1kW 200/400V dc/dc converter design tip

[webgiorgio]

Hello,
I would like to make a DC/DC converter to use it in my Photovoltaic plant, where I need to boost 200 Vdc 5A to 400 Vdc 2.5A, so that I can parallel a string of 6 PV panels with other strings of 12 panels. 1200W.
Vmpp for one panel is 33.3V (200V/6).

Is a simple boost converter with 50% fixed duty cycle enough? (inductor, N-mos to ground, diode)
Can you recommend any more energy efficient topology for this voltage and power level?
Shall I implement a check of Vi and Vo? (actually, Vo is imposed by the strings with 12 panels and the inverter).

I only need to make one, so I am not too concerned about the cost of it, it would be nice to make it efficient.

[Miyuki]

It is not a big deal, as you have low current

You can use SiC Diode and transistor if you want high efficiency / high frequency 
Not a big deal

Just simple boost converter

[Benta]

At 1 kW I'd choose a full-bridge converter with transformer coupling. The inductors in boost (this includes flyback) converters of that size tend to be unmanageable large, as they have to store the full energy.

A forward converter could also be an option.

[TrickyNekro]

A forward converter could also be an option.

Yup, I like that thought also, you could do it also center-taped style, push-pull. Which is pretty much a variation of the forward converter.

[coppercone2]

Someone should write a treatise on their experiance with magnetic elements in such smpsu of their own design. Its always been the x factor that makes me not want to venture into this area. I heard bad things about big magnetics even from seasoned university professors.

[Benta]

A forward converter could also be an option.

Yup, I like that thought also, you could do it also center-taped style, push-pull. Which is pretty much a variation of the forward converter.

At 1 kW, it's not a good option. The problem is, only one half of the primary is active at a time, meaning increased copper loss. Also, the transistors need to be rated at 2 x V_IN plus leakage inductance, meaning at least 800 V types.

[Benta]

Someone should write a treatise on their experiance with magnetic elements in such smpsu of their own design. Its always been the x factor that makes me not want to venture into this area. I heard bad things about big magnetics even from seasoned university professors.

You're absolutely right, I've done my own magnetics and it's a steep learning curve. The issue is, that there is no "closed loop" design process, it's purely iterative. Good data books, like the old Siemens "Soft Ferrites", are big help, but not easy to source these days.

[coppercone2]

Yes its incredibly inconveniant to get into that field.

[Mazo]

Standard boost topology but spread out to 2,3,4 or more phases looks like a compromise between custom magnetics and unmanageable inductor size?

[Miyuki]

Of the shelf inductor 1000uH; 8A; 155mΩ (0.3A ripple)
Simple 100kHz boost
Output capacitor rms current 2.5A
Transistor rms current 3.6A peak 5.5A

It is not a big deal for parts, no need for something more complicated in no need for isolation
Even can lower frequency as ripple is low
Just take care of board layout

[jbb]

I'd like to muddy the waters 

As you're doing this for a PV installation, a fixed gain of 2 is probably sufficient.  After all, the PV inverter will have Maximum Power Point Tracking (MPPT), which could interact with the controller for a boost converter in strange ways.

So how about a switched capacitor voltage doubler?  That offers a fixed voltage ratio and doesn't need fancy control or large inductors.  (Texas Instruments has an introductory post...)  Switched capacitors are quite efficient when you want an exact voltage ratio and can be made in kW sizes.

Note: some small EMI filter inductors are likely required on the input and output to prevent stray radio interference.

In terms of magnetic converters:

At 1 kW I'd choose a full-bridge converter with transformer coupling. The inductors in boost (this includes flyback) converters of that size tend to be unmanageable large, as they have to store the full energy.

A forward converter could also be an option.

I like isolation as much as the next person, but wouldn't designing the transformer (+ filter inductor for forward converter) be more difficult than designing an (admittedly large) boost inductor?

Here's what I would consider for a boost converter:

• Tight PCB layout is critical.
• As Mazo suggested, interleaving is helpful.  Currents aren't that high, and duty cycle will be 50% (ish!), so I'm thinking 2 phases.  This will help a lot with capacitor selection.
• Using a SiC MOSFET (or Gallium Nitride, GaN) + SiC Schottky diode to get high frequency; frequency around 100 kHz
• If efficiency is critical, replace SiC Schottky with another SiC MOSFET for improved efficiency (extra for experts)
• This yields inductor values in the order of 500 - 800 uH, which you might be able to hit with something from the WE-HCF range (Wuerth Electronics)

[Benta]

I like isolation as much as the next person, but wouldn't designing the transformer (+ filter inductor for forward converter) be more difficult than designing an (admittedly large) boost inductor?

In my opinion: no.
The transformer design for full (and half) bridge circuits is straightforward. The same goes for forward converters. The filter inductor can be bought off-shelf.

[webgiorgio]

I looked up the voltage doubler, the idea was good at first sight, but then I saw it takes too many switches.

To my little knowledge (I am electrical engineer, not electronic) a boost converter is totally fine with a voltage ratio equal 2, which can be done without use of transformers. For this voltage ratio, any topology with a transformer will be less efficient than a single inductor because there is twice the copper and not all the flux is concatenating with the secondary winding.

