|7th January 2012, 09:37 PM||#21|
Join Date: Dec 2009
Sorry eem2am, i read you start thread with step,then i confuse "offline" or other topology that use one mosfet.
yes,halfbridge/full bridge may be a good SMPS. remains the problem of RFI / EMI and transient response of the amplifier.
efficiency is inverse to get a very clean output signal.
|9th January 2012, 03:07 AM||#22|
Join Date: Oct 2008
Location: Plano, TX, USA
These topologies have been running standard ATX/Server power supplies for years. Take a look at the On Semiconductor site and their resonant controllers (this is a voltage domain controller). This controller has been used in the ATX space for several years with a stacked winding approach using secondary side trailing edge modulation to improve the cross regulation (and in case you are wondering, I did help them develope this architecture). That means there were no post regulators on the windings. The 3.3,5,and 12V rails were AC stacked on one core. Try doing a 240W, 3 output tightly cross regulated flyback that has to meet all the ATX transient requirements and cross regulation requirements. It is pretty much impossible on a flyback. It is possible using a coupled inductor forward topology which is the prevalent low end ATX platform topology.
As to the split rails… By definition the +/- voltage of an LLC is the best regulated of any topology because the 2 windings are coupled together and wound bifilar over the primary (ideal transformer) so that both see the same flux lines as the primary. Since they are also coupled together they keep almost exactly the same voltage differentials across the windings (similar to coupled inductor approach but much tighter). Your least regulated topology would be a Flyback because the +/- voltages do not have the same current requirements at any one point in time and are totally dependant on the PWM duty cycle. The next best topology would be a forward type topology using a coupled inductor to achieve better cross regulation.
Your scrunched secondary comment concerns me. I don’t know anyone who is trying to deliver real power (in commercial application) that would design the primary/secondary using a segmented approach where you have the primary at one end and the secondary at the other with the spacer in the middle. Typically you may see it referred to in reference designs because that is the easy method but none of the major PSU houses would design production transformers that way. That is the least efficient winding technique. Typically if you need a higher leakage than you can get in the transformer then you design the transformer as a low leakage transformer and add an external inductor to serve as the leakage. As to winding the transformer as long as you bifilar wind the secondary over the entire primary (using tape between the layers to generate your leakage) you get windings that have symmetric current share and are well coupled. There is no need for a secondary L since you are using the primary side Lk for the mutual inductor.
I believe that the above addresses both you items 1 and 2 in reverse order. The flyback at 500W would be my last choice for a topology. From an EMI/Cost standpoint I believe the LLC topology would be my first choice, followed by an active clamp forward with a coupled inductor, and last would be the flyback.
From the standpoint of ripple, at +/-48Vdc I can expect about 300mV ripple for a 1KW design (500W per rail) and can do 0-10A step with <1V sag on the output at 1A/us slew rate. These numbers are from a 1KW, 48V design I did and not a +/-48V). Please try that with a flyback and tell me what you get. These are real numbers, not something that I generated from a simulation somewhere.
That being said, this is just my personal experience. I am not trying to convince anyone that there is a right/wrong way of doing something. All I offer is a bit of insight from a long career of doing this for a living at both the high volume/low margin (adapter) end of the business and at the low volume high margin (ATX/Server). Trust me on this when I say, I do not look here for my answers to the power supplies I design. There may be some here that do, and if so then I say good for them. At least they are asking questions. And since this is about the only forum I know of that offers newbies some free advice on power supplies I commend them for seeking out knowledge and acknowledging that they may not know everything.
As to the last section of your post about the power guys invading this space. Wow… That’s all I can say.
If you look through the plethora of posts here in the forum most of your solutions come from the very people you are ranting against. If I were an audiophile DIY person (and I am), I would absolutely love to have someone that does this for living take a look at my stuff. If for no other reason than he/she might spot something that will either save me weeks of time trouble shooting or save me thousands of dollars worth of equipment from blowing up because the loop was unstable during a transient response and it put 200V on the output instead of 48.
I am an audiophile EE who just happens to do power supplies for a living and am willing to contribute a bit to the forum. Please don't attack those who are willing to try and offer a bit of their sage experience to those that might be able to use the help.
|10th January 2012, 08:25 PM||#23|
Join Date: Jun 2011
Thanks for this answer, i'm sure i speak for others too....i'm on the move now, so have copied it and will read it later as im getting sporadic web access only for the next week.
|10th January 2012, 09:07 PM||#24|
Join Date: Jan 2008
can you please clarify on your transient response test outside the loop bandwidth?
Normally I test it at about 10Hz from 10-100% load (or 0-100% load).
If my smps has a fxo of 1kHz what are you expecting to see when toggling the output outside the loop BW?
About the 48V 1kW flyback I have designed one and I can tell you that the ripple at full load was around 80mV and overshoot/undersoot around +/-600mV from 2-20A load.
I had tons of low esr capacitors a the output and an output lc filter to reduce the ripple.
Its is 2 switch quasi resonant design powered from 3-phase mains at 400-500v and it has a fan for cooling. The efficiency is around 93%.
Its sounds crazy to make a 1kW flyback but at the end it works well and the cost is reasonable.
Thank you very much.
|10th January 2012, 10:15 PM||#25|
Join Date: Jun 2011
Though you kindly spoke of the operation at the upper resonance point.....however, over 90-265VAC , you cannot assure being at this point over that entire range.....the flyback is better at the wide input range........................
I notice that a number of off-the-shelf SMPS's for audio usage are LLC resonant type. They have no PFC.
IMHO, there are problems with this, because LLC resonant converters are are known for not having good line regulation......as the following explains under the "frequency and gain" section.................
Using quasi-resonant and resonant converters
I am not too sure how you are equating LLC resonant with "ideal transformer"....the LLC converter has a leakage term which you spoke of...........as you well know, the current in an inductor cannot be quickly changed and i fear that this will put the downer on your anticipated fast transient response.
Though your angle on transient response is totally new to me........the concept of the output being regulated without having to wait for the feedback loop because of transformer action......
....this sounds utterly marvellous.......i cannot understand why none of the books out there , and none of the application notes out there, even mention it.
The only advantages i hear of about LLC is that at the upper resonance frequency, you get zero voltage switch on of the fets, Zero current switch off of the fets, and zero current commutation of the diodes.........also , i regularly hear of the EMC advantage of the smooth sinusoidal current...........
.......outside those advantages, i have never heard a squeak about the kind of brilliant transient behaviour that you mention.......are the salespeople of the semiconductor companies just a bit slow? -is it that they just dont understand?, because this advantage that you mention in LLC converters is not mentioned anywhere in their literature.
Regarding the forward converter with coupled output inductor, do you mean that the output inductors are coupled with each other, and also coupled with the transformer too?
In a flyback, the output inductors are in the transformer, as you know, so they are nicely coupled in there.
Also, from what you have kindly described, the LLC is THE converter for Class D amplifier supply..........so why do powerint.com offer the following 200W *flyback* solution for there Class D audio amplifier supply....?...............
...powerint.com know all about LLC converters...they have their own LLC controllers.... but in the dedicated audio section of their website they have *not one* single SMPS that uses the LLC topology.......are they missing something?
Anyway, Mag seems to be getting excellent results from his high power flyback, so i am still not convinced that the flyback is a dog for Class D power supply.
However, i am not sure what Mag means when he speaks of doing a transient response test "outside of the loop bandwidth"?
Anyway, lastly i have a guilty confession..............
I have taken apart a Class D audio amplifier for guitar of 350W. (It says "350W" on it , but i am told that is apparent power, and that its 150W at the load)
The class D amplifier was supplied by an SMPS.
I was curtailed in my guilty investigation......but i got far enough with the screwdriver to notice that the SMPS was an SMPS with a single FET.
I am sure it was NOT an LLC converter....there was no resonant capacitor, and no "external" leakage inductor.
The output diodes appeared to be surface mount and seemed to have a heatsink over them.......................i was desperately trying to investigate it when i was pulled away.
