The U.S. has recently changed the rules regarding EMC compliance requirements. Previously, products required accredited test lab measurements in order to claim compliance. Under these rules, the delegation of EMC testing to test labs was understandably standard procedure. Now, however, accredited test lab measurements are not mandatory, enabling the practice of self-testing, self-certification by manufacturers.
This has been the situation in the European Union (EU) since 1996.
The experience gained since 1996 in the EU is now relevant to the situation here in the U.S. Self-testing brings many benefits, not all obvious, including:
- Reduced costs paid to the test labs;
- Testing in-house right from early development prototypes through to final product ensures quicker time-to-market;
- Avoidance of expensive re-design phases; and
- Increased in-house expertise, not just in test techniques but in appreciation of how EMC behaves and can be tamed.
EMC measurements are thought of as either a “black art” that requires years of specialist experience to address or something that can be achieved with very minimal equipment (e.g., near field probes, etc.). Both are far from the truth. Unfortunately, such myths are perpetrated by practitioners and equipment suppliers alike. We had exactly the same happening in the EU. It has taken many years before reality has been established. It is possible that the same “learning curve” will be experienced in the U.S.
Some common myths include:
- Compliance measurements can be judged/interpreted using near field probes.
Completely untrue. - A standard spectrum analyzer will provide a good tool for measurement of emissions.
On the contrary, an analyzer can deliver very misleading results, often higher than they should be. - Screened rooms are required for measurement of conducted EMC emissions.
Absolutely not. - Screened rooms are required for radiated emissions measurements.
A screened room can be useful, but NOT for measuring emissions!
To understand the reasons for the above, an understanding of just two processes are required. These are very simple and basic, and you need nothing more than Ohms law. (I promise!)
Once understood, the EMC measurement techniques specified in the standards become self-evident.
For example:
- Why use an open area test site (OATS), instead of a screened room?
- Why does the antenna have to be at least 3m from the DUT?
- Why use height scanning?
- Why do we measure conducted emissions for frequencies below 30MHz, and why only radiated for 30MHz and above?
- What is the importance of wavelength?
- How is electrical energy actually transformed into radiated RF?
- And, once radiated, how does RF behave?
It is easy to become swamped in the math (Maxwell’s equations, etc.) but unless you want to become a specialist in RF, propagation, antenna design etc., there really is no need. Most of us simply want to be able to “see” or visualise and understand the basic principles. These basic principles are surprisingly simple and entirely adequate to explain all you need to know.
On many occasions, I have outlined these simple principles to “EMC experts” at larger companies and consultancies, and have been shocked by their “Aah, now I understand it” response. Clearly, they had simply been following procedures without any real understanding. This means that when confronted by a non-compliant device under test (DUT), they struggled to logically work out how to alleviate the problem.
In this article, I’ll focus on radiated rather than conducted emissions. Conducted emissions are more straightforward, but it is important to be aware of and appreciate the importance of dynamic range and detector characteristics since it is these factors that prevent the use of conventional spectrum analyzers. (I’ll more fully address conducted emissions in a future article.)
At the beginning… how is an electrical signal flowing within a conductor magically transformed into an RF signal (airborne emission)? And, once “airborne,” how does it behave?
Creation
Imagine a very simple circuit, a battery connected to the power pins of a CMOS digital integrated circuit (IC). The return wire takes a different path, thereby forming a loop. In its dormant state, virtually no current flows. An input from a clock causes the IC to switch at (say) 16MHz. Every time the IC switches, a tiny pulse of current is drawn from the battery.
This current pulse will create a magnetic field around the battery feed wires. We know that current causes magnetic fields around the wire because this is how transformers and electric motors work. The field will be created and then collapsed once the current pulse has passed. (at a 16MHz rate).
Some of the energy involved in creating this field will radiate outwards as magnetic energy. We can think of this as a flow of energy akin to a current flow through a conductor which we call “free space.” If free space has an impedance, then Ohms Law dictates that there will be a volt drop.
Indeed, free space has an impedance of 377ohms, so at some distance from the source, we will have a “voltage” (electric field) component related to the “current” (magnetic field) component. This is our electro-magnetic wave (see Figure 1).
Close to the source, the wave is entirely magnetic (current) with little voltage component. Again, using Ohms Law, this implies a low impedance wave. If the source was an open circuit stub with a high voltage signal applied to it, an electric field would be created. Its impedance would be high (lots of volts but no current). Again, this field will radiate away into free space and (thanks again to Ohms Law!) a corresponding magnetic (current) flow will be generated, related by the 377ohm term.
3 Meters?
The transition towards the 377ohm wave impedance is gradual, and there is no step change. For practical purposes, this transition is assumed to be at either 1/3rd or 1/6th of a wavelength (depends which textbook you read).
Why is this important?
