Help with power electronics

[MarkoAnte]

Hi,
I'm looking to start a project where I would require a adjustable power supply from around ~100 V to ~1KV (from ~600V 3 phase rectified). The upper voltage limit is 1.2 KV where the transistor will crap out. The prototype converter should be able to handle ~20 KW. The ripple is not important (tho I would not go for a thyristor + transformation approach), however the efficiency is. Also the polarity of the output is not important and it doesn't have to be isolated. I have worked with a buck converter ~ 15KW before.
My questions are:
- Any body know any good books on the subject. I have found some linear/ analgo devices datasheets for design of ~15V at ~2 A switchers. I have a feeling that it does not scale up to 20 KW  Maybe some /wide/ output voltage/efficiency graphs.
- I'm looking at the buck-boost converter and SEPIC converter. Does anybody know of any other topologies that I should take a look at?
- Would I be better of first using a 60 Hz transformer to boost to ~1KV and then just use a buck for down converter? (cost/efficiency)
- I'm also looking at a  isolated full-bridge topology, but I don't think I can achieve the voltage swing I need (the efficiency would be bad and ripple would be large)- right?
- Would building a scaled down test model help me at all? With the buck I have build, I had a problem that only appeared at 600V impute voltage and I fixed it by rewire the power part of the electronics.(Got ride of some interference.)

Any thoughts are appreciated.

P.S. All above voltages are DC.

[oldway]

3 phase breaker / 3 Phase variac / 3 phase transformer with fuses / 3 phase bridge rectifier

[MarkoAnte]

Thats a lot of copper and iron = expensive, but the main problem is that I need to adjust the output DC voltage quickly as part of regulation. So a switch mode power supply is a must.

[T3sl4co1l]

I'd suggest,

Input PFC (optional)
600-700VDC link (or ~800 if PFC'd)
Multiphase Forward Converter
Average current mode control
Each phase designed as a transconductance stage, so they can be paralleled easily.
Voltage error amplifier
Adjustable current/voltage limit (optional)
Assorted safety/management/startup/interface controls

For each phase: IGBT H-bridge, probably phase shift PWM, or a quasi/resonant configuration (e.g., LLC)
Consider using snubbers to save on switching loss and EMI
Quasi-resonant / lossless snubbers may be of interest as well

For extreme efficiency, use huge MOSFETs (CoolMOS or SiC) and synchronous rectification.  For even extremer: consider inductive gate drive ("lossless", minus internal gate resistance that is).

I'd suggest making a module of modest capacity (2-5kW?) and stacking them.  Bring along a sync signal, so they can be phase shifted, just like cylinders in a V8 engine (or whatever).

Good luck, and don't kill anyone.  Arc flash is nasty.  Respect it.

Tim

[oldway]

Thats a lot of copper and iron = expensive, but the main problem is that I need to adjust the output DC voltage quickly as part of regulation. So a switch mode power supply is a must.

The first thing to do before you start a project is to give the product specifications.
Your description contains no serious specification.

NB: if isolation is not required, transformer may be a low power auto-transformer to increase voltage up to DC 1KV (rectified)
It will be a low impedance power supply and regulation is perhaps not needed.
You can use a 3 phase motorized variac to make the regulation.
The great advantages of this solution are its simplicity and reliability
I have used a motorized three-phase variac of 120kVA for testing transformers and busbar.

1KV 20KW variable output power supply is a professional project.
If you need to ask help on a forum, you better not to try to deal with such voltages and power 

From another topic "Hi, what would you suggest for a first Spectrum Analyzer"
You said:

I'm a student and a hobbyist...

That's not a project for a student nor a hobbyist, it is far too dangerous !

 

[MarkoAnte]

@T3sl4co1l
PFC would be nice, maybe as a later upgrade. I afraid I don't follow you completely with the Multiphase Forward Converter. Is this what you meant?

Do you mean I should make (e.g.) 4 5 KW forward converters, with the transformer designed so that at ~90% duty cycle it would give 1KV. Then parallel them, phase shift their PWM so that the ripple is improved whilst making sure whit current sensing that all are equally loaded?

How do you make a forward converter with a H-bridge? Wouldn't that just make it a h-bridge converter (or a resonant, zero voltage || zero current switched converter).

@oldway
I did not give serous specification because I don't have the whole project worked out 100%. And a lot can happen in 2 years. Finished my studys, worked for 1 year. When I worked I made a induction heater from 0 and I have worked whit multiply Mhz induction heaters (using power tubes rated @12KV and @200KW) and developed a process of matching impedance of 3 different LC resonant circuits to minimise heating of components. So 1KV @20 KW. 

It doesn't matter if its 600V at 0.5KW or 1000V at 20KW. You have to respect it the same way. E.G. Powering it up with a variac, form a safe distance after you have finished a voodoo dance to the goods of power electronics. 

[oldway]

I did not give serous specification because I don't have the whole project worked out 100%.

This is nonsense...You should at least have previous specifications...Or you don't know what project it is ?

It doesn't matter if its 600V at 0.5KW or 1000V at 20KW.

Yes, it does matter...Both are dangerous but there are things that happen with high energy circuits that does not happen with lower one.

