I'm troubleshooting a system that is being bombarded with EMI. So far I've found a few basic design no-nos that should help mitigate the circuit's susceptibility to the EMI, but, I'd like to confirm some other items that I'm a bit fuzzy on.

Measurement setup:

Scope with 4GHz SE probe
reference tied to regulator output reference
signal tied to regulator output
leads as short as possible, difficult due to location, but i'll accept some noise due to this limitation.
Probe offset set to 3.3V

During EMI:

VCC swings between 6.6V and -6.6V.

Investigation shows nothing to clamp the output of the regulator, so my first thought is a uni-TVS.

My question is, how can I be sure this is "real"? Also, unfortunately I cannot probe in parallel close to the load, so it's LC filtering may be slowing the energy induced and be somewhat stable. Without being able to see what the load sees, would you agree a TVS at the load would be a good idea to ensure whatever noise is present is clamped to the best of our ability?

 

It is very possible that you're seeing an error in your test setup. I suspect you're trying to measure noise on a 3.3V regulator, correct?

Is your probe 10x or 1x? This makes a big difference when measuring this type of signal. Use a 1x probe if you have it, if not... an acceptable way to measure noise is to RC filter with a 50ohm resistor and terminate into a 50 ohm oscilloscope input (with a 50 ohm coax, of course). This allows you to not need to measure on the most sensitive settings on your oscilloscope reducing susceptibility to noise in several different ways. Also keep you probe leads as short as possible. Don't use those 5" ground probes... they are an AM antenna and a huge inductor. Remove all unneeded test connections (DMM's, function generators, etc). 10x probes can have up to 20mV of noise (on a digital scope - depending on the quality of the A/D converter).

Keep all wires as short as possible. Twist all wires with their return wires.

Turn all electronics off in the vicinity... cell phones, computers, fluorescent lights sometimes cause problems, etc. Just the other day my cell phone was giving me problems in a noise test.

Another trick I use is to put a ESD bag over my test fixture. This will do a fair amount of shielding, but use a good bag because old bags can short nodes.

If you are really having problems with radiated susceptibility put the device in a metal box with well grounded connectors (BNC's) this will block electric fields from getting into the box. magnetic fields are more powerful and need special types of metals to shield properly.

Look at your grounding and make sure you don't have any ground loops. That another big one that is overlooked quite a bit.

 

It is very possible that you're seeing an error in your test setup. I suspect you're trying to measure noise on a 3.3V regulator, correct?

Is your probe 10x or 1x? This makes a big difference when measuring this type of signal. Use a 1x probe if you have it, if not... an acceptable way to measure noise is to RC filter with a 50ohm resistor and terminate into a 50 ohm oscilloscope input (with a 50 ohm coax, of course). This allows you to not need to measure on the most sensitive settings on your oscilloscope reducing susceptibility to noise in several different ways. Also keep you probe leads as short as possible. Don't use those 5" ground probes... they are an AM antenna and a huge inductor. Remove all unneeded test connections (DMM's, function generators, etc). 10x probes can have up to 20mV of noise (on a digital scope - depending on the quality of the A/D converter).

Keep all wires as short as possible. Twist all wires with their return wires.

Turn all electronics off in the vicinity... cell phones, computers, fluorescent lights sometimes cause problems, etc. Just the other day my cell phone was giving me problems in a noise test.

Another trick I use is to put a ESD bag over my test fixture. This will do a fair amount of shielding, but use a good bag because old bags can short nodes.

If you are really having problems with radiated susceptibility put the device in a metal box with well grounded connectors (BNC's) this will block electric fields from getting into the box. magnetic fields are more powerful and need special types of metals to shield properly.

Look at your grounding and make sure you don't have any ground loops. That another big one that is overlooked quite a bit.

I am using active probes, 5x. I tried shortening the connections, still the noise is present. Sniffing with a near-field E probe confirms 32 kHz frequency riding along cabling back to base location. I measured this with the system off as well.

Noise source is high voltage low current device that has a pulse repetition interval of approximately 32 kHz. So no surprise that it's present.

As far as grounding the shielding, it is tied to a hybrid ground at the base location, floating at the remote location. Additionally, a secondary shielded twisted pair is connected at both ends to signal gnd for high frequency signal shielding.

Perhaps it is magnetic field inducing the noise, this could explain why our shield is not effective. I suppose I should measure with an H field probe to determine if there is a LF magnetic field that's causing the headache.

Is it possible for the noise to have both a electric and magnetic field component?

 

Is it possible for the noise to have both a electric and magnetic field component?

That would be why they call them electro-magnetic fields! I missed that you know the source of your noise in your first post. Hi voltage low current means that you'll be dealing mostly with electric fields.

I don't know what you mean by hybrid ground. That's a new term to me. Do you mean a floating ground? I'm going to assume you're using active single ended probes, meaning that your ground is the earth ground. When you connect your probe to the circuit that floating ground then becomes earth ground. This really shouldn't be a problem if your ground is truly floating. But double check all of your grounding.

http://www.interpoint.com/product_documents/DC_DC_Converters_Output_Noise.pdf - Figure 3 of this document shows what I mean by making your probe connections as short as possible.

