NOTE: There appears to be quite some debate as to whether this NGP800 PSU is actually a resonant converter topology, and yes, I need to take a closer look at this. But just go with the flow and pretend it is for the purposes of this video.
i can excuse the crudity of the models because there is no scale BUT, i have to argue that you obviously did mspaint it! I wonder if LC power supplies will become the norm as more and more power is required from smaller supplies, there is companies boasting 100 watts from a cell phone charger sized wallwart now with usbc out!
I doubt if they use LLC, I cant see the resonant Cap, and with inductor that big, its more likely to be a phase shift full bridge. The CoolMOS CFD7 can be used for LLC and PSFB as well, as long as you dont need body diode force commutation capability
Hi Dave, once again, thanks a lot for this great video! I do power supply design for quite a while now, and I think in the R&S it's done this way: since there are two current sense transformers, two transformers which look like they do gate driving, two power transformers, four FETs and four diodes, I guess it's two separate LLC stages. For one stage with two power transformers one current sense would be sufficient. I think the two stages are phase shifted. This makes sense, to counter for the LLCs big output ripple currents. Also, in nearly all my LLC designs, I integrate the resonant inductor into the power transformer. Because LLC stages are very bad at wide input to output voltage ranges (they work best at a fixed in/out ratio), I would suspect the large inductor with the spare transistor down right forms a buck post-regulator. Maybe the missing free-wheeling diode is below the board... Anyway, I hope you well have time to investigate this further, I'm very curious how they have done this. Greetings from Germany!
Excellent video! I learned alot here, mostly about why I don't use resonant mode converters. Too much math! (I can see it taking a team weeks to get the math for something like that bench supply sorted out.) Anyhow, not useful for small MCU projects of mine that typically consume
It's been 25 years since I worked on one of these! It was for a US Navy submarine reactor control system, where space and heat were both huge issues given the power density we needed to meet (there was no room for conventional supplies). After a failed prototype followed by a failed tweak to it, we got a successful design only after we gave up trying to use pre-approved MIL-spec magnetics and made our own. Literally: We hand-wound the first two sets. It was a design competition between two companies, and the Navy was impressed by how much smaller and cooler our supplies were compared to our competitor. In the end, we didn't get the contract. But we sold our supplies (and other tech we pioneered) to our competitor for an eye-watering markup. Turns out we made the profit we hoped to earn on the entire project just by selling some parts and licences. The engineering sadness at the loss was soon overcome by the financial win, and the sense that we had ducked a bullet: We likely would have lost money had we won!
@@edinfific2576 The power supplies were the only hardware we sold on the project, which was about 5% of the revenue. The other 95% of the revenue came from other tech we developed then licensed (it was a nuclear reactor control system, after all).
@@combin8or I was actually a software engineer at the time, but I had been a Navy and commercial nuclear technician before that, so I worked on the project in multiple roles. When the supplies started smoking, I was pulled in to help troubleshoot (because the supplies were controlled by an 8-bit microcontroller I wrote the code for, and we all know who gets the blame when the smoke leaks out), and then helped on the root-cause analysis. I did no electronics or magnetics design, but when it came time to make the prototypes, I was the only one with winding experience, so I got to lead that team. (Well, it was me, a technician, and an intern. But still.) The original design did direct-from-AC conversion, to avoid rectifier losses and minimize the power factor. Once we had the better magnetics, we were able to use a slightly more conventional design and still meet those goals. I can remember bits and pieces (it was a quarter century ago, after all), but nowhere near enough to reconstruct anything workable. I do recall the microcontroller interfaces and their signaling, but that's about it. In the final design, the microcontroller did little control, and lots of monitoring and reporting. That was an interesting project overall. I was a junior software engineer who was given some impressive chunks of the system to design. Other software engineers were senior to me (and older than me), but had fewer high-level responsibilities, which made for an "interesting" work environment. Basically, I sucked at that aspect of it, meaning management had to back-fill my lack of people and leadership skills. Ugh. But the real key turned out to be me improving my writing skills, so I could generate design documents that created less discontent because I learned to better explain myself. Trying to do that face to face was not optimal... Only three Seawolf subs were built (out of 22 planned), which was why we wound up being glad to have lost the competition: We never would have made the kind of revenue or profit needed to stay in that business. But the supplies were an unexpectedly good side-line for a while, until the engineers moved on to other projects and the designs were sold for a pretty penny.
These application notes are better than some books! I imagine an engineer putting all his teaching skills into the papers since none of his younger colleagues is interested anymore into how the stuff works they are making.
There's no where else on Earth that a hobbyist can get content this damn good! I paid $40 for a Beginner's Electronics class for my nephew and the content in that class isn't even remotely as good or educational as the information on this channel. Dave, you sir are a legend! It can't be overstated how much we hobbyists appreciate the information you provide.
I beg to differ. Not only hobbists. I'm a practicing engineer myself, and I'm still waiting for a bad video from Dave. (April fools bullshit are the exception, he should knock it off)
My team and I designed a bidirectional LLC converter for an EV, getting it working both ways was crazy. Custom magnetics, custom math. You end up with LLC because of EMC and efficiency targets, and they suck at regulating fast moving targets.
Dave, you remind me of all the old hat professors I had in college, worn smooth from years of experience in industry. You are a wonderful and entertaining teacher, and a treasure amongst makers and engineers. Thank you for everything!
