Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

What's up with the secrecy around input filters for switching converters?

Status
Not open for further replies.

David_

Advanced Member level 2
Joined
Dec 6, 2013
Messages
573
Helped
8
Reputation
16
Reaction score
8
Trophy points
1,308
Location
Sweden
Activity points
12,217
Hello.

I have for a while been designing a step-down dc/dc converter, I have no education to speak of nor experience and all I know I have learned online.
That leaves me in many cases unable to use information I find, the following is not 100% serious but there seam to be a safe guard in place allover the application notes related to amongst other things switcher input filter design.

I find papers saying "here we will show you how to design a dc-dc converter input filter" and then goes on to tell me the shocking truth that the filter consists of coils and capacitors! But not a word of how I work out which coils and caps, that's not very useful.

And then I find a few documents that do contain equations and descriptions of those BUT, they always leaves out one or two variables.
So when I try to work it out for my circuit then half way through the final equations there appears variables in the equations that has no explanation of what they are or where they come from and it all falls apart and this happens to me every time. I have not found a single complete solution other than that the second LC output filter if used should be set to 1/10 the switching freq... After input filter I also am searching for output filter design that is more complex than a LC filter, perhaps a two stage LC filter, But I would think that I can solve that my self when I have slayed this input filter monster.

Or that might be incorrect, I've might have found a more or less complete solution for a very basic input filter but it has flaws that can be "easily" cured. And its performans is weak next to a more complex filter.

I have searched for filter design solution for weeks and I have spent so many hours on it that it makes me feel ill, for real. I can't at all count the hours, its a enormous amount of time anyway.

Can someone help me?

I have a mains transformer which outputs 28VAC that's rectified and filtered with a 10,000µF electrolyte. I can not give any info about the regulators input impedance which I know is of vital importance in that the filter output Z must be lower than the regulator input Z at all times and frequency's in order to avoid negative resistance which can leads to oscillations. But I have seen that one can estimate the input Z with some info. I know its about damping and I would really like to be able to use a multi-stage input filter.
The input voltage might be 37V and the output will go between 3V and 33V, that is the thought anyway and I have been made aware that my transformer voltage is somewhat low but lets forget that for know.
The regulator output current is 3A max.
The design uses LT8612.

I know that this might be a big ask but I have run into a wall and I can't think of anything other than to ask here, any guidance what so ever is much appreciated.

With regards
 

A better Bang for the Buck (intended pun) is to design a low voltage drop linear regulator after the buck SMPS which maintains the constant low voltage drop, while the linear regulator minimizes the ripple and has the benefit of current limit control down zero with wide BW ripple rejection and low step-load regulation error.

In the old days Lambda would use phase controlled Triacs to pre-regulator the drop for an efficient Linear Regulator. Now it can be done easier with less filter issues and heat sink requirements.

I found this example.

https://caxapa.ru/thumbs/541661/LTJournal-V24N2-02-df-BenchSupply-Szolus.pdf

this may not answer your question regarding pole zero complexities , wide dynamic range potential instabilities and filter losses, but it may yield an easier and better solution.

The cost drivers that come with variable voltage , current and power loads to zero add a complexity factor. In old school designs, all SMPS designs required at least 10% pre-load. Now there are options for hiccup mode, burst control, zero valley current synchronous switching and synchronous multiphase convertors.

There are many books on this subject. My favourite has the PSU guru, Keith Billings, in the list of credits in the book. There are many others.
 

Ah thank you, I have read that article before quite a long time ago and I have been searching for it without finding it since I could not remember any specifics to it.
It does actually help quite a bit in other areas.

It is in concept the same as my circuit, I do have a linear regulator stage controlled by a microcontroller.
But the things I am concerned about regarding input filter circuit is:
Not spreading any high frequency content to other equipment connected to the mains.
Keep the input from oscillating.
And ehmm, something to do with EMC... I am reading to much to wide a topics so I can't remember but id had something to do with EMC:)

Can you explain what pre-load is?
I have read that switchers may have problems during light load conditions, might it mean the minimum load required to keep the regulator/controller stable...

