1kW LLC converter

Glad to see that this thread produced some benfits,It has been a education for me also and after all of the simualtions and proto work I realized that the benfits of this topology are geared more toward the dc to dc converter(pfc front end) than a true offline supply and the cost per watt ratio is higher based on the additional parts required also the magnetics required does not promote an easy diy project.At this time hard switchers or varations of them seem to be the best for diyer's in mho.Thanks to all who have replied to this thread and I hope someone with a background in llc will continue this thread.
chas1
 
I agree that this converter is not optimal for sth with wide input voltage range, so PFC front end is a best fit. And it can do buck conversion much better than boost.

For boosting up to higher ratio, may need lower magnetizing inductance, hence higher circulating current, much lossy,may fall into capacitive region too.

For higher wattage, a separate inductor is unavoidable.

So I guess for offline. the phase shift technique may perform better; the best application for LLC is following PFC stage and step down ( or up just a bit).


Glad to see that this thread produced some benfits,It has been a education for me also and after all of the simualtions and proto work I realized that the benfits of this topology are geared more toward the dc to dc converter(pfc front end) than a true offline supply and the cost per watt ratio is higher based on the additional parts required also the magnetics required does not promote an easy diy project.At this time hard switchers or varations of them seem to be the best for diyer's in mho.Thanks to all who have replied to this thread and I hope someone with a background in llc will continue this thread.
chas1
 
Details needed

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??



Can we see the schematics? Transformer construction?

As promissed?
 
schematics and magnetics

Most of my work was based on 500watt LLC and simulations using LTspice and Onsemi appnotes along with Fairchild and I will post the results along with transformer construction that I used, as for schematics they are what you will find in the appnotes with a few minor component changes based on the frequency I used for my proto's.I have to review my project files before posting the useful one's.Already mentioned is one drawback for the power required for most audio amps you will need added inductor and pfc.The disavantages that have to be overcome are input voltage varations even with pfc and along with that load varation.To sum up my findings,fixed input along with well defined load excellent results outside of that and very difficult to come up with a suitable regulated design over 500watts.
chas1
 
Stepping the load of smps

Luka
Glad to see you still monitor this thread,you should expect those results from your design
while a pfc frontend is a more or less a requirement in some countries for other reasons; it should have little effect on output ripple unless there are problems in the design as the output power is regulated while the dc to dc converter following it can have ripple and other problems if the feedback loop is not closed properly(output inductor properly sized,output cap esr accounted for and opamp compensated properly.A network analyzer is the normal method used to detemine this.



A well designed feedback should stablize any topology within the design requirements but the testing of a supply sometimes requires special test equipment to provide usable results and the parameters of the test need to adhere to adopted guidelines and are normally carried out in a lab.

chas1
 
Last edited:
LLC simulation using LTspice and models by Chris Baaso

Since the new 8 pin controller chips are stocked by Newark and Digikey I am
getting back to LLC conversion and I have been using the files that I am attaching to simulate conveter design info obtained from the XLS that is also
attached.If you follow the design's in AND8255/D you can see the simulation's
are very accurate and should help with component choices.
 

Attachments

IRF dual LLC

For info and comments just in case you missed it and I am back to work sorry for the dead end thread.I will post LTspice simulation of this design for 50VDC Shortly.Note:This appnote like many others has been deleted from IRF website when I checked this is from my files and it is also on the web google is your friend.
 

Attachments

For info and comments just in case you missed it and I am back to work sorry for the dead end thread.I will post LTspice simulation of this design for 50VDC Shortly.Note:This appnote like many others has been deleted from IRF website when I checked this is from my files and it is also on the web google is your friend.

I haven't seen this application note, but I am pretty sad, it is one of IR's worst, IMHO.
LLC converter is okay for a class D amp, but for 60W average and 200W pulsed load? With <90% efficiencies under most operation conditions?
It is quite pointless. For these power levels a flyback supply is:
- less complicated - 100%
- can have better efficiency and can have same low EMI, specially if it is a quasiresonant flyback. (a QR-flyback is still less complicated than an LLC half-bridge).

Just look at the transofrmer design. There are 2 primary and 6 six secondary windings (okay 2 from it is aux). Why the hell?
Using a center tapped secondary (wound bifillary on a split bobbin), with a diode bridge rectifier must have been simpler. Also relative effective current / ripple current ratios of lines would have been much less.
(it is usually important for high current PCB layouts, since at high current it can be possible that different lines will not share current evenly, causing assimetry)

Also look at the thermal data charts. Output diodes are the hottest, it is clear. They are using 200V ultrafast diodes, having 1,25V voltage drop at 10A 25°C and 1,05V at 10A 150°C. For +/-30V supply? Why?
Using a simple, old MBR10100, pretty enough for the job (lots of similar items), will have a voltage drop of 0,85V at 10A 25°C, and 0,71V at 10A 125°C.

