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.

Phase and frequency locking multiple (9 units) LLC drivers that are each 10cm apart

Status
Not open for further replies.

rxpu

Member level 3
Joined
Mar 9, 2019
Messages
58
Helped
0
Reputation
0
Reaction score
1
Trophy points
8
Activity points
819
I am trying to find a best suitable way to synchronize the phase and frequency of multiple LLC converter modules to achieve better EMI and better voltage regulation when they are supplying the same load.

The physical arrengement: The LLC converter is an all in one PCB containing half bridge,magnetics and synchron. rectifiers. Each converter PCB is 18cmx20cm and they are stacked one on another. The distance between the pcbs are approx 6cm.

The total length of the tower is (having 9 boards) approx. 54cm

The current configuration:
Each board have its owm controller a dspic33 delivering a variabe freq. PWM
Freq range: 80kHZ-150kHZ. Power of each converter 2100W , Secondary output 3.6V..6.8V @300A
Each converter has a galvanically isolated input for its half bridge drivers.

What is the best suitable way to synchronize the phase and frequency of each converter to a reference phase and frequency. the maximum distance may be 50-60cm

I saw some LVDS chips for long distance transmission. Such as
https://www.ti.com/lit/an/snla165/snla165.pdf

But I dont have any experience for LVDS chips.



Any ideas?
 
Last edited:

LVDS doesn't have large common mode range, RS422 is probably better, common mode chokes and common mode termination may be useful nevertheless. Or use transformer coupling.
 

LLC regulate by freq changing, so syncronize means lost of regulation?

is any current sharing between units?
 

I sketched the following arrangment to show the modules.

Each Module is a PCB (18cmx20cm) .They are stacked on another. Each converter delivers 3.4V..6.8V @300A. The outputs of the converters are connected togethar to reach 9x300A=2700A. Each converter has a secondary synchr.rectifier with mosfets. Mosfets when switched on may conduct current bidirectionally, so a good synchronisation of the frequency between the modules is neccesary.

Each module is a half bridge configuration. But there is also a capactor-diode clamp.This guarantees a hardware based current limitation. In worst case the current of each module is limited in hardware regardless of the frequency.

You may imagine it as a tower , the distance between the pcb is approx 6cm, 9 modules can be stacked so that a tower is 54cm tall.

A host controller generates the suitable half bridge variable frequency signal and delivers to the slaves which have isolated half bridge drivers.

I am assuming that the resonant capacitors and inductors are matched so that each converters resonance frequency is as close as possibleto each other. The host can tolerate a missmatch of %1 by an algorithm that adds the currents of each node and if the current and voltage are not in range a slower loop copansates the over or under balance by changing the frequency.

I know that the deviations of each converters resonance frequency can be a problem , but I believe I get worse results if I connect the outputs of the unsynchr. converters. There will be random fluctiuations that I can not control.

Any ideas to transfer this signal from the host controller to the modules?


LLC-secondaryParalel-Conducting-1.jpg
 
Last edited:

It sounds as though you're trying to impose one resonant frequency where your various converters have different resonant frequencies.

I wonder if you ruled out the idea of interleaving several buck converters? That ought to make your project easier to design. You can apply an equal duty cycle to all the converters. It's not terribly serious if one inductor is an oddball value. With the right combination of logic gates you can make a control circuit which staggers 9 On-Off times, and alter duty cycle with one adjustment. Output has little ripple so that you don't need an output capacitor. The initial power supply has a steady input current draw.

And considering the amount of Watts going through your system, it might be wise to keep the outputs of your converters separate, so that if one malfunctions it doesn't affect the rest of the system as much.
 

It sounds as though you're trying to impose one resonant frequency where your various converters have different resonant frequencies.

I wonder if you ruled out the idea of interleaving several buck converters? That ought to make your project easier to design. You can apply an equal duty cycle to all the converters. It's not terribly serious if one inductor is an oddball value. With the right combination of logic gates you can make a control circuit which staggers 9 On-Off times, and alter duty cycle with one adjustment. Output has little ripple so that you don't need an output capacitor. The initial power supply has a steady input current draw.

And considering the amount of Watts going through your system, it might be wise to keep the outputs of your converters separate, so that if one malfunctions it doesn't affect the rest of the system as much.

The reason I try to get all the freq. synchronized is that I want to use variable freq LLC at the primary side and sync. rectification at the secondary side.

So phase shifting is not possible (each converter have different freq. to regulate their internal loops)

I also have a problem with the synch. rectification. Even though many in this forum told me that it can be avoided, my simulation results show me that the secondaries of the different converters short each other (while each at operating random freq and phases) when using mosfets for rectification. Mosfet is a bidirectional device when switched on. With diode rectifiers I do not have this problem.

So I have 3 chioces:

1- Do not use sync.active mosfet rectification. Every topology is easy to use and hassle free. (cost:very high power loses at 2700A)

2- Insist on using sync. active rectification but find a robust way to synchronize the freq. and phase of all converters to a reference. (or forget modular structure use a big half bridge stage and a big resonant stage which has also its own design problems. Small amd modular is the key for LLC)

3- Find a way that even under unsynchronised converters working with their own different freq and phase, the sync. active rectifiers of different converters do not short each other. Many told me that it is possible and give recommendations but I could not resolve them. My simulation insist that the secondaries short while they are working at different freq and phases.


The reason may be the sinus signal at the output.Rather than other topologies (ex:pSFB), the voltage do not rise fast to a common level as in square wave.

2 sinus-shaped signal having different freq. and phase have more shorting regions when switched with a sync. mosfet.rectifier.

This is my main problem.

I do not want to change the topology because the output of LLC is very stable and does not need snubbering. (so that i can use very efficient low rdson low Vf semiconductors.)

I thank all of you for your help.
 

To parallel effectively, you must have a master voltage feedback controller which dictates the average current from each LLC. Each LLC should have its own inner current regulation loop.

Frequency locking LLCs is tricky. TI app notes suggest using some duty cycle modulation to implement current sharing. Never tried it though.
 

To parallel effectively, you must have a master voltage feedback controller which dictates the average current from each LLC. Each LLC should have its own inner current regulation loop.

Frequency locking LLCs is tricky. TI app notes suggest using some duty cycle modulation to implement current sharing. Never tried it though.

I agree,but even if I have a master voltage feedback and local fast current loop on slaves, there is a physical problem because of the nature of the LLC output signal. Its shape is sinus. And with sync. rectification if a converter has different freq and phase reference to another converter, the secondaries short each other randomly with respect to the voltage difference between active nodes.

Damn that mosfet is a bidirectional device when switched on. I can not get rid of this short by using any typeof regulation because it is a fundamental problem of mosfet. (with diode rectification it would not be a problem as it is a unidirectional device)
 

The assumption is, with rectifying, that you have very large o/p caps to give a stable Vout - without this the paralleling you are seeking cannot succeed.

- - - Updated - - -

You are far better off going to a different topology with current mode control on the primary - and current doubler on the sec ... paralleling then becomes easy ...

We have paralleled 5kW units at 200A each to 1600A DC this way ...
 
  • Like
Reactions: rxpu

    rxpu

    Points: 2
    Helpful Answer Positive Rating
The assumption is, with rectifying, that you have very large o/p caps to give a stable Vout - without this the paralleling you are seeking cannot succeed.

- - - Updated - - -

You are far better off going to a different topology with current mode control on the primary - and current doubler on the sec ... paralleling then becomes easy ...

We have paralleled 5kW units at 200A each to 1600A DC this way ...

Why I insist on LLC:

1- Severe osscilations on the seconadry rectifiers.(especially at light load). Need heavy snubbers to damp.

2- Need 100V Mosfet or diodes for 10V output and 150V-200V class Mosfet or Diodes for 15V output, which raises cost significantly.(especially for 2700A output), LLC variant can be implemented with 40V Mosfet or diodes for 10V output.LCC have very calm, clean and natural secondary.

3- LLC half bridge can be implemented with capacitor-diode clamp current limitation, which adds hardware based short circuit protection. This arrangement is a very effective protection for switching semiconductors. Similar structure is hard to implement for PSFB

4- Half bridge is effective, no need for full bridge, less cost improved efficiency

5- Need for an expensive (10uF-22uF range) decoupling flux balancing capacitor in series.

Point 1,2,3 are the main reasons, If I could find an effective way for these osscilations (practical and not complicated snubbers), point 4 and 5 could be tolerated.

Point 3 is also very interesting feautre, with this feauture there is no need for software or switching element based short circuit protection
 

Why I insist on LLC:

1- Severe oscillations on the secondary rectifiers.(especially at light load). Need heavy snubbers to damp. Yes but its only low volts so the snubber losses are OK.

2- Need 100V Mosfet or diodes for 10V output and 150V-200V class Mosfet or Diodes for 15V output, which raises cost significantly.(especially for 2700A output), LLC variant can be implemented with 40V Mosfet or diodes for 10V output.LCC have very calm, clean and natural secondary. Current doubler needs only 50V fets

3- LLC half bridge can be implemented with capacitor-diode clamp current limitation, which adds hardware based short circuit protection. This arrangement is a very effective protection for switching semiconductors. Similar structure is hard to implement for PSFB - current limiting is easy for PSFB a CT in the input lead and/or in series with Tx

4- Half bridge is effective, no need for full bridge, less cost improved efficiency. You can have half bridge parallel loaded resonant converter with current doubler o/p 100kHz to 200kHz

5- Need for an expensive (10uF-22uF range) decoupling flux balancing capacitor in series. Not with current mode PSFB

- - - Updated - - -

Regrettably you will never get LLC to work in parallel the way you want - unless all the stages are identical with the same gate drive and you parallel the AC as per the other posting on this subject ....
 
