Electronics > Projects, Designs, and Technical Stuff
Design of SMPS - Not sure what these spikes are being caused by.
<< < (23/27) > >>
Zog:

--- Quote ---this is my job and you ain't paying me, after all - but it should get you much closer to your (wacky) goal of powering a hundred Nixie tubes or whatever daft adventure you've set out yourself.
--- End quote ---

No I am not paying you. But steady on now. "wacky, daft"? . Probably to some, but this project will bring me closer to world domination ! :box:
I am not really offended. You are right it is wacky and daft but I do have a serious plan for this.
Anyhoo... with that out of the way. Sniff sniff wipes tears away from eyes. :'(

Thank you for getting the thing in better shape. I still see that "ringing" in the waveforms but since it has passed your muster I guess it's acceptable "ringing". Possibly just an artifact of LTSpice ?


--- Quote ---Usually you start with a C value that is twice the switch output capacitance then select R to be somewhere between 0.5x and 2x the characteristic impedance of the resonant network between the total capacitance and total stray inductance.
--- End quote ---

Gotcha, sorta. Thank you for that pointer. The C value is the easy part. The rest is experience and track layout but can always be tweaked later. I suspect you need very expensive equipment to do that properly though.


--- Quote ---2) Added a gate resistor - always, always, always use a gate resistor when driving MOSFETs and IGBTs.
--- End quote ---
I swear to the SMPS gods that I will always add a gate resistor. Even if it's 1m \$\Omega\$. If I don't, may they take my first born, if they can suffer him. (kidding love the bugga).


--- Quote ---5) Changed the transient simulation parameters - note that if you have to skip initial operating point in LTSpice to get a circuit to simulate, especially with one of their IC models, then something is probably wrong. It usually helps to specify the maximum timestep, however; something in the range of 0.1x to 0.5x the switching period usually ensures LTSpice doesn't inadvertently skip over ringing and other such phenomena.
--- End quote ---

I don't know why I use the skip the initial operating point. It worked either way but won't in future. Thank you for that tip about specifying the max timestep I had not even considered that and I will do that too in future.


--- Quote ---7) Changed the K1, L1 and L2 values to something more realistic for 300kHz CCM operation. Note that L2 has 3x as many turns as L1, so its inductance is 9x higher (inductance is proportional to turns squared). It wouldn't be unreasonable to go up to as high as 6x L1 for L2, though damping the ringing will become progressively more difficult as the turns ratio of the tapped boost increases farther away from 1:1.
--- End quote ---

Yes get all confused about turns ratios being turns squared and larnt all I know from this video. But too thick to actually apply it in practice. :palm:

https://www.analog.com/en/education/education-library/videos/5579254291001.html

I am having a lot of trouble with magnetics in general. I mean the theory I learnt decades ago, but it still confuses me all the time. Getting old. (59).

I have been having more brainstorms. That are in the wacky category for sure.
Could one use a current transformer ? Probably not not closely coupled enough and huge leakage (guessing). But just thinking out loud here.
https://www.digikey.com/product-detail/en/pulse-electronics-power/PE-68383NL/PE-68383NL-ND/2265791
I like it because it's a tiny little bugger.

Moving on.
A proper transformer like this one.
https://media.digikey.com/pdf/Data%20Sheets/Wurth%20Electronics%20PDFs/750031353.pdf
Nice size, and probably what I should be hunting down.

Got hold of one of these from CoilCraft and this has got me stumped.
https://au.mouser.com/datasheet/2/597/da2032-463371.pdf


What sorcery is this. Each individual primary measures 10uH as per my LCR meter modified with kelvin clips at 100kHz. It's a good one not a piece of rubbish.
https://www.ebay.com.au/itm/DER-EE-DE-5000-High-Accuracy-Handheld-LCR-Meter-From-Japan-F-S/202765545258?hash=item2f35c4af2a:g:AAYAAOSwG-1ZxKJ0

Ok, fair enough but you can then put 3 of them in parallel and it still says 10uH ?. I did and it does. What ? wait ... err.. inductors in parallel are supposed to act like resistors. Not so according to the CoilCraft magnet gods.

Well if that's true. I am getting hold of a bunch of these to play with. I think these are the magic beans that will help me actually get a bit of flexibility so I don't have to remake a board every time.

https://www.coilcraft.com/pdfs/hexa-path.pdf
MagicSmoker:

--- Quote from: Zog on September 19, 2019, 09:39:19 am ---...
Thank you for getting the thing in better shape. I still see that "ringing" in the waveforms but since it has passed your muster I guess it's acceptable "ringing". Possibly just an artifact of LTSpice ?
--- End quote ---

No, the ringing is still a problem; I just got it better, not perfect. Tweaking the values of the snubber and even filtering the current sense signal more will be warranted, but I needed to leave some work for you to do, eh? Also, using a smaller FET (ie - higher on resistance/lower current rating) would help, as it will have a smaller output capacitance (Cjo in LTSpice).

