1kW LLC converter

After completing a fullbridge and halfbridge converter and using them to power my amp projects, I continued to investigate other topologies to overcome the disadvantages of the above:
1.EMI
2.Large inductors on outputs
3.Difficulties caused light loading
4.Hard switching and others

I found this forum's help essential to the success of my projects and I am again seeking your inputs and comments.

A big thanks for all help in the past ,I have been bit by the promise of the LLC converter and I want to scale one to a kW.I have read many appnotes from Fairchild,Onsemi and others and one thing stands out,PFC frontend required.. therfore a 1kW PFC will be needed for reliable operation.Fairchild,onsemi and STmicro have controllers dedicated for this along with excellent appnotes and tools.(AND8257,AND8255,AN2450 and AND8311/D.I will start with the LLC design first and to open the project I would like to present a simulation of a dual output llc using LTspice(free spice simualtor).The attached file set includes all models and libraries for the simulation.Included is a single output sim(see appnote AND8255).
 

Attachments

Some of the reasons why the PFC stage is required, besides the need for a high power factor, and universal mains input is that the converter efficiency depends on the supply voltage, being maximum at 400-450V level, since the switching losses are very low, and the conduction losses are predominant. another reason is that the PFC front end provides a pre-regulated DC voltage, with the associated 100/120Hz ripple, much lower amplitude that in case of direct mains rectification, which allows to design a converter with lower gain, tipically in range of 1.2-1.3, lower than the other case, where need 1.4-1.5 in this way, m can be increased, Lm becomes larger, this resulting in less circulating current, and less low-load losses.
The maximum power level which i could get using LLC converters was up to 2.8KW with the standard Half-bridge LLC converter and up to 6.4KW with full-bridge topology.
 
pfc frontend

Luka

The operation from line and pfc can be simulated,but first what is needed IMHO is a gold standard spice simulation based on an appnote with waveforms this is why I picked AND8255 but Cristi is correct the pfc does allow for a more robust design and also for most of us 2kW should be more than sufficent,another draw back with this topology is most designs I have encountered are designed for a fixed output voltage,load can vary. (xfmr turns ratio).The other is transformer construction that is recommended is intergal magnetics(sectional bobbin)which I plan on modifying a ETD core with a 3mm spacer one for Lp and the other for Lm. My thoughts are a L6599 control IC feeding an IR2181 that drives a simular mosfet used in the simulation.I will post schematic along with design equations used later.
 
regulation of output

Luka

Yes as the load changes the the freq will lower or raise depending if the load moves toward zero or open(no load) this is the major advantage for audio amps at no load regulation is tight,however high currents can be present but to protect against this the controller mentioned as most other can be programed to go into skip cycle or shutdown PFC.😎
 
what about core design? I'm getting used to, that freq is constant.
What about here, how do you design it, I think saturation is still possible here, right?
And no output inductors? how is this better then with normal HB? coz of the rectified sine, that amplitude is changing instead of duty
 
another draw back with this topology is most designs I have encountered are designed for a fixed output voltage,load can vary.
the topology doesn't allow to get a large output voltage range, is limited at max. +-10-15% of the designated value. but it does allow to get constant output voltage from zero load to maximum load, and still keeping ZVS.
(xfmr turns ratio).The other is transformer construction that is recommended is intergal magnetics(sectional bobbin)which I plan on modifying a ETD core with a 3mm spacer one for Lp and the other for Lm.
there are 2 ways to get the leaking inductance, first is to use a separate series inductor (used just for high power levels) or to use sectional windings where the primary winding is on one side and secondary winding is on the other side of the transformer. there are few more methods, like winding the secondary over the primary, similar as a normal transformer, but with a thick insulation layer, to reduce the coupling, and even to shift the windings to a side and another. another method is to use a large EE core and place the windings on the outer legs, and the central leg must be grinded to get the desired Lm and Lp. the i used this approach for a high power converter, and one more method, with UU cores, were i could get 3.6KW
most of the LLC converters use special coil former, with sectional windings, which allow large leaking inductances, and very important, tight inductance tolerances compared with the other methods, of winding. the leakage inductance comes as a bonus, without need for an external series inductor, and what's more important, it's saturation current is equal with the current of an air-core inductor.
My thoughts are a L6599 control IC feeding an IR2181 that drives a simular mosfet used in the simulation.I will post schematic along with design equations used later.
There are at least 15 dedicated suitable IC's for this converter, most of the power management IC manufacturers have at least one. L6599 has integrated drivers, no need external one for power up to several hundred watts, but need to be buffered for more power. an interesting IC is the UCC25600 from TI, this doesn't have drivers, and need an external transformer, which makes it suitable for both HB and FB topology. i used this one for the 3.6KW converter. ON semi has few, NXP, champion micro, and few more other companies.
since the LLC topology wasn't invented yesterday, it was quite used in Tv sets in 1990's such as SONY which used discrete parts for controller, later some dedicated ic (can't remember name now, just search for some Sony TV service manual) and i saw in some they even used bipolar transistors for switches. in telecom industry, there are few companies which use this topology, and allows them to get very high efficiency, (see Eltek-Valere Flat-pack 2 HE >96%) and some of them use either discrete solution or even DSP/MCU.

