EMC compliance question: how can I fall in 30MHz directive with a 16MHz MCU?

[mcinque]

Together with a friend of mine, I've designed an elementary device based on an ATMega328.

Since we want to sell it commercially, we need to produce a "CE" declaration of conformity.

As you know, the declaration is only the final paper after a series of checks about the compliance of one or more directives.

After asking to more than one consultant, we come to the conclusion that the the only mandatory requirements to sell the product are:

EMC compliant for radiated emissions(EU)
Rohs compliant (EU)
RAE compliant (LOCAL, about electronics waste disposal)

LVD (Lov Voltage Devices) doesn't apply, since it runs *only* from 14.8V battery and LVD take in consideration only devices that runs *over* 75VDC or 50VAC.

Rohs is easily achieved, we've all the Rohs certificates from our provider (RS Components, Farnell etc).

RAE requires "only" that we must manage free of charge the end-of-life returns and dispose them properly.

We've some doubts about the EMC. In EU, you can:

declare the EMC compliance since you've followed all the technical prescriptions that the normative says (so you get a presumption of conformity, until an authority proves the opposite) or
assuming that you don't have followed the prescriptions, do a full EMC test.

Of course, since the device is floating, we don't need:
conducted emissions tests
harmonic current emissions tests
voltage fluctuations & flicker tests
electrical fast trasient / burst immunity test
surge immunity test
immunity to conducted disturbance tests
voltage dips, short interruptions and voltage variations immunity tests

We must compy to the radiated emissions.

Even if we've designed the circuit taking in considerations what the normative says, we don't have the instruments to test them. We've done a "poor man test" with a DIY field probe connected to a DSO, comparing the readings with a couple of other similar devices that we know are EMC compliant: the readings are almost zero as we expect.

A full EMC test costs much money. But the fact is that the EMC directive takes in consideration a range of frequencies that starts from 30MHz.

We've some serious doubts about how a circuit where the most high frequency is 16MHz (clock) could fall in that range, even if it's poor designed.

My feeling is that we have to prove that a man over a bicycle can't reach 120 miles per hour on the highway.

How do you think about it? Can we safely assume that the device will be EMC compliant for radiated emissions without doing expensive EMC tests?

[poorchava]

You can expect harmonics from mcu clock and from switch-mode converter.

[PA0PBZ]

A picture's worth a thousand words:

[nuno]

Also, 38mA is just an average. Try measuring during a busy clock transition.

[lewis]

The radiated emissions from the MCU itself (assuming it's an SMD package) will be extremely low. It's possible to pick up emissions with a sensitive RF sniffer probe at a couple of cm from the MCU die, but emissions beyond that tail off into noise very quickly, and will very likely meet product standards. 600-mil DIP packaged parts tend to have slightly higher emissions because of the length of the bond wire between the pin and the die limits how close you can get a bypass capacitor to the die, but the emissions are still minimal.

But under some circumstances, you can make a little FM transmitter: http://spritesmods.com/?art=avrfmtx&page=2

If the MCU is not properly bypassed, total emissions can be higher because the switching noise bleeds out of the chip onto the supply rails, which will act as antennae and radiate. 1uF // 100nF (X7R dielectric) is a good idea right at the chip. As is a good ground plane under the package.

The most important concern with radiated emissions, by several orders of magnitude, is high frequencies (or more accurately fast rise times) outside of the MCU: external clocks or serial buses, for example. If the MCU's port pins have a slew rate control function, set them to as slow as possible. It makes an enormous difference reducing transition times from ~15ns to ~150ns. If you have any fast clocks external to the MCU, you might need to get the product tested. Otherwise, you'll very probably be OK.

You might need to worry about EMC susceptibility too. A quick and dirty approach is to operate your product right next to a mobile phone (during a call), then next to an operational wi-fi antenna, while confirming your product still works properly.

You'll probably also need to comply with the battery directive 2006/66/EC.

[AndyC_772]

How do you think about it? Can we safely assume that the device will be EMC compliant for radiated emissions without doing expensive EMC tests?

That depends on your definition of 'safely'.

Without actually seeing your product, it's impossible for me (or anyone else) to say with any confidence whether or not it's likely to be compliant.

You might decide that, since you've designed a simple product with a limited number of digital signals external to the MCU, and that the edge rate on each one of them is well controlled by design, then it's likely the product will be emissions compliant. However, if you take this approach, you must understand that you are taking a risk. That risk is that someone - a customer, reseller, competitor or regulatory authority - will test your product and find that it's not compliant after all.

You, and only you, can assess how bad that risk is. If you're making a handful of boards to sell in a niche market, the chances of it being picked up and tested by a local regulator are much lower than if you were making televisions or photocopiers by the million. It's up to you to decide whether you want a product which is definitely compliant, or a product which definitely bears a CE mark. The two are not the same.

If the worst happens, and someone in authority asserts that your product is non-compliant, then you could find yourself in trouble. One of the best reasons to have a formal test carried out by an EMC lab is that it gives you a good defence should you find yourself in court. Provided you can show that you had a solid basis on which to claim compliance, you're unlikely to end up receiving a criminal punishment even if a particular product (for whatever reason) turns out to fail. You'll still have to bear the costs of any recall or redesign, of course.

As well as radiated emissions, there's also immunity to consider, including ESD. I'm not aware of any basis on which you can reasonably expect to claim your product 'must' pass immunity 'by design' without testing it. That said, the test criteria are pretty subjective - basically your kit is 'immune enough' if you say it is, but there are limits. If, for example, a nearby mobile phone ringing causes your product to reboot, then that's a no-go. If it makes a bit of a buzzing noise on an audio line, though, then that's probably OK.

Nobody reputable is ever going to say 'no, you don't need to test', and if anyone ever does, then you should ask them whether they'd be prepared to take personal responsibility for your product's compliance.

I've personally spent enough time in EMC labs to know that any product can yield a surprise. I've seen hugely complex boards pass without a hitch, and apparently much simpler ones fail horribly. It can be down to the enclosure design, the cables attached (which are the source of 95% of harmful emissions in my experience!), or even just an unfortunate choice of component on the board which happens to generate way more noise than perhaps it should.

I'm sure if you wait a while longer, someone will come along and tell you what you want to hear, which is that you don't really need to test at all provided you've designed your product according to good engineering principles. What that means is that they'd be prepared to take the risk themselves, and deal with the consequences should they arise.

Only you can judge the risk that you are personally willing to take.

[dannyf]

We've some serious doubts about how a circuit where the most high frequency is 16MHz (clock) could fall in that range, even if it's poor designed.

Harmonics on the clock signal can easily exceed that.

[T3sl4co1l]

Expect harmonics up to 100MHz from the switching regulator, and maybe to 200MHz from the digital logic.

Tim

[yramgu]

How do you think about it? Can we safely assume that the device will be EMC compliant for radiated emissions without doing expensive EMC tests?

Never assume anything when it comes to EMC/RF

I've personally spent enough time in EMC labs to know that any product can yield a surprise.

Couldn't agree more. Another example to illustrate: my company designs RF modules, and at the end of a project the engineers tested the new module against european standards and it passed without problems. Then they decided to use it for another project by embedding it on a motherboard, but it didn't pass harmonics limitations anymore. They said "we'll modify the module by adding a very sharp SAW filter at the output of the transceiver" to eliminate the harmonics. But it didn't work. One of my first tasks when I got hired was to look into it. Turns out energy was leaking between the transceiver and the LP filter (we're talking a few mm of separation between the two components) to the ground plane of the motherboard through the ground connections of components, and the ground plane was actually radiating it.
By the time you realise that you can loose a significant amount of time/money. NEVER assume...

The only way to be sure, if you really want to respect regulations, is to do the tests

[mcinque]

The radiated emissions from the MCU itself (assuming it's an SMD package) will be extremely low. It's possible to pick up emissions with a sensitive RF sniffer probe at a couple of cm from the MCU die, but emissions beyond that tail off into noise very quickly, and will very likely meet product standards. 600-mil DIP packaged parts tend to have slightly higher emissions because of the length of the bond wire between the pin and the die limits how close you can get a bypass capacitor to the die, but the emissions are still minimal.

Its a DIL28 package and it's properly decoupled.

But under some circumstances, you can make a little FM transmitter: http://spritesmods.com/?art=avrfmtx&page=2

Fantastic 

If the MCU is not properly bypassed, total emissions can be higher because the switching noise bleeds out of the chip onto the supply rails, which will act as antennae and radiate. 1uF // 100nF (X7R dielectric) is a good idea right at the chip. As is a good ground plane under the package.

It's what I've done.

If the MCU's port pins have a slew rate control function, set them to as slow as possible. It makes an enormous difference reducing transition times from ~15ns to ~150ns.

mmm... thank you for the advice, this could be the case for the PWM modulation for the buzzer.

If you have any fast clocks external to the MCU, you might need to get the product tested.

It's not my case.

A quick and dirty approach is to operate your product right next to a mobile phone (during a call), then next to an operational wi-fi antenna, while confirming your product still works properly.

