7KW 3phase PFC

[ali_asadzadeh]

Hi,
This is my first time entering this SMPS power range, I wanted to know the existing 3 phase PFC solutions, any hints and Ideas are welcomed! 

Currently I have done a 2KW single phase PFC both with a Cortex M0 for the main controller and an UC3854 as a pure analog design. do we have any available parts and app-notes? and open circuits?

[ocset]

This is DIY territory only.

If i was you i woudl do a single phase pfc of 7/3 kw off each phase........then have an isolated supply off each one......then bring them together.........eg if its a battery charger, have the three isolated psu's putting one third of the total current into the load.

You may have to do interleaved pfc for  each of the 2.3kw pfc stages.........its good to have two channels.....and one conducts in one 10ms period, then the other one,  etc etc..ie, they take alternate mains half cycles to share the thermal.....this is good but you need two boost pfc inductors.

[ali_asadzadeh]

Thanks, it was one strategy that I have came along too if there was no better way, But I think maybe there is better and more clever way of doing it too.

[BNElecEng]

I'd suggest you try to find a used Meanwell RST-5000 or RST-10000. They've already solved this problem very neatly.

Alternatively, look up information on the Vienna rectifier or the Swiss rectifier. The choice will depend on whether you want step up or step down operation.

[CM800]

I'd suggest you try to find a used Meanwell RST-5000 or RST-10000. They've already solved this problem very neatly.

Alternatively, look up information on the Vienna rectifier or the Swiss rectifier. The choice will depend on whether you want step up or step down operation.

Those look nice... It's a pity they don't do PFC modules like that on their own... I could do with a unit that just gives out the rectified, PFC mains for 10kW or so. 400-800VDC.

[Warhawk]

with this hype of HEV/EV vehicles TI has some cool stuff. Google their web (ti.com/tidesigns)
e.g http://www.ti.com/tool/TIDM-1000

[T3sl4co1l]

I've seen the design for a large industrial PFC.  They used three single-phase channels, connected in parallel at the output.  A master-slave configuration ensures current sharing.

Tim

[oldway]

3 phases  PFC is a very different concept than 1 phase PFC.
Use a common 6 pulses rectifier and eliminate the harmonics and improve the power factor of the non linear current with a separate PFC.

[BNElecEng]

If size is not an issue, then a non isolated 12 pulse design can be used. It's extremely rugged, has low current harmonic content and the PF is very close to 1. Active circuitry can extend the operation to 24 pulse.

[oldway]

Another solution is to make three differents one phase PFC + inverter.

https://www.eetimes.com/document.asp?doc_id=1272370

[jbb]

The Vienna rectifier is quite well regarded, if a little involved to control. It’s a boost type, so you’ll get around 700 VDC out. The advantage over a classic 3ph active rectifier is that the switching losses are reduced by using 600V switching devices and the inductor volume is approximately halved (assuming same switching  frequency).

If you’re just getting your feet wet and don’t need an optimal design, I suggest the classic 3phase 6 switch inverter using IGBTs and 10-20kHz switching frequency. You can start with an L filter then move up to an LCL. The control is a bit simpler but you’ll still learn a lot.

The Swiss rectifier (this is an example of the academic sense of humour) is new to me, but Kolar et. al. are excellent, so I expect it works well.  It looks like a buck-boost type, which is interesting. However, the control will be even more complicated.

Any of the above topologies will definitely require microprocessor control.

Regarding the 3x single phase concept, it will depend a lot on your application. If you need isolation transformers already  then you can do it OK. If you don’t need isolation transformers, then be aware that combining the outputs of 3x single phase would force you to use them anyway, which is a waste of time, space and weight.

[ali_asadzadeh]

Thanks a lot guys for your useful suggestions, I think I should dig deeper into your ideas:)

But regarding that most of the power is used in the industry and the industry mainly use 3phase power, it's weird that there is no commercial IC available to solve the PFC problem

[T3sl4co1l]

Calibrate your sense of scale.   Industrial products are made in quantities of thousands, maybe millions for some things.  Not really enough to justify an IC, certainly not when the controller is a small part of the total functionality, and where existing solutions are more than sufficient for the purpose.

A single IC can't solve the challenges of isolated DC supplies, high power gate drives, isolated current sensing, isolated fault protection, and more.  An integrated module might: there are IGBT + driver + isolator + protection modules, for driving motors.  (But again, there are millions of motors out there, hence these modules are worth making.)

Plus there's a fairly strong culture of we-made-this, sometimes in pursuit of greater efficiency (e.g., a DSP or FPGA implementing a more optimal control method than the available ICs offer), but often just faulty psychology.

On the other hand, a single ASIC can handle the core functionality of, say, a PLC, which might be made in the millions.  That's a good deal right there, and indeed that's what they do.

