Discussion in 'Reviews' started by crmaris, Jul 19, 2013.
To read this review go to: http://www.techpowerup.com/reviews/HighPower/AGD-750/
Thorough review - as always
Interesting with the powermeter - but as you said - it should have been external.
First of all, great review, as always. You're really spoiling us!
On that note...
I already asked this of you elsewhere, and you didn't answer (no would've been fine, so I could stop hoping). Because I'm pertinent (saucy even) by nature, and have no shame, I'll simply repeat my question/request here as well:
High Power always seem to have troubles with ripple/noise "leakage". I believe that somehow it always appears to be induced, "over-the-air", in the output wires. Doesn't it look almost impossible for those levels of r/n to not get filtered out by solid polymer and wet electrolytic NCCs and decently-sized air-core cylindrical chokes? Low ESR? Check. High capacity? Check. High inductance? Check. So what gives?
Crmaris. It's probably a lot to ask of a man as busy as you, but could you get an EMI shield/screen (just an insulated copper sheet, or something along those lines) somewhere, and put it between the output wires and the PSU's secondary? It should pretty much half the noise, if I'm right (possibly ripple too, but I do believe it's high-frequency EMI that's the real problem). I believe you're curious about it too, so I hope you'll find some spare time to try it out...
Now it's not my intention to be rude, but I've been observing Sirtec/Sirfa/High Power struggle with r/n suppression for years, and believe I'm seeing a pattern. I actually tried the EMI screen thing myself once (when I had access to both a Sirfa-made PSU and a good o-scope) and it seemed to really do the trick (something like 20% reduction in peak, and even more, like 30% reduction in sustained noise, and about 10% in ripple). Unfortunately I don't have the scope logs/shots, and I didn't get to do it again (it was an old 80+ standard platform, in a Chieftec PSU).
first of all I appreciate it very much when highly experienced users like you share their thoughts through the forums with all the rest
This unit, as the crossload tests clearly shown, has a problematic area somewhere close to full load with high ripple mostly on the +12V rail. However the fact is that the PSU I had in my hands was most likely a pre-production model or an early one, so soldering was crappy and bad soldering can be also the cause of increased ripple (and in some cases it can be also the cause for increased EMI). Also I don't know if the resonant controller is somehow confused and makes this problem, however I doubt about this since the change from PWM to FM is done at much lower loads.
Also the PCB was in pretty bad shape already, due to the heavy soldering, so once I removed the secondary heatsink it was left almost useless. Once I assembled back the PSU it didn't operate properly so I have to find time now to fix it firstly, if I can since the PCB is a mess, and try what you propose. But even if I am not able to fix this one or can't find any time to do so (because this period I have a really hard schedule to follow including two NAS reviews which need long hours) I will try this in a next High Power unit review.
As a side note, in some books I read, they were mentioning that EMI is a real pain to suppress and some times they try even trial and error procedures to find the best solution. One guy was telling a story that once they had to install the CM chokes the other way around in order to get noticeably better results! Go figure!!
Also I will pass your remarks to a highly educated friend of mine who has great practical experience in this stuff, to shade some more light.
And sorry if I acted a bit childish there. It will surely happen again at some point, though
Yeah, ripple can be caused by bad solder joints because of both increased resistance and parasitic capacitance that occurs there. Bad PCB traces and through-holes are perhaps even worse, even causing overheating and further degradation through micro-arcing and increased resistance. Massive joints also have large self-capacitance, and all of this together works to throw off the frequency tuning of the Pi-filters. This seems to be a pretty big deal with resonant converter primaries and high switching frequencies. LLC is also usually responsible for not-that-good transient load regulation, but that's hard to combat efficiently. An ordinary double-forward is much more robust in that regard.
I wonder if it's feasible or even economic to include a couple of small ferrite beads in the cabling, even outside the PSU housing, similar to how capacitors are sometimes integrated (remember "PowerCache"?)... This, of course, if the EMI screen method proves unsuccessful or unreliable...
Well, anyway, I'm glad you're also interested in this, and that you're even willing to include additional knowledgeable people to consider it. I look forward to getting a few more opinions on this matter.
Separate names with a comma.