Ripple Current

[BigAlNZ]

Hi Guys,

So I borrowed a 100Mhz scope and decided to try and visually see the effects on voltage of bad caps. I got myself a old Piii mobo with some leaking caps, although the computer itself never blue screened.

This brings me to my first question: Will bad caps menifest themselves by excessive ripple current?

And my second question : If so should I measure ripple current with the scope on AC?

My method so far was to hook up the scope to a cap near the CPU. I have got the V/div setting on 5mV and the time/div on 20ns. I am using 10:1 probes.

As a result my trace is peak to peak about 1.5 div, so I calculate :

5mV = 0.005V = 1.5div * 0.005v = 0.0075v but now factor in the 10:1 probes gives 0.075v.

The actual volts coming out of the Cap is 1.63 and 0.075/1.63 is 4%. So I am not sure if this is the correct terminology (or hell even if my methods are correct) but I estimate 4% ripple current.



Now would I go to the Intel Piii spec sheet to see if this is within limits for that processor?

-Al

[PlainBill]

Re: Ripple Current

You are correct, you have set the scope on AC, not DC. And a minor quibble. You are measuring ripple voltage, not current.

PlainBill

[KeriJane]

Re: Ripple Current

Hi BigAINZ

I also have been learning to use the oscilloscope.

Your first question, will Bad Caps manifest themselves as excess Ripple?

Well, that depends on what the capacitor does. If it is one of the Low ESR filtering caps common to Switch Mode Power Supplies found in computer power supplies and motherboards, then the answer is "Yes".

The SMPS system lives and dies on the ability of the filtering Caps and Inductors to smooth out the abrupt changes of voltage delivered from the output rectifier or switching transistor. Any failure of the filtering caps will affect Ripple which should be apparent on the scope.

Your P3 system is a bit different than the P4 to present motherboards which use +12v to power the CPU with SMPS technology.
The P3 and earlier systems ran the CPU off of the 5v rail.
Did they drop the 5v down to vCore with SMPS? Or something less prone to Ripple? I don't know. My guess is SMPS. Probably lots of people here do know.


The big question is: What is normal Ripple anyway?
Like most things, that depends.

An open cap that has vented usually has an insane amount of Ripple. I found some blown Fuhyjju caps in an older Antec that had 12v Ripple around 750mv or higher. PSU "normal" is under 120mv on the 12v or 1%

At the other end, the vCore usually has very low Voltage with extremely low Ripple. Even a little ripple here will disturb the CPU so the goal is 0%. What's normal? I'm not sure, but it's probably pretty low.

So far as measuring on AC or DC, setting your scope to "AC" is supposed to merely remove the DC component so that you can observe the AC component only. This is convenient so that you don't have to set the zero level way below the bottom of the screen.

I've noticed that doing this can affect readings though.

For a vCore of under 2v, you could read DC and stay on the screen pretty easily.

For 12v, setting the scope to AC is an easier way to observe Ripple than cranking the base way off the bottom.


Still learning,
Keri

PS. JonnyGuru has a lot of advice on scope usage.
http://www.jonnyguru.com/forums/show...ghlight=ripple



http://en.wikipedia.org/wiki/Switched-mode_power_supply

[BigAlNZ]

Re: Ripple Current

Hi BigAINZ

The SMPS system lives and dies on the ability of the filtering Caps and Inductors to smooth out the abrupt changes of voltage delivered from the

SMPS = ?

Also did my calcs look correct?

I have got another motherboard here with venting, so its going to be the next one to experience my scope cruelty!!

-Al

[KeriJane]

Re: Ripple Current

Hi BigAINZ

SMPS= Switch Mode Power Supply.

There's a link to the Wiki at the bottom of my post.

The way I think of SMPS is:
You want a voltage, say 1.5v for example that is very steady and can stay steady during quickly and widely varying loads.

SMPS accomplishes this by slicing up a higher voltage over time. This is called a Duty Cycle. A duty cycle of 10% has the current ON for 10% of the time of the cycle and OFF for the remaining 90%

So, the vCore needs 1.5v and the SMPS delivers 12v for 10% of the time and 0v for the remaining 90%.

Rather than have a very messy 0-12v square wave, it goes past a Capacitor, through an Inductor and past another Capacitor. This averages out the voltage and smooths it down.
The Caps store the energy momentarily and act as a really fast battery.
The Duty Cycle controller uses the voltage after the final cap (vCore) to make on-the-fly adjustments to the duty cycle to maintain 1.5v.

If the current required goes up suddenly, the duty cycle rises to maintain 1.5v. It can go from near 0% duty cycle to 100% very quickly if needed.
If the current goes down, the duty cycle is reduced to maintain 1.5v.

If the caps fail, some or all of the very rough unfiltered 0-12v voltage gets through. You can see this on the scope.

Oops! gotta go for a few,

Keri

[KeriJane]

Re: Ripple Current

Hi BigAINZ

Back from the Pool!

Anyhow, here's the idea of SMPS as I understand it in drawing form.

The upper trace represents the 12v supply. It is "switched" on/off at a fixed frequency, but of variable duration.