Tim

[jbb]

Not really enough to justify an IC...

Plus there's a fairly strong culture of we-made-this...

Very true.  Plus, in this world of out-sourced everything, having your own 'secret sauce' locked in a micro controller can be very reassuring.

Playing devil's advocate, expending lots of effort to optimise your way from, say, 98.0% to 98.2% efficiency makes a 10% improvement in the thermal situation.  This could make your product smaller or cheaper or something...

So, turning back to the topology question:

• What's your input voltage range?
• What output voltage (approx.) do you need?
• Do you need isolated output? (Free choice of transformer ratio can be a big help in design optimisation)
• Do you want to be able to inject reactive power into the AC bus? (Very unusual feature!)
• Do you have size and weight restrictions? (Tradeoff vs. design effort)
• How many do you want to build? (Tradeoff design effort time cost and BOM cost)

[ali_asadzadeh]

Thanks jbb
the input voltage is 380VAC 50Hz, 3 phase

I think it would be better to get 700V DC output,
I do not need isolated output as I would add the Full bridge DC/DC converters at this stage
No I do not want to inject reactive power
For now, I do not have any limits on size and weight, but always smaller is better for me
I want to build around 150 units of them for now

I hope this would help, I was considering using separate single phase PFC's to solve the problem, if you came with a better Idea I will go that way

[jbb]

You're very welcome.  My current job is all 12V DC, so it's nice to keep my hand in with the real stuff.

For 380V RMS L-L, 700V output is quite reasonable.  This means 1200V power devices.

The input voltage is 380VAC 50Hz, 3 phase
I do not need isolated output as I would add the Full bridge DC/DC converters at this stage

If you're doing full bridge DC/DC, it sounds like you'll have an isolation stage to get the final output voltage anyway.  Therefore deploying 3x single phase units (with provision for power sharing) is not at all crazy. As both you and treez thought:

... single phase pfc ... then have an isolated supply off each one ...

I was considering using separate single phase PFC's to solve the problem

Deploying 3x single phase units does have some downsides:

• More power devices
• Reduced efficiency (assuming you use traditional diode rectifier + boost stage)
• More bus capacitance required to handle 2x line frequency ripple
• Need 3x DC/DC converters too
• Need a neutral connection

If you want to try a 3 phase design, you'll need to think about the following:

• Input EMC filter.
• Inrush current limit.  At this power level, you would typically be looking at resistors and a relay.
• Input filter.  Typically an LCL for the voltage-fed topologies
• Line voltage sense (you could try 2x diff amps, you can try for sensors later)
• Current sensors (typically 1 per phase, you can try for 1 on the DC link later)
• Switching stage (I suggest a classic 6x IGBT + 6x antiparallel diodes. Infineon FS25R12W1T4 might be along the right lines.)
• DC filter cap (likely 2x 450V electrolytics in series with bleed resistors)
• DC/DC stage

The control system will be a bit different to the usual single phase edition:

• All the usual over/under voltage/current protections (do this first  )
• You'll need a (software) phase locked loop to extract the grid phase
• Current transformations from phase domain abc to control domain
• 2-axis current controller with line voltage feedforward
• DC link voltage control
• Space Vector PWM (SVPWM) unit

My 'secret weapon' is to test the current control system with an external supply on the DC link, 0 line voltage (i.e. short circuit on input terminals), the PLL fixed to always run at 50Hz and forced steps in the current references. That way you can check the current control works before interacting with the mains.

I once connected up a system with the wrong polarity gain and then I had to replace all the IGBTs and gate drivers , so be careful at power. Also be careful at power down. And at steady state. And when you've turned it off; those DC link caps will hold lethal charge for some time. I suggest starting with big dump resistors to bring it down safely.

I expect a hunt through the Texas Instruments (or other usual suspects) reference design collection would turn up results. It's actually very similar to a 3 phase induction machine drive or solar inverter.
Here's a PLL app note.

[ali_asadzadeh]

Thanks a lot jbb, it helped a lot

[jbb]

You're welcome.

Ah hah! Paydirt! TI has a Vienna rectifier reference design.  There seems to be a good chunk of information here which could help you get started - including PCB design files - if you want to try something more compact & complicated.  If you're willing to use TI C2000 parts they also have pre-written software which might be a big help.

They are using CoolMOS MOSFETs and Silicon Carbide (SiC) diodes to get the switching frequency pretty high and the filter inductors pretty small (just L filters, not LCL).

There's also one on the six-switch AC/DC stage, but it's not up to their usual standards.  They do talk about line current control in vector form, which is a very useful concept for this field. Then, for some reason, they suggest per-phase current control, which I would not recommend because all three phases interact with each other.