OC plays with an oscilloscope (and a mobo!)

bigben2k · 12-06-2002, 03:25 PM

check it out

Not sure if they (or even I) fully understand the limits of the scope.

It looks like this guy has a poor PSU, but it clearly demonstrates what some of us have said: the 12V line is noisy.

Puzzdre · 12-06-2002, 04:34 PM

I dunno if I'm saying something stupid, cos' I don't know much about this stuff, but:

the noise in 12 V is clearly visible on scale of 1 V per division, that's for sure not good.

On the first page, the ground plane looks flat at 0.5 V division, but not so flat in more sensitive scale...

The 3.3 V and 5 V rails are also measured (and flat) on 0.5 V scale, but they aren't shown in more sensitive scale...

As far as I can see, 12 V is very bad, but 3.3 and 5 V are good for 0.5 V scale...What about the more sensitive measurements? Like, same as ground, 20 mV?

Or it doesn't matter?

Feel free to hit me onna head with some soft sponge stick or a cup of coffee, but I had to ask...:shrug:

bigben2k · 12-06-2002, 05:05 PM

The point is that there is always noise. How much noise may or may not affect one's ability to overclock.

The important point (I think)is that the 12V line has a lot more noise, and that's probably OK.

Strangely, the noise is induced by the mobo, and would affect peripherals.

Sponge thrown.

The type of noise is also important: a square wave can really throw off a component.

gmat · 12-07-2002, 05:03 AM

20-30mV scale noise usually represents EMI. You cannot get rid of that, other than adding layers to the PCB - expensive. Actually it can be noise in measuring probes. So actually the 30mV noise on the ground plane is normal.
5V and 3.3V should have been shown in AC mode (on the 'scope) at a 100mV scale, just to make sure that noise stays under those 30mV.
12V ripple is quite 'normal' as well, though it could be better - some PSU add more stages after the MOSFETs to 'flatten' that. But since it's the 'main power' line, adding components is expensive as well.

BB2K: hum i don't see any square noise... Actually i see ripples at switch points, that show an over-compensated switching output. Over-compensating is normal, this is to ensure there's no 'drop' in the signal. Active PFC can 'flatten' them a bit, though it's only a side effect and not a goal of PFC.

Stepback · 12-07-2002, 07:35 AM

Here is a copy of an e-mail i sent to Brett ,

Quote:

Hiya Brett,

When were your O-scope and meter calibrated last? I've been work in the contract electronic business for about 14 years. At the last company I worked for I was the test department manager. I believe the difference your seeing in your readings are do to the scope or meter being a little out of calibration. There are many companies out there that do NIST traceable calibration on just about any test equipment you can think of. It's not that expensive to have them calibrated. You may also have a bad filter or switch (most O-scope probes have a 1:1/10:1 switch) in your O-scope probe, have you tried using a different probe? Just some thing to think about.

Since87 · 12-07-2002, 07:40 AM

The scope itself only has a bandwidth of 100 MHz. Much worse than that though...

It doesn't appear that the 2246A scope has the connections to detect the multiplication factor of the probes. This implies the author was likely using 1X probes to do these measurements. 1X probes typically have a bandwidth of 10 MHz.

If these measurements were done with 10 MHz bandwidth, then the traces shown only provide hints at what the real signals look like.

Puzzdre · 12-07-2002, 03:32 PM

Thx for clarifying some things guys!

12-06-2002, 03:25 PM	#1
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	OC plays with an oscilloscope (and a mobo!) check it out Not sure if they (or even I) fully understand the limits of the scope. It looks like this guy has a poor PSU, but it clearly demonstrates what some of us have said: the 12V line is noisy. __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

12-06-2002, 04:34 PM	#2
Puzzdre Cooling Savant Join Date: Aug 2002 Location: Croatia Posts: 969	I dunno if I'm saying something stupid, cos' I don't know much about this stuff, but: the noise in 12 V is clearly visible on scale of 1 V per division, that's for sure not good. On the first page, the ground plane looks flat at 0.5 V division, but not so flat in more sensitive scale... The 3.3 V and 5 V rails are also measured (and flat) on 0.5 V scale, but they aren't shown in more sensitive scale... As far as I can see, 12 V is very bad, but 3.3 and 5 V are good for 0.5 V scale...What about the more sensitive measurements? Like, same as ground, 20 mV? Or it doesn't matter? Feel free to hit me onna head with some soft sponge stick or a cup of coffee, but I had to ask...:shrug: __________________ 'Out of cheese error... ...please reboot the universe (press the GBL to continue)'

12-06-2002, 05:05 PM	#3
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	The point is that there is always noise. How much noise may or may not affect one's ability to overclock. The important point (I think)is that the 12V line has a lot more noise, and that's probably OK. Strangely, the noise is induced by the mobo, and would affect peripherals. Sponge thrown. The type of noise is also important: a square wave can really throw off a component. __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

12-07-2002, 03:32 PM	#7
Puzzdre Cooling Savant Join Date: Aug 2002 Location: Croatia Posts: 969	Thx for clarifying some things guys! __________________ 'Out of cheese error... ...please reboot the universe (press the GBL to continue)'

12-07-2002, 05:03 AM	#4
gmat Thermophile Join Date: Mar 2001 Location: France Posts: 1,221	20-30mV scale noise usually represents EMI. You cannot get rid of that, other than adding layers to the PCB - expensive. Actually it can be noise in measuring probes. So actually the 30mV noise on the ground plane is normal. 5V and 3.3V should have been shown in AC mode (on the 'scope) at a 100mV scale, just to make sure that noise stays under those 30mV. 12V ripple is quite 'normal' as well, though it could be better - some PSU add more stages after the MOSFETs to 'flatten' that. But since it's the 'main power' line, adding components is expensive as well. BB2K: hum i don't see any square noise... Actually i see ripples at switch points, that show an over-compensated switching output. Over-compensating is normal, this is to ensure there's no 'drop' in the signal. Active PFC can 'flatten' them a bit, though it's only a side effect and not a goal of PFC.

12-07-2002, 07:40 AM	#6
Since87 Pro/Guru - Uber Mod Join Date: Sep 2002 Location: Indiana Posts: 834	The scope itself only has a bandwidth of 100 MHz. Much worse than that though... It doesn't appear that the 2246A scope has the connections to detect the multiplication factor of the probes. This implies the author was likely using 1X probes to do these measurements. 1X probes typically have a bandwidth of 10 MHz. If these measurements were done with 10 MHz bandwidth, then the traces shown only provide hints at what the real signals look like.

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)