The first assignment in my resonant converters class was to design one of these, and you got it spot on. And you're right about it being hard, it was painfully difficult, especially if you are designing controls as well
I’m speechless) The most complete analysis of an LLC architecture what so ever! There are two transformers because they took two standard half-bridge transformers and connected primaries in series and secondaries in parallel (there is no need to calculate a new transformer, primary L is doubled, secondary L is cut a half win-win)
LLC resonant converters are actually getting pretty common in the PC power supply space. Nearly all 80 Plus Gold/Platinum/Titanium models are use a LLC resonant primary. Many also use synchronous rectification on the secondary side. Even mid-range models like the Corsair CX450 (which is "only" 80 Plus Bronze) have moved to LLC resonant topologies.
those Gold and up PSUs uses LLC for efficiency, i guess?. there's not much bronze or 230V standard rated PSU uses LLC even today, sadly. did CX450 uses LLC? i thought it's only the 2nd gen CX550, 650 and CX modular series uses this topology
Actually, it is the high-current switching transients that generate high-energy noise. Resonant mode supplies switch at near zero current and thus generate much less noise. Resonant mode supplies DO vary the driving frequency. This is how they change the current that flows from a constant input applied voltage amplitude. This does not present a noise problem and, in fact, is equivalent to dithering a microprocessor clock to spread the noise over a broader spectrum
Nice review. Yes, this is actually my day job designing Power Supplies. I’m actually doing my first LLC design now, so I found this video in my research;)
Normal switches' drains begin at the input power voltage and must conduct current until they are open again at the input power voltage. At the beginning and end of these time intervals, they also have non-zero drain-source voltages. Since P=VI, there are substantial power pulses at the beginnings and ends of these pulses. In resonant designs the series capacitor stops the current every half cycle and allows the switches to change state without dissipating these pulses of power.
These videos right here are worth the price of admission. A+ for guys like me who are looking to learn. Thank you Dave for all the work you put into all of this as well as your enthusiasm. It really means a lot.
16:37 What're the odds those two are both transformers, but the secondaries are wired in parallel for the 32V outputs, and in series for the 64V outputs? 16:55 A big fat inductor (on a *black* core) with a MOSFET next to it, takes my mind towards a mag-amp type regulator. Can those be made with an adjustable / variable output voltage, or do they need to be designed for a fixed value?
I have some power supplies from Delta Electronics that do this as well. They have a secondary transformer set that can be switched between series and parallel depending on the load and regulation.
fantastic video, been getting my feet wet with switching supplies (just built some basic buck/boost stuff) but always interested in learning more about what's out there...
MOSFET turn on..volts across it drop to low value..the current through it rises from min to load demand max.. & vice versa at turnoff.. THE MOSFETS NEVER SWITCH INSTANTANEOUSLY... BUT WITH DEFINITE RISE & FALL TIMES.....ramp up/ down of V & I each in a few microsecs... times that KHz sw.freq. rate...& it is something for the designers to worry about... That AREA UNDER THE CURVE(slopes) ARE THE POWER SWITCHING LOSSES... ...the device "R..D-S on" times the load current sq...... { ( R× I sq. = W) for "one switch on" period } are the " conduction losses.....
Back in 1990's we used the " blocking oscillator " type power supplies.. rated around 80W.. +5V@6-8A...+/-12V @2A....these ran rather cool.. and had good line & load regulation..... don't know why is lost out.. but we had to upgrade to higher power later...
Resonant LLC is the topology of choice for just about every AC/DC power supply from 100-400W these day. It’s also really not that expensive at this point, and I’d argue that it might be even cheaper.
Maybe I missed it but I didn't hear anything about the disadvantages of llc compared to hard switching topologies in the video. I'm referring to the increased size due to the resonance network and extra required output capacitance.
Could also be a programmable current sink, to help with overshoot and the output voltage regulation, or to dump energy when you command a drop in voltage, or when your load imposes current on the output for some reason. Lose energy in the saturation of the inductor, or in the diodes and transistor, till the input side feedback can compensate for it.
The final MOSFET you found may be to have a linear regulator after the switching one. This would allow very precise output control and could be made to take only a few volts across. Makes sense?
There are also LCC converters having a second capacitor in parallel with the transformer primary instead of a second inductor in series with the transformer primary.
It's not my day job, but I just joined a research group that solely designs resonant converters, and I am watching this video as sort of an introduction to the topic... it seems quite overwhelming! Are there any other introductory resources you may recommend?
I don't know if anybody else has mentioned that the LLC converter output voltage can only be varied from 100% down to about 60% and therefore incapable of providing control down to zero output voltage .I suggest that the LLC converter feeds a buck converter - which is that FET that you are not sure about and the large filter inductor .
Did some testing a number of years ago on a hyper efficient power converter, of several KW output. Used in Solar car race, Darwin Adelaide and from memory efficiency ran at 99.8%. Quite certain it was a resonant design as output harmonics were also extremely low.
@@simontay4851 sure was a surprise when evaluating the results. Had to build a 3KW adjustable load, at minimal cost. Would have liked more time to design a electronic load, but made do with large 'gold' power resistor chain on a long tube heatsink an fans forced air flow. Had a hurry as decision on using these converters was critical to be ready with race setup. Contributed to success of the local car that year!
I believe that the resson of why there are two transformers is for not saturating the transformer (create larger Ae value on the primary side in series and parrallel the secondary side to reduce the condtuction losses).
Please pay attention to the fact that the LLC converter does not have an output LC filter due to the ZCS at the output. Nevertheless, a solid inductor is installed on the board, so I would bet it is PSFB topology. Please let me know if I am wrong and there are variations of the LLC with the output LC filter.
I would consider the possibility that there are two transformers in 2x Lm is used at resonance, and as suggested the secondary is either a centre tap output or as suggested by another comment used series and parralel for 64 V and 32V respectively
More video's on this subject needs to be covered. practical one as well. I know of someone who used a PIC16F177x/PIC16F178x part and designed a LLC controller in firmware exceeding features found in jelly bean parts.
Unless it’s for educational purposes, designing a resonant controller by yourself seems like a huge waste of time. NXP and TI already make extremely reliable controllers that work up to 500KHz for
@@Eng_Simoes 400KHz? maybe yes, but what does that frequency got to do with anything?. Fully resonant mode is not related to some high frequency standard only a factor related to core size reduction that's it. its all to do with the transformer core material used + LC tank. We can achieve full power at resonance with frequency as low as 63Khz using that PIC.A cheap modern design can be realized with EPCOS material N96 operating at 150KHz with cheap mosfet switching devices and a tight layout.
