# how to build a high-power (4kW) switched-mode DC-DC converter?

I would like to build a switched-mode DC-to-DC converter which should support up to 4000 W of power. It should be able to act as a CCCV power-supply, so supply a constant current Imax up to the point where voltage Vmax is required to have that current, and then supply a constant voltage Vmax with a diminishing current.

I would like it to be modular, to be able to swap-out a few components to change the Imax and Vmax. I do not want to have knobs to change the Imax and Vmax like on a general-purpose lab power supply, as these parameters will be changed very rarely, when I change the battery configurations, maybe every few months, and I do not want the risk of me or someone else accidentally turning the knobs, and then connecting it to the batteries with wrong parameters which could kill very costly batteries - as I want this to support up to 4kW, you can imagine the batteries will be large. No, I don't have batteries supporting such current yet, but within years I could have, so I want this device to be easily expandable as my battery park grows.

EDIT - I modified my expectations, which I hope will make it easier to build:

The configuration I would like to have:

• Input: rectified 230VAC mains (that will be something around 325VDC?)
• Output: CCCV with exactly Imax=15.0A, Vmax=230.4VDC

Parameters:

• input and output electrically isolated
• low ripple (1% is OK, 2.5% is too much)
• efficiency at least 90%, I am willing to spend up to 10% more for every extra 1% efficiency that I gain over 90%.

In the future I would like to be able to install modules to be able to accept a different voltage input, for example from solar panels, so a design that will accept a wide voltage input range (say 100-400VDC) would be a plus.

I would be grateful for some basic pointers into how to build such a beast, possibly using some ready-made blocks, but I don't want to buy the whole device, as I want it to be modular and expandable (unless some DIY kit is available).