The switched 12v charges the Filter capacitors until the microcontroller sees the output voltage is high enough (1.5v in this example). Then, the 12v is shut off and the caps discharge into the circuit until the next cycle begins. Since the caps store a charge, they supply current until the next cycle begins and the output voltage only drops 10mv.

The lower trace represents a sample 1.5V circuit with 10mv ripple in a nice sawtooth pattern. The microcontroller senses the output voltage and adjusts the 12v duty cycle to maintain the desired voltage average.


If the load increases, the average output voltage goes down and the controller increases the duty cycle. opposite for reduced load.

Bad caps do not store a good charge, so the output ripple goes up to maintain the average. If the caps go totally "open" the Inductor will still provide a little filtering but the ripple can increase to very high levels and cause instability or damage circuits. The controller will often shut down in such a case but not always.

Still learning,
Keri

[BigAlNZ]

Re: Ripple Current

Also did my calcs look correct?
-Al

What about my calculations?

Cheers

-Al

[BigAlNZ]

Re: Ripple Current

Hi BigAINZ

I also have been learning to use the oscilloscope.

Your first question, will Bad Caps manifest themselves as excess Ripple?

Well, that depends on what the capacitor does. If it is one of the Low ESR filtering caps common to Switch Mode Power Supplies found in computer power supplies and motherboards, then the answer is "Yes".

I was really talking about the caps along the edge of the CPU, not the ones in the SMPS. Will I still see ripple as a result of these ones failing (as opposed to the SMPS ones) ?

[BigAlNZ]

Re: Ripple Current

Remember I am really talking about trying to see ripple on the VRM Caps next to the CPU.

[PCBONEZ]

Re: Ripple Current

Grain of salt.....
P3 is not 'guaranteed' to be powered from +5v.
Though few manufacturers did so until P4, Intel was recommending VRM be powered from the +12v rail since the i810 chipset.
-
See Figure 7-1 on page 7-2.
http://download.intel.com/design/chi...x/29067501.pdf
-
I don't remember coming across any P3 boards with VRM on +12v but some -could- have been built like that.
-> So, don't -assume- a P3 VRM is on +5v. -> Check it!!!

~~~~

BigAINZ,
The caps along the CPU are in the VRM which IS an SMPS.
+5v or +12v in -> Switchers -> Vcore out.
.

[BigAlNZ]

Re: Ripple Current

Ok, but what gets me is I tested a modern Socket 939 board with good caps and one with bad caps. There did not seem to be any difference?

VRM=voltage regulation moduel right?

[PCBONEZ]

Re: Ripple Current

Yes -> VRM = Voltage Regulator Module.
It's not really a 'module' anymore but the name remains.
Intel likes to call them VRD [Voltage Regulator Down] now to indicate they are soldered onto the board but Intel is the only one jumping on Intel's bandwagon.

Without being there to see what was going on I can't say about your tests.
First thought is that bloated doesn't mean they are operating out of range [yet!].
Also caps can fail numerous ways. A partially shorted cap isn't going to be letting much ripple through but those MOSFETs are gonna get HOT trying to keep voltage in spec.
.

[BigAlNZ]

Re: Ripple Current

Yes -> VRM = Voltage Regulator Module.
It's not really a 'module' anymore but the name remains.
Intel likes to call them VRD [Voltage Regulator Down] now to indicate they are soldered onto the board but Intel is the only one jumping on Intel's bandwagon.

Without being there to see what was going on I can't say about your tests.
First thought is that bloated doesn't mean they are operating out of range [yet!].
Also caps can fail numerous ways. A partially shorted cap isn't going to be letting much ripple through but those MOSFETs are gonna get HOT trying to keep voltage in spec.
.

Ok - but (a) you agree that I should be able to hook the probe up to the leg of a cap, and the clip to gnd and then use the v/div setting to zoom in on the trace and work out the peak to peak voltage ripple? And (b) that a good VRM system should have no ripple?

Trying to get my head around all this..

-Al

[PCBONEZ]

Re: Ripple Current

Yes that's how you check.
Adjust scope for the frequency you expect to see.
200kHz is a good place to start for SMPS.

At other [higher] frequencies there will also be a small amount of RF noise but that is handled by the tiny caps in the CPU socket or in the CPU itself depending on what kind of CPU it is. [Off the top of my head, for those I've looked up they were ~22uF and were 20+ of them.]
- People don't know or forget those are there. Those caps are on the Vcore voltage and in parallel with the VRM output caps.

[Real world]
There is always -some- ripple because there is always -some- resistance or reactance [or both] in the circuit. The only way to actually have zero ripple would be to have zero reactance and zero resistance and that would be a dead short.
.

[BigAlNZ]

Re: Ripple Current

Thanks for all this. It really sounds like a scope my not help much when looking at Vcore . I really should pull my finger out and get one of Atlas ESR meters.

-Al

[PCBONEZ]

Re: Ripple Current

Scopes can help a lot but you have to use a little math and figure out what your looking for to set them right.

On the other hand you still need an ESR meter because caps in Vcore are in parallel and if only some or one are bad you'll never figure out which one(s) without removing them and checking ESR.
.