@@yaghiyahbrenner8902 energy is 1/2 C×Vsq. .. and/or 1/2 L× Isq... AND POWER IS ENERGY times FREQUENCY.... so higher the Freq. ..higher the pwr. for the same L/C & V/I also higher the operating Freq. .. smaller physical sizes of L/C components.. The limitations earlier being reliability & capability of the then AVAILABLE mosfets.. dedicated controllers.... & low economical sense to change over from the well established SMPS " industry standard "... Today we have SiC mosfets which are way ahead of conventional insulated gate mosfets..and as they become more acceptable & price competitive.. ..EXPECT TO SEE KW. OF POWER handling at UPTO 5-10 MHz in resonant modes...
Is it true that resonant mode converters are less noisy? I think I read somewhere because they are soft-switched, there is less switching noise on the output.
I'd like to know, too! I've heard this, but never from an authoritative source. If true, resonant mode converters also might be useful for sensors that are sensitive to switching noise!
A LLC-Convert alone isn't able to full fill the Output voltage requirements from 0 to 250V. As the Gain Curves indicates, they will struggle at low voltages and light loads withou large frequency variations. I can't see a resonant capacitor. These Caps are locking more like input filters for the H-bridge. A third stage is common to achieve the needed voltage range but there must be a beefy diode anywhere. It could be also some kind of protection circuit for the load. I would choose a different topology like a phase shift push pull converter. But how knows (=
Jesus H Christ. I must be dumbing down by being in home office for all this time. I was looking and looking at this simplified schematic datasheet page, that refers to LLC everywhere - wondering what LLC was an abbreviation for xD I stopped wondering, glanced over the schematic itself, and then it dawned on me... Good job, me - good job.
Just take a look at a modern, higher efficiency (Gold or Platinum) ATX Power Supply. Most of them are LLC-Rensonant mode Converters. Though they usually don't have a big coil behind the Transformer... A cheaper one might be better. Something like Silverstone SST-ET550-HG, as that's a single layer PCB one. And cheaper...
Despite the fact that I usually design other devices including microcontrollers (I love C++ and C#) I also fell in love with DC/DC converters design many years ago. I began with buck converters, then forward, and half-bridge. I'd like to build LLC converter but it seems to me it would be beneficial to design low voltage LLC converter first (powered from 24V instead of 230VAC) so that would be safer at the beginning. LLC converters are much more difficult and any error in your design may result in component explosion, especially at high voltages such as 230V AC for instance.
I too am an EE and bot MCU and destop apps programmer, and design all sort of PSU's and small control sysems with microcontroller for which I use mostly C, C++, and whenever possible Pascal. And yeah, get the circuit wrong, and mods/parts explode, which for me is a complete loss because I do a redesign - expensive, but being a full time contractor, I must have happy customers. On a tangent, and only for Windoze desktop GUI apps. Try Delphi, you'll learn to HATE C++. C# is a mixed bag for me, even though it is designed by the same guy that did Delphi before he sold his soul to M$. It has the CRAP syntax of C++, but has a lot of nice abilities, BUT it is suffering from "project creep", i.e. too many not too useful but COMPLEX abilities (which Delphi is now following - BAD and STUPID MOVE - KISS IS BETTER) that are better off not being there ever! I started with many Assembler, then a bunch of HLLs like Basics, PL/1, Pascal etc. etc. Then C, then C++ and finally Delphi. All this over four decades, and for productivity, READABILITY (brackets are EVIL becaus the brain processes words like begin/end MUCH MUCH faster due to context not being required to recognize them), there is NOTHING that comes close to Delphi, ...but it is an unlearning curve from C++, to which I was addicted to for a very long time, went to rehab :), got much wiser and after hating Delphi at first after it was pushed onto me, several several years I good at it and now NOTHING comes close to how fast and ultra readable I code compared to C++ that I was so in love with before ...NOTHING! Because of the FALSE LYING POS M$ propaganda about Delphi, as M$ does with better products that are not theirs, most programmers never looked at it in the last decade! VERY SAD but my gain, especially when it come to high response GUI's. Try doing GUI's under any DotNET language without spending too long on it (apart from the cookie cutter and limited wizard stuff). I mean truly custom schmik designs my customers cannot get anywhere else. I designed a very complex multi monitor system for $50k price ($30k + $20k later extensions), took two and half months combined. Two decades later, due to M$ evil ways, they wanted it converted to C#. I told them it would has laggy GUI (because 6 level deep layered GUI with huge inter module updating), which DotNET would comes to crawl with. Anyway they got an initial quote of $350k from a C# house ...yep, and the quote was about right too, because I also quoted for it at $400k becaue DotNET just could not do most of the GUI stuff well so many Custom Controls would have been required. The DonNET GUI is only good if you stick to common GUI needs that 90% of projects need. Anything out of the ordinary and you are FUCKED for performance and programming time! ...so here is my triggered rant! :):):)
@@stevenbliss989 Well, I used Delphi mamy years ago, starting with version 1.0. I use C/C++ for embedded systems. C++ is complicated but I'm getting used to it. However in many cases is better than pure C especially when I need to write complex firmware for devices that include TFT displays with rich GUI, FAT file systems and so forth. In my company some engineers use Python but it looks weird to me and I don't want to use it in my designs.
@@rk-dj4sg Not a fan of Python either, it has it's place, and is very useful for quick and dirty automation and/or data analysis, but I would NEVER write a real program in it, for many reasons, including it being an interpreter.
@G E T R E K T Before Delphi there was only C++ or VB (BLAAH!!!) and Turbo Pascal, plus some oddballs like PL/1 & Modula 2. Sadly I never used Turbo Pascal, BUT SHOULD HAVE back then, it would have saved me a lot of time and agro! :( ...I despise M$ for essentially killing Delphi! ...NO FORGIVENESS!!!!!!!!!!!!!!!!
I understand this is being applied to a power supply per say, but what this same H bridge mosfet topology were being used for a 'transformerless' inverter & output? Would this LLC design work or be applicable then?
it seems very odd that "all" that power would be completely directed thru that 1 inductor thats a lot of power just thru those winding for all of the outputs . Am i incorrect to think that ?
hello, i'm simulating this LLC resonant convertor in simulink ,how can i implement pi controller to achieve both constant current and constant voltage at output?
did that have one of those self made opto isolator like the electric fence??? with the UV led & the light sensitive transistor in heat shink.. Or am i seeing things(resistor) TOP LEFT???
