DC-DC converter isolation, questions from a newbie.

[David_]

Yes I hope so, I have now placed orders for all parts needed except the diodes.
I have spent quite some time scavenging while taking things apart learning as much as I can and I could avoid ordering many things but I like to do things with certainty to avoid spending a lot of time searching for faults caused by bad components so I prefer to spend a small amount of $$$ to be sure.

But I have gotten worried about which diodes to use where because stuff I've read which appeared to be seriously telling me I HAVE to use schottky diodes and if not schottky there where one other type that would do, I had planed to use schottky diodes for the secondary's rectifiers(maybe a couple in parallel, we'll see) anyway but for the diode that goes across the primary clamping the voltage to the rectified line voltage(320V) I had thought to use any diode I can find that has high enough voltage/current rating.
For those diodes schottky is not even a option and seeing as in a one-switch flyback you use a snubber to dissipate the energy there could not matter what diode I use as long as it conducts in the right direction...
When going online to buy diodes I find types that I haven't even heard of before so that makes me ask since SMPS seems to be highly specific in regards to components used.

If there are no trouble with the diode I will probably not write here again for something like 3-4 weeks while waiting for parts and making and tweaking my prototype but I will return and present my results.

Regards.

 

[Warpspeed]

The clamping diodes on the primary can be the "ultrafast" type.
These come in higher voltage ratings than shottky diodes, and are readily available and are not expensive.

 

[David_]

Hello.

I've bought cores to use for a low power version but I thought I acquired cores that could handle 97W but apparently it is more like 23W, by the way, do you ever solve a potential size problem through extra cooling of the core?

Anyway I recently scavenged a big old/but not terribly old TV for its high voltage flyback transformer and found on the PCB the biggest ETD ferrite core I have found so far in my scavenging and I want to reuse it but I have never ever succeeded in dissembling a SMPS transformer without cracking the core in more than 3 pieces.
Seriously I find SMPS transformers to be impossible to separate the core intact from, do any one have any technique to manage to without breaking it break loose the ferrite E-style core halves?

I have used a scalpel to cut lose any amount of glue/epoxy or whatever it is that I can but the bigger part of the amount of epoxy still fast on the core, you know how ETD cores outer leg is curved on the inside, that hole area is impossible to get to without removing the windings but for now I'd like to keep the windings on the bobbin if possible.
I have read room one guy online how he does it but he had built a hot air blower designed in a way that he could place the blower right in front of him so he had plenty of space to work with the core while at the same time it was heated enough to necessitate thick gloves and tools not to get burned.
However any such device is above my economical resources and I have not found any useful device while scavenging but I have secured access to a couple of "grov soprum" translated into (the opposite to fine...)trash room filled with electronic junk.
Might I maybe be able to use a toaster owen + some powerful fan or if luck hair drier to produce a hot air blower..., what do you think?

On the subject of flyback design I have some trouble grasping the question of DCM or CCM, for one thing there seams to be a very large difference between the required primary inductance. Small inductance for DCM and large inductance for CCM and the whole transition concept makes me want to ask if I really need to set the maximum duty cycle to 50%. I do get that in a forward or half-bridge topology below 50% is critical but in a two-switch flyback is that necessary?
That question relates to the question about if it would be possible to design a flyback spanning 10-55V at 275W and have it stay in DCM always?

I have aquired all components exept for the PWM controller but I have begun researching the difference and similarity between 3825, 3525, 3843, tl494, and one or two more, I am having a hard time with the broadness of the functions of these controllers and its very scary that I experience a lack of advancement is how wide a concept a can grasp with my lack of any real attention span to speak of. It simply does not expand and my C/C++ skills is laughable, after 3 years I have covered the basics but µC programs very quickly becomes to complex and past maybe 3 A4 pages my attention dissolves into nothing.

I believe that it is paramount that I stay with the step by step design procedure but part take so long to arrive and there are so many thoughts and want to knows so I loos my way, but I have layed out a PCB for 3825/IR2110/IRF840/ETD34 but I have trouble finding a procedure to design the transformer but I just need to find my documents.