That said, you're never going to eliminate all ringing in a hard-switched (that is, non-resonant) converter... Not without seriously degrading efficiency, anyway. And LTSpice does a really good job of getting such behavior correct - better than some professional SPICE packages with prices in the 5 digit range - so no, the ringing isn't just an artififact of the program, it really will occur (now, whether the ringing occurs at the exact frequency and amplitude depends on how accurate you were at determining all of the stray Ls and Cs).


--- Quote from: Zog on September 19, 2019, 09:39:19 am ---Gotcha, sorta. Thank you for that pointer. The C value is the easy part. The rest is experience and track layout but can always be tweaked later. I suspect you need very expensive equipment to do that properly though.
--- End quote ---

Some empirical testing will likely be required, regardless, so just plan on trying a few different capacitor values on the prototype board. Snubber values are rarely critical (as evidenced by the frequent use of suspiciously generic values like 100pF + 10k or 10nF + 10R [seem familiar?], etc.)


--- Quote from: Zog on September 19, 2019, 09:39:19 am ---I swear to the SMPS gods that I will always add a gate resistor. Even if it's 1m \$\Omega\$. If I don't, may they take my first born, if they can suffer him. (kidding love the bugga).
--- End quote ---

There are two reasons for the gate resistor: 1) it limits peak current charging/discharging the gate capacitance; 2) it suppresses ringing between that capacitance and stray wiring inductance.


--- Quote from: Zog on September 19, 2019, 09:39:19 am ---Could one use a current transformer ? Probably not not closely coupled enough and huge leakage (guessing).
--- End quote ---

For measuring current, sure, though usually not worth the expense until the peak current is at least 10A or so and the sense signal is scaled to 1V peak.

Otherwise, yes, a proper transformer for L1/L2 would be better, but note that you can get pretty good results with a commercially available toroid with the right L2 value then adding the 1/3rd to 1/6th turns required for L1 on top of it as that will likely get leakage down into the 0.5% to 1% range. It's hard to do that well with any of the E cores (ETD, EFD, ER, EER, EC, etc.), actually.

EDIT  - forgot to add that the reason paralleled windings on a common core don't decrease in inductance is because they are all tightly coupled together. It's really no different than if the winding was made of multiple strands of insulated wire if you think about it. For the inductance to decrease in parallel - as happens with resistors - then you need the windings to be on separate cores with no interaction between them (that is, little to no mutual inductance).
Zog:

--- Quote ---EDIT  - forgot to add that the reason paralleled windings on a common core don't decrease in inductance is because they are all tightly coupled together. It's really no different than if the winding was made of multiple strands of insulated wire if you think about it. For the inductance to decrease in parallel - as happens with resistors - then you need the windings to be on separate cores with no interaction between them (that is, little to no mutual inductance).
--- End quote ---

Ohhh. Ok thanks for clearing that up I thought the laws of physics had been changed while I was on the toilet.

Just going through cores on digikey.
This mob seem to have a material at the right sort of frequency so should not get too hot from core losses right ? https://www.ferroxcube.com/upload/media/product/file/MDS/3f3.pdf at least compared to plain old ferrite ?
I really would like a nice cool end result. As cool as possible anyway.

Digikey have this thing available.
https://www.digikey.com.au/products/en/magnetics-transformer-inductor-components/ferrite-cores/936?k=&pkeyword=&sv=1&v=1779&pv70=1069&sf=0&FV=1d480002%2C1f140000%2Cffe003a8&quantity=&ColumnSort=0&page=1&nstock=1&pageSize=500
Looks like it might be a bugger to wind though.
And they indicate inductance factor of 4.5uH So that is the max I will get out of it ?

I looked here but got a little lost.
http://www.encyclopedia-magnetica.com/doku.php/al_value

MagicSmoker:
AL is inductance per turn squared. If it says 4.5uH then a single turn will have an inductance of 4.5uH, 2 turns = 18uH, 3 turns = 40.5uH, etc.

Ferroxcube 3F3 (and TDK/EPCOS N87) are both excellent ferrites for 300kHz operation.

I ain't reviewing random cores, okay, but the pot core you pulled up is just about the worst choice possible, really.

Instead, look at part number 2100LL-180-H-RC on DigiKey which is an 18uH toroidal choke rated for 10.1A saturation current using a relatively low-loss material (Sendust). You could use the existing winding as the primary then wind on top of it, say, 4x as many turns, tying the start of the new winding to the end of the existing one to make a 1:4 autotransformer. Et voilla, a semi-custom tapped inductor for about $4 and a few minutes of your time. It will likely run hot and you'll have to restrict the peak current well below the claimed 10.1A saturation rating (maybe 5-6A peak), but it will be good enough to get something working and teach you a bit in the process.


Zog:

--- Quote ---AL is inductance per turn squared. If it says 4.5uH then a single turn will have an inductance of 4.5uH, 2 turns = 18uH, 3 turns = 40.5uH, etc.
--- End quote ---
Great. That makes it easy.

Yes I am already using a "fairydust" core. They still run a bit hot for my liking but will play around with it some more till I get the inductance right then proceed to order a super pixie dust toroidal.

I have enough information now to ask my CoilCraft rep what he can do for me I think.
Navigation
Message Index
Next page
Previous page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod