Mean Well RSP-2000-48 Teardown deep going, ends with EMC stuff

[oz2cpu]

this is my longest teardown video so far.

sorry it is over 30mins, but there was quite a lot of details to explain.
questions are welcome

SORRY youtube crashed and F--- my file... maybe i re-make it....
right now feeling depressed

NEW FILE:     https://youtu.be/bIP53wtVbcU

[T3sl4co1l]

Yeah, weird isolated converter section!

I wonder what effect all the leakage has there.  The wiring alone (loose cables from transformer to rectifier) incurs a good 200nH already, maybe more seeing as it's litz.

I imagine most of the emissions are from the output side.  Have seen plenty of supplies that need a CMC and "Y" cap on the output side to behave.  The primary side isn't usually too bad, having ample reference to ground (Y caps), but the output, between being isolated, and cheaping the hell out on filtering (enough to meet the ripple spec, and no need to do anything >20MHz), can be pretty messy.

SORRY youtube crashed and F--- my file... maybe i re-make it....
right now feeling depressed

Oh, I guess that explains the some minutes of black in the middle.   Hopefully a reupload or whatever does the trick.

Good voice by the way, reasonably good mic, clear narration, just a tinge of Germanic accent?

Tim

[oz2cpu]

FINALLY, i had to remake the missing minutes
i really hope it works now

https://youtu.be/bIP53wtVbcU

[oz2cpu]

here is a little bit better explain of the filter i made

all things that juse single-point to ground / chassis as well,
is the most quiet, since this way all cables refer to that single point = no ac difference = low emi

:-)

[thm_w]

2kW in that tiny case? incredible.

[2N3055]

Nice work Thomas!

I agree with Tim, good, clear narration, not to slow. Nice analysis, both circuit and EMC.
It was nice and interesting to watch.
Interesting PSU, some quite exotic choices.

Best,

[NiHaoMike]

Mean Well does have a reputation of being electrically noisy, I once worked at a place where a specific Mean Well wall wart was the standard for a "noisy PSU".

2kW in that tiny case? incredible.

Pretty much the same as what rack server PSUs can do, just with a higher output voltage.

[oz2cpu]

funny you mention the noise, and reputation, where i come from, most people thinks : if there is written mean well on it,
it is well below limits, and all problems are only implementation and user error..
and all the magic i need to do, is to solve that :-)

fell free to re-do the radiation experiments in EMC labs, with cables connected to the unit,
of course you need to transport the energy to something a few meters away.
else it is not a realistic setup, if you power this up, with out cables.
Lets say 2-4 meters, on input, and on output.
and the cables needs to be on the test table too..

here another picture, from another test lab, same curve, same level..

[oz2cpu]

thanks for the comments on the video..

i made a few other teardown videos, but most are made in Danish, the plan is to do some more in English too, specially if they get more likes :-)

this design is old and funny, I am sure you get the point even if you dont understand the talk.

https://youtu.be/qVu1LMr85fY

[NiHaoMike]

of course you need to transport the energy to something a few meters away.
else it is not a realistic setup, if you power this up, with out cables.
Lets say 2-4 meters, on input, and on output.
and the cables needs to be on the test table too..

Having the input directly connected to a cord that leaves the end product is reasonable, but having the output directly connected to a length of cable going outside the enclosure of the end product is not the intended use case for that kind of PSU. It's intended to be mounted inside the same enclosure as the load. Most high power 48V loads are electrically noisy so it would be wasteful to try to reduce the noise on the output.

[oz2cpu]

good input there mike, in my use : the 2kw power go somewhere else, external of the main controller and main case with the powersupply..
so of course I need a very effective filter..

NEWS : the 2.2uF red cap that handle ALL input peak currents,
when i wanted to desolder it, to test its internal resistance and current capeabilities,
one pin just fell off ! all by it self, way too easy
and i was very carefull with the desoldering, so maybe this was the main cause of death !!
ac current overloading !!
maybe we can find full datasheet over this cap ?

[oz2cpu]

here that capacitor is, picture added

ok it is from HJC

partnumber must be

MFTD225*0450DB*22*

https://www.hjc.com.tw/admin/product_en/images/file/2021-11-09/618a2666a4791.pdf

no info about AC current capeabilities at 110kHz

[joeqsmith]

At 25:24 where you talk about the two extra diodes, think about it.  You are charging these caps with the boost converter (PFC).  Imagine what would happen if the FETs turned on directly shorting those caps.   

While earlier versions were conformal coated, I suspect if you purchased a current version you would find it was removed (along with other changes).   I suspect you would also find the conducted emissions have gone up for each revision.