Regards

 

[David_]

I now see that it became a quite large post so I will now summarize sparing anyone who do not want to get through all that.

1.
I have found a big ETD ferrite core in an old TV and I want to reuse it but so far I have never succeeded in disassembling such a transformer without cracking the two halves into multiple chunks.
Do anyone have any good ideas on how to take apart a SMPS core without breaking the bobbin or any core half?

2.
Do anyone have any thought on creating a hot air blower(like a hot air gun but built as something that stands on a table) from scavenged toaster owens and fans/hair drier?

3.
I am facing problems grasping the concept of flyback inductor-transformer 'complete energy transfer mode(DCM)' and the transition into 'incomplete energy transfer mode(CCM)' if I remember the terms correctly.

3,a
In a two-switch flyback converter, is there any good reason why not enable the use of duty ratios greater than 50%/0,5?
3,b
Would it be possible(and is it a good or bad idea...?) to design my flyback transformer to cover a potential voltage range of ≦10V to ≧55V while being able to deliver 275W(a little less during lower output voltages) and all the time stay in DCM?
So as to avoid the low bandwidth of the feedback amplifier and to avoid the hole transition stuff.

Regards

 

[BradtheRad]

3,a
In a two-switch flyback converter, is there any good reason why not enable the use of duty ratios greater than 50%/0,5?

You have some room to increase it, but there is a limit because when you lengthen the first half of the cycle, then you must shorten the second (output) half of the cycle.

In practice you would wind a transformer according to your best ability, then experiment with the duty cycle as you adjust for desired output voltage and Amperes.

3,b
Would it be possible(and is it a good or bad idea...?) to design my flyback transformer to cover a potential voltage range of ≦10V to ≧55V while being able to deliver 275W(a little less during lower output voltages) and all the time stay in DCM?
So as to avoid the low bandwidth of the feedback amplifier and to avoid the hole transition stuff.

DCM tends to be associated with a short duty cycle.

It could be feasible. Here is a recent thread about designing a flyback with a short duty cycle.

https://www.edaboard.com/threads/338595/

 

[FvM]

3,a + 3,b

The duty cycle is in fact limited to 50% because the flyback voltage is clamped to the input. Above 50 %, all energy is recirculated to the input bus and the inductor flux can't be reset anymore.

This is the price you pay for the convenience offerred by a two-switch flyback converter, and I guess the reason, why the topology isn't used very often. Your intended voltage variation of 1:5.5 involves respectively an minimal duty cycle of 15 % and rather high peak currents.

 

[treez]

David_, if you can tell me how to send you a zip file on this forum, ill send you a basic file full of ltspice simulations of isolated and non isol. smps

its ok done it now..here..

- - - Updated - - -

In current mode, ref DCM or CCM, -if you get it stable in CCM, then it will be stable when it goes into dcm at higher vin.
DCM is easier of the two to get stable though.
Running it on the sim is a good way to get your head into it....trying to mentally visualise stuff is a long way round it.

- - - Updated - - -

here is some quasi resonant flyback ltspice simulations.....these are in boundary conduction mode.

 

[treez]

What topology(s) are you considering?
Give me a shout if you want any more simulations of the others.
Regarding forwards and bridges, some of the sims have expressions so you can visualise the magnetising current separately from the power current for the forward/bridge type...that always stumps people at first, when they come from the flyback to look at the forward/bridge. People don't realise that bridges etc only saturate on the manetising current...the power current, the fields of it from pri and sec, cancel out, so no saturation from power current in a bridge/forward.

The flyback problem at high power centres around the leakage current, and the reflected secondary voltage which increases the off state voltage on the fet, upping turn off/on losses, the leakage makes this worse, and when the leakage starts discharging after an on cycle, it unfortunately pushes some power current into the rcd clamp aswell.

