Bidirectional LLC

Hi,

Every converter with synchronous rectification should be capable of 
bidirectional operation. But there seem not to be any practical 
realisations of bidirectional LLC, except for some vague conference 
papers. And even then it is not a pure LLC, but a CLLC or split-rail 
LLC. Why?

     Best regards, Piotr

On Tue, 22 Sep 2020 08:22:46 +0200, Piotr Wyderski
<peter.pan@neverland.mil> wrote:

>Hi,
>
>Every converter with synchronous rectification should be capable of 
>bidirectional operation. But there seem not to be any practical 
>realisations of bidirectional LLC, except for some vague conference 
>papers. And even then it is not a pure LLC, but a CLLC or split-rail 
>LLC. Why?
>
>     Best regards, Piotr

Suggest you try it. Just because you can draw a topology on 
paper and give it a name, doesn't mean it has a practical 
application. Not all of the relevent Cs and Ls operating 
in an 'LLC' converter appear in its nomenclature. Not all 
operating states that are possible for the topology are 
of practical use at any specific time..

Bidirectional tri-ports, multiresonant or otherwise, isolated 
or otherwise, have been used in UPS systems since the 80s, 
with varying success. All have pros and cons, depending upon 
the voltage ratios, range of regulation and direction of power 
flow.

Suggest you examine the vague conference papers more closely.

The problem is efficient, versatile bidirectional power flow 
 - not 'why can't I do it?'.

RL

legg wrote:

> Suggest you try it.

Seems to work OK in Spice. In the reverse direction it is a push-pull 
pumping a resonant tank. The half-bridge then becomes a synchronous 
voltage doubler or even can be entirely deactivated if the transformer 
ratio is sufficiently high. For 30:1 the parasitic MOSFET diodes do the 
job just fine with an amp or two.

The waveforms look fine, the converter reacts to frequency changes as 
expected. So if it looks OK, so why there are no practical 
implementations widely available? In the best case it can be described 
as a curiosity.

Theory says it's OK, Spice says it is OK, practice says verboten. Why?

> Just because you can draw a topology on
> paper and give it a name, doesn't mean it has a practical
> application. 

A high-efficiency converter of such a simple structure should have 
plenty of practical applications. A DAB requires twice as many switches, 
a bidirectional PSFB is about the same and can be considered a DAB 
driven in a fancy way.
> Suggest you examine the vague conference papers more closely.

Sure, but there are not many of them, which raises my concerns even 
without reading them. And hence comes the question.

> The problem is efficient, versatile bidirectional power flow
>   - not 'why can't I do it?'.

The efficiency should be expected to be high in at least one direction, 
while in the other it is at least satisfactory. Very good approach if 
power flows mostly in one direction.

	Best regards, Piotr

On Tuesday, September 22, 2020 at 4:22:54 PM UTC+10, Piotr Wyderski wrote:
> Hi, 
> 
> Every converter with synchronous rectification should be capable of 
> bidirectional operation. But there seem not to be any practical 
> realisations of bidirectional LLC, except for some vague conference 
> papers. And even then it is not a pure LLC, but a CLLC or split-rail 
> LLC. Why? 

There may not be all that many practical applications so far..

Domestic roof-top solar power operating into a home battery might be one. There isn't that much of it around, but it is certainly becoming more popular in Australia.

The grid won't pay you anything like as much for the power you feed into the grid as they charge you for power they sell you from it, so it makes sense to store the power you generate during the day and run your house on it overnight (while it lasts).

For large scale solar farms you want convert the low voltage direct current from the arrays into high voltage AC as close as possible to the panels - you can use much thinner and cheaper wire to connect the arrays together and back to the grid, and the power is only going to go one way.

With roof-top solar the wires aren't long enough for this to be much of an issue.

If  domestic roof-top solar with battery becomes a mass-market item, you may see high volume production of devices aimed at this market.

Electric cars and trucks with regenerative braking might be another such market. That might be closer to maturity.

-- 
Bill Sloman, Sydney

On Tue, 22 Sep 2020 15:08:03 +0200, Piotr Wyderski
<peter.pan@neverland.mil> wrote:

>legg wrote:
>
>> Suggest you try it.
>
>Seems to work OK in Spice. In the reverse direction it is a push-pull 
>pumping a resonant tank. The half-bridge then becomes a synchronous 
>voltage doubler or even can be entirely deactivated if the transformer 
>ratio is sufficiently high. For 30:1 the parasitic MOSFET diodes do the 
>job just fine with an amp or two.
>
>The waveforms look fine, the converter reacts to frequency changes as 
>expected. So if it looks OK, so why there are no practical 
>implementations widely available? In the best case it can be described 
>as a curiosity.
>
>Theory says it's OK, Spice says it is OK, practice says verboten. Why?
>
Suggest you breadboard your spice circuit. If it works as required 
under the range of input and output conditions needed, then you may 
have a viable application.