- - - Updated - - -

I have come to understand that all the "good stuff" is found in real books, though my ADD does make the task of reading books something almost meant to torment me but I will choose a suitable book and try really hard to keep at it.

- - - Updated - - -

Am I correct in thinking that I could achieve a physically smaller and perhaps more cost effective design if using more than one stage in a filter?
Through the use of smaller values in each stage in comparison to larger values in a single stage configuration.
 

Step back a moment....

Your input supply is via a honking great mains power transformer (With a response falling off a cliff above a KHz or so), with a honking great smoothing cap after the rectifier bridge?

Your switcher is going off somewhere above 100KHz, I assume?

In that case, stick a few tens of microfarads of X7R MLCC right across the switcher input, in parallel with maybe a 100uF electrolitic (Not too low ESR just a jellybean electrolytic, this is only there to damp the resonance due to the lead inductance, ESR is good here), and stop worrying about it unless testing reveals you still have a problem (It will be fine in all probability).

Seriously, the only interesting thing about buck converter inputs is making sure the ripple current rating on the caps is large enough and that there is enough cap (And that is not much with modern high frequency converters).

If you have one of those once in a blue moon situations where the rules of thumb break down, the easiest thing to do is model it (Remember the parasitics), spice is your friend.

The effective input impedance (If you consider the input caps part of the filter) is clearly given by observing that the input power is substantially constant with varying input voltage, so you can calculate input impedance from input power and input voltage (Yes, the impedance will change with voltage), at least for frequencies well below the switching frequency.

Regards, Dan.
 

I think the technical stuff here is good, and I'll offer some
conjecture about "secrecy".

As far as real secrecy, the power supply business and the
components business attending it are all overserved and
hyper-competitive. Standards compliance and EMI are part
of that competition. Doing really well at this bears on design
and production-release times and costs. So people in the
business have a vested interest in not helping competitors
get smart. You will have some professors and some industry
types who just don't care about that, or care more about
publishing, but most down and dirty experienced types are
neither. Trade secrets and tribal lore are closely held/

And then, there's the broad variety of topologies, voltages
& currents, frequencies and edge rates which preclude a
one size fits all guidance (such as would make a good
concise article). Perhaps an analytical expression would
serve a broad group of people, but as I observe the input
filter is often more empirical than theoretical - you may
get gross values from theory, but probably then it's bench
selection of (say) 1 10uF vs 2 4.7uF ceramics, or 3 3.3uF...
a lot having to do with ESR/ESL and best placement in
the board level reality. Not real conducive to final answer
on a plate, with neatly folded napkin.
 
So you are saying that my transformer is not capable of transferring high frequency content back on to the mains due to the cores frequency characteristics, but what about the secondary-primary winding capacitance?

I will address a answer to all other points made in post here but I can't manage that tonight, I will answer tomorrow. Thanks for your inputs.
 

My switcher will operate at 200kHz perhaps a little higher.
I have come to the conclusion that the kind of filter I have been pursuing is way over my head anyway and I will take the advice given here and ether add a ceramic cap with a series resistor or as suggested a electrolytic with not to low ESR, cost and PCB real estate will determine which. And then I suppose I will ad a standard EMI filter right after the main input connector(common-mode choke and a couple X and Y caps), or do you mean that such a filter would be unnecessary to?
I long for the day when I find a job that would enable me to test my own circuits with real quality instruments, I guess if one worked in a lab one would be able to find at appropriate times a opportunity to do such things.
 

Quality instruments make measurements more accurate,
repeatable or easy. They (or their lack) seldom are going
to prevent you doing things the hard way and learning
more as a result. In fact in the interest of learning
fundamentals, a minimal setup is better than a big shiny
box with an answer button on the front panel (or worse,
Ethernet enabled).

Do not let a low hardware budget dissuade you from any
adventure.
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top