Dave showed them in the previous video. They are current transformers for current sensing. The giveaway is that the "secondary" winding inside them is only a single turn.
The ones with the thin wire and exposed green core are current transformers, as was mentioned ten months ago. The ones with yellow wire are gate drive transformers.
Didn't want to release two videos in one day, but made it available to those who bothered to follow the link. Zero views might just be stat lag, or you are the first.
@@EEVblog An LLC is just a particular type of resonant converter, not every resonant converter is an LLC. Full bridge LLC is also not as common as half bridge LLC which can be used effectively up to more than 1kw power level. This seems not at all an LLC but more a Phase Shifted Full Bridge converter (PSFB). An LLC won't have an inductor on the secondary side while a PSFB needs it. An LLC will work with good performance at fixed input and output voltages. Here this stage needs to regulate from 0v to 32v (or 64v).... Good luck with an LLC converter, the needed frequency variation to keep the output voltage regulated will be impossible to achieve. I suspect this is a PSFB with 2 identical transformers wired with the primaries in series and the secondaries in parallel or in series to differentiate the 32v and 64v models using the same hardware. For more information on PSFB check TI app notes about UCC3895. Great content as usual.
Do you new Guys what kind of smps I need to use for a tube/amp hybrid amp? Can I use one supply for the 65 x 2 15 amps, 12,8 volts and 6.3 volts 5 amps, 350 volts 0.10 amps, 2 x 120 volts 0.25 amps. all on one transformer. LLC is nice, but did read LCC do better here., for feedback, open loop the best way? the transients from a audio amp can never be corrected, feedback to slow, a possible way, current feedback use, injection on chip triangle oscillator, did read some about that. I can use also fase shifting and other topologies but as fat LCC has soft switching, and I need as low as possible emi because it is a tube input/driver amp, filtering is not that difficult with the low amperages for them. I can also use separate smps for the 65 x 2 15 amps and the other voltages make it more easy. THanks for advice.
Daves explanation is a little incorrect in a number of places (LLC converters are not fixed resonant frequency for example). For a better explanation, go to Ben Yaakows youtube channel: th-cam.com/video/tCEqm-RoP20/w-d-xo.html
10:45 Absolutely false. Typical CFL drivers use BJTs, and they operate almost exactly as an LLC converter, just don't transform the voltage. And the board you have shown is extremely unlikely to be an LCC converter. There is no sign of the resonant tank, and the output inductor after rectifiers is not just unneeded in LLC, but ruins it.
Is there someone making noise near you? I'm in the middle of nowhere and it sounds like someone is on my roof. I should have stuck to the cheap headphones instead of upgrading.
gah, i still cant quite get my head around this. everyones chasing regulated outputs and stuff like that. im chasing current. plain and simple. as much as possible. i cant stae my output voltage as technically, there isnt one... or its dictated by the load, coupling, temperature, blah... i dont give a damn what it is as long as its being fed lots of current. induction heater. i think im getting a grasp on it, run through various calculations, all looking good... but damn i am releasing a LOT of smoke in the testing phases! and none of its coming from steel heating up. if it runs unloaded, it blows up loaded, and vice versa... meh, another two week wait for new components... weee!
..the Sanjay Manektala of ex. Crompton Greaves Mumbai1990-92...ex....NSC.... with that chain smoking habit...??? he had written (at least)a book on p.s...???
Infineon data sheet: "Competitor A", "Competitor B"... Why not name them if you're confident about your numbers? I don't like that. Either disclose the names or mind your own business!
NOTE: There appears to be quite some debate as to whether this NGP800 PSU is actually a resonant converter topology, and yes, I need to take a closer look at this. But just go with the flow and pretend it is for the purposes of this video.
I live infineons application notes. You can teach a college course from them lol
i can excuse the crudity of the models because there is no scale BUT, i have to argue that you obviously did mspaint it!
I wonder if LC power supplies will become the norm as more and more power is required from smaller supplies, there is companies boasting 100 watts from a cell phone charger sized wallwart now with usbc out!
I doubt if they use LLC, I cant see the resonant Cap, and with inductor that big, its more likely to be a phase shift full bridge.
The CoolMOS CFD7 can be used for LLC and PSFB as well, as long as you dont need body diode force commutation capability
Hi Dave, once again, thanks a lot for this great video! I do power supply design for quite a while now, and I think in the R&S it's done this way: since there are two current sense transformers, two transformers which look like they do gate driving, two power transformers, four FETs and four diodes, I guess it's two separate LLC stages. For one stage with two power transformers one current sense would be sufficient. I think the two stages are phase shifted. This makes sense, to counter for the LLCs big output ripple currents. Also, in nearly all my LLC designs, I integrate the resonant inductor into the power transformer. Because LLC stages are very bad at wide input to output voltage ranges (they work best at a fixed in/out ratio), I would suspect the large inductor with the spare transistor down right forms a buck post-regulator. Maybe the missing free-wheeling diode is below the board... Anyway, I hope you well have time to investigate this further, I'm very curious how they have done this.
Greetings from Germany!
Excellent video! I learned alot here, mostly about why I don't use resonant mode converters. Too much math! (I can see it taking a team weeks to get the math for something like that bench supply sorted out.) Anyhow, not useful for small MCU projects of mine that typically consume
It's been 25 years since I worked on one of these! It was for a US Navy submarine reactor control system, where space and heat were both huge issues given the power density we needed to meet (there was no room for conventional supplies). After a failed prototype followed by a failed tweak to it, we got a successful design only after we gave up trying to use pre-approved MIL-spec magnetics and made our own. Literally: We hand-wound the first two sets.
It was a design competition between two companies, and the Navy was impressed by how much smaller and cooler our supplies were compared to our competitor. In the end, we didn't get the contract. But we sold our supplies (and other tech we pioneered) to our competitor for an eye-watering markup. Turns out we made the profit we hoped to earn on the entire project just by selling some parts and licences. The engineering sadness at the loss was soon overcome by the financial win, and the sense that we had ducked a bullet: We likely would have lost money had we won!
