PSU Voltage Noise

[Scottz]

All PC switching power supplies seem to generate some noise on the output voltage lines as it's inherit of the device itself.
How well it is filtered out is based on the design of the PSU itself.

My question is how much noise (in mV) is "normal" and does load play a factor in it?

If I have a 600W power supply, but only really use 320W of it, does the noise characteristics of your average to "high quality" PSU go down, higher, stay the same on the voltage lines?

 

[wouse101]

the magnitude of the noise should be in mV tops. Any higher and then you should worry.

if your overload the supply, you get bad ripple, your smooth DC lines start to look like directly rectified AC, plus the PWM unit controlling the boost/flyback circuit will be going like the clappers trying to back off the current demand. but your PSU should cut out by then if it has overload sensing.

as far as noise from the PSU is concerned, under normal conditions, like you've stated, it shouldn't be a problem, as there are plenty of chokes and filters cleverly placed (and even lengths of track on your motherboard act as filters), so that any noise from any source (hard drives, etc) is suppressed.

 

[Scottz]

Yes, the noise level is in mv and while motherboard electronics maybe designed for such impurities, it seems to affect other components such as video capture cards.

The various motherboards I tested do show some variance as to how well they filter this out, but it seems to go by "what's good enough" which I understand is allowing up to 120mv for ATX motherboards.

Depending on quality of the capture card itself, they do try to do some additional filtering via caps and have regulators going to the ADC's. This suggests the designers are expecting some noise. Unfortunately the vast majority of capture devices (including USB/Firewire based) do not try to isolate their onboard "filtered/regulated" voltage from system ground or other potential areas that can get the PSU noise introduced again.

For instance, my tests are showing that the "noise" is still getting into some of the components such as the "video switch" (typically a cmos 4052 based chip) used for going between different inputs on the capture card.

This mv noise can heavily affect video luminance (i.e. voltage or amplitude if you will) which is what I'm seeing on more than a few test beds.

Of course there are many possibilities when it comes to "noise" and video capturing and believe me, I've gone through the gambit of possibilities and researched each one of them.

However this question is still on the list.

Does a PSU designed for a max 620W load generate more noise (albeit acceptable to the system board) if say only 200-300W is used?

 

[RetiredChief]

Wouse101 please correct me if I insert foot into mouth

I think you will find the least noise at mid load, This my not be true for all PSs. The old series regulated Power Supplies had the lowest ripple are the lowest load. These PS would be great, except the would be the size of a ATX case - by themselfs.

Switching PS typically step the freq up to approx 20 KHz and regulate by changing the on/off duty cycle. This is why they are noisy.
- At low current. (1) The filtering is more effective, Longer RC time constant, However, this is offset by the longer "OFF" time of the charging circuit.
- At full load current, the "Off" time is lowest; However you now have a shorter RC time constant. (The filter Cap has a much harder time to maintain the "E" level.

(This noise is generally classed as Low to mid freq, ie 20KHz)

This is simular to Effiency - Best at mid load, lower at both reduced and max loads.

The second Noise souce is the load itself and is a result of the "Clock" generators (High freq.). How much of this is coupled onto the supply lines is a function of the quality of Power Supply (Not High Capacity caps are poor High freq Filters (Note this is contridictory of 1/2piFC - but a fact). And a function of the Load filtering. As Scottz pointed out the design of the traces (ie thichness, lenght and distance from other traces) have an effect.

Based on this, if I was a betting man, I would pick Lowest noise at mid load. However it is too dependent on both Load and PS to state definatively.

 

[Scottz]

Thanks for the further clarification. That helps a lot.

wouse101 pointed out the motherboard design can help in cleaning up the noise and I confirmed I see this as well.

Based on PSU design I'm guessing the individual "rail" load supplies could also affect this (i.e. If I'm not loading all "rails"). Or do they typically have their own individual filtering that doesn't "leak" noise across other supplies?

I purchased a "Watts Up?" line monitor and a bit surprised to see the system in question draw 140W under load (at the AC input). Typical is about 105W (i.e. system on but idling, no devices in standby).

I still have some stress tests to go through but would have guessed a higher input wattage being drawn under normal circumstances.

I know there isn't a 1:1 correlation in terms of input/output wattage (due to PSU efficiency) but can assume the input wattage is typically higher than the total being delivered on the output?

Based on the info so far, it seems a 600+ watt power supply is overkill for this system which again surprises me a bit. I haven't bought a PSU rated under 450W (typically purchase 520W or more) in years.

For a test, I plan on getting a Seasonic high efficiency 330W supply to see if that helps.

 

[RetiredChief]

Not sure what your system is and are you planning on using this P/S for an upgrade.

Not surprised at your power measurements. My system (see computer icon) draws 216 W durning boot, 192 W at idle, and 312 W with graphics card under load.

not sure how you loaded system, I used 3dmarks06, also ATItools and enable 3d window.

I shoot for twice my Max power consumption, BUT big problem is you must insure you have enough Power for the 12 Volt rails (primarily for higher end gaming graphics cards. Don't take manuf. spects as gosple, while they are normaly correct, some are very misleading.

 

[Scottz]

It's more of a test than anything else. I think my second post probably explains it best.

The system that I took the quick measurement on is:
abit IP35 Pro
Intel E6850
4x2Gig OCZ2VU8004GK
Nvidia 8600GT
36Gig Raptor
250Gig Maxtor
WinXP x64
Various video capture hardware.
Various PSU's from 520-680W.

I have more than a few combinations of hardware I've been working with and will settle on this one for the test with a lower wattage PSU.

Yes, I do the same for my gaming rig in terms of picking a PSU.

In this case it was a hunch that a higher than necessary wattage PSU on a system that doesn't need it could potentially create a minor issue in certain circumstances.

 

[croc]

All PC switching power supplies seem to generate some noise on the output voltage lines as it's inherit of the device itself.
How well it is filtered out is based on the design of the PSU itself.

My question is how much noise (in mV) is "normal" and does load play a factor in it?

If I have a 600W power supply, but only really use 320W of it, does the noise characteristics of your average to "high quality" PSU go down, higher, stay the same on the voltage lines?

www.jonnyguru.com does excellent psu reviews, he covers 'noise' quite extensively. Good reads, generally excellent commentary.

 

[Scottz]

Looks like a great resource!
Thanks for pointing it out.