Oscilloscope probing paradox

[Noise0]

hi, i have often an issue that i am unable to solve by myself, i need some help..
The pattern is this:
1) I found some unexpected noise or oscillations in a circuit.
2) After many tests i ask myself is the issue is due to some bad probing practice, maybe am I catching some EMI witt my probe ?
3) I disconnect the probe prom the circuit and i start moving it around to find if i read some disturbance coming from the underlying conductors, but i got no signal.
4) I touch the circuit with a short-circuited probe and i got the noise.

For my limited comprehension this seems a paradox because:
- The noise have to be conducted because happens only when i touch a conductor
but
-The noise have NOT to be conducted because the probe is short circuited

This situation throws me into total despair because it makes even my most elementary beliefs fail..
Moreover I am not able to understand in which direction to move to solve the problem.

I am just an hobbist, if you could help me, with simple words I would be very grateful

[Vovk_Z]

Typical common-mode noise.
You may start from here and go on: EEVblog #279 - How NOT To Blow Up Your Oscilloscope!

[Vovk_Z]

Here is a video closer to the question itself: EEVblog #441 - How To Track Down Common Mode Noise

[David Hess]

I agree with Vovk_Z.  That is a classic example of common mode noise.

That means there is noise between the circuit and probe ground so when the probe ground is connected to the circuit, current flows through it.  The common solution is to use a differential probe, which might be two probes with the oscilloscope in add and invert mode, which will subtract the common mode noise from the measurement.  An isolated probe would also work.

Check that the oscilloscope is powered by the same AC circuit that is powering the device under test.

[Manul]

Your scope is mains grounded (at least it should be) and your device is grounded, or at least capacitively coupled to the ground (this is inevitable). So DC or at least AC path exits. It means, a current can flow between your device and scope through the probe cable shield.

For a simple example, look at the first attachment. Thats a DC model of 10x probe attached to a scope. Series resistor R1 is probe input and R2 is probe cable shield. Probe is shorted. Scope is measuring voltage on R3. The 1V source, which is ground referenced is dumping 3.3A into the shorted probe. Almost all the current is passing through the cable shield and because it has resistance, this current creates a voltage drop. Ground potential at the probe tip and scope end is different. This voltage difference is shown by a scope as DC input voltage. So a shorted probe does not mean that a voltage cannot appear at scope input. If there is current flowing through the cable, the grounds will shift and voltage appears at R3.

While the above is true, it is not so simple with high frequency AC. The effect is dominated by a bit different things. Look the AC model. In the rectangle area is a simplified model of transmition line. R3 is scope input impedance, R4 is signal source impedance. You short the probe input with a ground clip. It is shown as inductance L3 and resistance R5. Resistance almost does not matter. What matters most is that inductance which is not part of transmittion line. The AC current, which goes into your shorted probe is passing L3 and creates a voltage drop because of it's impedance. This is what your scope sees. You can prove the theory by shorting the probe with a shorter wire - the amplitude will decrease. Also, if you short the probe as you did, but touch with a ground ring (which is used for a ground spring), you will likely see almost no effect. It is all because of that ground clip wire which is not part of cable.

Contemplate these ideas in your head and you will go out of mystery.

[Noise0]

Thank you, your explanation is crystal clear and halp me a lot  in understanding the problem better.
I have found that the source of common mode noise is my PSU. It's a 30A switching PC PSU modified by me in the attempt to achieve a ground isolated PSU. Of course my PSU hacking design is faulty.

I replaced that PSU with a battery and now i am able to probe my widget like a charm , even my old linear psu works good.

But this bring me to a new issue.
I have done the same test (with the short-circuited probe) on an Arduino connected to the PC USB (main earth). And  i got common mode noise again.
It seems that connecting together two or more earth referenced circuits is not good as far as I understand from your explanation and from my measurements....
As I told you, I am just an hobbist and digging deep into understanding electronics and buying expensive equipment is beyond my abilities and interests BUT in the light of these new discoveries I have to make some changes in my laboratory setup in order to get rid of those sneaky issues.

I have done a sketch (picture 1) representing the worst case scenarion that can happen on my workbench. Worst case means that all those blockls have to be connected together via common ground.

If I could just keep the oscilloscope connected to main earth and everything else at isolated ground, would my life get better?

In order to achieve this: an optoisolated usb hub (picture 2) and a commercial switching psu (picture 3) could works fine for my needings?

Thanks in advance

[Manul]

Generally it is ok to connect two mains earth referenced things. The problem is all these switch mode power supplies. Welcome to the modern world. Full of garbage flying through the wires and in the air. Listening to low frequency radio bands (LF, HF) also became impossible. Pollution everywhere. Serious analog scoping is hard and the lab should be prepeared for that specifically. Gladly, in a lot cases we do not care about small noises and just ignore them.

Short ground clip and scope bandwidth limit to 20Mhz (if possible) reduces noise by a lot. When I use ground spring on my scope I get very little noises, so I just ignore them. You may play with feritte rings on the cables.