Topologies do not always scale - their relationship to the quality 
and cost of pracical components (with their limits and strays) often 
restricts applications to a pretty narrow range of power, voltage, 
current and frequency.

>> Just because you can draw a topology on
>> paper and give it a name, doesn't mean it has a practical
>> application. 
>
>A high-efficiency converter of such a simple structure should have 
>plenty of practical applications. A DAB requires twice as many switches, 
>a bidirectional PSFB is about the same and can be considered a DAB 
>driven in a fancy way.

Practical LLC converter circuits may appear simple. I suggest that 
you try to design one from scratch for an application not previously 
illustrated in a cookbook or app note. Obtaining magnetics or other 
parts may be an issue.

>> Suggest you examine the vague conference papers more closely.
>
>Sure, but there are not many of them, which raises my concerns even 
>without reading them. And hence comes the question.
>
The LLC designation is one of convenience, applied to specific 
applications. This naming 'convention' is one of convenience, 
or intent, rather than accuracy or definition. It is probably the 
last thing in mind (topological classification), when developing 
practical circuitry. Whether you're performing a patent search, 
or just googling, you will have to extend your terms of search 
to find useful examples.  

It's just a variety of series resonant converter - no shortages 
of papers or references there. The third or fourth term make it 
multi-resonant - for specific reasons. Is there a difference 
between LLC and LCL, if the last two elements are parallel-resonant 
anyways?

There are always more papers than practical and honest examples 
of working applications. Always more chaff, than wheat. Try 
searching for articles by the same authors, or tracking down 
references to authors, companies or educational institutions that 
they mention, if they smell right in the first place.

You might check some of Johannes Kolar's (large power) output,  
Ashoka K.S.Bhat,or the research papers made available by Virginia 
Tech (low power). Examples earlier than Y2K will involve 
semiconductor and magnetic types and structures that you will 
likely have no interest in exploring.

Contemporary semiconductor app notes only show up after they 
have already achieved significant market penetration. Manufacturers 
and customers don't care whether it's LLC, upside-down, or an 
apple (well, possibly if it's an Apple . . .).

>> The problem is efficient, versatile bidirectional power flow
>>   - not 'why can't I do it?'.
>
>The efficiency should be expected to be high in at least one direction, 
>while in the other it is at least satisfactory. Very good approach if 
>power flows mostly in one direction.
>
These are trade-offs that can make or break a bidirectional 
power transfer application, just as they do battery or other 
energy storage or transmission applications. They have to result 
in advantages to justify their development or use.

RL 

legg wrote:

> Suggest you breadboard your spice circuit.

It is better to ask before breadboarding. It can save a lot of time and 
smoke. I tend to assume that I may be missing some detail that would 
ruin this not very typical approach.

> Practical LLC converter circuits may appear simple. I suggest that
> you try to design one from scratch for an application not previously
> illustrated in a cookbook or app note. Obtaining magnetics or other
> parts may be an issue.

I have already done so several times, up to a kW. I use Octave for the 
task. I prefer integrated magnetics, which implies custom parts. So I 
would rather not discuss a unidirectional mid-power resonant converter 
design, because I already can do that. I have no experience with 
something really high power (30kW+), but I have no need for that either.

> You might check some of Johannes Kolar's (large power) output,
> Ashoka K.S.Bhat,or the research papers made available by Virginia
> Tech (low power). Examples earlier than Y2K will involve
> semiconductor and magnetic types and structures that you will
> likely have no interest in exploring.

I think I have read most of the Kolar's papers, but his all efforts
related to bidirectional power transfer I know of seem to be 
DAB-centric. I believe that DAB is too complex for my needs (even just 
counting the parts: 8 switches instead of 4).

> These are trade-offs that can make or break a bidirectional
> power transfer application, just as they do battery or other
> energy storage or transmission applications. They have to result
> in advantages to justify their development or use.

Of course it always boils down to assumptions and design constraints.
Mine are very high power transfer efficiency (~97%) in one direction and 
decent one (92%) in the other. The power level is about 1kW again.
A bang-bang charge control LLC should handle the typical transfer 
direction pretty easily, but I am tempted to implement direction 
reversal with exactly the same structure. The sims say it should be 
doable, but google says that if I succeed, it will be a pretty unique 
construction. Being unique is not always an asset, hence my concerns.

	Best regards, Piotr

legg wrote:

> Topologies do not always scale - their relationship to the quality
> and cost of pracical components (with their limits and strays) often
> restricts applications to a pretty narrow range of power, voltage,
> current and frequency.

So I have finally found one example:

https://www.hpe.ee.ethz.ch/uploads/tx_ethpublications/06397282.pdf

 From ETH Zurich, so can be trusted.

	Best regards, Piotr