Yes, I've done that kind of tests with no issues on the circuit while its running.

You'll probably also need to comply with the battery directive 2006/66/EC.

Thank you for the suggestion, this is achieved providing a compliant battery pack.

if you take this approach, you must understand that you are taking a risk. That risk is that someone - a customer, reseller, competitor or regulatory authority - will test your product and find that it's not compliant after all.

If the worst happens, and someone in authority asserts that your product is non-compliant, then you could find yourself in trouble.

Is what I want to avoid.

As well as radiated emissions, there's also immunity to consider, including ESD.

mmm... right.

If, for example, a nearby mobile phone ringing causes your product to reboot, then that's a no-go.

That's exactly the same kind of "poor man immunity tests" that I've done on it... mobile phones in call almost in contact with the board, a couple of 433MHz and 446MHz transceivers with the antenna close to the pcb and in various positions, an old switching transformer nearby, a neon tube... I've tried many ways to make it reboot or crash, but luckily all went fine.

I've personally spent enough time in EMC labs to know that any product can yield a surprise. I've seen hugely complex boards pass without a hitch, and apparently much simpler ones fail horribly.

I understand this. There is so much variables.

Only you can judge the risk that you are personally willing to take.

You're damn right.

We've some serious doubts about how a circuit where the most high frequency is 16MHz (clock) could fall in that range, even if it's poor designed.

Harmonics on the clock signal can easily exceed that.

Indeed, but they should be weaker than the 16MHz tone, don't you?

Expect harmonics up to 100MHz from the switching regulator, and maybe to 200MHz from the digital logic.

The switching regulator is a single component and is EMC certified, so that one is maybe the only certain EMC compliant part of the circutry.

Never assume anything when it comes to EMC/RF
Another example to illustrate[...]By the time you realise that you can loose a significant amount of time/money. NEVER assume...
The only way to be sure, if you really want to respect regulations, is to do the tests

Understood.

[mcinque]

So probably the best approach is to do some kind of pre-compliance tests to try to be into the directive at the first try, just to avoid a 2nd test.

For the immunity test, I've used (and you suggested) mobile phones, wi-fi antennas, ham radios... do you have some other suggestions?
For the ESD probably a piezo ignitor (those used in lighters) should do the job... what do you think about it?

And what for the radiated emissions? At the moment I've only a 50MHz DSO and a home made field probe (almost useless setup I guess), but in the next days I should have an RF Explorer 15-2700 MHz to play with. What kind of setup do you suggest to do in a home lab? What kind of antenna should I use?

[ejeffrey]

We've some serious doubts about how a circuit where the most high frequency is 16MHz (clock) could fall in that range, even if it's poor designed.

Harmonics on the clock signal can easily exceed that.

Indeed, but they should be weaker than the 16MHz tone, don't you?

Not necessarily.  The original square wave has less power at the harmonics than the fundamental but you can have LC resonances that make your filtering ineffective at certain frequencies.  Parasitic capacitances will more easily conduct high frequency noise out of your will filtered/bypassed MCU and into an area that isn't designed for high frequency such as the power leads.  Finally, high frequency radiates more easily than low frequency.  With a square wave, harmonics only fall off as 1/n so you can easily have the 21st harmonic play a significant role if it happens to hit some resonance that isn't well damped.  Triangle waves have harmonics that drop off with 1/n^2 which is why limiting slew rates is so effective, they make the waveform look more triangular.

[Neganur]

You could also ask a local university if they'd stick your device into their GTEM cell / anechoic chamber to give you an idea of what your emissions are like. You'll probably be surprised

Absolutely do not just assume that EMC will not be necessary there is always something you forget to take into account.

Don't underestimate conducted emissions either, that stuff can spread all over the place with shitty cheap prefabricated PSU's. (you mention that you don't need it, so disregard this sentence)

[yramgu]

You could also ask a local university if they'd stick your device into their GTEM cell / anechoic chamber to give you an idea of what your emissions are like.

Exactly. You need an anechoic chamber to make any relevant measurements. Don't expect to be able to do this at home.

The switching regulator is a single component and is EMC certified, so that one is maybe the only certain EMC compliant part of the circutry.

Wrong. Pointing to my first post, just because your component alone is EMC certified doesn't mean it will be when integrated in a larger product. Also, don't trust silicon manufacturers. Test everything you can

[Neilm]

We've some serious doubts about how a circuit where the most high frequency is 16MHz (clock) could fall in that range, even if it's poor designed.

The switching regulator is a single component and is EMC certified, so that one is maybe the only certain EMC compliant part of the circutry.