- - - Updated - - -

If you are slightly struggling, then you are not alone, I fairly recently worked for a huge electronics/consumer company…..(you will definitely have heard of it). They had a 100W offline flyback made for them by a electronics consultancy. This flyback was full of design errors. It nearly got put into production. It also had a 2W housekeeping flyback on it, which had exactly the same RCD clamp components as the 100W flyback. (just basically an SMB size TVS with an UF , SMA size diode going into it.).
The TVS had been put right beneath the flyback transformer, so the thermal camera scans taken did not show how hot it was getting.

Also, the boost PFC was clamped up to its overvoltage limit (418V) all the time, instead of smoothly regulating to the 394v (average) that it had been set to by its divider.
The BCM flyback (100w) was too deeply in DCM….despite having a proper bcm driver ic. –there was about 2-3us of dead time, way too much.

There were also several other faults, including just 100n capacitors on the pfc and bcm flyback gate drive supply pins. PFC sense resistor too far from chip, and with switching power currents going between them. Gate drive OV zener getting activated every time the fet switched on, etc etc.

I was utterly shocked that such a huge company had not been able to get a decent job done for themselves, and had nearly put this thing out into the market.
-------------------------------------------------------
Also shocking to hear of was a vast number of flybacks for the new power meters, failing due to design faults and huge batches having to be recalled.

Also, I came into a job once where a very good, promising smps engineer had just been sacked. He had been doing a 20W offline flyback, on the output of which was an enormous energy storage capacitor (not just the usual output capacitance, much much more than that). He couldn’t get it working because as the controller charged up the output at start-up, it took a long time due to the big cap….so the controller “thought” it was in output overload, and kept shutting down before getting into regulation. That was the only thing he did wrong, but it cost him his job. Poor chap never got to work in SMPS ever again. What a waste…all they needed to do was let him find out his simple error. Just imagine, he would have made a great smps engineer, and gone on to produce work and help make loads of jobs, but not now…he was sacked, and ended up in some place in north London not doing SMPS…and from what he said, he wasn’t doing much of anything, what a waste .(Also north london was worse because he then had to commute round the M25 every day, and he’d just had a new born.)

 

[David_]

Thanks for all answers, I have had a hard time with my router but it is fixed now.

I have actually not considered any other topology than a flyback, my first electronics project ever was a LM317 power supply and this is the evolved version which has gone from using a mains 50Hz transformer followed by a linear regulator, to a mains transformer followed by a buck converter followed by a linear regulator and when that was done in theory I set out to make it a offline flyback instead followed by a linear regulator stage controlled through digital means.

But I am having a hard enough time with a flyback and feel a forward or any other topology is over my head, I am just about to implement my first prototype circuit based on ether SG3525 or UC3825 followed by a IR2110 that drives a two-switch flyback inductor-transformer but if I understand FvM right then my current version demands a very narrow and small duty cycle to manage it which goes along with high currents through the transformer?

I am not very good with LTspice but I am working with it to learn more, the files found in the flyback_offline.zip files appears very valuable, thank you very much. The pdf seems very interesting indeed.

I knew from the start that my power level, which is 275W output at the most is ideally to high for a flyback. Literature tells me that at least, and that a forward converter would be a better fit but I am very weary in regards to if I can manage a forward design.

I had no idea that the market was so rough, your story really seems as a big waste of potential engineering resources.
I will go over the files and see what I might find and learn.

Regards

- - - Updated - - -

Maybe I should say that as this is my first design of a offline flyback converter I do consider efficiency not very important, there are a few things I can live with that you could not live with in a professional environment but that mind set is only meant to easy up on the level of difficulty designing such a circuit.
I know am beginning to think that CCM is the better alternative to design for in order to bring down the peak currents, it might be very relevant that I will not use the output of this flyback directly. It will be feeding a linear regulator stage and the thought is to regulate the flyback to keep a constant or sort of constant voltage drop over the linear regulator(linear regulator + bjt's to enable higher current than any low noise linear regulator can manage).