Do you know that an TL494 or a SG3525 can be used for a LLC controller ??? 😀
 
the duty cycle must be locked to almost 50% (50%-deadtime) then the soft start which increase the PWM duty cycle should not be used, but the converter must have a start-up circuit which allow to start at 2-4 times the res. freq then decrease till get the operating freq. to do that, the Rt and Ct pins must be controlled with a simple current source feed from the start-up circuit initially then the feedback loop takes control and regulate the output voltage by changing the freq. remember, the duty cycle should be locked, any loss of duty cycle will decrease the eficiency and can lead to converter failure because possible lose of ZVS condition.
 
things

Cristi
Many thanks for response and very correct I might add,even though this topology has been kicked arround for a long time it is revisted by all the major IC manufacturers because its use in the consumer electronics market has increased(flat screens,flat screens,flat screens...) and yes there are many IC's that can be used as the controller(I used the MC33025 for prototype)but IMHO it would pay to design in the new controllers because of the protection circuits and other intergrated features that are offered.As for the magnetics I plan on using the ETD cores and bobbins for the ease of winding,a 3mm seperator can be fabricated easily and with super glue placed where needed.If you add a couple of extra turns to the secondary and make small changes to the feedback you can vary output like Cristi implies.This can be proven using LTspice or other spice software.AND8257 an appnote for the Onsemi NCP1395 llc converter uses a ETD core with no sectional bobbin.
 
Will you post data and finding, pictures of proto too?
So you will go for split bobbin, I must say I didn't look for one if the are ready made, but would be too hard to have "one fits all", it depends on design.
Will you go for 24v output or something higher (didn't had time to check simulation yet)
 
output

Luka
Output +/-50VDC@5amps prototype and yes all schematics along with pics as well as notes which we will supply will be posted.If I have time I will post transformer contruction step by step.The reason is that the PFC is available or can be contructucted with ease(plenty of cheap parts)The caps and inductor needed are not very expensive.During testing I want to take it to the limits and not light bulbs for a few minutes.Transformer for this project at frequency of around 100kHz should have an Ae of about 150mm or 1.5cm^2(ETD44)HUH??
 
I am waiting to see your schematics with bated breath.

I just got my driver transform in, when you posted this project. Reading the linked articles and viewing the fairchild presentation stirred up my engineer challenge and reminded me of a project from quite a few years ago. The varied load with 4 input sources was never really resolved, but this LLC would have made it a lot easier.

Thanks for sharing looking forward to viewing the schematic and I just might parallel develop with you, but maybe just do something less sophisticated to re-wet my feet.

I like the part about using a MCU, maybe like a Arudino to control this. That was a direction I would like to take. I have some programing skills, experience and a couple of Arudinos laying around. I really want to put a zero-crossing power on soft-start on the input AC.

Good luck and such.
 
I like the part about using a MCU, maybe like a Arudino to control this. That was a direction I would like to take. I have some programing skills, experience and a couple of Arudinos laying around. I really want to put a zero-crossing power on soft-start on the input AC.

Good luck and such.
Uhh that would really be something to talk about and do too, but performance could probably be lower, but less parts and easy to change parameters, hope you decide to go this way