EMI: I'm in or out the directive? (edit: poor man's testing method)

[dannyf]

Get an LED  + loops of wires. Put the led + loop next to a few known compliant devices and your device and see how bright the LED is.

Cell phone flashers will do the same.

[Precipice]

Get an LED  + loops of wires. Put the led + loop next to a few known compliant devices and your device and see how bright the LED is.

Cell phone flashers will do the same.

I'm assuming you're kidding, yeah?
By the time you're even remotely comparing your unintentional radiated power with a 0.5W (or more) intentional transmitter, you've got some pretty serious EMC issues!

[Tabs]

For what its worth:
test houses charge around £1200+/-200 per day or £3-4k for a full suite of tests.
It depends on what product family standard your product is covered by and what tests it tells you to do.

see
http://ec.europa.eu/enterprise/policies/european-standards/harmonised-standards/electromagnetic-compatibility/index_en.htm

for european EMC harmonised product family standards. If you dont think any apply then the generic residential,commercial and light industrial standards apply and you will have to decide which ones to apply.

Being battery powered with only internal cables connecting to a front panel makes things a lot easier for you so I would expect the test house to charge much less (since there are no surge/bursts/EFTs and no conducted immunity or conducted emissions to worry about)

The EU allows you to self declare but it would be in your interests to do this based on evidence.
BTW
Even when you do have evidence from a test house, you still only have the presumption of conformity and if your product still fails in real use then its your problem, but at least the courts will see that you did due diligence.

That being said, it looks like you only need radiated emissions and radiated immunity.
The test house will charge much more for radiated immunity because they need amps, couplers and power meters. The setup and run time is longer because you have dwell times at each frequency.

The most important one for you in this case is radiated emissions because this has the potential to interfere with other people equipment.
Problems with immunity will give you customer relation problems and will only cause legal issues if one of them takes it that far.
It sounds like this is a 'hoby/enthusiast' product so the above may not happen. (If this is meant to be a proper consumer product then you need to test).

If test houses still aren't cheap enough then hire a spectrum analyser and antenna. If your basement is quiet you should see the SA noise floor. If you do see something then make a note of its characteristics so you can distinguish it from spikes caused by your product. Don't take the difference between the two readings. If your basement is not quiet then go to the middle of nowhere (get a SA with a battery option) and use it as though it was an OATS.

If still too expensive then then use the near field techniques just to see what frequencies are floating around your product.
IMHO: your'e more likely to fail in sub 50MHz region. These will mostly be down to common mode imbalance in your cables.

 

[mcinque]

Thank you tabs for you detailed post.

However, since this will be an hobby/enthusiast product for a very small niche, I often ask myself if I really need to do all this. I'm putting a big effort in this and sometimes I think that I wasting my time (and money).

I strongly believe that most niche-products like arduino relay or display shields, crap digital multimeters, rubbish power supplies modules, disgusting tx-rx modules, all sold by various European online shops have not even saw from a mile a piece of paper that talks about EMI and EMC. And no authority is preventing or acting against this products.

Take for example this module produced by this company and by Sparkfun also in Europe.

I've one of this crap modules and I've seen many differences comparing to a serious Aurel tx module like this (I know it's not the same frequency but this is not the point: both frequencies are well regulated here.

The sparkfun module does not have any declaration of conformity nor meets any EN requirement, it hasn't even the metal crystal oscillator case connected to ground... judging by its price you can easily imagine what kind of crap it radiates when turned on, or its frequency stability. The Aurel instead shows a crystal clear razor-centered main tone with basically nothing in the surroundings. As a serious tx module should do.

We have a chinese crap toy tx module vs a serious EU-made tx module, fully certified and compliant.

Even if the manufacturer of the crap module is in asia, the reseller is still responsible for putting on the market something that is not CE compliant. But if this was true, noone here would sell that garbage. Neverthless, almost every hobby/enthusiast electronics european online shop, sells that filth with no issues.

This makes me think that if you're not in a big market, you can sell whatever you like: noone will catch you. But this is not fair.

