1kW Boost

1W Boost Post #1
I am evaluating some solutions to have a step-up for a 1kW three phase inverter.
I have an input mains voltage of 165-265Vrms and I need a "stabilizied" at least 300Vdc for the inverter or better I could accept the low frequency ripple voltage of a simple bridge but I need at least to increase of 100V when I am at 165Vrms.
I prefer to have higher efficiency than better regulated output.
I do not need a PFC controll so I can have a direct feedback from the optput voltage.
At the moment there are two solutions in my mind:

-Low frequency Boost
a simple boost where the switching frequency could be not higher that 1kHz and the switch is driven for short time only some times before the normal charging of the bridge to have the requested 100V increase; the feedback is realized with the DSP used for the inverter, having a sensing of the ouput voltage and perhaps a sensing of the switch current.

-High frequency Boost with ZVT
I see that for this power many projects use the ZVT method in order to reduce the switching power dissipation allowing use of high frequency like e.g. 100kHz. This solution allow to have higher efficiency, better regulated output a small magnetic component but I need to pay attention of frequency ripple voltage that can disturb the current sense of the inverter.

Anyone have suggestions? Are there other interesting solutions?
Thank to anyone can help me, i do not want ready solutions but only some suggestions and possibilities to discuss about.

Allanon77
 

allanon77

Member
2004-08-04 8:42 pm
Italy
OK the simulation of Low frequency version gave bad results: it is needed at least 33mH (16Apeak 9Arms) to achive the target with only one commutation every half wave. Having more commutations during an half wave probably should reduce the value of inductor needed but I think we are far from reasonable.
No comments?

Allanon77
 
I found a very interesting thread (2005) where ChocoHolic, Eva and N-Channel gave their best :D
I started to think to a Boost but, sincerely, at the end it is better to have a PFC because I do not need the Input inductor and I am not worried about the LF ripple voltage at the output (obviously in a range of 10-20V).
So I change the target to:

1kW PFC
165-265Vrms
low cost and compact solution

So I found an interesting post http://www.diyaudio.com/forums/showthread.php?s=&threadid=54440&perpage=25&pagenumber=1
....
I will be back after a deep immersion on that, probably with many questions :)

allanon77
 

allanon77

Member
2004-08-04 8:42 pm
Italy
I feel alone :bawling:

However if someone is reading this thread, I am starting the simulation with spice but I have not the controller model.
Probably in the Forum someone had the same problem with his/her project, how do you simulate PFC without IC model? do PFC controller models exist?
Thank you in advance

Allnon77
 

fernando_g

Member
2007-11-11 2:48 am
I've been reading the thread, but had not been able to respond. At least not with an intelligent suggestion.

My suggestion would be to use, as you have correctly figured out yourself, a PFC-corrected circuit.

Although there are many vendors for those, I would suggest to use a controller from Texas Instruments (formerly Unitrode), as they have literally dozens of devices and dozens of very thorougly explained application notes. I'm sure that they will have SPICE models for that.

One final suggestion: because you require such a high power level, I believe that you must operate in Continuous Conduction Mode. Otherwise your peak Mosfet and diode currents will be very high. Having said that, CCM has a major disadvantage with the diode recovery time. You may want to consider Si-C diodes from Ixys or Cree.
 

allanon77

Member
2004-08-04 8:42 pm
Italy
Thank you, the first contribution!

I agree with you, I need CCM it is impossible with DCM to reach so high power level without having too high Current Peak.

I know that I have to take care about reverse recovery peak because I made in the past a prototype of PFC with NCP1650 with a Mosfet and a very fast diode, however I had a very strong recovery peak. Is it true that I have to consider the energy of recovery to be dissipated mostly by the switch?

Talking about PSPICE model, unfortunately I have not found any model also in Texas Instruments site (I have also the complete collection of Unitrode seminaries); I have not found any kind of PFC IC model nor Average models neither Transition models.