Parts and licenses? You mean, the cores with windings and innovative circuit design?
@@edinfific2576 The power supplies were the only hardware we sold on the project, which was about 5% of the revenue. The other 95% of the revenue came from other tech we developed then licensed (it was a nuclear reactor control system, after all).
Schematic or it didn’t happen. In all seriousness, it’d be really interesting to see a schematic of this topology. TiA :)
@@combin8or I was actually a software engineer at the time, but I had been a Navy and commercial nuclear technician before that, so I worked on the project in multiple roles. When the supplies started smoking, I was pulled in to help troubleshoot (because the supplies were controlled by an 8-bit microcontroller I wrote the code for, and we all know who gets the blame when the smoke leaks out), and then helped on the root-cause analysis. I did no electronics or magnetics design, but when it came time to make the prototypes, I was the only one with winding experience, so I got to lead that team. (Well, it was me, a technician, and an intern. But still.)
The original design did direct-from-AC conversion, to avoid rectifier losses and minimize the power factor. Once we had the better magnetics, we were able to use a slightly more conventional design and still meet those goals. I can remember bits and pieces (it was a quarter century ago, after all), but nowhere near enough to reconstruct anything workable. I do recall the microcontroller interfaces and their signaling, but that's about it. In the final design, the microcontroller did little control, and lots of monitoring and reporting.
That was an interesting project overall. I was a junior software engineer who was given some impressive chunks of the system to design. Other software engineers were senior to me (and older than me), but had fewer high-level responsibilities, which made for an "interesting" work environment. Basically, I sucked at that aspect of it, meaning management had to back-fill my lack of people and leadership skills. Ugh. But the real key turned out to be me improving my writing skills, so I could generate design documents that created less discontent because I learned to better explain myself. Trying to do that face to face was not optimal...
Only three Seawolf subs were built (out of 22 planned), which was why we wound up being glad to have lost the competition: We never would have made the kind of revenue or profit needed to stay in that business. But the supplies were an unexpectedly good side-line for a while, until the engineers moved on to other projects and the designs were sold for a pretty penny.
@@flymypg cool. I’ve had a direct from AC topology on the back burner for awhile, so that’s pretty interesting. Thanks for the story.
These application notes are better than some books! I imagine an engineer putting all his teaching skills into the papers since none of his younger colleagues is interested anymore into how the stuff works they are making.
There's no where else on Earth that a hobbyist can get content this damn good! I paid $40 for a Beginner's Electronics class for my nephew and the content in that class isn't even remotely as good or educational as the information on this channel. Dave, you sir are a legend! It can't be overstated how much we hobbyists appreciate the information you provide.
I beg to differ. Not only hobbists. I'm a practicing engineer myself, and I'm still waiting for a bad video from Dave. (April fools bullshit are the exception, he should knock it off)
@@Eng_Simoes can't be all work and no play
also th-cam.com/users/rv4hv is good, but in russian only...
My team and I designed a bidirectional LLC converter for an EV, getting it working both ways was crazy. Custom magnetics, custom math.
You end up with LLC because of EMC and efficiency targets, and they suck at regulating fast moving targets.
Dave, you remind me of all the old hat professors I had in college, worn smooth from years of experience in industry. You are a wonderful and entertaining teacher, and a treasure amongst makers and engineers. Thank you for everything!
The first assignment in my resonant converters class was to design one of these, and you got it spot on. And you're right about it being hard, it was painfully difficult, especially if you are designing controls as well
my final undergrad project for uni was based around one of theese, it was hell trying to both design the converter and the control loop for it
I’m speechless) The most complete analysis of an LLC architecture what so ever! There are two transformers because they took two standard half-bridge transformers and connected primaries in series and secondaries in parallel (there is no need to calculate a new transformer, primary L is doubled, secondary L is cut a half win-win)
LLC resonant converters are actually getting pretty common in the PC power supply space.
Nearly all 80 Plus Gold/Platinum/Titanium models are use a LLC resonant primary. Many also use synchronous rectification on the secondary side.
Even mid-range models like the Corsair CX450 (which is "only" 80 Plus Bronze) have moved to LLC resonant topologies.
those Gold and up PSUs uses LLC for efficiency, i guess?. there's not much bronze or 230V standard rated PSU uses LLC even today, sadly. did CX450 uses LLC? i thought it's only the 2nd gen CX550, 650 and CX modular series uses this topology
Of all the old ATX PSUs I've taken apart this was immediately recognizable
I'd love for dave to take a look at one of those high end passively cooled 600W models
Actually, it is the high-current switching transients that generate high-energy noise. Resonant mode supplies switch at near zero current and thus generate much less noise. Resonant mode supplies DO vary the driving frequency. This is how they change the current that flows from a constant input applied voltage amplitude. This does not present a noise problem and, in fact, is equivalent to dithering a microprocessor clock to spread the noise over a broader spectrum
Nice review. Yes, this is actually my day job designing Power Supplies. I’m actually doing my first LLC design now, so I found this video in my research;)
You are a legend, Dave! I should start studying EEVBLOG instead of going to uni!
Nice! Brings to mind some of the complexity of a dual-resonant solid state Tesla coil, requiring a lot of careful design and tuning.
I guess I’ve graduated from Fundmentals Friday to Somewhat-higher-than-fundamentals Sunday.
Normal switches' drains begin at the input power voltage and must conduct current until they are open again at the input power voltage. At the beginning and end of these time intervals, they also have non-zero drain-source voltages. Since P=VI, there are substantial power pulses at the beginnings and ends of these pulses. In resonant designs the series capacitor stops the current every half cycle and allows the switches to change state without dissipating these pulses of power.
These videos right here are worth the price of admission. A+ for guys like me who are looking to learn. Thank you Dave for all the work you put into all of this as well as your enthusiasm. It really means a lot.
Thank you, I always wanted to learn this stuff from a good teacher (----> Prof Dave)
Very well explained! Now i know where losses come from. Forget the square wave!
That bigass inductor is probably a secondary filter and that MOSFET next to it is to replace the freewheeling diode that you'd normally see there.