Okay it's resonant and the diodes are swithcing ZCS mostly, so you could even use slower ones, but the capacitance of diodes have a negative effect at low load conditions. An LLC converter without burst-operation, wil increase the output voltage at no load conditions, mostly because of the capacitance of output diodes.
IR's ICs has no burst mode (at default, but it can be easily implementd), so they use some loading resistors (R12 and R13). Using schootky diodes would decrease the previous effect.

Also it must be kept in mind that with an LLC converter MOSFET turn off is costly. MOSFET's must turn off under current. If the converter has a wide input voltage range, the relative magnetizing current and the switch off current will increase. So MOSFET's should be turned off fast. If the current caused by the miller plateau, under turn off is bigger than the gate turn off current supplied by the driver, then the MOSFETs will turn off pretty slowly, caused high losses, or they can re-turn-on after turn-off (this is rare with good layout and bootstrapped FET drive). I would reccomend increasing turn off gate drive current (lowering 10R turn off resistor) in this design.

Look at TI's application note. A similar reference design for a Class-D ampifier supply, for TAS5630. It is also a complicated design (PFC, with PWFC-switch off for low loads, reswitch with clip indicator pin), and has about the same efficiency but with a PFC.

IR's ICs are much more competetent (itegrated half bridge bootstrapped drive), so why not prove it in a competent design?
 
lorylaci

Your reply covers all the reason's for the post and I think most of these is why IRF pulled the appnote as you pointed out most of the pitfalls that DIYer's have to overcome for a workable LLC design. The short of this is why would anyone build it.Thanks for the through insight.The simplest topology for DIY in power ranges 1kw up to 3kw is the two switch forward converter in IMHO, less efficient, but easy to realize a working design and no PFC required.
 
Last edited:
I believe the appnote use an ordinary ATX transformer ready made, ie they built the app note around this. Otherwise, why use an external resonance inductor for an LLC converter ? Even if Lr is noted with jumper in the BOM.

Anyhow good points Lorylaci. You are Diyaudios L-LLC expert 🙂
(local LLC expert)
 
lorylaci

Your reply covers all the reason's for the post and I think most of these is why IRF pulled the appnote as you pointed out most of the pitfalls that DIYer's have to overcome for a workable LLC design. The short of this is why would anyone build it.Thanks for the through insight.The simplest topology for DIY in power ranges 1kw up to 3kw is the two switch forward converter in IMHO, less efficient, but easy to realize a working design and no PFC required.

Well, IMHO half bridge is about the same difficoulty as two switch forward (there is a need for high-side MOSFET drive).

LLC converters are not that hard to design. Most manufacturers do it a bit complicated. Howerver if you do not need PFC, then you also do not need an LLC converter for audio, really.
For an audio amp the regulation of rail voltage is not so mandatory. It is only mandatory that tha rail voltages should no drop below a certain limit (the limit is quite subjective, it is up to you).
For Class-D amplifiers, it is usually good to use the lowest voltage MOSFET needed, up to its highest voltage limit.

Normally a series resonant converter does the job for the audio amp supply, if you do not need PFC and an 220-240VAC input voltage is enough, specially if it has a burst mode operation. it will help to regulate rail voltages so, that it does not exceed MOSFET rating, but will not loose so much peak output power.

However PFC is highly needed for high power amp. Not really for the 1,00 Power factor, but for other: an audio amp can drove very high peak currents, and very small average currents.
At one ocassion my band had to perform outdoor on diesel generators. You could see the light blinking to the beat of the bass drum.
Amps with PFC will include a 10-20Hz low pass filter in the PFC feedback network, so the current inrushes will be limited. But you either:
- need lot of capacitors both at primary and secondary
- need medium lot of capacitors at priamry and a converter with regulation at wide imput voltage range. So here bumps on LLC converter

Also there are differences:
- most of the time the converter will operate from the mean input voltage, so efficiency can be maximised for it. At the sides of input voltage range it onyl enough if it tolerates it (can regulate output), the fficicency can be smaller.
- without PFC the converter must operate with high gain at high loads (input votlage is decreased), and with <1 gain at low loads.

To rikkitikkitavi (soory if i spelled it bas)

It is sure not using ATX transformer. ATX supplies are mostly PWM half-bridge converters. They do not have split bobbin, because every minor leakage inductance will come up asw "duty cycle loss".
(this is why I also do not favorise ZVS full-bridge phase shift converters)

I haven't been able to come up any logical explanation why they did that appnote that way, other then lazyness ("hey, we put this piece here, but it is not to good - just leave it there, it is **** but will be good".)

The easiest supply topology is clearly flyback (for 230VAC). They use it form mobile chargers (one BJT, one MOSFET self-oscillating flyback 🙂 ), up to 100-300W supplies. It can regaulate multiple outputs withouth compromise.
A flyback can be easily desgined, so that it can put out 100-200W for short periods wihtout overheating (transofrmer is large enough not to saturate at the extra power, MOSFET and diodes has enough heatsink), but has a good efficiency for 40-100W range. 40-100W is an optimal flyback range. An audio amp supply does not have to have the best efficiency at max loads.