I could only see the ONE TO220 FET on primary side......it had a tall (3 inches) heatsink attached to it...................-but there was another TO220 package device with no heatsink, and i was pulled away before i got a good look at it.....it wasn't terribly close to the single FET with the heatsink.....at first i was wondering if it was an active clamp FET, but as i said, i was pulled away..........maybe it was an active clamp flyback?
I also did NOT see any coupled output inductors, so that rules out Active clamp forward.
So it must have been a single switch flyback.......ahhh i remember now, there was NO RCD clamp resistors (unless of course, they were hidden on the bottom layer of the PCB) , so it perhaps could NOT have been single switch flyback, so i suppose it must have been active clamp flyback.........or perhaps active clamp forward with the coupled output inductors actually inside the transformer?
This amplifier was top-of-the-range.
I dare not tell you the name of the manufacturer because they will sue me, but i can gaurantee you that you will have heard of them....they are a big player.
Oh, and by the way, the mains input connector had 220VAC written on it, and there was no PFC, and the input electrolytics were two series 200V, 680uF electrolytics.
It really was a very sparse, small area SMPS..............i was pulled off before i could unscrew the PCB from the enclosure top.....there would have been more clues on the bottom side of the PCB but alas those did i not see.
The transformer was a vertically mounted ETD type......it looked like it was well "bloated" with layers.
Anyway, if the LLC converter really has these great properties, then why has this big player in the audio class D guitar world NOT used an LLC?
Also why are coldamp not using the LLC if the transient response and coupling of LLC really is that good?
Last edited by eem2am; 10th January 2012 at 10:26 PM. Reason: spelling
|11th January 2012, 11:26 AM||#26|
Join Date: Oct 2008
Location: Plano, TX, USA
Actually you can guarentee that the operation is always at resonance. If you look at the On Semi part you will see it is a fixed frequency part and you use an intermediate stage to muce the bus votlage up and dow to compensate for the IR loss.
That's why I said what I did about running in voltage domain. I did not mean it for the loop rather is was the fundamental operation of the LLC.
As to PowerIntegration, I have a lot of respect for these guys. That does not necessarily mean I agree with their design approaches.
You do not need coupled inductors in a Flyback because there is no freewheeling diode acion during the reset operation. For multiple output flybacks you usually apportion a certain amount of current to each output and sum them into the TL431.
For the forward I meant that the output windings are coupled together using a coupled inductor (see the old TI app notes slup059.pdf by Lloyd Dickson) so that they cross regulate.
I am not saying one way or another what topology you should use. I simply described why there are so many others using HB topologies which I believe is the original question.
This is not an attack ee2am. This is just a discussion.
I am sure what you was was a flyback ee2am. Good luck my friend it has been nice talking with you.
|11th January 2012, 11:30 AM||#27|
Join Date: Oct 2008
Location: Plano, TX, USA
You test outside the crossover frequency to prove to yourself that everything is still well behaved in the case of fast transients in the system. It is not uncommon to get very fast high frequency transients come through the system and if you aren't careful the loop can set up a sub-harmonic oscillation and this can go very wrong in a hurry.
Again, I have no problem with using a flyback. I was just responding to the question of why so many flybacks.
|12th January 2012, 06:43 PM||#28|
Join Date: Jun 2011
"At resonance a voltage domain LLC will always have a better transient response than a Flyback. "
......This statement is confusing me because considering a light-load-to-heavy-load transient.............well, the LLC converter would not be switching at resonance at light load, it would be far above resonance, so how can you say that
"At resonance a voltage domain LLC will always have a better transient response than a Flyback. "....................it won't be at resonance when the transient occurs.
LLC converters also cannot manage the large input voltage variation of 90-265VAC (even with a voltage doubler link)......LLC converters need a PFC front end.......and nobody is going to use a PFC for a 500W flyback to supply a Class D amplifier because its average power is 500/8 = 62W.
I hear you say that flyback is your last choice for 500W.
You have put forward EMC as one argument...but EMC testing for SMPS to supply Class D amplifier is done at peak/8 = 62W.