16:37 What're the odds those two are both transformers, but the secondaries are wired in parallel for the 32V outputs, and in series for the 64V outputs?
16:55 A big fat inductor (on a *black* core) with a MOSFET next to it, takes my mind towards a mag-amp type regulator. Can those be made with an adjustable / variable output voltage, or do they need to be designed for a fixed value?
TTi made a series of PSUs which used magamps in the secondary side regulator, so it is doable
@@jaycee1980 link pls ☺️
@@productivityatmost TTi EX354 series
I have some power supplies from Delta Electronics that do this as well. They have a secondary transformer set that can be switched between series and parallel depending on the load and regulation.
Resonant switching always fascinated me.
This had me go down a wee rabbit hole, thanks
The resonant circuit filters out the higher harmonics of the square wave drive so only the fundamental is a pretty good approximation.
Thanks for the explanation. I was wondering how that worked, that important detail was glossed over a bit.
fantastic video, been getting my feet wet with switching supplies (just built some basic buck/boost stuff) but always interested in learning more about what's out there...
As for the topology:
Maybe they use a Phase Shift Topology perhaps? SOmething like the 1350W Enermax Gold/Platinum units.
But also Huntkey X7/1200W.
5:43 The area under the voltage and current waveforms most definitely is *not* power dissipation. What the heck?
Yes, how is that supposed to be power dissipation? Where is the connection? I dont see it.
MOSFET turn on..volts across it drop to low value..the current through it rises from min to load demand max.. & vice versa at turnoff..
THE MOSFETS NEVER SWITCH INSTANTANEOUSLY... BUT WITH DEFINITE RISE & FALL TIMES.....ramp up/ down of V & I each in a few microsecs... times that KHz sw.freq. rate...& it is something for the designers to worry about...
That AREA UNDER THE CURVE(slopes) ARE THE POWER SWITCHING LOSSES...
...the device "R..D-S on" times the load current sq......
{ ( R× I sq. = W) for "one switch on" period } are the " conduction losses.....
So cool, I love learning stuff. And I understood it! Will I build one myself? Yeah, nah.
Back in 1990's we used the " blocking oscillator " type power supplies.. rated around 80W..
+5V@6-8A...+/-12V @2A....these ran rather cool.. and had good line & load regulation.....
don't know why is lost out.. but we had to upgrade to higher power later...
Nice overview of LLC, thank you.
Resonant LLC is the topology of choice for just about every AC/DC power supply from 100-400W these day. It’s also really not that expensive at this point, and I’d argue that it might be even cheaper.
17:19 lmao. thanks for the application note and video. going to help a lot with my homework.
Maybe I missed it but I didn't hear anything about the disadvantages of llc compared to hard switching topologies in the video.
I'm referring to the increased size due to the resonance network and extra required output capacitance.
Since this is a variable supply my guess is the big inductor and the single MOSFET is a secondary buck converter.
Could also be a programmable current sink, to help with overshoot and the output voltage regulation, or to dump energy when you command a drop in voltage, or when your load imposes current on the output for some reason. Lose energy in the saturation of the inductor, or in the diodes and transistor, till the input side feedback can compensate for it.
The final MOSFET you found may be to have a linear regulator after the switching one. This would allow very precise output control and could be made to take only a few volts across. Makes sense?
There are also LCC converters having a second capacitor in parallel with the transformer primary instead of a second inductor in series with the transformer primary.
It's not my day job, but I just joined a research group that solely designs resonant converters, and I am watching this video as sort of an introduction to the topic... it seems quite overwhelming! Are there any other introductory resources you may recommend?
I don't know if anybody else has mentioned that the LLC converter output voltage can only be varied from 100% down to about 60% and therefore incapable of providing control down to zero output voltage .I suggest that the LLC converter feeds a buck converter - which is that FET that you are not sure about and the large filter inductor .
I am very far from "well versed" in this, but I have such a hard time explaining this sort of stuff, thanks for this video.
This seems a good candidate for a vacuum tube power supply. I'm thinking for mic preamp and EQ stages.
Could it be mixed (LLC + linear) power supply? Last mosfet could be a linear regulator, with precise voltage / current limiting...
Why would you ever do that??
Did some testing a number of years ago on a hyper efficient power converter, of several KW output. Used in Solar car race, Darwin Adelaide and from memory efficiency ran at 99.8%. Quite certain it was a resonant design as output harmonics were also extremely low.
99.8% efficiency. Wow, thats amazing!
@@simontay4851 sure was a surprise when evaluating the results.
Had to build a 3KW adjustable load, at minimal cost. Would have liked more time to design a electronic load, but made do with large 'gold' power resistor chain on a long tube heatsink an fans forced air flow. Had a hurry as decision on using these converters was critical to be ready with race setup. Contributed to success of the local car that year!
I believe that the resson of why there are two transformers is for not saturating the transformer (create larger Ae value on the primary side in series and parrallel the secondary side to reduce the condtuction losses).
Please pay attention to the fact that the LLC converter does not have an output LC filter due to the ZCS at the output. Nevertheless, a solid inductor is installed on the board, so I would bet it is PSFB topology. Please let me know if I am wrong and there are variations of the LLC with the output LC filter.
I would consider the possibility that there are two transformers in 2x Lm is used at resonance, and as suggested the secondary is either a centre tap output or as suggested by another comment used series and parralel for 64 V and 32V respectively
Resonant Mode Converter
"Straight to the pool room" 👍😎
When operating at resonant frequency, R=0, so IL goes to infinite values, sorry I didn't understand.
More video's on this subject needs to be covered. practical one as well. I know of someone who used a PIC16F177x/PIC16F178x part and designed a LLC controller in firmware exceeding features found in jelly bean parts.
I seriously doubt it. There's no way one could achieve ressonant performance at, say, 400 kHz switching frequency, with one if those puppies
Unless it’s for educational purposes, designing a resonant controller by yourself seems like a huge waste of time. NXP and TI already make extremely reliable controllers that work up to 500KHz for
@@Eng_Simoes 400KHz? maybe yes, but what does that frequency got to do with anything?. Fully resonant mode is not related to some high frequency standard only a factor related to core size reduction that's it. its all to do with the transformer core material used + LC tank. We can achieve full power at resonance with frequency as low as 63Khz using that PIC.A cheap modern design can be realized with EPCOS material N96 operating at 150KHz with cheap mosfet switching devices and a tight layout.