I could take four Chinese 48V (adjustable to 58.8V) 15A CCCV supplies for less then 200$each and connect them in series to do the job, but then I have "black boxes" which if a single component inside brakes, not having the schematic will be difficult to fix. So I would prefer to spend even twice as much for a modular, self-made design, so if something breaks I can easily exchange the faulty element, and if my desired voltages change I can just rebuild part of the device without buying a whole new one. • I'm not sure what you're doing with this, but maybe it could benefit from parallelization? Presumably you have multiple batteries. Maybe it makes more sense to have multiple power converters? – endolith Mar 29 '11 at 15:02 • I would be interested to see how you draw 4kw from a single mains socket – BullBoyShoes Apr 4 '11 at 20:47 • 240V * 20A = 4800W, 240V * 30A = 7200W. Its not that difficult, even in the US you have 240V available in your house(two 120V legs), its just not used for common outlets but rather things like your dryer or electric stove. – Mark Apr 4 '11 at 21:06 • 230V * 16A = 3680W is a standard European socket rating, and I wrote 4kW so to have some overhead and not exploit the DC-DC converter at 100% of its capability, but slightly lower. – miernik Apr 4 '11 at 21:16 • @FakeName no, I want to fast-charge a LiFePO4 battery in my car. – miernik Nov 1 '11 at 5:36 ## 5 Answers Sadly, what you are asking for is impractical for a one-off design. Someone may possibly offer you a complete DIY design, but you overall specification is so high that you would find it hard to build unless you are already expert in such matters. More on that below, but your most likely chance of success is to look for top quality surplus units that either suit your need exactly or which can be adapted relatively easily. Doing it yourself, one mistake could lead to the destruction of electronics (say main MOSFET switches) worth most of your budget. 95% efficiency at 2 kW means 100 Watts total losses. At 65 A output, for resistive losses alone to be less that 100 Watts requires I^2.R <= 100 or resistance to be under about 20 milliohm. This includes winding resistance, tracks, connector resistance and more. This is doable but with vast difficulty. But, that doesn't leave any margin for any other losses. At 28V out, 0.5 volt rectification loss will by itself result in 0.5/28 ~= 1.8% losses. A large Shottky diode MAY get down to 0.5V in that application - but probably will drop a larger voltage so lose even more than 1.8%. Synchronous rectification is about the only superior option - but even to equal a 0.5 V Schottky performance requires a genuine on resistance in your synchronous switch of R= V/I = 0.5/65 = about 7 milliohms. Again, that is doable, but you require top specification parts and vast amounts of skill and care. And reasonable quality batteries (presumably deep cycle and presumably rated for some hours of operation at this power level) will cost you far more than your inverter budget. Overall, a more realistic approach is to gain some experience by buying the best surplus equipment that you can get for your budget and seeing how well it works, and investigating where the main losses are and what may be able to be done to reduce them. You should be able to buy surplus equipment for far less than you can build something equivalent yourself. If you are capable of building something like this and adequately understanding what needs to be done, then you could buy surplus equipment that has been damaged and rebuild or repair it. If that sounds too hard then doing it yourself from the start is liable to be harder. • The batteries won't be so large and expensive as you expect, because the energy will go to A123 LiFePO4 cells which can accept charge at 4C, so only 16Ah of them are needed to max out a 65A supply, and 1Ah 18650 cells are now 2 USD each at 3.3V, so only 128 of them (at a total cost of 256 USD) are needed to max out the mentioned Vmax=28.8VDC, Imax=65A. So the batteries won't cost more than my inverter budget. – miernik Mar 29 '11 at 5:57 • That's an encouraging cost ifthose are A123 cells at that price. However, battery capacity required depends on how long you want to charge them for. As I noted ... presumably rated for some hours of operation at this power level. The A123 array that you mention will charge in about 15 minutes. To provide one hour of storage would cost about$US1000 :-(> – Russell McMahon Mar 29 '11 at 15:39
• 0.5V rectification drop at 28V is a very conservative estimate. Ultrafast rectifiers will be higher than this. Synchronous rectification is the right approach but at voltages >12V you're outside the 'sweet spot' where the silicon has very low Rdson. – Adam Lawrence Mar 29 '11 at 16:40
• I modified my expectations, see edit (I will use a 64S3P A123 battery pack - cells will cost total 400$, charger should not cost more), does that make it easier to build? – miernik Apr 3 '11 at 17:03 Designing a power supply is not a trivial task. I've spent more than 10 years in the employ of professional companies which do this sort of work. Some thoughts: 1. Input-side - that's huge power you'll be drawing from the mains. I'd estimate that you'll need at least 20A service. You'll also need some hefty EMI filtering to avoid polluting the mains with the switching noise from the converter. Don't forget about inrush limiting, surge suppression, fuse sizing and X-capacitor discharge circuitry. 2. PFC - Most designs need some sort of power-factor correction to ensure that sinusoidal current is being drawn from the mains. You'll most likely need a multiphase PFC to handle this sort of current effectively. 3. DC/DC converter: You'll definitely need a soft-switching topology to achieve any reasonable sort of efficiency on the primary side. Look into the zero-voltage-transition full-bridge converter, aka phase-shift full-bridge (Intersil, Texas Instruments and others make controllers for this sort of topology) and look into really rugged MOSFETS (I've used IRFPS40N50Ls for 3kW designs). You'll need really good transformer core material (consider Ferroxcube or Nicera) to keep the losses low. You may want to even consider running two 2kW converters, and sum the currents on the secondary-side with a current sharing circuit. 4. Rectification: There isn't an efficient way to rectify such high voltage. You won't be able to take advantage of synchronous rectifiation (in my estimation) so you'll be burning power in Schottky diodes. 5. Thermal management: You need to ensure that your magnetics and switching devices are all in their safe operating areas, and design heatsinks / install fans to ensure that they all stay out of any potential thermal runaway conditions. 6. Protections: Over voltage, over current, over temperature, short-circuit, line surge, ESD, EFT ... all things that you need to design protection against. 7. Regulatory / Safety: You'll need to ensure that you meet proper creepages and clearences, that your touch current is safe, that all the critical magnetics and power train devices do not result in the power supply becoming unsafe during any single abnormal event, that your thermal management is keeping the parts within their safe operating areas, that you're using regulatory-approved parts for safety-critical functions (X- and Y-caps, MOVs, optoisolators, etc.) Are you sure you really want to try and design such a thing yourself? There are books out there that explain how to design, but becoming proficient is a life's work in itself. • I would be more then happy to buy the device instead of designing it myself, provided 1) its at least a bit "open hardware" so I get a schematic of what's inside and not just a black-box that I would need to reverse-engineer to get it repaired (and don't tell me the word "warranty" - I want the thing I spend so much money on to work much longer then any warranty) and 2) it is at least a bit modular, so if my desired output voltage changes a bit, for example I decide to use 172.8V and not 230.4V I can swap a few components instead of buying a whole new device. Any suggestions? – miernik Apr 4 '11 at 22:23 • You'll not find many off-the-shelf supplies that offer full schematics these days. The return/repair business is too lucrative for the manufacturers. Back in the 80s and 90s, you could get full service manuals and schematics for HP power supplies, loads, etc. ... try getting the same information for any new Agilent product today. At my employer we use Elgar SW series sources up to 10kW - they can deliver AC as well as DC. They're not cheap but they're rugged (we abuse them badly) and are still supported. Datasheet – Adam Lawrence Apr 5 '11 at 12:52 • The company I work for builds a charger that seems to meet your specifications, excepting isolation, which can be had via an easily-available (if large) 60 Hz transformer. Bonitron M3528AC-L020. bonitron.com/m3528.html Voltage and current limits are adjustable by potentiometers. You won't get the schematics, though. If one dies and you want repairs, you'd have to ship it back to us. – Stephen Collings Sep 12 '12 at 21:17 I'd tie four >1000W computer power supplies together, and add a PIC or (insert-favourite-microcontroller-here) to control the CC/CV stuff. A computer power supply can run off high voltage DC, and it can be made adjustable by modifying the feedback loop. A microcontroller could modify the feedback resistors by using a digital potentiometer and it could monitor the current through a hall effect current sensor. However, 95% efficiency is going to be next to impossible to achieve. • This sounds like a recipie for disaster! – BullBoyShoes Apr 4 '11 at 20:48 For very high power 98-99% efficiency is very possible to achieve with motor-generator. May be for power levels of few KW 90% is possible as well. With good motor-generator and some rotating mass you can get very clean and safe design. You won't do it for a few hundred$ 200VDC input would be OK for most universal-input mains SMPS units (if it doesn't confuse the PFC circuitry) so you might luck out on a surplus unit.