At 62W it is not going to be hard to manage EMC for a flyback.
If anyone wants to get a screwdriver and open up a 350W Fender Mustang IV Class D guitar amplifier you will see that the Class D amplifier is supplied by a single switch flyback.........also, you will see that the heatsink for the FET is not really all that big.......the heatsink is 3 inches tall but is narrow and slender.
(i would admit that ive been told that the 350W is the apparent power)
So why is flyback Fender's first choice?
I put it to all that its because the flyback is the best.
If they are using it , then it must be because it has a perfectly satisfactory transient response, becasue Class D audio amplifiers need a very good transient response from the SMPS.
You may not believe this, but i once worked with one of the best audio amplifier engineers in the world.....he was with the best in terms of his musical skill, as well as electronics experience in amplifiers....................when he spoke to me about SMPS for Class D amplifier supply, he stressed repeatedly that it was transient response that was the key point.
I cannot accept that a single stage LLC converter (which is obviously in voltage mode) has better transient response than a current mode flyback.
The flyback is working perfectly well in the aforementioned Fender amplifier.....so obviously the flyback has got what it takes.
...given this fact, why would you want to use an LLC converter which needs a PFC front end for handling universal mains?
...and why would you want a forward converter with coupled output inductors when a flyback is perfectly satisfactory.
500W Max power means 62W average power with Class D amplifiers.....thats flyback territory.
I hear what you say that people can do what they like .....and they can, but i reckon that most diyaudio types have an "Engineering mentality".....they simply wont want to implement a type of SMPS thats more complex than it needs to be, and more expensive.
I hear talk of the LLC with intermediate stage but why use an intermediate stage when a simpler way will do.
I think any amplifier hobbyist who wants an smps for their amplifier project can rest assured that up to 500W, they will be fine with a flyback.....they neednt have to become a diysmps enthusiast just to support their diyaudio desires.
Reading this forum in general (not the posts of the present correspondants), you would think the flyback was a dead duck for <500W audio SMPS.
|13th January 2012, 02:24 AM||#29|
Join Date: Oct 2008
Location: Plano, TX, USA
There are 2 fundamental methods and a hybrid method for doing an LLC controller.
Method 1: Frequency Domain Control. This method sweeps the frequency to match the output load impedance to the tank impedance on the primary side. It does suffer many of the drawbacks which you describe.
Method 2: Voltage domain uses a PFC boost followed by either a buck or a boost stage. This method runs the LLC controller at a fixed frequency all the time (usually at resonance). This intermediate stage is used to move the bus voltage up and down enough to compensate for whatever power train losses you have due to power requirements. Meaning at light load you might be at 373V and at full load you might be at 390V. The transformer is working as an ideal transformer in this mode. As long as your bus does not sag on the primary it will not sag on the secondary. Since the primary is a CV^2 issue than the secondary transient response is excellent. The DC/DC converter (buck/boost) stage can be current mode control if you want. So… you get whatever additional current you need as soon as there is demand and the loss you suffer is only the IR losses in the power train devided by the turns ratio. As an example for a HB with a bus range of 370-390 you would have a maximum sag on the output of (390-370)/(2/n) where n is the turns ratio. For a 20V bus using schottky’s and n=10 that would be 20V/(2*10)=1V drop. The buck/boost loop would then move the bus up to compensate out the 1V drop in the output. Its advantage over other topologies is that duty cycle never enters into the equation for the LLC.
Hybrid Method: This method uses both frequency domain (transient response output ripple rejection) and Voltage domain control (DC bus moves to hand the average power needs of the supply).
As to why Fender uses the method… If I had to guess it is cost only. As long as you are below 72W average power then you are not required to have a PFC front end.
I have absolutely nothing against using a flyback. I use them all the time. However, if I have a topology that has poor PSRR in the amplifier then I may choose a power supply that has inherently low ripple and harmonic content. Method 2 is VERY good at that. Especially if I am driving >72W continuously. If I were below 72W and still need +/- rails and inherently low ripple I might choose a forward with a good coupled inductor because I know that I will have better cross regulation and a lower harmonic content because of the inductor in the path to the output caps.