  • Like
Reactions: rxpu

    rxpu

    Points: 2
    Helpful Answer Positive Rating
Why I insist on LLC:

1- Severe oscillations on the secondary rectifiers.(especially at light load). Need heavy snubbers to damp. Yes but its only low volts so the snubber losses are OK.

Yes, it is manageable, should not be a critic problem.


2- Need 100V Mosfet or diodes for 10V output and 150V-200V class Mosfet or Diodes for 15V output, which raises cost significantly.(especially for 2700A output), LLC variant can be implemented with 40V Mosfet or diodes for 10V output.LCC have very calm, clean and natural secondary. Current doubler needs only 50V fets

I want to avoid active snubbers to reduce complexity and coponent count, so I am not sure if a simple RC-snubber will be enough for no load or light load condition



3- LLC half bridge can be implemented with capacitor-diode clamp current limitation, which adds hardware based short circuit protection. This arrangement is a very effective protection for switching semiconductors. Similar structure is hard to implement for PSFB - current limiting is easy for PSFB a CT in the input lead and/or in series with Tx

I agree ,but I was talking abaout a naturally and analog current limiting scheme that does not need any digital control. I am not sure if it is possible to implement a fully analog clamp for current limitation for PSFB, the only solution I may think is to limit the power capability of the transformer ,and using overdimensioned semiconductors which can deliver this current saturating current (in case of a secondary short)



4- Half bridge is effective, no need for full bridge, less cost improved efficiency. You can have half bridge parallel loaded resonant converter with current doubler o/p 100kHz to 200kHz

Thank you for the idea , I will investigate parallel loaded resonant converter with current doubler o/p 100kHz to 200kHz


5- Need for an expensive (10uF-22uF range) decoupling flux balancing capacitor in series. Not with current mode PSFB

Unfortunatelly I hate fast current mode control, I think it relies on precision measurement of current through ADC.This may make the things more complex and unreliable. It tend to use voltage mode,

- - - Updated - - -

Regrettably you will never get LLC to work in parallel the way you want - unless all the stages are identical with the same gate drive and you parallel the AC as per the other posting on this subject ....

I agree, I see the drawbacks of the LLC, I just wanted to investigateand and I am unhappy that the LLC does not meet the needs of the paralelization especially for paralel connected unsynch. converters.

Until now the I was tending to insist on LLC but the following limitation that I could not predict let me switch backto PSFB:

For synchronous rectifiers if the secondary output is a square-wave shape I may connect these outputs in paralel even if they have different freq. and phase. Because the overlapping parts of the signal is either 0V (and the mosfet is switched off) or a common voltage that is very close to each other.

*BUT* , if the output signal is a sinus-shape such as in LLC ,there is a severe problem for multiple unsynchr. converters which are paralel connected. Because if the phase and freq of a sine-wave signal are different, there exist a very long period where the voltage difference is high.and This causes a short current between the various seconderies of the converters. If the output is square, the outputs (mosfets) are switched off, and only switched on if the voltage has rised to a common voltage which avoids a short. Maybe there is a very tiny period while the mosfet is switching but this effect is really minimal.

As my secondary currents are very high and voltages are relatively low, I can not sacrifice the synch. rectification. The LLC topology (when connected paralel and works unsycnh) has severe drawbacks also in this implementation.

I go back to PSFB.

Thank you again for the detailed info.

 
Last edited:

Post #5 shows a parallel connected LLC with output capacitor, using this topology, there's no need to synchronize the converters. Load balancing may be necessary for variable frequency (voltage varying) LLC, in case of fixed frequency LLC in series resonance you can expect fair balancing without controller.
 
  • Like
Reactions: rxpu

    rxpu

    Points: 2
    Helpful Answer Positive Rating
Post #5 shows a parallel connected LLC with output capacitor, using this topology, there's no need to synchronize the converters. Load balancing may be necessary for variable frequency (voltage varying) LLC, in case of fixed frequency LLC in series resonance you can expect fair balancing without controller.

I am sorry that I repeat the scheme. But this is because I did not fully understand it. When we look at the following sketch ,I see that there is current flowing through Point A to Point B. (magnitude: Ia-Ic) . This means the secondaries of converter1 and converter 2 short each other.

I know that I miss something , but I dont know what. I assume when mosfet is switched on it behaves like a very small resistor and bidirectional current flow is possible.

What is specificly wrong with my understanding?

LLC-Paralel-Output-short.jpg
 

The synchronous rectifiers are acting as full wave rectifiers and sourcing DC current. The pulsating current output of both LLC has ripple with possibly different phase, but always same polarity.

I'm further assuming that the filter capacitor is large enough to reduce the voltage ripple to a small percentage of the DC voltage, in this case there's effectively no interference between both LLC outputs. With smaller capacitor, you have still positive output voltage and currents, but the voltage ripple waveform varies depending on the LLC phase relation.


https://en.wikipedia.org/wiki/Rectifier#/media/File:Fullwave.rectifier.en.svg
 
  • Like
Reactions: rxpu

    rxpu

    Points: 2
    Helpful Answer Positive Rating
Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top