EMC immunity

[2dye4]

I have designed a card to go into a machine that has a noisy PWM drive near where my card is going to go.

Card was designed with this in mind to create as much immunity as possible.

The product is costly to set up and test in whole.

My problem is that i have been requested to measure the safety margin against my card failing due to EMC emissions from the PWM drive. Somewhat like if the source was X times stronger what is X where a failure would occur.

I am not sure how to do this as the source is not easily quantifiable nor is the vulnerability in the card.
And i can't make the source go higher than max.

My thoughts were to set up a few and test thoroughly by just running through the operating range and looking for failures. How do you attack a problem like this??

Thanks

 

[VEBill]

Maybe scope around the circuit and measure the noise level. You'd have to be careful to minimize the effect of the measurement technique on the results. And ground references (use two probes differential). Then compare the induced noise level to the specifications for the circuit to see how close you are to triggering an unwanted reaction.

At a system level, they're often looking for a minimum 6dB margin.

For some tests, the approach is to replace the DUT with a data-logging dummy (this assumes the susceptibility is conducted into the DUT).

Stand by for others' input.

 

[Skogsgurra]

There are three kinds of compatibility: Theoretical, formal and real. (This is not to be found in any paper or standard, it is the way I use to talk about it).

You obviously have Real Compatibility. And that is OK IRL. But, if you are in the EU, you also need Formal Compatibility. That can be achieved by submitting your device to different levels of radiated and conducted HF emissions and see what your immunity is.

If your device withstands the highest emission levels and also ESD, then I think that you cannot do much better.

The problem, as I see it, is that someone has asked you to perform some kind of testing without knowing that there are standardized tests that should have been specified when the product's specifications were written.



Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[VEBill]

This might be a case where the immunity requirements exceed any of the standardized tests. I've seen cases where compliance needs to be demonstrated that there exists a 6dB margin in situ.

 

[benta]

There are laboratories doing EMC testing and they are not even that expensive.
In this case you'd need them to find your board's suceptibility and to quantify the emissions from the PWM area.
I'd expect them to be able to do this for 1...2 k€.
But you need to define the failure criteria for the board.

Benta.

 

[2dye4]

Thanks for good posts.

Let me clarify a little bit.

My management is only concerned about reliability.
I voiced my concern some time ago about the proximity to the noisy high power PWM circuitry operating at 480 VAC which could cause not component failure but misoperation.

I was in favor of building units and running them and watching for mis-operation. With enough hours certainty in the reliability can grow to desired levels.

The source is hard to characterize and the vulnerability is hard to characterize and my management wants numerical safety region type of data.

VE1BLL
Your close to what i need but i don't have vulnerability info for the board. I designed it so I do know what levels of voltage would create mis operation, so I could monitor some key points but they want a well designed experiment showing the environment intensity as well as what it takes to create failure. They are not knowledgeable about electronic design.
I think its very hard to design an experiment to prove reliability given the wide variance in the emissions of the radiating circuitry.

I would think this type of question comes up in design every once in a while for high reliability items.
The only sure path i can think of is redesign to further enhance expected isolation, more shielding and filtering.

I dunno.

 

[VEBill]

Or go backwards. Reduce the immunity of the design until you see failures, and then use that information to calculate the margin of the full design.

 

[Noway2]

Keep in mind that susceptibility and emissivity are flip sides of the same coin. In other words, if your product radiates a lot, then it stands to reason that it will be susceptible to interference from other sources.

Since it is easier to measure the emissions in quantifiable terms, this may be a "back door" method to get the confirmation you are looking for.

 

[IRstuff]

You can rent RF probes and instruments.

You can then measure the environment, rent EMI generators and generate a scaled-up version of the measured environment.

TTFN

FAQ731-376

 

[itsmoked]

Here is what I'd do for your case. I would get an isolated scope. Like a battery powered type. My Tek is isolated even when run with a wall wart.

Go thru all lines of interest hooking the scope probe ground to your board's ground. Hit the signals of interest and the power.

Then turn on the PWM noise maker. You will see the noise change it causes. This is the normal noise your board will be seeing. Methodically run thru all the same signals you did previously. Make written notes. Better would be photos or screen captures.

Now the next thing you do is temporarily replace one of the existing PWM lines with just a wire that can generously reach your board.

With the scope hooked up you can easily see the noise increase dramatically as you bring this loop of wire over to your board or the leads coming to your board. I would suggest that you actually bring it to a point where your board malfunctions. Either way you will be able to see a quantitative result. Either the noise reaches some high level and your board continues to perform or it chokes at some point and you will be able to see that point and easily duplicate it.

I use this method to test a lot of my product designs and it is quite illuminating. I discovered that those fluorescent magnifier lights give off hideous amounts of noise. You can often crash a board or communications with the use of one of them. If you bring it down close to your DUT you can see the strange triangular noise waveform that shows up on your long signal wires(>1"). If you what a problem locator use a small screw driver that you hold by the shank in your fingers. Use it as a noise probe. You can touch anyplace you want to inject noise. If you want mega-noise touch the actual bulb with the fingers of your other hand while probing. By moving an elbow closer or father to the light, you can control, rather accurately, the injected noise.





Keith Cress
kcress -

http://www.flaminsystems.com

 

[zeusfaber]

It's probably a good idea to separate your problem (at least conceptually) into questions of conducted and radiated susceptibility, and deal with them one at a time.

One place you might look for clues about how a complete EMC specification and testing regime can be put together is Def Stan 59-411, available at

http://www.dstan.mod.uk/data/standev/59e.php

You probably won't want or need to invest the time and money to do everything the Standard suggests, but it should give you some pointers for coming up with a defensible regime of your own.

A.

 

[zappedagain]

keith, I should have known you'd have a calibrated elbow!

I'm curious, did you see this with a magnetic ballast (50-120 Hz fundamental) or an electronic ballast (much higher frequency)?

Z

 

[itsmoked]

Zapped; Magnetic. But if you touch the bulb you get the arc noise which is almost a white noise. HF

Keith Cress
kcress -

http://www.flaminsystems.com