I wrote all this down couse someone told that just becouse the computer is in a metal box, there is no EMI inside.
Thats a bullshit.
someone who has no clue about electrnoics, and says LIES (physicaly impossible things, wich any technician could easyly verify) should never tell me I talk bullshit. While I did post background materials related.
"I'm a certfified aviation electronics dunno what, and I mesured no voltage changes on my psu" is equal to : " I'm a car engineer suspension specialist, but I didn't notice stearing treshold". Read up to find out why.
should I let someone like this say I'm talking bullshit? no... I would hate my self.
I did tell in -i think - all my responses -maybe 1 didnt include it- that these mods are not for the avarage persons.
problem is that for a nice EMI mesureing
1. a good enviroment is needed --> probably most of don't have it
2. computers EMI are usualy way over 1 ghz, most scopes don't mesure them, and price for those that can are up in the skys... special "probes" //basicly sensitive antennas for oscope// may do the job, and in most case they are used. But for this high freqvency.. I don't think. oscopes (affordable for home / small laboratory) are good choices to mesure a simple liner transformers EMI, or .. in audible ranges, so checking for an hi-fi amplifyers "noise" they will do. Just think of the lowest freqvency You find in a computer.
Thats the PSU-s 25 khz. And its harmonics.
Anything else starts at the ghz range, not to mention harmonics. And the point is, to take into account the product of these harmonic distorsions.
I wrote this simple article set to show that there IS significant amount of noise/EMI/RFI inside any computer. And showd links that manufacturers DO battle with them.
In IT these noises are considered as "data". Sometimes they mess up the communication between devices in the computer, than error corretion or re- requesting data is what computer do. Thats where You loose work cycles.
allso about PSU modding.
Notice that I don't just filter the psu, but allso buffer the outputs.
Thats usefull, when You load Your cpu You will notice a nice voltage drop.
q6600 is one of the best overclock wise.
at 9x333 --search for forums, I only come up with linkable info-- the vcore setting is : 1.2875
windows idle: 1.23v
windows load : 1.18---1.20
what happens is, as soon as Yours cpu starts to work, it will lover its imeadnce / demand more power.
low output impedance will not let it fall thismuch,
and a buffercap will not let it drop suddenly.
so, what would happen if vcore drop would be reduced?
say.. it would drop down under load to 1.25?
You got it. You could have higher overclock.
Vcore drop is a common effect, it can not be elliminated.
But can be reduced. The guy told he mesured no voltage drop with his PSU (well.. its impossible, in best case there would be some drop that does not count, and for at least a short time a slightly bigger drop..)
To achieve a small vcore drop, a good psu is indeed needed.
It has to have a low output impedance, and proper filtering.
thats why separate rails are usefull. If an other device starts to get more and more amps at the same time as the one wich supplys vcore, the "sudden" drop will be larger. If after the drop the minimum wich is needed for the cpu to work is not present, the system will freeze.
When You run a staility test, basicly You get a view of by how temperature / load changes affect the system.
Keepeing a balance between components is important, start with a quality psu, becouse even if You have a stable motherboard and other stuff, and a good cpu, and powersupply does meet power output rates, if it can not compensate fast and enough of voltage drops, Your system will not be a good overclock platform. And may even freeze in normal operation, 140 watt quad core amd is likely to do that with say.. a codegen PSU.
the noises affect PSU freqvency of switching, and it affects the resistance of the coils in the transformator.
(2 * pi * f * L) + dc resistance is the impedance of the coil.
if freqvency is lowered, the input impedance of the transformer is allso lowered. The power drawn from the wall outlet is basicly vac * Vac / ((2 * pi * f * L) + dc resistance) and this apply to the secunder part of the transformer too.
loading effect for more info.
load canges as vcore demands more power, so the voltage changes too.
(verry basics of electronics, it is one of the verry basic calculations. Thats why there is no psu that does not have voltage drop..)
the outcome is, power supplys that have ferrite beads at the input, and have a high output power (wich can be done with a low output resistance) and have separate power rails at the output are the ones to go for.
Then You get a decent board, and then balance cpu /gpu / blabla other parts.
All this will not work well (specialy high power applications) if the power supplys voltage has flickers, or can not handle sudden load changes.
//to addres this I do use ferrite beads, separate power rails, and BIG efficient heatsinks so power transistors don't change temp. too much --> ensureing they have stable properitys. And less thermal noise, offcourse..
clearly, if someone reads some electronics will find out that all this makes sense, but -for the last time- is not needed for avarage people.
but, I made my posts becouse i don't like when people state its bullshit. When all info is available for anyone intrested. And I made clear points.
Aaaaand, if someone states he is an engineer, specialy in an area that requries EXTRA attention to noise/EMI/RFI filtering, I would only wish I knew wich aircraft is he related too. Becouse thats a flight not to take.
// for other people a verry simple example are the USB soundcards. They are made becouse of many reasons , but one is directly related here:
inside the computercase there is SO MUCH NOISE its more worth it to make it external. It is an audible lvl of noise.
a qoute from there:
"Internal sound cards are subject to a lot of interference and that can cause distortion or background noise. An external device eliminates much of that."
//notice, in the artice it says its rubbbbbbbish, but its one of the first devices, and problems are mainly with linux not realy supporting this hardware. nowdays even soundblaster offers external sound devices.