@@SolidStateWorkshop don't agree with that statement, see CIP Hybrid Power Starter Kit
Part Number dm164147, there are lots of really great benefits.
@@yaghiyahbrenner8902 energy is 1/2 C×Vsq. ..
and/or 1/2 L× Isq...
AND POWER IS ENERGY times FREQUENCY....
so higher the Freq. ..higher the pwr. for the same L/C & V/I
also higher the operating Freq. ..
smaller physical sizes of L/C components..
The limitations earlier being reliability & capability of the then AVAILABLE mosfets..
dedicated controllers....
& low economical sense to
change over from the well established SMPS " industry standard "...
Today we have SiC mosfets which are way ahead of conventional insulated gate mosfets..and as they become more acceptable & price competitive.. ..EXPECT TO SEE
KW. OF POWER handling at UPTO 5-10 MHz in resonant modes...
Is it true that resonant mode converters are less noisy? I think I read somewhere because they are soft-switched, there is less switching noise on the output.
I'd like to know, too! I've heard this, but never from an authoritative source. If true, resonant mode converters also might be useful for sensors that are sensitive to switching noise!
You are right!😊
A LLC-Convert alone isn't
able to full fill the Output voltage requirements from 0 to 250V. As the Gain
Curves indicates, they will struggle at low voltages and light loads withou large frequency variations. I can't
see a resonant capacitor. These Caps are locking more like input filters for the
H-bridge. A third stage is common to achieve the needed voltage range but there
must be a beefy diode anywhere. It could be also some kind of protection circuit
for the load. I would choose a different
topology like a phase shift push pull converter. But how knows (=
Jesus H Christ. I must be dumbing down by being in home office for all this time. I was looking and looking at this simplified schematic datasheet page, that refers to LLC everywhere - wondering what LLC was an abbreviation for xD I stopped wondering, glanced over the schematic itself, and then it dawned on me... Good job, me - good job.
Just take a look at a modern, higher efficiency (Gold or Platinum) ATX Power Supply. Most of them are LLC-Rensonant mode Converters.
Though they usually don't have a big coil behind the Transformer...
A cheaper one might be better. Something like Silverstone SST-ET550-HG, as that's a single layer PCB one. And cheaper...
Designed a 500W one for ebike charging. Prototype coming soon
This kinda reminds me of induction heater circuits up to the inductor.
Despite the fact that I usually design other devices including microcontrollers (I love C++ and C#) I also fell in love with DC/DC converters design many years ago. I began with buck converters, then forward, and half-bridge. I'd like to build LLC converter but it seems to me it would be beneficial to design low voltage LLC converter first (powered from 24V instead of 230VAC) so that would be safer at the beginning. LLC converters are much more difficult and any error in your design may result in component explosion, especially at high voltages such as 230V AC for instance.
I too am an EE and bot MCU and destop apps programmer, and design all sort of PSU's and small control sysems with microcontroller for which I use mostly C, C++, and whenever possible Pascal. And yeah, get the circuit wrong, and mods/parts explode, which for me is a complete loss because I do a redesign - expensive, but being a full time contractor, I must have happy customers.
On a tangent, and only for Windoze desktop GUI apps.
Try Delphi, you'll learn to HATE C++. C# is a mixed bag for me, even though it is designed by the same guy that did Delphi before he sold his soul to M$. It has the CRAP syntax of C++, but has a lot of nice abilities, BUT it is suffering from "project creep", i.e. too many not too useful but COMPLEX abilities (which Delphi is now following - BAD and STUPID MOVE - KISS IS BETTER) that are better off not being there ever! I started with many Assembler, then a bunch of HLLs like Basics, PL/1, Pascal etc. etc. Then C, then C++ and finally Delphi. All this over four decades, and for productivity, READABILITY (brackets are EVIL becaus the brain processes words like begin/end MUCH MUCH faster due to context not being required to recognize them), there is NOTHING that comes close to Delphi, ...but it is an unlearning curve from C++, to which I was addicted to for a very long time, went to rehab :), got much wiser and after hating Delphi at first after it was pushed onto me, several several years I good at it and now NOTHING comes close to how fast and ultra readable I code compared to C++ that I was so in love with before ...NOTHING! Because of the FALSE LYING POS M$ propaganda about Delphi, as M$ does with better products that are not theirs, most programmers never looked at it in the last decade! VERY SAD but my gain, especially when it come to high response GUI's. Try doing GUI's under any DotNET language without spending too long on it (apart from the cookie cutter and limited wizard stuff). I mean truly custom schmik designs my customers cannot get anywhere else.
I designed a very complex multi monitor system for $50k price ($30k + $20k later extensions), took two and half months combined. Two decades later, due to M$ evil ways, they wanted it converted to C#. I told them it would has laggy GUI (because 6 level deep layered GUI with huge inter module updating), which DotNET would comes to crawl with. Anyway they got an initial quote of $350k from a C# house ...yep, and the quote was about right too, because I also quoted for it at $400k becaue DotNET just could not do most of the GUI stuff well so many Custom Controls would have been required. The DonNET GUI is only good if you stick to common GUI needs that 90% of projects need. Anything out of the ordinary and you are FUCKED for performance and programming time!
...so here is my triggered rant! :):):)
@@stevenbliss989 Well, I used Delphi mamy years ago, starting with version 1.0. I use C/C++ for embedded systems. C++ is complicated but I'm getting used to it. However in many cases is better than pure C especially when I need to write complex firmware for devices that include TFT displays with rich GUI, FAT file systems and so forth. In my company some engineers use Python but it looks weird to me and I don't want to use it in my designs.
@@rk-dj4sg Not a fan of Python either, it has it's place, and is very useful for quick and dirty automation and/or data analysis, but I would NEVER write a real program in it, for many reasons, including it being an interpreter.
