Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Resonant SEPIC converter for LED driver

Status
Not open for further replies.

Meri96

Junior Member level 3
Joined
Dec 3, 2020
Messages
28
Helped
0
Reputation
0
Reaction score
0
Trophy points
1
Activity points
456
Hello everyone, I am waiting for your help for my university project. I am working on the LED driver. I have no experience in this field, please excuse me if this is an absurd question. The LED specification I will use is 1A, 2.85-3.20 V forward voltage. I will use a li-ion battery for supply voltage. So my input voltage will be 3.7-4.2 V. The switching frequency will be close to 2 MHz. The circuit topology I should use is a Class E converter based resonance SEPIC converter. Can I use controllers produced for SEPIC converters such as LM3410 or LM3421 to control this converter? Or should I use a resonant mode controller? Or should I use a separate circuit with an oscillator and a gate driver instead of a single controller?

resonant-sepic.PNG
 

Many current mode controllers can be used to control a sepic...what about LTC1871...i think this has a Vin down to 2.5V.(?)
But yes, i suppose if you can provide the min vin, then a LM3421 can also do the job...i think its a constant off time controller(?)
Class E converter based resonance SEPIC converter
...I dont think there's any such thing as that

Also, i dont see how your circuit acts as a "resonant" SEPIC converter.

A resonant converter avoids switching loss in its FET(s) by having a resonance which forward bias's the FETs intrinsic diode just before the fet gets switched on....i cant see how this happens in your schematic. Resonant converters also have high circulating currents, which is part of the thing that sorts out the intrinsic diode action that i spoke of...i see no place for high circulating currents in your schematic.

By the way, a SEPIC converter, especially a coupled SEPIC converter, using an offtheshelf coupled inductor, would be a good choice for your LED driver...coilcraft do a good set of coupled inductors which you could use. SEPIC is good due to its low side fet drive and low side fet current sense....and also its ground referenced load.

If you insist on 2MHz f(sw), then a GaN FET based SMPS solution could be relevant to you.
 
Last edited by a moderator:

..woops, beg your pardon..never say never
...resonant sepic.

But for your case of 3.5 Vout, you could improve efficiency by just using a low Vf schottky.

Its a matter of whether the extra conduction loss's that you will get from resonant operation, outweighs the extra switching loss's that you would get with a hard switched sepic.
As i understand it though, this is for an LED that goes in someone's body whilst they are being surgically operated upon? (so the surgeon can see what they are doing)....and you want low noise so it doesnt interfere with the prolifera of medical electronics modules that you always get in an operating theatre?
--- Updated ---

...ok, so you want to use LM3421.....but LM3421 uses peak current control....true resonant converters do not use peak current control......think of an LLC....they have overcurrent limiting but cycle to cycle involves no current sense shut off of the switching fets. So i doubt that LM3421 could be used for a true resonant converter.
Quasi resonant converters can use peak current control......but they really are not truly resonant and have hard switching edges, often the off-going is a hard switching edge. In fact, even true resonant converters often have a hard switching edge......eg the shut off of the magnetising current in an LLC...thats a hard switching operation. So resonant converters are often not totally "quiet".

I think for low noise, i would just use a standard SEPIC, have a common mode choke at its output and input, plus enough diff mode filtration......then feed the current to the LED remotely by a length of flexible coaxial cable, and make it a low noise system like that. Staunch the switching edges so that the switching transients were as soft as possible, without causing too much switching loss's.
 
Last edited by a moderator:
While researching I saw these articles and thought I could use a Class E based resonance SEPIC converter because my teacher wanted the LLC or Class E topology. I thought that with the Resonans SEPIC, I could obtain the necessary input-output values for the LED and also provide ZVS as my teacher wanted. LM3421 was just a suggestion for SEPIC, I don't necessarily need to use it, I just wanted to consult with those who know the business here on the situations I saw while researching. As you mentioned, the LED will enter the human body to illuminate the operation site in diagnosis or treatment, there should be less noise and light flickering, and a remote connection will be provided with a cable. Some control techniques are explained in these articles, but I wanted to learn the alternatives, if any, because they seemed a bit complicated to me.
 

Attachments

  • hu2012.pdf
    1.2 MB · Views: 124
  • madsen2013.pdf
    890.8 KB · Views: 129

When i get time ill try and simulate that resonant sepic in the free LTspice and report back here. Matter of fact, i always feel a bit suspicious when people present a paper without providing a baseline LTspice sim with it. Because no matter how many words or pictures you put of smps's, there's no substitute for having the thing running on the sim in front of you so you can probe it and check it out. (obviosuly real hardware and a scope is best, but they cant provide that to everyone). I'd like to see the vds of that sepic fet going down, its intrinsic diode coming on, and then the gate drive going high for a zvs switching.
 
That is a resonant class-E sepic - if some readers can't see it first off - is no reason to show ignorance by saying it is not - that ckt is well known in literature. ( class E is defined as switch on for 90 -> <180 deg approx )

Good control at 2 MHz is tricky - at least it is low power - so less RFI - research all the available IC's - often a combination of ASIC control IC and high speed logic devices will get you there - happy designing ...
 
Attached is a SEPIC which fulfills your spec. It is in the free download LTspice.
But its not resonant...yet...
I am working on trying to define which L and which C are the "resonant pair", whose resonant frequency must be same as the switching frequency, in order to give the resonance. Somehow we must make the current lag the voltage such that the fet's intrinsic diode comes on before the gate switches the fet on.
I am wondering if i can change it to a coupled sepic and add a resonant L and C afterwards, as the coupled sepic is easier.
--- Updated ---

Hi,
Hacked the resonant SEPIC about a bit.....considered that the cap across the diode and the series sepic cap were in parallel effectively...and made their resonant frequency with the RHS inductor just a little higher than the switching frequency.....and thats it done...resonant sepic.
Ive attached the LTspice sim and the schem and the waveforms.

As you can see, the fet intrinsic diode is made to conduct just before the fet is switched on.

I hope this will give some electrical-noise-free surgical hours to those surgeons!
.......
I must admit a bit more work is needed, since there's some fairly high di/dt's in there, eg when i look at the current in the series sepic cap. Mind you you could lay out to make this a tight loop.
The input current isnt too sinusoidal, but its not too jagged either. More work is needed though.
I have to say though that the diode switching is also nice and quiet...the diode voltage comes down to zero just before diode current starts to flow.
 

Attachments

  • sepic resonant.zip
    815 bytes · Views: 102
  • sepic resonant_works.zip
    831 bytes · Views: 91
  • SCHEM_Resonant sepic.jpg
    SCHEM_Resonant sepic.jpg
    70.4 KB · Views: 249
  • Waveforms.jpg
    Waveforms.jpg
    109.7 KB · Views: 172
Last edited by a moderator:
Thank you very much for your simulation and sharing, below is the circuit when the switch is on and off.

Ekran Alıntısı.PNG
Ekran Alıntısı2.PNG


Here, while the diode is going to cut, L2 and CD resonate until VL2 reaches Vout. In the cut-off state, the voltage across the capacitor is equal to Vout and IL2 begins to decrease.

L1 and Csw go into resonance when the switch is off. Vds rises, Csw is charged, resonant inductor L1 current increases. The Vds voltage continues to rise until the Csw current is zero. When Csw drops to zero, switch is open, Csw becomes short circuit. ZVS is provided.

When I look at the design parameters, the capacitor was added parallel to the switch and diode to provide ZVS, making calculations like a traditional sepic. In other applications, two designs are made as rectifier and inverter and combined to provide zvs.

I guess the first method was applied in your simulation. How did you determine the design parameters and how did you choose the frequency? I have to use PSIM for simulation, I couldn't understand how you drive the switch in LTSpice, Can you help me, how do I set up the PWM generator to drive the key in the PSIM application?
psim.PNG

--- Updated ---

As you can see, the fet intrinsic diode is made to conduct just before the fet is switched on.
And how did you make this inference? How do we understand this?
--- Updated ---

RHS inductor
What is RHS inductor here?
 
Last edited:

RHS inductor is literally that..the right hand side one of the two
The inference........it is simple...if diode voltage goes positive from anode to cathode then the diode is conducting......but even if it isnt properly fully conducting, the fact that the fet drain_source voltage has gone to zero just as current is starting to flow in it is satisfactory enough.
Zero x value = zero, so thats nice...no switching loss.

I guess the first method was applied in your simulation. How did you determine the design parameters and how did you choose the frequency?
I just fiddled about with the sim till it was right.....if i remember right, there was like a "camel hump" half sine resonance, which you had to make sure could fit fully inside the fet off interval...then the vds of the fet was zero just at the time that the fet was about to switch on again.
...so use w^2 = 1/LC and make out the resonant capacitor and inductor from that.
 
  • Like
Reactions: Meri96

    Meri96

    Points: 2
    Helpful Answer Positive Rating
Hi again,
When I look at the literature, firstly, a traditional SEPIC converter is designed, then resonance and soft switching are provided by connecting a parallel capacitor to the diode and switch.

so use w^2 = 1/LC and make out the resonant capacitor and inductor from that.

The value of the resonance components is calculated from the formula C = 1 / (16 * pi ^ 2 * f ^ 2 * L) as you mentioned.

The design you made at 100k switching frequency is pretty good. As I see it, a lot of changes need to be made on the design parameters for the simulation to be correct.
Below is the LTspice file of my design for f=5MHz. In the design, I first made the SEPIC converter design and checked that it works correctly, then I added resonant components according to the formula. but I could not get the resonance and soft switching. Even though I played on the simulation, I couldn't fix it. You are right, I must make the current lag the voltage for soft switching. How can I achieve this? If you could help with this situation, I would greatly appreciate it.
 

Attachments

  • resonantSEPIC.rar
    836 bytes · Views: 105
  • reSepic.PNG
    reSepic.PNG
    20.6 KB · Views: 147

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top