About the inductor for a boost converter, I looked on Mouser and it is possible to buy for about 10-15 €. Ok.
Out of curiosity I measured one from a computer power supply. About 500 uH. Wire is 1.32 mm2. So, I would make the prototype with two of those inductors in series. The core is green, with blue side. I am not sure how much flux can take before saturation, but if the wire is fine with the current, I guess that the core is fine with the flux.

For higher efficiency, the use of a transistor instead of a mosfet came like a surprise to me. Why is it more efficient? Converters use mosfets most of the time...

[Amper]

Just for comparison so you see that the problem ist not very difficult here is a picture of a 1975ish laser power supply. Its a Buck converter using only bipolar transistors, two or three OPs and one NE555 for generation of the switching signal. It will generate  200V 30A from 400V three phase.  Frequency was around 30khz through an iron core inductor.

[coppercone2]

what efficiency does it have?

The zip-tie density is interesting. I thought that would be too much for terrestrial use. I woulda thought 1/4 of that would be good and 1/3 more then enough.

[Amper]

The ziptie use is a trademark of equipment from this period, archeologists of the future will use this fact to date back machinery they found in the ruins of the old world.

I dont know the exact efficiency but the black heat sink you can see is just a bent piece of 4mm sheet metal, no fins attached anywhere so the loss cant be to great or it would overheat.

[coppercone2]

i wonder why though?

I HATE repairing stuff like that. The only thing worse is wire nylon lashing with wire that is soldered to the PCB. IMO its like avionics grade. https://en.wikipedia.org/wiki/Cable_lacing

I don't even know why. I saw a HP supply roll down a hill on a steep road bouncing up and down (extreme abuse due to flying off a car roof). The only weak point was the card edge connectors and the thinner metal on the transformer bracket. It must be related to weird vibrational frequencies. That won't help with gross impact IMO. If you made the transformer bracket 1mm thicker and put some screws on the cards (rather then just press fit into the connectors) it would be perfectly fine. I always wondered if you put some steel zip ties around the card edge connectors if they would hold it without popping the ends off.

Even if you had the wires all super lashed your gonna probobly resolder it anyway if it takes a really stiff beating.

Does this have something to do with automotive/turbine vibration ?

Laser labs typically have seismic provisions..

Does anyone know how the vibration changes when its lashed together that tight?

[Amper]

I guess it was just easier to put more than to think where to place them. Its a pretty dense cable tree in there with thick wiring that is usually never replaced or modified. They may even have pre manufactured the cable tree externally on a fixture and kept in shape by ziptying the living crap out of it. Some times its even useful when you take out some assembly, take off all the cables and for reassembly you just need o connect them in the place they point to because they are fixed in position that nicely.

[Benta]

Yep, those wiring harnesses were made on a fixture and mounted as a complete assembly at the end. Standard practice in the 70s...80s

[Amper]

Yaaay, guessed right 

[David Hess]

I would not use a switching converter at all unless I had to have a regulated output.  Simpler would be a high frequency inverter to just double the input voltage.

[nctnico]

Hello,
I would like to make a DC/DC converter to use it in my Photovoltaic plant, where I need to boost 200 Vdc 5A to 400 Vdc 2.5A, so that I can parallel a string of 6 PV panels with other strings of 12 panels. 1200W.
Vmpp for one panel is 33.3V (200V/6).

I don't think this will work. You'll need to mimic the maximum power point tracking (mppt) as well. I'd create 3 strings of 6 panels each. Preferably each on their own inverter DC input channel to maximise mppt effectiveness.

[webgiorgio]

A string of 6 panels would not have enough voltage in relation to the input range of the inverter (333V-500V).

I think that the inverter does a good job to make the 12 panels operate at the voltage that corresponds to the maximum power point.
If my DC-DC keeps Vin=Vout/2, this Vin corresponds to the MPP of a series of 6 panels.
I neglected the losses, as I expect about 5W out of 1000W, so 0.5%.

[Benta]

I neglected the losses, as I expect about 5W out of 1000W, so 0.5%.

Are you talking about the losses in the converter?
If so, that's totally unrealistic, it corresponds to an efficiency of 99.5%.
Expect a power loss of 50...100 W.

[webgiorgio]

I meant losses in my dc-dc converter.
Mhh, you are right, 0,5% losses sounds too optimistic.

Using this Sic mosfet C3M0280090D
https://www.wolfspeed.com/media/downloads/825/C3M0280090D.pdf

Rdson=0,28 ohm, switch losses are 50 uJ per period.
At 2.5A average current:
Conduction losses= 0,28*2,5²=1,75W
Switching losses= 50*0.100=5W (from graph 23)

I hanven't look up a transistor, which was suggested as more efficient than a mosfet. Can it be?

This diode STPSC1006D
https://www.mouser.com/datasheet/2/389/stpsc1006-956763.pdf
would have 3,56W losses at 2.5A

If the inductor has 150 mOhm, dissipates 1W.

13,3 W so far. So, 1.33% losses. I still have to quantify the losses in the gate driver and capacitors.
I could also reduce the switching frequency and increase the inductor to reduce the switching losses.

The whole 7 kW PV inverter has 2-3% losses to do MPPT and AC conversion (according to the datasheet, it is a SMA SMC 7000 TL)
I hope that my simple boost converter can have around 2% losses.

Can I reduce the switching frequency at partial load?