Finally if I needed cheap, fast, and I have an AMP with a good PSRR then I will use a flyback. In general… I would have designed the Amp first (with a good PSRR) and then chosen a topology to match the required PSRR.
As you can see I choose the topology from a system level first, cost second, compliance (EMC/Safety) next, and finally design complexity. However, I do this for a living so what is easy for me is not necessarily easy for others so that is why I leave the topology choice to others and add my 0.02 when asked.
I hope my comments helped.
|13th January 2012, 08:19 PM||#30|
Join Date: Jun 2011
OK Thanks Dtproff,
Regarding the Comparison for cross-regulation between forward-with-coupled-output-inductors and flyback, the following ti.com article speaks on it.....
..This article merely says that coupling the output inductors of a forward makes the cross-regulation better compared to a forward with non-coupled output inductors.
....Its notable that this article does not say that the coupled-output-inductor-forward is better for cross regulation than the flyback.
This article makes many comparisons but does not make that comparison.
I would say that this is because the coupled-output-inductor-forward and flyback have similar cross-regulation.
However, i would say that the flyback is better because the coupled output inductors of the forward.....well....the coupled output inductor principle no longer works in lighter loading when the inductor current goes discontinuous.
My only problem with the single switch flyback of the 150W Fender Mustang IV that i have is that i never got to see the bottom side of the PCB.
I am also foxed that i never saw any RCD clamp resistors.............assuming that they weren't on the bottom side of the PCB, well, it must have been a "resonant reset flyback"
.....i would have said that it was an active clamp flyback, ....however, the active clamp flyback requires a heatsink on the main primary fet aswell as on the clamp fet.....and there was definetely only one fet on a heatsink at the primary side.
The major threat to the hard-switched, single switch flyback for power of <500W is that the FET switching losses are high.
-obviously this can be mitigated by slowing up the switching transition of the fet drain........but it must not be slowed up so much that the fet needs a fan as well as a heatsink....because a fan can make audible noise as well as being expensive and a further failure risk...........so the fet drain voltage needs to have its switching transition slowed up just enough to enable the flyback to pass EMC at peak/8 Watts.
.......and then, how do you know what is the FET drain transition time which will allow you tp pass EMC at peak/8 Watts?
........i believe the answer lies in damping the FET drain switching transition just enough so that a fan is not needed.......you then surely will be able to pass EMC at peak/8 Watts.
The alternative is a two switch flyback
.......surely the two-switch flyback is gauranteed to not need a fan for <500 W audio usage (for class D amplifier supply).
The problem with the two switch flyback is that it brings the expense of the high-side pulse transformer fet drive.....which isn't that bad, but its another custom wound part that you're going to have to myther over to make sure its being wound with the correct low leakage inductance.
Yet still the Fender Mustang IV Class D amplifier foxes me......its SMPS....
1. On the primary side has just one FET on a heatsink.
2. Has no output inductors
3. Appears to have no RCD clamp resistors (unless they're on the bottom side (unlikely)
.....so it must be a flyback, but howcome no RCD clamp resistors?......or if active clamp, howcome no 2nd heatsinked primary fet?
.....thats why i now say "resonant reset flyback" for the Fender Mustang IV ...would you agree?
Last edited by eem2am; 13th January 2012 at 08:25 PM. Reason: spelling
|Thread Tools||Search this Thread|
|Thread||Thread Starter||Forum||Replies||Last Post|
|Hypex SMPS's?||mr_push_pull||Power Supplies||0||22nd April 2011 05:00 PM|
|Small SMPS's for preamps/DSPs||fb||Power Supplies||0||28th December 2010 03:10 AM|
|The difference of Push-pull,Half-bridge,Full-bridge||digi01||Chip Amps||0||8th September 2006 02:10 AM|
|Whats the difference between full bridge and half-bridge SMPS ?||skaara||Class D||6||3rd February 2005 07:23 PM|
|New To Site?||Need Help?|