Questions you asked in your first post prove that you don't have knowledge nor experience to deal with such a project...but after all, if you want to risk to kill yourself, that's your problem.

[MarkoAnte]

Don't worry I put my safety squints on and have my dear old mother on speed dial. So its k.

[oldway]

Don't worry I put my safety squints on and have my dear old mother on speed dial. So its k.

You have been advised, in my opinion, you are irresponsible and unconscious.

At a minimum, you should specify:
- 20KW power is the maximum power (1000V 20A to 100V 20A) or it is the power for all output voltage levels (1000V 20A to 200A 100V)?
- Current limiting? Adjustable?
- Control: Static: How much %    Dynamic: special requirements?
- Minimum eficiency  at full power ?
- Mains current: special requirements? Power factor ?, harmonics ?
- Dimensions? special requirements?

Why not use a SCR 6 pulses bridge with an auto-transformer ? It would be a very simple project.

[MarkoAnte]

- 20KW power at all output voltage levels.
- Adjustable voltage, current limited only as a fault condition.
- Adjustable voltage form 100V to 1KV. No special requirements on ripple on the current or voltage. Also the speed of regulation is not super important but it has to be faster then a motor controlled variac.
- Minimum eficiency  at full power - hard to define as depending on the conditions it may draw 20KW at 100V or at 1KV. So that the efficiently  will be all over the place.
- Mains current no issue will put a mains filter in before the converter.

I don't like the idea of a 6 pulsed bridge as I have made it and the efficiently and the ripple on the current or voltage are terrible at eg 100V. Also to filter it you need big metal chokes whit large caps. With the buck converter I build I was way happier.

[oldway]

20KW power at all output voltage levels.

That make the project almost impossible to realize with SMPS technology, as output circuits will have to be sized for 200A !
Ferrite transformers and inductors will be very big.
I would do this with a 3 Phase transformer with primary triangle, secundary hexaphase (or zig zag) with neutral with 3 taps and 3 x 6 SCR's all with commun cathode.
Lower Taps and 6 associated SCR's would be phase controlled from 100 to 400V with 200A current limit, the mid taps and 6 associated SCRs would be phase controlled from 400V to 700V with 50A current limit and the higher taps with associated 6 SCR's would be controlled from 700V to 1000V with 29A current limit.
For protecting primary (as power of 20KW may be exceeded in some conditions) I would provide a 3 phase thermal trip in the primary.
I do not recomand to make a not isolated power supply with such a high voltage.
Output waveform will stay very good even with 100V.
When SCR's from mid tap begin conducting, scr's from the lower tap stay fully conducting, and when the scr's from higher tap begin conducting, all other scr's stay fully conducting.
Current information coming from TC's.
Lower tap: 3 TC's 200/1A (each TC with 2 Wires from 180° phases passing in opposite direction to generate AC information)
Mid tap : 3 TC's 50/1A
higher tap : 3 TC's 50/1A

[DanielS]

20KW power at all output voltage levels.

That make the project almost impossible to realize with SMPS technology, as output circuits will have to be sized for 200A !

Hence the other posters' suggestion to make it a load-sharing modular design with each module providing a much more manageable 40A or so.

[MarkoAnte]

@Oldway
I have never seen something like that. Sounds elegant, but I'm not quite sure how to draw a schematic from your description.

Shematic of 1 phase
http://shrani.si/f/14/rT/2TOLxyXg/6phase.png

In the 1 options all SCR have the common cathode. But at the higest tap the bottom 2 scr should be off?
The 2 options is just wrong isn't it?

[oldway]

Hence the other posters' suggestion to make it a load-sharing modular design with each module providing a much more manageable 40A or so.

But I would like to see how he would make a SMPS with variable output voltage from 100V to 1KV even with only 40A....

[T3sl4co1l]

@T3sl4co1l
PFC would be nice, maybe as a later upgrade. I afraid I don't follow you completely with the Multiphase Forward Converter. Is this what you meant?

Do you mean I should make (e.g.) 4 5 KW forward converters, with the transformer designed so that at ~90% duty cycle it would give 1KV. Then parallel them, phase shift their PWM so that the ripple is improved whilst making sure whit current sensing that all are equally loaded?

Yes!

How do you make a forward converter with a H-bridge? Wouldn't that just make it a h-bridge converter (or a resonant, zero voltage || zero current switched converter).

H-bridge isn't a complete converter, it's a kind of inverter.  It becomes a converter when paired with another on the secondary side, whether passive (diodes) or active (synchronous rectification).

Half bridge could be used as well.  I wouldn't suggest other topologies such as push-pull (peak voltage too high) or the current mode equivalents (requires constant current driver).

Speaking of, it might be worthwhile to consider a rectifier that's a different style than the primary side inverter.  (That is, H-bridge in diodes == FWB.)  Full-wave doubler would be good, but has the problem of being a cap-input filter.  This would require an LLC resonant, or something like that.

@oldway
I did not give serous specification because I don't have the whole project worked out 100%. And a lot can happen in 2 years. Finished my studys, worked for 1 year. When I worked I made a induction heater from 0 and I have worked whit multiply Mhz induction heaters (using power tubes rated @12KV and @200KW) and developed a process of matching impedance of 3 different LC resonant circuits to minimise heating of components. So 1KV @20 KW. 

