Hi TomC, rf-loop, lemon and whoever's interested,
last weeks I dug into denoising these switchers and finally I'm able to come up with a new 'installment' to use TomC's expression.
I would call this,
(and please get me clear in that I use upper case further down only to indicate a headline - not shouting even though I would like to shout a bit at all this professional designers of and writers about how to denoise SMPSs! But you never know. Certainly not all of them because I read somwhere on a page of a supplier of sealed and small converters that his ones are free of noise and well suited to supply sensitive circuitry. So I knew in front of my experiments that such thing must be able to be designed somehow. And not to mention all those books on amazon and ebay - and believe me I didn't search for them but these advertisement algorithms became quite advanced nowadays, almost mystical - about SMPSs at prices I thought, wow! there must be lots of fine information - quite above 100$/€ many of them - bu didn't buy one of them.)
so here it is:
HOW TO DENOISE A SWITCHING DC-DC CONVERTER PROPERLY
-------------------------------------------------------------------------------
Once you know how to do it it's quite easy but it took me a while and a few redesigns of my prototype to find out.
Please understand that I didn't document all stages I went through including the performance of the original circuit on OWON's adapter PCB. I think that everybody who's got aware of these problems should be able to understand everything from what I post and write. It' just a lot of work to take and prepare the pics and it took me many days.
Setting out from my capacitors test tounament - TomC you certainly remember - I thought that that should give the basic idea how to denoise these converters.
So I designed a little PCB also with much better layout than OWON's design for to generate as little as possible spikes in the circuit itself according to some ideas given in the papers I found on the web and also got from you here (TomC, thanks again).
And when testing found again these unpleasing spikes on my Tek 475A.
So went into further thinking. What I had done was, as my capacitor tournament showed, to arrange several ceramics beginning with 10/20uF 3.2mm going down to 100nF 1.5mm on input and output all connected in a straight chain - no shortcuts - and in between a ferrite bead.
The failure in my design was that I used a too wide ground area in between the circuit and the outside (in-/output). That there should be only one ground connection between the switcher and the outside was clear to me before (and no vias at all inside the circuit) but not that there must not be any space in between the opposite filter Cs also on the ground side.
So I redesigned my PCB and finally arrived at the trick:
That is that the ground run must be also as thin as possible almost like the positive in/out lines!
I would call this 'Straight Line Back To Back' Design!
What it means is to arrange the in/out filter capacitors not only in a straight line but also back to back one after the other and pretty symetrical.
With a broad ground connection and the filter Cs on both sides HF currents would build up inside the ground area making it only possible to denoise either the input or the output but never both of them at the same time.
And as you can see on the pics below it worked.
So again in one sentence:
There should be only one quite narrow ground connection to and from a switcher (in one line!) with it's filter Cs arranged straight down/up this line symetrically just as the hot lines should connect the other sides of the Cs one after the other also by thin tracks.
And It didn't need any beads anymore they are actually useless in this configuration and don't work much anyways. The two toroids I used on the in- and output are for to filter out remainders of the switching frequency itself and even more also the remainding regulator loop's feedback oscillation, not the spikes and the HF ringing.
I used two 100nF ceramics as a final HF killer after the toroids and soldered them directly to the back of the BNC connector and again there where little spikes to be seen. Only after i moved them close together on one side of the BNC ground sleeve also these spikes where gone.
So the whole design has conducted me to the understanding that expresses itself best through the fact that even a BNC connector will not be free of internal HF voltage drops as long as you do not connect it the right way!
And we almost could have known before only, that even on the ground plane just a few millimeters count so much was not really clear to me - thanks a lot rf-loop for pointing in that direction time and again.
(And be aware that there is a big difference in between conducting metal and shielding metal. Conducting metal causes HF radiation due to the voltage drops of the conducted HF in accordance with the in the metal inherent complex resistances, whereas the only shielding metal reduces radiation. Still to think about further why exactly...)
And btw I used a tantalum chip also at the output after the toroid as it performed marginally better concerning inherent noise what means - an I haven't examined that closer yet - that I guess that these very dense ceramic Cs - e.g. 10uF 3.2mm - create also some kind of broadband noise inside - not very high up rather below HF. You can see that kind of noise on the output screen shots as it is the same with both capacitor types only the tantalum seems a bit lower.
The tantalum at the input is just a replacement fot the 220uF reservoir C on the original adapter PCB.
And I have used double sided epoxy for my little PCB and the lower side is connected to ground only at the BNC connector.
(Some basic ideas to this where also influenced somehow by my oil-can test you probably remember?)
So to make a long story short here are the pics and hopefuly I haven't forgotten anything important but I'm sure and also thankful in advance that you would tell me if so.
Just take the file names please for description so far.
And not to forget also to write it down, there is neither HF noise nor spikes anymore on the in- and output of the redesigned +3.3V buck converter.
(And finally I do call out to far China: PLEASE, OWON ENGINEERS , CHECK THIS!!!!!)
PS.:
...and I forgot to mention that these two schottky diodes across the output toroid are there for keeping the over-/undershoot caused by the toroid during power up under +- 100mv...
...and of course I connected the switcher directly to the scope using a BNC mail coupling for to have maximum sensitivity and bandwidth and not to have to use any probe for the noise measurement itself...
....and the probe hook you can see on the setup pics close to the PCB is for to pick up spikes radiated by the converter capacitivly - distance about 2-3cm for coupling capacitance far below one picofarad - and to use these via CHII as the trigger signal...