I have had a project for an offline switch-mode power supply for a multi-channel power amplifier in mind for a while
While I wait for my business to ease up so I can work on it properly, I have time to think it over, in particular how to make it electronically quiet.
Cuk converters are often discussed as potentially a less noisy option but not without problems, R. Half Plane Zero(es) for a start.
I don't think this is a major issue for an audio amp, transients are fairly slow compared to the switch frequency and the Power Supply Rejection Ratio at audio frequencies should minimise the problems.
But one issue is how much the circuit stresses the (transistor) switches, best expressed as the required switch capacity Vstress * Istress compared to the output power Vout * Iout.
Call this the utilisation, denoted U, typically a function of the duty cycle U(D).
I found a bit of discussion of this in "Fundamentals of Power Electronics" by Erickson, available on the Uni. of Colorado website.
But it seems to have a few mistakes, some of them probably just typos, others I'm not sure.
I have included a short extract, anyone know which edition this is from? Edit: Looks like it's the 2nd Edition
And if there is an Errata list, or if it's different in the later editions? Edit: Looks like it is different in 3rd edition. Not sure if just moved or edited and corrected.
The isolated Cuk looks fine for utilisation of the switch, actually a fraction better than the more common (isolated) Full B. or Half B. if the table is correct...
David
While I wait for my business to ease up so I can work on it properly, I have time to think it over, in particular how to make it electronically quiet.
Cuk converters are often discussed as potentially a less noisy option but not without problems, R. Half Plane Zero(es) for a start.
I don't think this is a major issue for an audio amp, transients are fairly slow compared to the switch frequency and the Power Supply Rejection Ratio at audio frequencies should minimise the problems.
But one issue is how much the circuit stresses the (transistor) switches, best expressed as the required switch capacity Vstress * Istress compared to the output power Vout * Iout.
Call this the utilisation, denoted U, typically a function of the duty cycle U(D).
I found a bit of discussion of this in "Fundamentals of Power Electronics" by Erickson, available on the Uni. of Colorado website.
But it seems to have a few mistakes, some of them probably just typos, others I'm not sure.
I have included a short extract, anyone know which edition this is from? Edit: Looks like it's the 2nd Edition
And if there is an Errata list, or if it's different in the later editions? Edit: Looks like it is different in 3rd edition. Not sure if just moved or edited and corrected.
The isolated Cuk looks fine for utilisation of the switch, actually a fraction better than the more common (isolated) Full B. or Half B. if the table is correct...
David
Last edited:
I'm afraid the proper topology would be HB or FB LLC, preferably with charge type control. There are Fairchild and TI parts for that. The main part one needs to perfect would be the LLC transformer besides proper loop and layout. Design/application notes to the rescue. There is a guy Nicola Rosano who made pretty good Exel tool to develop power part of LLC. Hope you are familiar with gate drivers and basics of SMPS. Unfortunately higher frequency would not allow one just to neatly arrange few parts on PCB/ chassis to get a proper result. Whether or not you need a PFC would be yet another story.
Do not try Cuk. It's a very peculiar topology.
Do not try Cuk. It's a very peculiar topology.
A resonant LLC FB is one of the more obvious options but has it's own trade-offs.
It pumps around reactive power so that the stress on the transistors is worse for a specified output power, yes?
Cuk is "peculiar" is kind of an uninformative statement.
It looks easy to do a Zero Volt Switched CUK to reduce the switch loses and EMI.
It would have the benefit that the switch transistor is low side so no driver ICs to worry about.
And only one switch transistor so potentially simpler and more reliable.
It looks so nice that I am a little suspicious that it's not more common, hence the thread here.
But I'd like to analyse it from solid fundamentals like power out versus power switched rather than just because it's called a bad name.
Best wishes
David
In short, no. The reactive power is needed to achieve ZVS. Newer devices have really low static losses allowing some more current to circulate without a penalty. LLC in particular made 92-96 per cent efficiency in serially produced PSU possible.
What can I say: You've been warned.
There's a tiny amount of reactive power to achieve ZVS, but the downside is that LLC is closer to a sine wave compared to the square wave of a conventional pulse width modulated converter. Compare the peak load of a sine wave with the peak load of a square wave with the same RMS value. The sine wave puts more stress on the switches, this is usually understood.
Hopefully the improvement from the ZVS is worth the cost of the increased peak volt stress but depends on the details of the application.