If I could just keep the oscilloscope connected to main earth and everything else at isolated ground, would my life get better?

In order to achieve this: an optoisolated usb hub (picture 2) and a commercial switching psu (picture 3) could works fine for my needings?

Even if you have full isolation (like isolation transformer) it might still be a bit of a problem with switchmode PSU. It is isolated DC, but some AC will still couple.

I dont know how bad is your noise, you might just live with that. Someone else might have more tips about this.

[S. Petrukhin]

I didn't really get it...
No one said that the loop of the ground wire works like an antenna? This is a coil turn.

If you want a low-level measurement, hide from interference, use the spring that comes with the probe to take the ground nearby and not have a receiver loop.

[radiolistener]

-The noise have NOT to be conducted because the probe is short circuited

This is a false claim. The short circuit works as a magnetic loop antenna.
So, it will produce a lot of the noise from a nearby noise source.

[Noise0]

-The noise have NOT to be conducted because the probe is short circuited

This is a false claim. The short circuit works as a magnetic loop antenna.
So, it will produce a lot of the noise from a nearby noise source.

The noise appear only when I touch the conductors.
If it were a radiated signal catch by antenna, it would be visible even in the absence of electrical contact with the source.
Am I wrong?

[Vovk_Z]

The noise appear only when I touch the conductors.
If it were a radiated signal catch by antenna, it would be visible even in the absence of electrical contact with the source.
Am I wrong?

Yes, you are right. So it is common-mode noise.

[S. Petrukhin]

-The noise have NOT to be conducted because the probe is short circuited

This is a false claim. The short circuit works as a magnetic loop antenna.
So, it will produce a lot of the noise from a nearby noise source.

The noise appear only when I touch the conductors.
If it were a radiated signal catch by antenna, it would be visible even in the absence of electrical contact with the source.
Am I wrong?

This may be the noise that flows through the probe shield to the PE from the circuit. The shell works as a resistor, from the circuit to the PE, and the scope measures the voltage drop on the shield. The ground jumps at the noise.

But it may be the noise from the resulting autotransformer, the winding of which is the loop of the gnd wire.

[S. Petrukhin]

It is possible that the PE contact of the power plug of your scope is not connected to the outlet, you have a floating potential on the ground of the scope.
Be careful to make a short circuit through the probe shield to the PE from your circuit.

[Manul]

Ideal coax transmittion line just carries power from one point to another. This is exactly the purpose. It does not radiate and does not receive. For example, if we have a coax cable running from radio transmitter to antenna, the antenna radiates power, but the cable itself does not radiate. This may seem a bit strange, because RF currents are flowing through the cable shield too. But the electric fields of shield and central wire are oposite and fully contained inside the coax. Fields do not reach outside and do not radiate. The transmittion line is like a water pipe, where water is flowing inside from A to B, and does not leak out. (Pic. 1)

Same is true for reverse action - receiving. Ideal coax cable will not receive common mode noise. The current will be running through the shield, but not affect a signal inside. Imagine spilling water over the outside of water pipe. Water just runs through the outer surface of the pipe and does not affect the water flow inside. (Pic. 2)

When you short the probe with a 10cm wire, you create a "hole" in this pipe (transmittion line). Things can leak in and out. (Pic. 3)

If you use same setup like in your first picture, but touch with the metal ring (probe.jpg), you will see very little noise apearing. The hole still exist, but the spilled water is not going over it, so no water comes into pipe. (Pic. 4)

Practical transmittion line is never perfect, it contains some small leaks, also shield has some resistance. There could also be some little holes in the scope input stage and so on. Still it does not mean, that common mode noise is always a big problem. Good ("hermetic") probing connection will reject almost all of it.

[Noise0]

In the attempt to investigatethe common mode noise, today, i made this useless test:
Refer to picture.
I made a dummy load circuit that switch current draw at 1Khz.
Then i powered it with:
-A lipo
-My linear psu
-My switching psu

Then i connected the short-circuited oscilloscope probe circuit to gnd.

I expected to read no signal with the lipo, a low common mode noise with linear psu, and the full common mode noise with switching psu.

Of course the resut is different, even with the lipo i pick up a signal; 
while linear and switching psu get the same signal.
 
it's clear that im just doing a mess.
Maybe it's better that i do a step back and I should post the circuit on  which I was working on at the beginning of this post, forgetting about this short-circuited probe thing, sorry

[Bud]

Johnson, Graham
"High - Speed Digital Design. A Handbook of Black Magic."

[Noise0]

Johnson, Graham
"High - Speed Digital Design. A Handbook of Black Magic."

thank you all for the reply
I am aware that a probe loop catch electromagnetic through the air
But this time I was amazed by the fact that a new strong noise appears only when i electrically connected the loop to the circuit.

Users suggested me to pay attention to the "common mode noise" topic.
But i fail to realize how relevant it can be in my case, how to test it. I don't even know if it is possible to check if a power supply is a source of "common mode noise". Does it make sense to speak in these terms?