To properly shield from E-fields you need to have the entire circuit completely encased in metal with no openings of any kind. Any connectors should be sealed entirely and cabling should be shielded as well.

I doubt it's magnetic... my money is on probing techniques, shielding, grounding, or a combination thereof (what else is there?)

Good luck, that's about all I can give ya without knowing significantly more (pictures, schematics, etc.)

 

That would be why they call them electro-magnetic fields! I missed that you know the source of your noise in your first post. Hi voltage low current means that you'll be dealing mostly with electric fields.

I don't know what you mean by hybrid ground. That's a new term to me. Do you mean a floating ground? I'm going to assume you're using active single ended probes, meaning that your ground is the earth ground. When you connect your probe to the circuit that floating ground then becomes earth ground. This really shouldn't be a problem if your ground is truly floating. But double check all of your grounding.

http://www.interpoint.com/product_documents/DC_DC_Converters_Output_Noise.pdf - Figure 3 of this document shows what I mean by making your probe connections as short as possible.

To properly shield from E-fields you need to have the entire circuit completely encased in metal with no openings of any kind. Any connectors should be sealed entirely and cabling should be shielded as well.

I doubt it's magnetic... my money is on probing techniques, shielding, grounding, or a combination thereof (what else is there?)

Good luck, that's about all I can give ya without knowing significantly more (pictures, schematics, etc.)

Hybrid ground - shield tied to ground through cap, ground in this case is earth.

Anyways, I agree the amplitude of the noise i'm seeing is most likely due to probing error because I do have 1-2" tails on my leads, I am working to fix this (requires drilling and product modification).

As far as shielding goes the entire product is enclosed in an earthed shield, until the point of this hybrid ground, from there on, only the cable carrying this voltage and signaling is shielded (housing's plastic). The circuitry at the end of the cable is shielded (copper tape for now) with a ~1/2" aperture at the very end to allow for interaction with outside world. I'd think the 1/2" aperture would definitely be susceptible to frequencies of 1GHz and higher (1/20th approx), but 32 KHz? 1/20th approx is still something like 10 miles, .

So that's what lead me to think magnetic interference, need to get some muMetal if the h-field measurement shows a high amplitude magnetic field at 32khz

 

I suspect your hybrid ground cap is relatively large (>1uF). Hence, if there is noise on your ground, it will probably couple 32kHz easily. Have you tried removing the cap to see what happens? You only provide DC isolation with the cap, not AC. Even a 1uF cap gives you about 5ohms of impedance at 32kHz (Z=1/(2*pi*f*C))

It sounds like when you connect your probe to the circuit you short out that cap and cause a ground loop at high frequency! You need one of 3 things to make this measurement. 1) a good differential probe 2) a battery operated scope or 3) simply remove the hybrid ground cap. There are other ways to do the measurements too, but due to what you're trying to measure these are the most accurate and safest ways to do the measurement. I won't recommend any other method.

Another thing to point out is that it will take a lot of power to get 32kHz to propagate in free space. Unless the noise source is very high power then I doubt it's radiated susceptibility your having problems with. I should have keyed into this sooner. Sorry.

My money is still on the ground loop. I bet that you'll see a significant attenuation in your noise when you do one of the three above methods. You'll then have to improve the probing techniques to really measure your amount of noise.

 

I suspect your hybrid ground cap is relatively large (>1uF). Hence, if there is noise on your ground, it will probably couple 32kHz easily. Have you tried removing the cap to see what happens? You only provide DC isolation with the cap, not AC. Even a 1uF cap gives you about 5ohms of impedance at 32kHz (Z=1/(2*pi*f*C))

It sounds like when you connect your probe to the circuit you short out that cap and cause a ground loop at high frequency! You need one of 3 things to make this measurement. 1) a good differential probe 2) a battery operated scope or 3) simply remove the hybrid ground cap. There are other ways to do the measurements too, but due to what you're trying to measure these are the most accurate and safest ways to do the measurement. I won't recommend any other method.

Another thing to point out is that it will take a lot of power to get 32kHz to propagate in free space. Unless the noise source is very high power then I doubt it's radiated susceptibility your having problems with. I should have keyed into this sooner. Sorry.

My money is still on the ground loop. I bet that you'll see a significant attenuation in your noise when you do one of the three above methods. You'll then have to improve the probing techniques to really measure your amount of noise.

Thanks for your help, I'm looking into this and have started spinning our remote assembly into a more accessible evaluation module. This will allow us to try different shield grounding techniques as you mentioned.

For reference the hybrid ground is actually implemented with safety caps, so 200pF. Measurement of the noise source alone indicates the 32kHz is actually the modulation frequency additonally, this is high power device, it is capable of arcing to it's return circuit through air, it's actually the intent

Thanks again, I was thinking about what you had mentioned and will definitely be trying that.