What should I reply?...Ooh, scary?😉
To return to the reason why I started this thread, can you or anyone else explain the apparent errors in Erickson?
Best wishes
David
Scary would of been too easy.
You would not get the results you might of gotten otherwise.
You are correct, form factor of a 50/50 PWM is equal to 1.
However, as mentioned, conduction losses of newer devices like trench Si and SiC fets
are a small part of a total ones, thus small increase of conduction losses yields significant decrease of a switch ones.
That's said ZVS gives twofold gain:
1. Total losses are less, despite the "poorer" form factor of a current/voltage etc.
2. Harmonic content of a) voltage across the switch b) current in rectifier (sync switch) is better.
You may have no diode recovery losses as well. Lower noise, easier filtering and so on and so forth.
You would not get the results you might of gotten otherwise.
You are correct, form factor of a 50/50 PWM is equal to 1.
However, as mentioned, conduction losses of newer devices like trench Si and SiC fets
are a small part of a total ones, thus small increase of conduction losses yields significant decrease of a switch ones.
That's said ZVS gives twofold gain:
1. Total losses are less, despite the "poorer" form factor of a current/voltage etc.
2. Harmonic content of a) voltage across the switch b) current in rectifier (sync switch) is better.
You may have no diode recovery losses as well. Lower noise, easier filtering and so on and so forth.
Possibly, you have any data on this? simulations?
I think there is no problem to have a quasi resonant Cuk converter with ZVS switch-on.
A resonant circuit like the LLC buck can also have ZCS switch off and trade lower switch-off losses with worse form factor.
My suspicion is that at lower frequencies the trade-off may not favour the LLC, simulation will tell.
An LLC Cuk seems possible, is attractive intuitively.
Dr Cuk himself has done work on this and there's a bit of published analysis that I need to look at and think over.
The fact that it isn't commonly used may be a clue that there's problems, or perhaps just that the world is slow to catch up.
Best wishes
David
I would stay with KISS principle.
Which means forward/buck converter and derivatives.
Unless you do PFC where boost may be a good fit.
P.S. Dr. Cuk is not the person I would take advice from.
P.P.S. You insist on your solution time after time , so why bother?
Which means forward/buck converter and derivatives.
Unless you do PFC where boost may be a good fit.
P.S. Dr. Cuk is not the person I would take advice from.
P.P.S. You insist on your solution time after time , so why bother?
Sure, me too, if it doesn't hurt performance.
Let's compare
PFC is practically required for a substantial power supply so I will assume it.
And also that it's off-line and therefore needs to be transformer isolated.
Assume no synchronous rectification to keep it simple.
So your typical solution is a boost PFC before the LLC (HB or FB), then transformer and secondary rectification.
1 PFC boost transistor+ 1 PFC controller +1 boost rectifier diode.
Then the buck LLC section.
2 or 4 LLC transistors, half are hi-side so 1 or 2 hi-side drivers + 1 LLC controller + 2 or 4 secondary rectifier diodes.
Cuk can do boost and buck so has
1 switch transistor + 1 controller + 1 secondary rectifier diode.
Even the simplest LLC option (Half B. with centre tapped secondary) has 3x more transistors, 3x more diodes, 2x the controllers and has 2 hi-side drivers that Cuk doesn't need.
The transformers will be practically the same, as will most of the other stuff.
Cuk liked KISS too, that was kind of his point.
The other point is that all the extra transistors and ICs and controllers and diodes eat power and waste efficiency.
Well, just because he did his PhD in power converters.
Invented the State Space technique that was a fundamental break-thru.
Was professor of power electronics at CalTech for 20 years.
Literally "wrote the book" (in collaboration with Middlebrook) on the foundations of power converters.
And ran a power electronics company for about 40 years.
Doesn't mean he never made mistakes, of course.
But unless you have some evidence I am inclined to take his advice rather than yours.
I could be mistaken, and it helps me think about the problem: as I try to explain clearly to others I have to clarify my own ideas.
Also it keeps the thread active, I still hope someone will read it and reply to the question I asked.
Best wishes
David
Last edited:
Look it's going nowhere: you decided to make it personal. I do quit.
There are groups at LI and FB where one can directly interact with the person in question,
but you, for the reasons unknown, continue to argue with me.
In regard to advice: look which topology industrial PSUs are using the most and see
how many Cuk's are out there, pun intended.
Spoiler alert: At low power it would be flyback and LLC otherwise.
"I have to clarify my own ideas."
Which are?
It well might be that I do not have collateral interest in Cuk converter(s) but Dr. Cuk might.
There are groups at LI and FB where one can directly interact with the person in question,
but you, for the reasons unknown, continue to argue with me.
In regard to advice: look which topology industrial PSUs are using the most and see
how many Cuk's are out there, pun intended.
Spoiler alert: At low power it would be flyback and LLC otherwise.
"I have to clarify my own ideas."
Which are?
Not at all, you said you wouldn't follow Dr Cuk's advice. I pointed out his excellent credentials have earned him credibility.
If you have evidence to support your opinion, or a CV to support your credibility then I'm happy to learn it.
This is a reasonable point, and part of the reason I started a thread.
I expect it's partly because Cuk had a patent on his idea and it was cheaper to use a non-protected circuit.
Once that starts it can self perpetuate, everyone does what everyone else has done because "that's the way everyone does it".
But the lack of take up on some of his later ideas does make me wary.
1. It looks pretty simple to use an LC resonant circuit to have ZVS turn on for the switch transistor, which would reduce losses and EMI.
This could usefully include the transistor capacitance and stray inductance as a feature, not a problem.
2. I think it's possible to stick an LLC resonant circuit into a CUK to have similar benefits to the usual HB or FB LLC, but simpler.
That would have both ZVS turn on and ZCS turn off, for even less loss and EMI.
When I doodle it, there is a close resemblance to a circuit Cuk released a few years back, which is no surprise.
His is even more tricky and I don't quite follow the details yet, but as I "reinvent the wheel" (or his wheel anyway) I understand how he arrived there.
3. Control loop of the usual LLC is fairly clumsy, just frequency variation. A time domain approach would allow independent control of both time on and time off for better optimisation. Quite feasible now with fast micro-controller,, there are papers and application notes on this.
Not simulated yet, still too busy at work.
Best wishes
David
Thank you, I think the ripple cancellation trick is a nice bonus if the converter is reliable and efficient.
It is not a crucial feature in my application:- to use the output to drive Class H amplifiers which switch the rails anyway.
So low ripple hardly matters to me, however nice to have for other uses.
But the presentation is quite informative about other issues too.
No, I studied Science.
So call it data or evidence, in either case my profession (and my habit) is to try to find out what's true in reality, not what people's opinion are.
I know this is not a popular idea but I hope it may catch on one day.
The company is out of business, you have any information about their tech.?
I don't really care about compactness, but it's a useful hint to power efficiency.
Best wishes
David
Last edited:
Off topic
You've been told this is not the place to consult with Dr.Cuk and that neither I nor anybody else owe you "evidence" or other wise but you insist. Oh well, ignore list
Off topic end
P.S. How come the bunch of MIT professors failed to deliver freaking ac/dc stick being beaten by Chinese mass manufacturers. Get it a thought instead of picking on diyer(s).
You've been told this is not the place to consult with Dr.Cuk and that neither I nor anybody else owe you "evidence" or other wise but you insist. Oh well, ignore list
Off topic end
P.S. How come the bunch of MIT professors failed to deliver freaking ac/dc stick being beaten by Chinese mass manufacturers. Get it a thought instead of picking on diyer(s).
I had a closer look at this and noticed that the author discusses how to select the main capacitor value to exploit resonance, just as I proposed.
Nice, thanks!
It looks like a PDF of a presentation, a bit hard to work it out without the spoken explanation but a clue to the correct direction.
You have any more information or ideas on this?
Best wishes
David
Yes, that is the plan, more or less.
It's for a home theatre, with bi-amped active speakers, so a lot of of amplifiers.
That makes it less practical to put a separate SMPS on every amplifier, to follow it's specific audio.
There is simplicity and economy of scale to do all mains related functions in just one module.
So I plan a central SMPS to handle turn-on, soft start, current overload limit, input transients/spikes and EMI, over volt protection, power factor correction, mains isolation, etc.
That's quite a lot of functionality taken care of, so each amp module can have a simple(r) switch-mode rail control.
The rail tracker is self-oscillatory, like the best class D amplifiers, maybe call it Class HD.😉
I have done LTSpice sims for this part and it looks very nice.
But not yet happy with the SMPS, hence my enquiries, thanks for your contribution, I did indeed find it helpful.
Best wishes
David