Perhaps you could tell us a bit about the 100V end of the problem, and what it's powering.  I'm guessing it's not powering a tube oscillator, at least.

Tim

[oldway]

In the 1 options all SCR have the common cathode. But at the higest tap the bottom 2 scr should be off?
The 2 options is just wrong isn't it?

Not at all...It does not matter if the SCR's of the lower tap are conducting or not when the higher tap scr's are beginning conducting as the middle tap scr's are allready fully conducting...but for easier control circuits, it is a better solution.
No need of any king of logic, only the control voltage and a voltage divider.
Each scr's tap begin to conduct from a voltage level of the control voltage.

[MarkoAnte]

Its is a HF induction heater project. Where I would always work in near resonance and adjust the power via DC link. I have made a prototype that works form ~ 100V to ~ 550V adjusted with a buck converter. And I got nice results with it. The problems comes if you want to make it universal and it has to run at resonance with an empty inductor + if you make inductor a funny shape (required for e.g. heat treating a funny part). In testing the existing design I have found that the voltage span is kinda really needed.

So the 1 option is correct. I'm not following you with the voltage divider. The only SCR I have seen were in a triac configuration on the primary of a 3 phase transformer and were being triggered form with a adjustable alpha so the secondary whent form 0 to 15 KV.

[T3sl4co1l]

Oh, is that all?

It's my opinion that a wide range, fixed tuned system is a dumb idea.  For, well, basically the entire purpose of this thread, right?  For all the kVA of excess inverter and supply capacity you need, you might as well direct drive the inductor (which isn't an exaggeration for about half the operating range!).

I normally figure on a factor of 2 (total, or +/- to each side), for a modest 2-4x increase in inverter/supply capacity.  This directly determines the tap spacing, and the center of the desired operating range (frequency, inductance and Q) determines how many taps are needed, and where.

It's way the fuck easier to wind a transformer, and tell the user to change a bolt and lug every so often, than to design and build, and attempt to sell, all that extra capacity.

(Speaking of -- I've worked on a direct-drive supply before.  The added selling point there is, not only is tuning kind of out of the picture to start with, but frequency -- and overtones -- can be adjusted in real time to control heating depth and pattern.)

Tim

[oldway]

If you want to have 200A at 100V and 20A at 1KV, you absolutely need to use a transformer with taps.
Primary control with triacs is not possible in this case...
NB: nor the solution of the variac !
Everyone who has field experience with power electronics will say to you that phase primary control of transformers is not a reliable solution (risk of saturation of the transformer by DC current), even worse if you use triacs as they have a low dV/dT.

[MarkoAnte]

Everyone who has field experience with power electronics will say to you that phase primary control of transformers is not a reliable solution (risk of saturation of the transformer by DC current), even worse if you use triacs as they have a low dV/dT.

That is why I didn't like the SCR suggestion at first.

@ T3sl4co1l
I don't understand what you are trying to say.

fixed tuned ... direct drive the inductor ....  tuning kind of out of the picture to start with....

All the heaters I was followed the resonant frequency some more real time then others. Then the power was adjusted with frequency or with pwm of the main frequancy.

[oldway]

This problem of risk of saturation does not occur when phase control is in the secondary of the transformer as I suggest.

[MarkoAnte]

What about a ball park estimate:
- off efficiency?
- transformer size?
- cost compared to SMPS?

[oldway]

What about a ball park estimate:
- off efficiency?
- transformer size?
- cost compared to SMPS?

What do you mean with "off efficiency"...With no load, efficiency is allways = 0 ...With full load, you loose 2V on SCR, that's 2% with 100V...Efficiency of the transformer is high, let's say more than 90% (97% for a high quality transformer)
Transformer size: more or less 50KVA's 3 phase transformer.
I don't think it is possible to make such a project with SMPS , then no way to compare the cost.

[MarkoAnte]

@oldway
Ty for the insight. I will think about it and simulate it.

*off was a type. estimate of efficiency.

[DanielS]

But I would like to see how he would make a SMPS with variable output voltage from 100V to 1KV even with only 40A....

With a modular design, there is the option of series/parallel switching and each module only needs to be variable enough to cover the range between interconnect configuration changes. For a six modules configuration, you get 166V max per module at 1kV with the six modules in series, at 83V while still in 6x series for 500V, each module needs to provide 40A max. After switching in three pairs in series, each module is back to 166V but can now back off to 20A each. At 333V in the three pairs topology, each module needs to provide 30A at 111V which becomes 20A at 166V after switching to two triplets in series. At 166V in the serial triplets topology, we are at 40A and 83V per module which becomes 20A at 166V in full parallel configuration. At the final 100V/200A, each module only needs to provide 33A.

There, a much more manageable 2:1 range for individual modules to cope with.

Since his application is induction heating, there should not be much of an issue with power having to cut off for a few milliseconds while relays or other mechanical switches are being reconfigured. Or just put a bunch of high-voltage MOSFETs in parallel for shorter dead-time between range changes.