@G E T R E K T Before Delphi there was only C++ or VB (BLAAH!!!) and Turbo Pascal, plus some oddballs like PL/1 & Modula 2. Sadly I never used Turbo Pascal, BUT SHOULD HAVE back then, it would have saved me a lot of time and agro! :( ...I despise M$ for essentially killing Delphi! ...NO FORGIVENESS!!!!!!!!!!!!!!!!
@G E T R E K T You got it man, but I hate them for removing floating form designer!
You're producing a much better approximation of a sine wave.
Trying to design a resonant dual active bridge for my dissertation. Due may 4th. Bit off farr more than i can chew ehehehe
Top notch content Dave!
Dave can you reference some good tutorial/doc on basic LC discrete oscillator design and how to do some am modulation with that
I understand this is being applied to a power supply per say, but what this same H bridge mosfet topology were being used for a 'transformerless' inverter & output? Would this LLC design work or be applicable then?
it seems very odd that "all" that power would be completely directed thru that 1 inductor thats a lot of power just thru those winding for all of the outputs . Am i incorrect to think that ?
hello, i'm simulating this LLC resonant convertor in simulink ,how can i implement pi controller to achieve both constant current and constant voltage at output?
dear, u may show the schematic and circuit board so we may better understand.
Thank you.
Well done
did that have one of those self made opto isolator like the electric fence??? with the UV led & the light sensitive transistor in heat shink.. Or am i seeing things(resistor) TOP LEFT???
Input fuse.
@@v8snail was thinking that V8Snail!! coz Dave wouldn't have missed it if it was one!!
Maybe explain EV on-board charger?
For more information search for Kasyan TV for DIY POWERFUL Inverter for every need.
Question: What are those little green and yellow toroids? Are they inductors? Are they not part of the circuit?
Dave showed them in the previous video. They are current transformers for current sensing. The giveaway is that the "secondary" winding inside them is only a single turn.
The ones with the thin wire and exposed green core are current transformers, as was mentioned ten months ago. The ones with yellow wire are gate drive transformers.
two or four output diodes have nothing to do with half or full bridge
I'm not gonna do a full tutorial on that; it'd be like an hour
*films an hour of tutorial anyway*
A numbered video is unlisted with 0 views?
Didn't want to release two videos in one day, but made it available to those who bothered to follow the link. Zero views might just be stat lag, or you are the first.
@@EEVblog An LLC is just a particular type of resonant converter, not every resonant converter is an LLC. Full bridge LLC is also not as common as half bridge LLC which can be used effectively up to more than 1kw power level. This seems not at all an LLC but more a Phase Shifted Full Bridge converter (PSFB). An LLC won't have an inductor on the secondary side while a PSFB needs it. An LLC will work with good performance at fixed input and output voltages. Here this stage needs to regulate from 0v to 32v (or 64v).... Good luck with an LLC converter, the needed frequency variation to keep the output voltage regulated will be impossible to achieve. I suspect this is a PSFB with 2 identical transformers wired with the primaries in series and the secondaries in parallel or in series to differentiate the 32v and 64v models using the same hardware.
For more information on PSFB check TI app notes about UCC3895.
Great content as usual.
Verry Interesting!
biphasic for smaller ripples?
Do you new Guys what kind of smps I need to use for a tube/amp hybrid amp? Can I use one supply for the 65 x 2 15 amps, 12,8 volts and 6.3 volts 5 amps, 350 volts 0.10 amps, 2 x 120 volts 0.25 amps. all on one transformer. LLC is nice, but did read LCC do better here., for feedback, open loop the best way? the transients from a audio amp can never be corrected, feedback to slow, a possible way, current feedback use, injection on chip triangle oscillator, did read some about that. I can use also fase shifting and other topologies but as fat LCC has soft switching, and I need as low as possible emi because it is a tube input/driver amp, filtering is not that difficult with the low amperages for them.
I can also use separate smps for the 65 x 2 15 amps and the other voltages make it more easy.
THanks for advice.
Toooooo complex..... I'm happy with my flyback and push-pull designs..... !! :)
it looks like Two-pipe forward
It looks like those induction heater driver
Daves explanation is a little incorrect in a number of places (LLC converters are not fixed resonant frequency for example). For a better explanation, go to Ben Yaakows youtube channel:
th-cam.com/video/tCEqm-RoP20/w-d-xo.html
Now im more confused than before watching the video :)
10:45 Absolutely false. Typical CFL drivers use BJTs, and they operate almost exactly as an LLC converter, just don't transform the voltage. And the board you have shown is extremely unlikely to be an LCC converter. There is no sign of the resonant tank, and the output inductor after rectifiers is not just unneeded in LLC, but ruins it.
Is there someone making noise near you? I'm in the middle of nowhere and it sounds like someone is on my roof. I should have stuck to the cheap headphones instead of upgrading.
gah, i still cant quite get my head around this.
everyones chasing regulated outputs and stuff like that.
im chasing current. plain and simple. as much as possible. i cant stae my output voltage as technically, there isnt one... or its dictated by the load, coupling, temperature, blah... i dont give a damn what it is as long as its being fed lots of current.
induction heater.
i think im getting a grasp on it, run through various calculations, all looking good...
but damn i am releasing a LOT of smoke in the testing phases! and none of its coming from steel heating up. if it runs unloaded, it blows up loaded, and vice versa...
meh, another two week wait for new components... weee!
Please note teory
Is it similar to a "ZVS converter" ?
I think it is ZVS converter
PHOOL BRIDGE RECTIFIER !!!
I'm a soft switcher guy
So yeahhhhhhh I hopr I can get away with just the design tool from on-semi
Here we have a God, his name is Sanjaya Maniktala
..the Sanjay Manektala of
ex. Crompton Greaves Mumbai1990-92...ex....NSC....
with that chain smoking habit...???
he had written (at least)a book on p.s...???
How can I contact you?
..... Carl Sagan used the word 'Quasi' to describe the intelligence of our planets dominant bi-pedal placental mammals.
Infineon data sheet: "Competitor A", "Competitor B"... Why not name them if you're confident about your numbers? I don't like that. Either disclose the names or mind your own business!