Over 200-300W it is cleary LLC converter if you need multiple outputs IMHO. (if you do not, you can choose between forward or PWM half-bridge or other)

However over 500W LLC converter gets bulky. You need a hell lot of LOW-ESR capacitors at secondary for current ripple. The turn-off currents (and losses) get huges. And a gain range of 0,7-1,4 is impossible to maintain efficiently.

For 1-2kW levels I choose two-phase 3rd order series resonant converter. This is the one I am experimenting with, right one. But I am still sucked by controller board.
 
Although I did design several off-line flyback over the last decade, I would not say "The easiest way is the flyback".
It circuitry looks easy, but is not.
The transformer always gets hotter than expected, often you have to choose a bigger size for your power needs.
Optimal transformer design requires strongest magnetic coupling that can only be achieved with interleaved winding. As a result coupling capacitance increases, giving emc problems.
With improper coupling stray losses increase, and these must be dumped in the snubber. I have seen 70W notebook-adapters with small smd resistors as a snubber - I disassembled the transformer and I recognized this technique far exceeds the skills of normal DIY-people.
Another topic I would mention is so called QR-ZVS. In most cases, primary reflected flyback voltage is a fraction of primary supply voltage. Thus the amplitude of the resonant ringing after demagnetization is far too small to swing to zero.
Assuming 230V mains and a flyback voltage of 130V there is a turn-on at about 180V, far away from ZVs, just valley switching.

EMV-signature is better than fixed frequency PWM with continous mode, but far from being superior to LLC topology.

Ripple currents in QR-ZVS flybacks are the worst, and they stress excessively the primary power-MOSFETs, the transformer windings (litz wire required), secondary rectifiers and secondary bulk capacitors.

Blocking voltage at secondary rectifiers is, as a rule of thumb, rated 3 times the DC-voltage.

Using multiple secondary windings provides multiple outputs - but the tracking vastly depends on load distribution and winding sequence of the transformer.

That is way imho the flyback is a beast!
 
You may ask why I use QR-ZVS flyback converters in my designs?
-With current control, dynamic response to load changes is fast and rock solid.
-Short circuit proof, no need for a softstart
-simple loop compensating scheme
-being discontinous, there is no RHPZ-Problem
-no storage chokes required
-multiple outputs available
 
two switch forward

Drive upper switch with single GDT, lower can be driven with silicon.Using silicon to drive upper switch can present problems and the solution adds extra cost over GDT(gate drive transformer).refer pg 851 (SWITCH-MODE POWER SUPPLIES)Christophe Basso.
 
Drive upper switch with single GDT, lower can be driven with silicon.Using silicon to drive upper switch can present problems and the solution adds extra cost over GDT(gate drive transformer).refer pg 851 (SWITCH-MODE POWER SUPPLIES)Christophe Basso.

Well I have experimented with gate drive transofrmers in LLC, and I was not satisfied. There was a time, when I wanted to change to UCC25600 from IRS2795x, because of the burst mode, and OC protection method. I used Rdson sensing on low side MOSFET and a gate drive transformer.

I used a GDT designed for UCC25600 from coilcraft, with an 2x4A drive buffer before the GDT. Due to the leakage nductance of the GDT the driving was not fast enough (specially turn off), so that when turnin of MOSFET they re-turned-on due to the dV/dt at the half-bridge.
(I managed to solve this by adding ZVS caapcitance, slowing down dV/dt to 320V/200ns, but I was not satisfied)
Then I tried driving an other GDT directly form UCC25600 but adding PNP turn off transistor to secondary. In this way the dead time plateau was splitted due to the mangetizing current of GDT and the output inpedance of UCC25600. So there was no real dead-time.
Also winding capacitance of the GDT matters!

So with GDT there is need to use pre-GDT buffer (or GDT must have a very small magnetizing current). But with big magnetizing inductance there is usually bug leakage inductance (bifillary windings with triple insulated wire can make a solutio if you need 2-3kV dielectric strength) so you must use post-GDT buffer.

Now in my two-phase converter I am using GDTs, only because IR2110 would have a difficoult job at >200kHz and >200V.

But for low power (<kW), bootstrapped half-bridge in IRS2795x is pretty good. I usually use an additional PNP transistor (FMMT549A, i can get it for <10 eurocents) for fast turn-off, and an RD for turn-on.
 
An advantage of GDT is the ability to deliver neg gate drive during off-period.
Good GDTs are trifilar, triple insulated windings on very hi-mu toroids. And they are expensive - several Euros for the GDTs from VAC.
If you have access to hi-mu toroids, you can wind them yourself.

And yes, coupling capacity of trifilar windings might be about 100pF - allowing significant discharge currents at sharp transients.