I would like to use IGBT instead of MOSFET and for this reason I am studying Zero Voltage Transition PFC that allow to use high frequency without having high switching dissipation; the other option that I am evaluating is a "lower" frequency (40kHz) with IGBT and a lossless snubber for diode recovery peak (solution that I have seen from Eva in the thread I mentioned before).

Allanon77

P.S. I see a "lapsus langue" in the last post, "I do not need the input inductor .." obviously it should be capacitor instead of inductor ;)
 

fernando_g

Member
2007-11-11 2:48 am
allanon77 said:
I would like to use IGBT instead of MOSFET and for this reason I am studying Zero Voltage Transition PFC that allow to use high frequency without having high switching dissipation;

I have seen IGBT-Mosfet "hybrids" in high power PFC circuits. Essentially, the IGBT carries the most of the current, and the Mosfet is only used to carry the current during the IGBT's turnoff phase. As you know IGBTs turn on reasonabily fast, but turn off slowly, and that is where all the losses are. The Mosfet can really improve the turn-off period.

What type of lossless snubber are you planning to use?
 
I cannot answer ... I saw a magnetic snubber in the thread I mentioned before but to have a first evaluation of different solutions I need to simulate with Spice because I have no time to build every solution with breadboard ... this for the first selection, at the end with the best solutions I will surely build a prototype.
If I do not find any models probably I have to write ones ... very simple only to close the feedback loop but ..
CONVERGENCE ... the nightmare of every switching simulation!

Allanon77
 
I'm using the IGBT and MOSFET approach and the magnetic snubber too and it produces up to 5KW at 230V AC input with quite good efficiency (2.5KW at 120V AC).

For 1KW you can consider an hyperfast diode like 15ETX06 from IR and chose an inductance and a switching frequency that result in discontinuous mode during part of the time (say up to 500W).

My favourite PFC controller is L4981 from ST.

Don't try to simulate the switching and the control loop at the same time. You can get quite good results by simulating control loops with ideal power stages instead. Forget about simulation to evaluate switching performance. Also, you don't need any control IC models, you can model the control loop with ideal elements quite easily.

BTW: Don't forget the isolation transformers and the 100Mhz oscilloscope...
 
Thank you Eva.
I have experience on smaller SMPS so I know how to move (I have isolation transformer and also oscilloscope).
I have no experience with high power SMPS (the bieggest at the moment is a Flyback 50W).
I know that is difficult to simulate, i know that i cannot take account of all parasit element to eximate switching power but I use simulation to have the first look to a solution and to have the first selection of values for all component.
I want to try PSIM that is born for SMPS simulation, have you ever tried it? Does it work?
It is strange ... the olny Diode I have in hand are 15ETX06 but I searh the datasheet on IR web site and they discontinued product :eek:
Infact in the PFC with NCP1650 I used one of this diode.
Which kind of IGBT do you suggest?

Why do you prefer L4981?, it seems to be an old IC NCP1650 seem to have more features for overpower protection ....

Thanks a lot

Allanon77
 
It is true I read too fastly! OK it seems OK for my PFC I have already tested but I have also seen the peak recovery that is high.
Talking about snubber I am studying the solution proposed by Eva in the post mentioned before, is it with saturable core? unfortunately I have not time to study yet :(

Allanon77
 
notes:
ZVS/ZCS are ok, but have issues of their own. regenerative snubbers and full resonant systems can basically have losses associated with more classic "power factor" issues -- eg, some energy is transfered from the supply then back to the source, with ohmic losses on the round trip.

interleaving can also reduce peaks.

CCM is slower and harder to control, and responds worse to peak current mode control.

There is an ESBT and equivilents, basically, a low voltage (20V) fet is placed at the emitter of a BJT (maybe an IGBT). the fet can break current quickly.



there are many other topolgies that can give a voltage gain, and boost derived designs. you can place an inductor in series with a full-bridge topology (without output inductor). In this design, having all 4 FETs on will provide a short, charging the inductor. turning off two fets will allow the current to flow through the transformer -- an isolated boost design. there are a few others -- two inductor boost for one, that can work.