlling modulator than you may may feel inclined to build into it.

Slower parts (with controlled avalanche characteristics) might even work to your advantage...as well as using the lowest practical conversion frequency. You're shooting for 'simple' - don't forget.

Simply hanging a sign on it, saying 'don't do that', is asking for trouble in the field....or in specific combinations of battery input or application output voltage, as windings are series'd or paralleled for 'type' variation.

RL

Reply to
legg

I love the background, but it doesn't seem to illuminate the present project. How do they relate?

For low voltage inputs, push-pull is often better. You don't spend the voltage drop of two switches in series, as you do with an H-bridge, and the extra winding loss is quite tolerable as the primary halves are only a few turns, often of copper strap or something like that.

Note your schematic is missing an inductor. You can put it on the primary (like L3, but larger, and /after/ C2), or on the secondary (after the diodes, before C1). Without, turn-on dI/dt is limited by LL only, most likely defeating any subsequent attempts at current mode control and current limiting.

Don't follow the trap of the automotive inverter. They use such a circuit, typically an open loop TL494. The only reason the poor things start up at all, is a big fat capacitor on the soft-start pin. They're a house of cards on a wagon... start it moving slowly enough and it might not fall, but that's a far sight from being reliable.

Also, being open loop, they aren't regulated, and can't be controlled to regulate (at least not in any reasonable manner).

Story warning:

Watch out for parasitic winding capacitance. The last high voltage forward converter I made had a combination of ills:

- Poorly wound transformer. Ran out of space on the bobbin, so the primary and secondary weren't well coupled. Lots of leakage.

- The leakage caused large voltage overshoot at the rectifier (FWB, choke input filter). This necessitated 2kV rectifiers -- which is to say, pairs of UF4007s in series.

- The diodes still got very hot. R+C dampers were added to the winding, and caps piggybacked on the diodes. An awful hack, at this point, and still insufficient. (Diodes tend to behave poorly when connected in series, because they don't recover at quite the same time, thus the early bird wins the high voltage worm, and burns power in avalanche.)

This was for about 100W and +/-200VDC. Quite some years ago. It would've definitely benefitted from a more holistic design process. Hacking doesn't work out very well!

For the project, I later smashed the whole converter section, and rebuilt it (point to point, like a real man!) with flyback topology. The new transformer was wound on a similar sized core, but it was taller, so had more winding width. I got the primary and split secondaries in single layers, same number of turns, phased so that no AC voltage appears between layers, only DC. (HV flyback transformers are designed this way -- single layers, with a diode into the next layer, stacked all the way up. It's the only way to get HV, at variable-up-to-100kHz operation, in a high resolution CRT monitor!)

Result? Much more efficient. All the passive stuff (transformer, diodes and filtering) ran quite cool. (I also tossed in some 1200V 4A SiC schottkys, just because.) The switching transistor had one flame-out, which was solved by buffering its gate drive with a proper driver chip (the UC3842 controller wasn't quite powerful enough on its own).

So, that's 100W. At 500-1000W, you will be better off with a forward converter. At that ratio, you certainly won't be able to use a transmission line transformer approach (which, after all, is kind of what I constructed above), so you'll still need to mind the windings. 500V 1A is 500 ohms, pretty high, so the parasitic capacitance will easily show up. Try to balance LL and Cp so that 500 ohms ~= sqrt(LL/Cp), secondary referred that is.

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website:

formatting link

Reply to
Tim Williams

The topology is the same, with no added inductor. My original prototype transformers had 25 uH leakage inductance, or about 6% impedance. So that would provide enough to limit the start-up transients. The layer wound prototype had 2 uH leakage inductance, so it may be more of a problem. I might want to add some inductance in series with the primary, as I did for a subsequent simulation (see reply to legg).

Here is more information on my small DC-DC converter:

formatting link

I built a push-pull DC-DC converter a few years ago, and used it to provide

250-300 VDC to a VFD driving a 2 HP 3 phase motor on a small riding mower / tractor. It used a toroidal transformer with tape-wound silicon steel core. It worked for a while, but eventually blew up:

formatting link

That was probably a combination of problems, but mostly the heavy capacitive load and the rather low frequency (about 800 Hz).

I added the inductor to a subsequent simulation. See my response to legg. It looks like 4 uH would be OK. I also tried adding a 10 mOhm resistor in one leg of the bridge to see how imbalance might affect things. I will probably add some sort of current limiting and PWM shut-down using voltage on the shunt common to the legs. The ON times at 50 kHz are about 20 uSec and I should be able to shut down the PWM within about 1 uSec or so. That should be fast enough.

I actually have a couple of automotive inverters on which I brought out the DC bus voltage. They are 220V inverters and the bus is only about 240-260 VDC. But it did drive the VFD and motor, although not under load:

formatting link

I might be able to make a soft start using variable PWM from the PIC. But I don't want to fool around with output feedback and regulation. The VFD allows for a wide range of DC link bus voltage, typically 200-400 VDC. Most of the time it will probably be operating at about 1/2-1 HP (375-750W), and about 1/2 top speed, which will be about 120 VAC and 30 Hz.

[snip - for brevity - but I will consider your experience and observations and caveats]

If this simplistic approach is unsuccessful, I may adopt the SiC LLC design presented by Tony Bogs:

formatting link

His design is for an isolated charger, where output will typically be lower than the input, but the same principles should apply. My original idea:

formatting link

I also presented an explanation of magnetics from what I found on-line and adjusted to how I understand the concepts (and magnetics theory is admittedly a weak point in my repertoire).

formatting link

Thanks,

Paul

Reply to
P E Schoen

Would a NiZn-core common-mode choke be usable in that application?

--sp

--
Best regards,  
Spehro Pefhany 
Amazon link for AoE 3rd Edition:            http://tinyurl.com/ntrpwu8
Reply to
Spehro Pefhany

CM chokes use the same high permeability material as signal tfx. Lossy for power apps. Low curie temp. Available in a host of shapes. Wish low loss matl had some of those shapes.....

RL

Reply to
legg

I also did not see the relevance.

Magnetic characteristics don't scale linearly. Energy storage, intentional or otherwise scales with I^2 and V^2, while power throughput is a direct relationship.

Surface area for cooling degrades with part volume.

Parasitics compound with layer dimensions and xsectional area.

You can get away with murder some times at low power levels, simply because the components are over-rated for the app (smaller parts or heat-conducting structures being impractical/unavailable).

RL

Reply to
legg

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.