[mcinque]

Ah... so, you have a 2 layer board and some internal cabling.

Without a solid, unbroken ground plane, you're just asking for trouble.

I appreciate your thoughts, seriously.

But I wonder how many simple commercial products (like alarm clocks for example, or cheap bench power supplies) can be EMC compliant using  internal cabling and 1 or 2 layer PCB without any ground plane

[AndyC_772]

They may well be compliant, but they'll have been tested to be sure. They may well not be rev 'A' PCBs either.

Not having a ground plane doesn't mean your product won't be compliant, but it does mean there's a greatly increased risk that it won't be compliant on its first iteration compared to a board which does have a ground plane.

2 or 3 PCB revisions to identify and fix EMC failure points is not at all unusual, especially if you're not experienced in EMC design to begin with.

[qno]

A 4 layer board is hardly more expensive than a 2 layer board.
Maybe 20 % when ordering 100 pc.

[mcinque]

They may well be compliant, but they'll have been tested to be sure. They may well not be rev 'A' PCBs either.

Imho, anything that is compliant shows with more or less pride a declaration or a certificate. Or at least, you can find it if you search. There is nothing to find about that products.
Many power tx modules or cheap DMMs has not even a known manufacturer. They could not even sell them...

especially if you're not experienced in EMC design to begin with.

That's what I am!

[Tabs]

mcinque,

I understand the point you were trying to make, its not fair to compete against cheap shit from asia sold through dummy retailers who dont care about CE or FCC etc.

In your particular examples 433MHz tx modules operate in an unregulated ISM (Industrial, Scientific, Medical) band (70cm i think). This band is used by radio hams and RC plane hobyists and doent require a licence in the US [Can an american ham confirm please?). The start and stop bands are different in each country but I think 433 is safe. The module is also limited to +14dBm which I think is below the limit.
They still have to comply with EMC in other bands.

Check where you are, but I think these bands are internationally agreed and are part of an ITU-R regulation. Again check and confirm for yourself.
Also, transmitters fall under the R&TTE directive which means they have to be licensed (if operating at a regulated frequency above the limit)
 
Just doing a very quick google search showed:

http://www.arrl.org/news/fcc-proposes-to-fine-texas-radio-amateur-7000-for-malicious-interference
http://www.ce-mark.com/euen.html
http://www.compliance-club.com/archive/old_archive/980816.html
http://www.southgatearc.org/news/may2010/ofcom_prosecutions.htm

It seems big corporations and the little guys are being prosecuted.

Just a couple of months ago I was checking emissions problems with a new product. Spent ages thinking it was the PSU, I got to end of my tether and proved it wasn't the PSU. In the end I got desperate and removed the PSU board and powered the unit off Duracell batteries. The product stopped failing which kind of says it was the PSU. I put the PSU back in, cut the power tracks on the secondary rectifiers and powered from the battery. The unit failed. Turns out it was SPI on my shitty little microcontroller. The slew rate was high (this is where the emissions come from, not the slow datarate).

If you've got anything that remotely looks like a square wave then I'd test. Especially if you have SPI talking to a display on the front panel over those cables. The frequencies may not be high enough for the cable to act as an antenna but it depends on slew rate. Thats the thing with EMC, you can plan for it, but ....

[mcinque]

I understand the point you were trying to make, its not fair to compete against cheap shit from asia sold through dummy retailers who dont care about CE or FCC etc.

Exactly. And please mind that I know Italian retailers that aren't dummy, but well regulated ltd. companies, with their responsibility. The CE directive is well clear: the manufacturer and the retailer ARE responsible for what they put on the market. So I ask: if they sell crap things, shouldn't they be prosecuted by law? And since the answer is YES, why they continue to operate without any fine?

Without a solid, unbroken ground plane, you're just asking for trouble.

just for test, I've tried to route the PCB with 4 layers, 2 inners with the signals and 2 outers grounded like the literature suggest to keep all the emissions inside.
Forgetting that I should modify some trace witdh, the result is that I need another 2 layers since the pre-routing fails (many ratlines impossible to route). Apparently there are too much components and nets on the board for only 2 inner layers. So I should use 6 layers... Of course in this case I can't manage the costs, at least for a niche product like this.

Back to the main thread, I quickly summarize my goal: pre-test my device for radiated EMI

If I was a firearm manufacturer, and had to deal with power limit of 2000J for my new handgun, I would obviously need instruments to measure wich energy my new product has.

Without having a chronograph to measure the bullet speed and a scales to know its weight it would be hard to calculate its energy in numbers.

But I need to know ONLY if I'm UNDER the 2000J limit, I don't need to know the value.

I can measure the power of a "compliant" firearm and compare it with mine. I can shoot to layers of cardboard until I find how much layers stops the bullet of the "compliant gun". At that point, if my gun, using the same projectile, DOES NOT penetrate more than 3/4 of layers than the number the compliant gun does, it's mathematical to have a compliant gun.

I cannot declare how much energy my gun has, but I can state safely that its under 2000J.

Applying this logic in my case I've thought this method:

1. use a cheap, uncalibrated spectrum analyzer (RF Explorer 27-2700MHz), near field probes and diy log periodic antenna (with preamp) in a "RF quiet" basement
2. compensate the uncertain calibration of the instruments by collecting radiated emissions from similar EMC-certified devices, setting the lowest value as emission limit
3. use this values as a threshold for radiated emissions
4. test my device with the setup and see if something goes over the threshold.

So now I have:

1. a testing setup
2. a threshold value to be observed - I understand that I have NOT a calibrated hardware but his spectral response MUST have the same linearity with my product both with CE compliant product, so knowing the "limit value" to observe, it should be idiot-proof to know if you're exceedings the emissions.

And at this point in my mind the REAL point is the following:

if I find that my product emissions are well below (I said WELL BELOW!) the threshold, even with an uncalibrated cheap setup like this I cannot be wrong: the device is EMI compliant.

At that point, how much sense does it have to pay € 900 for a certificate that states it's compliant when every product, by law, can benefit of the "presumption of conformity" (where the unconformity must be proven by the authority)?!?! And even if the autority tests my device, how can they prove that is not EMI compliant when it's measured that it radiates less than CE approved products?
I'm well documented in this and even a EMC center consultant that I've asked to told me that there is no law that prescribe mandatorily an EXTERNAL EMC test with a report. It's a choice every manufacturer does. You can do it internally and if you're sure, you can declare it as compliant (and take all the responsibilities, the same is if you pay for the EMC test, it's ALWAYS your responsibility). I know "nothing ventured, nothing gained".

I know and I'm sure that is something that I'm missing, otherwise we wouldn't have so many EMC test centers and all manufacturers could save much money doing thinks like I say So could someone please explain where my lack of knowledge makes me feel so safe?

EDIT: with his great uCurrent sold all over the world (so even in EU) Dave has gone thru all this annoying EMC testing?

[mikeselectricstuff]

Applying this logic in my case I've thought this method:

1. use a cheap, uncalibrated spectrum analyzer (RF Explorer 27-2700MHz), near field probes and diy log periodic antenna (with preamp) in a "RF quiet" basement
2. compensate the uncertain calibration of the instruments by collecting radiated emissions from similar EMC-certified devices, setting the lowest value as emission limit
3. use this values as a threshold for radiated emissions

This is not a useful approach, as unlike the firearm example where you are measuring a single one-dimennsional value, each of the other devices will have their own spectrum of emissions, and at many frequencies, the level will be below the noise floor, so this will correspond to an unnecessary "strict" limit. You also don't know if they actually meet the standards any markings claim. 
Testing outside of a shielded anechoic chamber is pretty much impossible as there is so much (continuously changing) ambient signal. 
Your basement won't be much quieter unless it is very very deep.
Even in a quiet environment, a non-anechoic area will have all sorts of reflections which means that very small changes in position of EUT and antenna can make huge differences in readings.

[dannyf]

A more quantitative approach is to use a RF probe - they are fairly simple to build.

The LED approach I outlined earlier will work as well and is more visual.

Since you don't know the absolute value, you will need to use a known compliant device or devices for "calibration".

[AndyC_772]

It's a common misconception that putting a noisy track on an internal layer, sandwiched between planes, is needed to suppress emissions from it. Putting it on an outer surface with a solid plane on the layer below is also effective, and although you'd need a physical prototype or 3D field solver software to actually prove it, there are several 'mental models' you can use to help understand why this might be the case.

You have:

- return current immediately underneath the noisy trace, therefore a very small loop area and, therefore, an inefficient radiating antenna

- very high quality capacitance between the signal and ground, formed by the PCB itself, with minimal series inductance which leads to good high frequency (100+ MHz) performance

- electric field lines terminating on the plane, rather than radiating off into space. Putting a solid lump of copper right next to the antenna is exactly what the data sheets for things like Bluetooth modules tell you NOT to do, for just this reason.

Turning a 2 layer board into a 4 layer is, therefore, pretty trivial. Keep the top and bottom layers just as they are already, but put ground on layer 2 and power on layer 3.

It sounds as though you might be using an auto-router, right? Don't.

I understand your frustration when it comes to comparing your equipment against 'known compliant' samples, but a couple of reasons why it's a poor way to demonstrate compliance are:

- the near field spectrum of a piece of equipment is not the same as the far field spectrum. The range of frequencies that are detectable will be the same, but the amplitudes vary dramatically.

- just because a piece of kit is claimed to be compliant, doesn't mean it will be. The emissions spectrum from a product depends on its physical orientation, cables attached, reflections from its surroundings, and variation from sample to sample.

Arguments based on total radiated energy don't stand up, because the test limit is a limit on emissions in any particular direction, not just an overall total. You can't assume that your product radiates equally in all directions, and if it radiates predominantly in one direction, it can still exceed the limit even if the total radiated power is low.

It's entirely true that you don't have to have your product formally tested by law. It is, however, the case that the only way to actually be sure your product complies with the relevant standards is to do so.

If the authorities test your product - which is, incidentally, quite unlikely - then they will do so according to the prescribed methods in the relevant standards. If your equipment passes, then that's fine, but unless you've carried out that same test yourself beforehand, you have no way to know that it will do so.

[mcinque]

This is not a useful approach, as unlike the firearm example where you are measuring a single one-dimennsional value, each of the other devices will have their own spectrum of emissions

hmm... this is damn true. I didn't consider it 

You also don't know if they actually meet the standards any markings claim.

I've learned to not trust a compliance of a product only by looking at its markings. It's better to find its declaration of conformity to have a chance to know if it's compliant.

Testing outside of a shielded anechoic chamber is pretty much impossible as there is so much (continuously changing) ambient signal. 
Your basement won't be much quieter unless it is very very deep.

mmm.... understood. What about building a "mini" RF anechoic chamber with a shielded metal box and rf adsorber inside? It could be helpful or I haven't any chance?

Even in a quiet environment, a non-anechoic area will have all sorts of reflections which means that very small changes in position of EUT and antenna can make huge differences in readings.

Damn, reflections are really important and it's a behaviour that I've never take in consideration.

It's a common misconception that putting a noisy track on an internal layer, sandwiched between planes, is needed to suppress emissions from it.

Well, I've just tried what I've found on literature; for sure the practice and experience are definitely better than theory

Putting it on an outer surface with a solid plane on the layer below is also effective, and although you'd need a physical prototype or 3D field solver software to actually prove it, there are several 'mental models' you can use to help understand why this might be the case.

Thanks!

It sounds as though you might be using an auto-router, right? Don't.

Oh, often I use autorouting only to see if the all the nets are routable: if the software miss a couple of nets it's easy to understand that you can do better manually, but when he miss something like 15 nets I've learned that I can do very few things to complete the board.

- the near field spectrum of a piece of equipment is not the same as the far field spectrum. The range of frequencies that are detectable will be the same, but the amplitudes vary dramatically.

A very good point to explain why my idea is not good.

A more quantitative approach is to use a RF probe
Since you don't know the absolute value, you will need to use a known compliant device or devices for "calibration".

That's exactly what it was on my mind.


So, to end this thread, even with my cheap hardware and a DIY calibration based on compliant devices there is no doubt that I can't do to do a pre-compliance without an RF anechoic chamber.

[mikeselectricstuff]

So, to end this thread, even with my cheap hardware and a DIY calibration based on compliant devices there is no doubt that I can't do to do a pre-compliance without an RF anechoic chamber.[/size]

Correct, but what you can do is identify where any peaks are, and see what can be done to reduce them, you just can't get any absolute figures.

[nctnico]

However, since this will be an hobby/enthusiast product for a very small niche, I often ask myself if I really need to do all this. I'm putting a big effort in this and sometimes I think that I wasting my time (and money).

I strongly believe that most niche-products like arduino relay or display shields, crap digital multimeters, rubbish power supplies modules, disgusting tx-rx modules, all sold by various European online shops have not even saw from a mile a piece of paper that talks about EMI and EMC. And no authority is preventing or acting against this products.

They don't have to. Modules are parts and therefore not required to have a compliance certificate. Even modules with approval may fail to comply when built into a product.

[nctnico]

It's a common misconception that putting a noisy track on an internal layer, sandwiched between planes, is needed to suppress emissions from it. Putting it on an outer surface with a solid plane on the layer below is also effective, and although you'd need a physical prototype or 3D field solver software to actually prove it, there are several 'mental models' you can use to help understand why this might be the case.

I agree. A while ago one of my designs a 6 layer board with a 1GHz SoC and lots of switchers went through EMC testing. There was no difference between the readings with the board inside the metal casing or not. All the radiation came out through the wires.

Also something interesting happened during the test: there was a 90MHz spike which caused some confusing since there is no clock on the board which could produce that. A quick look around revealed the lab technician forgot to put a filter clamp on the cable going into the EMC shielded room. The 90MHz was a radio station.

[mcinque]

They don't have to. Modules are parts and therefore not required to have a compliance certificate. Even modules with approval may fail to comply when built into a product.

The parts that don't need to be tested must be sold only to tech personnel, not to hobbyst.

[mikeselectricstuff]

They don't have to. Modules are parts and therefore not required to have a compliance certificate. Even modules with approval may fail to comply when built into a product.

The parts that don't need to be tested must be sold only to tech personnel, not to hobbyst.

Citation?

[Niklas]

Exactly. And please mind that I know Italian retailers that aren't dummy, but well regulated ltd. companies, with their responsibility. The CE directive is well clear: the manufacturer and the retailer ARE responsible for what they put on the market. So I ask: if they sell crap things, shouldn't they be prosecuted by law? And since the answer is YES, why they continue to operate without any fine?

The company that imports goods into the EU is the one that is responsible to make sure that the product complies with applicable standards. The main problem is that many trading companies are just only trading companies and therefor lacks the technical knowledge. To hire a consultant to do the verification cost money and may consume all the profit margin of the product. A Chinese manufacturer can put whatever marking(s) they like just to boost their sales. They don't have to face possible fines, recalls etc and got their money in the bank before the shipment has left the country.

I have seen both electronics and water boiler for a steam sauna produced by an Italian company that was everything else than CE compliant. A customer wanted to integrate our control panel (the user interface) with parts from another producer (hidden for the user). During our review of the electronics we found several safety related issues, where a weak PE connection was the worst. Also the wires that went to the heating element were standard PVC insulated ones not rated for the high temperature. The result was that we adapted the SW in our power electronics box to suit the application and we also replaced the wire harness attached to the heating element.

As long as the producer and/or importer can just slap a CE mark onto the product, without a significant risk of getting caught, the problems will continue. The local regulatory agency in Sweden only starts to look at a product after customers report to them. How many customers know about this agency and the possibility to file a report? Not so many I guess. The life cycle for many imported products is often very short and the regulatory agencies are usually quite slow.

One more thing about Italian products even if it is a bit off topic. If I ever was in the situation to purchase trains, I would never even consider something that was designed or produced by Ansaldo Breda. Their sales team might be good, but their finance or engineering teams are quite disappointing...

[sandy]

Exactly. And please mind that I know Italian retailers that aren't dummy, but well regulated ltd. companies, with their responsibility. The CE directive is well clear: the manufacturer and the retailer ARE responsible for what they put on the market. So I ask: if they sell crap things, shouldn't they be prosecuted by law? And since the answer is YES, why they continue to operate without any fine?

But who is comming forward to make a complaint against them and pursue a case, are you  ready mcinque ? . I think the answer is No, because once you get a compliance certificate your anger will be over. So who will give a sacrifice and fight international cases and in return become bankrupt? . They will bribe authorities and start a fresh business? isnt it .

[dannyf]

And since the answer is YES, why they continue to operate without any fine?

Because we have competing priorities and limited resources to pursue them all.

There are unlimited number of things that we should do but we don't do, because of that.

[Precipice]

mmm.... understood. What about building a "mini" RF anechoic chamber with a shielded metal box and rf adsorber inside? It could be helpful or I haven't any chance?

This is harder than you might think, I'm afraid.
The amount of absorbing material you need, and the distance between your unit and the antenna, plus some way to move the antenna (or unit), means that you end up with a big chamber that looks a lot like a real EMC chamber.
You can investigate TEM cells, but they're more for susceptibility than emissions.

(I've got a 1.2x2.4x2.4m shielded chamber that I use when I want to investigate stuff at home. (It's built out of sections of an ebay chamber - these things are often modular, so you can make them whatever size you like). It's big enough to get me, a test unit, an antenna and a spectrum analyser in, reasonably comfortably. It's nice because there's no external noise - but reflections mean that if anything moves, amplitudes change by tens of dB. It'll be interesting to see how much improvement packing all the spare space with absorbent cones will make - at the moment, it's more useful to contain emissions than it is as a precompliance chamber.)

Also, while I feel your frustration - I think there's another level yet to come. It's when you just fail compliance testing. One peak half a dB over, or a tiny disturbance during susceptibility, that you might not have noticed, but you did...
 
I'm now building a much bigger proper anechoic chamber, so I can get some proper results. Again, second hand off ebay. Had to rent an industrial unit to build it in

[mcinque]

Citation?

Guide for the EMC Directive 2004/108/EC, see attachments.

If a component/part is for the end user (like the hobbist for example), it must comply with the directive. The end user should not care about the emissions or any else concerning the emissions or any else related to the directive.

If a component is sold to tech personnel that will certify the EMC of the final installation it can be excluded from the directive. It's who installs the hardware that must produce a certificate of conformity and compliance. The guide tells clearly that to be excluded the component/part must not be available to the end user.

Take for example a PC video card. It's only a "module". It's a part that cannot absolve a specific function without be installed into a PC. But its available at the end user, and must comply the EMC directive.

I'm confident that I cannot produce and sell to the end user a video card that once installed interferes with tvs, radios and mobile phones and then justify me saying "it's only a module, it's not my problem".

[AndyC_772]

It's a whole new can of worms if you're selling something like a video card.

Even the simplest card, with no active components at all, can pick up noise which exists inside the chassis of a compliant product and radiate it to the outside - thereby rendering the system non-compliant.

Of course, your add-in card isn't allowed to make a compliant system non-compliant, but the difficulty is that neither item is necessarily in any way deficient. Unless your add-in card has superlative shielding, there's always a risk that someone will plug it into a particularly marginal (but compliant) host system - but you get the blame for making it non-compliant with your card.

In this case, a good approach is to be able to present a formal test report for some other system, with your card installed in it, so you can demonstrate that your card isn't responsible for the noise.