 

[treez]

SURELY I CAN TEMPT YOU INTO A TWO TRANSISTOR FORWARD DESIGN, WITH SIMPLE BOOTSTRAP HIGH SIDE DRIVE? (sorry about capitals). As attached
I suppose the flyback can be doable at 275w, watch it if your reflected voltage to the primary is bigger than the primary side input voltage, because it will blow up if its a two switch flyback and that happens.

ltspice sim attached...just change it to .asc and you can run it in ltspice

 

[David_]

Okey so I have now begun studying the forward converter topology and it appears as I can change my design into a forward by rearranging some components(not a must though) and change my transformer design which was not even done for the flyback.

But FvM's comment in this thread as well as in another thread tells about a uncomfortably small duty cycle but forward converter seems to enable options that can enable duty cycles over 50% but does that really help in increasing the needed duty cycle range?

The two-switch forward is limited to below 50% D(duty cycle) but as I see it if using a single switch I could make the reset winding with more turns than the primary thus shortening the reset process of the core?
Or change to a active clamp circuit.

I am doing a offline converter and since I don't know much about forward jet have read things I can't recall although the result of those characteristics made a forward topology difficult to implement I'd like to show you the hole concept for my circuit so that maybe you can tell me where I will encounter problems.
I will make a drawing and post it here soon.

 

[treez]

if duty cycle is small, that's ok if your primary current is less than your secondary current.
Also, if in current mode, and your duty cycle is less than 50%, then you don't need slope compensation , which is a slightly good thing.
Attached is a two tran forward simulation (ltspice) with an expression for the magnetising current . If you go "add trace" and paste that in, you will see the magnetising current, -it is this that can saturate your transformer, or cause you too much delta B loss if it has too high swing at too high frequency.

 

[David_]

Okey, I will check it out.

I wanted to ask about an interleaved dual forward converter, is that anything that I should be interested in?

Could that not help to keep the primary current(s) down and easy up on the input capacitor?

 

[treez]

good point, input cap....you need a pfc stage ahead of your converter. as attached

- - - Updated - - -

Dr Ray Ridley is the person who I have heard talking up the dual interleaved two transistor forward converter, I his book.
Anyway, for your power you shouldn't need it, and for the input caps, the pfc will take care of that.

 

[rikkitikkitavi]

I now see that it became a quite large post so I will now summarize sparing anyone who do not want to get through all that.

1.
I have found a big ETD ferrite core in an old TV and I want to reuse it but so far I have never succeeded in disassembling such a transformer without cracking the two halves into multiple chunks.
Do anyone have any good ideas on how to take apart a SMPS core without breaking the bobbin or any core half?

2.
Do anyone have any thought on creating a hot air blower(like a hot air gun but built as something that stands on a table) from scavenged toaster owens and fans/hair drier?



Regards

Depending on resin type, boiling hot water might suffice. Or placing in a oven, up to about 130ish centigrade should soften the resins. Possibly the bobin might not survive though.

Reusing old ferrites might be tricky if you are planning to go high in frequency.

 

[David_]

I am just starting out in SMPS but I think there will be quite some time before I would go above 100kHz, with the controllers available today that seems pretty low. Although what high and low frequency's actually means depends on the subject and context I suppose.

If the bobbin does not survive then there is not much point to it all, I guess I could try boiling the hole thing.

Up until now I had thought to skip a PFC stage, I suppose I should get involved in learning something about PFC. While still aiming for a flyback I started looking at a 3835 PWM controller and it will still do but the PFC, does one need a separate controller for that or how do you implement a PFC most simple?

Regards

- - - Updated - - -

Ahh you supplied me with more of those wonderful spice files, thank you very much.

Frankly I find LTspice very difficult to learn, I'm really not suited for programming but anyway I find it hard to find out spice directives to use and I have realized that without being able to add lots of different things through SPICE directives you can't really do very much fun.
Sure I can make and research lots and lots of thing but if it wasn't for you files I would still not have been able to simulate any offline converter since I can't figure out how to produce a signal source that is equal to 230Vrms mains.