I have seen things that were certified fail when put into a system - especially switching regulators, due to poor layout, light loading or using it in a way the supplier did not expect. Usually, PSUs are only tested at a specific loading (80%). If they operate at very low powers, they can fail EMC testing.

[mcinque]

The original square wave has less power at the harmonics than the fundamental but you can have LC resonances that make your filtering ineffective at certain frequencies.

This is so damn true. Sometimes I wonder how I can ignore things like that. Seriously, I didn't take in consideration resonances.

You could also ask a local university if they'd stick your device into their GTEM cell / anechoic chamber to give you an idea of what your emissions are like. You'll probably be surprised

I will try to ask but I fear that due to the typical italian mentality the reply will be "what can we gain if we allow you to do that?" 

Exactly. You need an anechoic chamber to make any relevant measurements. Don't expect to be able to do this at home.

This is frustrating 

just because your component alone is EMC certified doesn't mean it will be when integrated in a larger product. Also, don't trust silicon manufacturers. Test everything you can

Thank you for the advice.  I would test everything if I only have the right instruments and setup.  EDIT: and knowledge!

[qno]

One simple trick I use to se if you have a radiating circuit is  to use a high frequency current clamp.

You can build a current clamp  from a ferrite toroid.
To make life easy there are types that have 2 halves so you can position the clamp over a wire.
 
Wind one pick-off winding on it to connect it to your scope.
Then lead both power supply lines true the center and measure with your scope the signal of the pick-up coil.

In theory the current flowing in the + wire of the power supply should flow in the - wire.
However if any energy is escaping as radiation or is coupled in a nearby conductor these currents will not be the same.
Then you will measure something on your scope.

If these current exceeds the magic number of 5 uA you will  not pass the radiation test.
Because you build the current clamp yourself you will not know how many mV on your scope is 5 uA.

You can get a ballpark figure by analyzing some equipment known to pass CE radiation requirements.

If you designed the PCB according to EMC guidelines you will have you signal lines close to or above a ground or return path and have no trouble with the radiation test.
 

[mikeselectricstuff]

EMC certified[/b], so that one is maybe the only certain EMC compliant part of the circutry.

Not necessarily. Power supply type components are usually tested under best-case conditions, usually with a simple resistive load. Their EMC characteristics can vary hugely under different loads, and are also likely to have some dependence on mounting, decoupling,  grounding etc.
And how much margin is built into their emissions specs ? If you're unlucky, it could be that a low level emission from your circuit combines with that from the regulator to push things over the edge.

 

[AndyC_772]

What Mike says is absolutely true and is worth re-iterating. A power supply can get a CE mark if it passes on full load when connected to a resistor. Legally, that's all that the supply manufacturer needs to do in order to sell their product.

However: to actually be usable in a real product, a power supply must *also* be compliant when powering the specific load to which it's connected, and when that load is generating some noise of its own. Many times I've seen a "compliant" PSU fail because it's noisier at half load than full load, or because the thing it's powering isn't completely silent. I've had lengthy arguments with PSU manufacturers over this; just because their product is compliant on its own doesn't mean any product that uses it will be.

Usually the test that fails is conducted emissions - and if it fails that test, then there's not generally a lot you can do other than use another power supply.

[nctnico]

To answer the OP's question: no. EMC emissions are measured to at least 6 times the highest frequency in a circuit. 16*6=96MHz.

When I have to deal with EMC compliance testing it usually goes like this:
1) Do a scan for emission
2) Fix the emission problems
3) Do a test for real
4) Fix the other problems
5) Retest the bits that failed earlier.

It is not only emission which needs to be tested but also susceptibility and ESD. The latter can be challenging as well. The reason there is so much focus on emission is that it be very costly to redo an entire PCB design. OTOH most emissions and susceptibility problems on a multilayer board come from where cables enter the board. Adding series resistors + capacitors, ferrite beads or common chokes helps a lot.

[mcinque]

One simple trick I use to se if you have a radiating circuit is  to use a high frequency current clamp.

You can build a current clamp  from a ferrite toroid.

Because you build the current clamp yourself you will not know how many mV on your scope is 5 uA.

You can get a ballpark figure by analyzing some equipment known to pass CE radiation requirements
 

 

[R_G_B_]

If you fail to get a CE certificate. You can always sell your product in Bargains.  Seems the garden Gnomes are happy that their not  CE certified. They wolf wistle at anyone who walks past holding a mobile phone.


A lot of these so called CE products are not  CE end up in shops like tesco wilkos etc... Another example a glue  gun I got.

I plugged it into the wall and the switch next to the handle start arching and burnt its self out.