EEVblog #594 - How To Measure Power Supply Ripple & Noise

[EEVblog]

Fundamentals Friday
Dave explains what the ripple and noise specifications on a power supply is and how to measure it using different methods on both analog and digital oscilloscopes. From bad techniques through to good, showing the effect of each one. Traps for young players aplenty in this one.
How do you detect common mode noise issues and ensure that the signal you are measuring is really coming from your device under test?
Single ended & differential measurement, DIY coax solutions, termination, analog vs digital oscilloscopes, bandwidth limiting, and even oscilloscope probe coax construction issues. It's all here.
Mysteries of X1 oscilloscope probes revealed:

How to track down common mode noise:

Opamp Noise voltage tutorial:

[apelly]

Nice t-shirt!

[orion242]

Fundamental Friday on Thursday.  Must be back to the future!

[WattSekunde]

Great didactical way from the "normal" probe to the DIY differential probe.

Big That's for sure.

[mariush]

In ATX specifications they recommend measuring the ripple by placing a 10uF electrolytic as close as possible to the probe tip and ground and setting the scope on 20 Mhz as you did.

I guess it would be as a sort of filter for the noise in the leads from the power supply to the connector at the end of the leads?

If you have time, it would be great if you could make a video about how to measure the current ripple a capacitor would be subjected to.. lately I've seen a blog on Corsair's site saying that their primary capacitor is often subjected to current ripples much higher than the rating of the capacitor, but they use them like that anyway due to size constrains. It's not very clear to me how one measures the current ripple in a capacitor.

[robrenz]

Post of a Preamble 1855 (now lecroy) diff amp feeding a Tek 2225 here measuring PSU ripple.

Post of a Tek 7A22 diff amp in a 7603 mainframe measuring <40µV PSU ripple here

Another good technique is to connect all your probing and connections to the PSU with the PSU turned OFF. Whatever you are seeing on your scope is ambient noise getting into your measurement setup. This allows you to tweak your setup without getting fooled.

[EEVblog]

Another good technique is to connect all your probing and connections to the PSU with the PSU turned OFF. Whatever you are seeing on your scope is ambient noise getting into your measurement setup. This allows you to tweak your setup without getting fooled.

Yes, but not foolproof. That likely wouldn't have caught that 142Hz from the electronic load.

[robrenz]

Another good technique is to connect all your probing and connections to the PSU with the PSU turned OFF. Whatever you are seeing on your scope is ambient noise getting into your measurement setup. This allows you to tweak your setup without getting fooled.

Yes, but not foolproof. That likely wouldn't have caught that 142Hz from the electronic load.

Completely agree.

[BravoV]

There goes all used differential probes at the market, all their prices will be sky-rocketed. 

[EEVblog]

There goes all used differential probes at the market, all their prices will be sky-rocketed. 

The EEVBlog, all care, but no responsibility taken 

[BravoV]

There goes all used differential probes at the market, all their prices will be sky-rocketed. 

The EEVBlog, all care, but no responsibility taken 

Lucky me, scored this "last year" under 50 bucks  , bragging my so called "Jim Williams" differential probes. 

Expecting a massive ebay snipers get armed for the word "differential probe" by the mob, hurry folks. 

[Jay_Diddy_B]

Hi,
I am going to contribute the technique that I use for measuring power supply ripple and noise. HP used to the term PARD, meaning Periodic And Random Disturbances.

This is the schematic:



The setup:



A photograph of the DC block, you can also see the 50 ohm series termination at the end of the coax. Tis soldered directly across the output capacitor:



This setup ensures a 50 Ohm measurement environment. The output from the power supply is very much less than 50 Ohms so we need the series termination resistor to prevent reflections in the cable. The spikes may have frequency content in 100-300 MHz range.

Jay_Diddy_B

[jaxbird]

For reference, the method recommended to measure ATX PSUs:





Source: http://www.formfactors.org/developer/specs/ATX12V%20PSDG2.01.pdf

(surprised they recommend very specific Tek scope/probes and caps)

[edit: added details]

[N2IXK]

These kinds of measurements are one reason I keep a 7000 series Tektronix analog scope around. With a 7A22 differential amp plugin, this type of stuff is easy to measure.

This plugin goes down to 10 uV per division, so it is hard to beat for looking at really low level signals. Only 1 MHz bandwidth, though.

I use it quite a lot for checking the output from strain gages when building transducers. No other signal conditioning needed, just a DC excitation source.

[jaxbird]

Lucky me, scored this "last year" under 50 bucks  , bragging my so called "Jim Williams" differential probes. 
...