So when it comes to simulating SMPS you better know some SPICE directives, you know there are lots and lots that I need to learn more about in electronics but I really can't afford any real literature for now and for some time forward and although the world wide web provides me with much(all I know I have learned by reading online) but I have discovered that the really good stuff is found in text books.

 

[treez]

to get mains, just click voltage source, then right click, and "go advanced" then select "sine", and type 50hz for frequency, type zero offset, and put in the peak value...(230*sqrt(2)).
Most simple PFC to implement is a "boundary conduction mode " pfc AKA "critical conduction mode" pfc.

Otherwise just use a continuous mode one, if you do above 200w then it is best done as continuous mode, unless your input is 230vac, then you can stretch it a bit above 200w for a "bcm" pfc.

 

[David_]

Okay, though I am somewhat confused regarding the following:

I've tried to produce rectified mains voltage with a voltage source and 4 diodes and the result has been unusable, then I have found in you files one approach and in a other file(can't remember where it's from) shows another approach and in both cases the look very complex compared to simply generating a sine wave 50Hz 230*sqrt(2).

I took a look at the PFC files you posted and yes it appears to be just that simple...
And in the others as well...
I do wonder how I managed to mess it up... Oh well.
I have at least one complex version with an inductor in series with the AC signal source and then as I interpreted it some SPICE directives doing something related to the 4 diodes in the full-wave rectifier, I have gotten the impression that the basic models of diodes for example can be somewhat simplistic and you can if you want give greater specifications to improve the realism of the simulation?
And I now see that the triangle like zigzag is simply how LTspice depicts sinewaves.

You don't happen to have a good idea for a easily adjustable(duty cycle) PWM signal source?
I have made a IR2110 component rigged up as my circuit prototype will be but I can't find suitable PWM generator, it would be a pain to do it with a Piecewise Linear function.

Regards

 

[treez]

theres some pwm generators in here, you just use the pulse thing

 

[David_]

I fail to understand what the difference is between a Single-ended Forward topology & a Double-ended Forward topology, it feels as that is something I really should be able to find out my self but I sure have tried and my best guess is that is has something to do with if the converter has one or two switches?
But the name convention does not make sense, not according to my understanding of the words.
Unless the names related to ether one side or both sides of the primary coil is being switched.

FvM has previously pointed out that my targeted output voltage range vs input voltage means a really small duty cycle range and that should not change for a forward converter so >50% duty cycle is not necessary so a two-switch forward seems the best choice, ether that or a resonant-reset forward to avoid the reset winding.

But since I face both a situation where a uncomfterbly small duty cycle and a minimum load condition(something about changes the gain drastically during light loads...?), is it right that what is tough is the rang of output voltage I want, (7-10V) to (55-60V) my final output after a linear regulator stage is 0V-50V so the max SMPS output will need to be something like 55V(I need at least 3V maybe more + some safety margin) and I previously set the minimum SMPS voltage to 10V to simplify the generation of the needed 5V rail and the following 3,3V rail.
Know that entails a higher power dissipation while outputting low voltage with high current but since no low-noise linear regulator can output 5A I still need to use some external pass-transistors controlled by a low-noise regulator.

But how difficult is such a small duty cycle range? Is it so difficult that I should consider a buck-converter at the forward converter output?
Though it would feel stupid not to use the forward converter as the pre-regulator since it can perfrome that function.

I just realized that I need to look at the minimum AC line voltage to find out what turns ratio I need to design for at least.

I have actually forgot how I determined I should supply the 5V/3,3V rails, I would love to have one set of 5V/3,3V for the analog section and then a separate isolated pair of 5V/3,3V for the µC/digital parts, I might not need more than a 3,3V for the digital supply but maybe I might just as well only filter the analog rail and decouple it properly from the digital parts.

Regards