I also can't imagine a situation where two connected to earth equipment can be joined together without generating common mode currents.
I made a schema to visualize my perplexity.
Let's say that device A transmits a certain current to device B through the SIGNAL cable.
The current will flow back baybe through GND or EARTH, you cant tell.
So if not all the current return to A through GND the consequence is that some common mode current appear on the SIGNAL/GND pair.
Is my reasoning correct?
So I see tree options:
-Common mode current is normal behavior
-Two or more connected to earth devices cant be joined
-I dont get the point

For example an Arduino board connected to PC USB hub (PE connected)  AND  an Oscilloscope  (PE connected) can be be joined (I mean connecting their GND together) ??
In this case the current coming form PC USB will flow back to PC not on the usb cable but on the earth cable since its bigger and maybe low impedence. Then again a comon mode current appears on the usb cable.
 

[Noise0]

hi, today I got a new 12V 30A PSU
I am testing noise open circuit and under 10A load
The picture shows 220mVpp

When I touch GND with short-circuited probe i got same 50Hz 68mVpp noise

Is that the so called "common mode noise"?
How to get rid of that?
Is there something wrong in my setup?

bye

[Vovk_Z]

I'm not sure but it seems to me you possibly have too much common mode noise in your measurements (it's hard to tell from just a two photos). So possibly you have some problem with a power supply of your oscilloscope (or with ground wiring) or something else.
Or you just don't get enough into a measuring with a high noise present.
That book 'Johnson, Graham "High - Speed Digital Design. A Handbook of Black Magic." is definitely one (of two or three) bibles on this topic.

[Noise0]

good new
i got big improvements using that ferrite core salvaged from usb cable
using 4 turns, the common mode noise fall from 70 to 28 mv
and the diffrential noise from 150 to 52 consequently
it seems that more turn i do more common mode noise is attenuated.

since that psu have to deliver 30A it seems that i need abigger toroid

[Manul]

Frankly speaking, what do you expect from such PSU? They are not build to be low noise. They are build to be powerful and cheap.

[Noise0]

Frankly speaking, what do you expect from such PSU? They are not build to be low noise. They are build to be powerful and cheap.

Good point!
At least i hope to learn something.
Since i have not a ferrite big enough to hold my 30 A cables , i have done the same test as before with a laminated transformer core.

It was found to be completely inefective in choking common mode noise.

Hmm.. probably iron is not the same as ferrite? 


Update:
Its not completely inefective, it shows some reduction of CM but only a little and only with much higher current: not good.

[radiolistener]

The noise appear only when I touch the conductors.
If it were a radiated signal catch by antenna, it would be visible even in the absence of electrical contact with the source.
Am I wrong?

when you touch conductors, you're improve coupling, it leads to RF current leakage through conductor.

[Manul]

i have done the same test as before with a laminated transformer core.

It was found to be completely inefective in choking common mode noise.

Hmm.. probably iron is not the same as ferrite? 

You want inductance for high frequencies. Iron core does not work for high frequencies. Above a few hundred Herz the inductance decrease rapidly because of Eddy currents generated in the core. Ferrite has many feromagnetic particles isolated from each to prevent currents. Also ferrites themselfs are not equal in that regard, they have different frequency characteristics and will not work equally good in your aplication.

[Wallace Gasiewicz]

You will also pick up noise from your USB. Ethernet cables also radiate. You can pick up signals from an ethernet cable on a short wave radio easily.
Most Switching Mode Power Supplies are really terrible.
You can just connect your ground clip of your probe to the tip to make a loop and you will see noise in lots of places. If you use a Spectrum Analyzer it will be easier to see than on a scope. Move the loop around some electronic devices and see. The tip by itself is not a good antenna but making a loop with the ground lead improves it as an antenna. Also connecting it to a wire or anything else that conducts.
This is a thought: why not get an old fashioned linear variable bench PS for Arduino work? They are cheap these days. If the Hamfests ever open again you can pick up a nice small one for maybe $5-10. At least you can get rid of some of the garbage. They are far less efficient but you probably don't care while you are developing something.
Maybe someone else will disagree, but I just switched to a modern digital scope(Agilent) and the old analog scopes have less "interference" (aliasing?), especially at low freq.

It is tough to get rid of this stuff, it is in the air (sorry, it is in the ether)

[Noise0]

This is a thought: why not get an old fashioned linear variable bench PS for Arduino work?

You are right.
The reason why i had to switch to a 30A Switching PSU is because recently i had to join my widgets with commercial boards capable to draw some current. Namely small quadcopters blushless motors ESC.

Anyway the project is a success!
I built an effective common mode choke using spare ferite beads then connected the PSU to my front end in the picture.
I got a nice 50mVpp noise, thet's fine.

Furthermore, most important thing, is that now the original circuit under test shows stability paired with this PS.
While before it was affected by nasty oscillations.

thank you all for the great advice, much love