Nice score .. according to the spec I posted earlier, these are The probes of choice for the job 

[BravoV]

(surprised they recommend very specific Tek scope/probes and caps)

If its the P6046 is what you're talking, I guess it is a proven design and was quite popular back then when CRT scopes still roamed the earth. So popular since it was 1st launched at late 60s, and was still actively produced until in the 90s, and a brand new probe with the amplifier was priced about $1300 back in year 1989, and its equivalent to $3500 in today money.  (Reference -> Tektronix P6046 history)

Probably its like a semi de-facto standard and was mentioned almost everywhere at old literatures when it comes to measurement at a scope using differential probe.

[BravoV]

Lucky me, scored this "last year" under 50 bucks  , bragging my so called "Jim Williams" differential probes. 
...

Nice score .. according to the spec I posted earlier, these are The probes of choice for the job 

Thanks, it was this reading below that sparked my interest in this particular probe for the 1st time.

It took me for a very-very long time and patiently waited & lurked for a good deal at ebay, it was not easy. 

[jaxbird]

If its the P6046 is what you're talking, I guess it is a proven design and was quite popular back then when CRT scopes still roamed the earth. So popular since it was 1st launched at late 60s, and was still actively produced until in the 90s, and a brand new probe with the amplifier was priced about $1300 back in year 1989, and its equivalent to $3500 in today money.  (Reference -> Tektronix P6046)

Probably its like a semi de-facto standard and was mentioned almost everywhere at old literatures when it comes to measurement at a scope using differential probe.

No doubt they are the reference model in the world of differential probes. I was just surprised an official document would note a recommendation on scope model and probes to use for a specific measurement

Would be great if you could do a comparison between the p6046 and analog Add/Inv and/or digital math sub. I think it would be interesting to see how much extra detail is obtainable with a proper set of diff probes vs. using 2 channels.

[darrell]

I'd tie the grounds of the two coaxes together at the measurement end since the return current has to go back to the scope otherwise. That is probably less important if 20 MHz is all you are measuring. The series terminations at the measurement end can be omitted as the scope should be a good match and not reflect much. The coax DC blocks are nice since you can get the ground loop even smaller on the board by omitting the capacitor.

For measuring power supplies on boards, I like to remove a decoupling cap and solder coax there. Or, alternately, do as Xilinx does on some dev boards and include a footprint for a coax connector to each power plane.

[han]

Instead differential amplifier why not using balun with DC block on input side?
there is 50kHz ~ 200MHz balun with quite flat frequency respond.

[i4004]

rather weird this was published today, the day i almost burned the ground clip lead of my hameg hm307 scope measuring the ripple on pc psus. 

also interesting in the context of

(EEVblog #279 - How NOT To Blow Up Your Oscilloscope! )

current did flow (to the point of starting to melt ground clip wire of the scope...luckilly i don't see any damage to the scope...or psu) when i (by mistake) touched ground clip to +12V output of psu.
wall outlet that was used (to power both scope and pc psu) doesn't have earthing contact.
that's what the eevblog 279 lacks, description of floating ground dangers... 
(when both devices are plugged in the wall socket, continuity exists between psu metal case and ground of the scope, which means that 12V output was shorted for a brief period via that scope ground clip...)

mesuring itself was much easier than i thought it would be, crappy/worn psus producing more ripple/noise than good ones even on light loads such as 12v car light bulb of 60w (or 21w, for that matter), as without load there's nothing to measure, it's always perfect dc...unless caps are totally blown and then feedback shuts down the primary switch circuit as soon as you turn the psu on....
much easier than i thought by reading (and participating) in the previous discussion on badcaps and jonnyguru forums...
(offcourse my aim was never to get a ueber vertical resolution of that noise..i was just wondering on how visible it is on psus that are starting to malfuncion, ie to damage motherboards they power...i just hooked the scope to the output of the psu with that bulb as a load)

[grimmjaw]

Hi,
I have two questions which somewhat  related to this topic.
How much is the noise and ripple effect to a mixed signal PCB.
Say a precsion ADC and a few digital logic block (SPI etc).

If I separated the Vddd (powered by SMPS) and Vdda (powered by Linear PS),
the noise and ripple  should't effect the ADC (50SPS)right?. How much an effect has the SMPS
to the digital block , SPI running at 20Mhz?

Secondly has the noise and ripple ex.200mVpp has an effect on a constant load?
For example 12 Vout through a 100 ohm ..so will 2mA of extra current will be flowing through the resistor?

 
 

[hammy]

This video also have a lot problems with the autofocus of your camera. 

You got a new camera, Dave? Or you changed something? Is the camera now with a new mount moved closer to the device?

Something changed, that's for sure!

[mariush]

I think it has more to do with that Ultravision feature of the scope, which fades the graph so the camera tries to focus on it from time to time.

[hammy]

Also in the video "#592 Mailbag" you can see it. The autofocus is slower now.