The “transducer” that we use to measure RF fields is an antenna. Antennas are electric field sensors. If we use an antenna close to the source, which will probably be a magnetic source (most sources are), then the antenna will not measure the field. We need the antenna to be in the far field where the magnetic and electric fields have achieved balance. At the longest wavelength of interest (i.e., that of the lowest radiated emission frequency we need to measure, which is 30MHz), the wavelength is 10m. One third of this is (around) 3m. Hence the standards specify a minimum distance of 3m when measuring emissions.
This also explains why you need two probes for near field work (magnetic and electric).
Some numbers to remember here:
Wavelength (m) = 300/frequency (MHz) in free space.
300MHz = 1m 1GHz = 30cm
100MHz = 3m 3GHz = 10cm
30MHz = 10m 10GHz = 3cm
Impedance of free space = 377ohm.
The Source?
The energy source of the emission will be a chip/clock/switching circuit or similar component. Although this is the source of energy, it is unlikely to be the source of any emissions. In order to radiate a signal, you need an antenna (as any radio ham or telecoms engineer will confirm). In our case, the antenna will be a wire, a cable or a trace on a printed circuit board. It is this conductor that is the real source of our emissions.
So we have a source of energy (the chip) connected to an antenna (length of conductor). These are typically separate items. The actual radiated emission as measured at 3m will be largely dictated by the antenna, not the energy source.
Near field probes are very good at detecting energy sources, but fail to take account of any antenna, which is why they cannot be used to judge compliance or estimate EMC fields.
To be an effective transmitting antenna, a conductor should ideally be ¼ wavelength long. Other lengths will still “transmit” but will do so with decreasing efficiency. This relationship between wavelength and radiating efficiency is fundamental to the understanding of EMC characteristics of any product. Sometimes, you can just look at a product and know what frequencies will be problematic.
I once tested a golf cart which had some electronics up near the handle and a motor at the base. The distance between the two was around 0.6m, and I amazed the customer by taking one look and predicting that they had a problem around 100 – 150MHz region. Sure enough, they did. (¼ wavelength @ 0,6m = 2.4m full wavelength = 125MHz.)
A golden rule for considering radiated emissions is to think in terms of wavelength…. NOT frequency!
If a product is small, and has no connecting cables, any issues with radiated emissions will be related to higher frequencies. If a product has connecting cables (e.g., mains power feed), use the wavelength criteria, apply to the cables and check the relevant frequencies.
Why Measure Only Conducted Emissions up to 30MHz and Radiated Emissions Above That?
A good question! What is so magical about 30MHz that causes this switch? I mentioned earlier that when a current pulse creates a magnetic field, some of that energy radiates off into space. BUT, some falls back into the conductor in such a way as to oppose further current flow. The shorter the time between the pulses, the more effective this “block” becomes. This is the self-inductance of a wire. If the wire is coiled, the magnetic field becomes more concentrated, and the result is higher inductance.
Reverting to just a plain wire, at DC, its impedance will be very low (mohm). As frequency increases, its impedance increases. So, at 1MHz, impedance may be in the region of some 10s of ohms. Energy flowing through this wire has a choice. Energy will always flow in the path of least impedance, so it will elect to stay in the wire. However, at 30MHz, a typical length of wire reaches an impedance around 377ohm, and above that frequency, the path into “free space” now offers the least resistance, so energy will radiate rather than stay in the wire.
All this is a very generalized “rule of thumb.” But it works!
The EM Wave
So we now have some emissions from our wonderful product and we place an antenna 3m away. Why 3m? Because that’s the minimum distance we are allowed by the standards. Why use the minimum distance? Because that will measure the highest level of DUT emissions and hence provide us the best S/N ratio over the background noise.
The standards will quote class B emission limits at 10m, but in the small print provide a calculation for adjustment of the limit levels for other distances. As a simple rule, changing the limit from 10m to 3m requires a 10dB increase in the limit level.
Officially, we should use an open area test site (OATS). But that creates two problems to address, as follows:
- Ambient noise
- Find a ‘quiet’ site for your OATS—Not so easy in this modern world. Broadcast and telecom services are everywhere!
- Use an ambient cancelation technique—Definitely an option which works, but only available from certain suppliers.
- Use a screened room—OK for locating frequencies of interest, but useless for taking measurements (see 2e below). Once you know what frequencies to look for, use an OATS and just select the frequencies of interest.
- Use a test cell that is completely screened from ambient noise—Good option for small products, but allows higher measurement uncertainties.
- Use an anechoic chamber—Best option, but expensive.
- Reflections
- EM waves reflect (just like light). They reflect from anything that is conductive. They will pass through any non-conductive and dry obstacles such as brick walls, but will be affected by wet materials.
- EM waves interfere with each other such that two coherent waves can cancel out each other, resulting in a much-reduced field strength level.
- All waves from a given source will be coherent, that is, they are locked in frequency and phase.
- If we have an ideal OATS with just the ground plane causing a reflection, it means that two signals will be received at the antenna, one via the direct path and the other via the reflected path. If the distance to the antenna is 3m, it is quite likely that the ground plane reflected signal will have travelled 3.5m, a path difference of 0.5m. If the emission frequency was 300MHz, wavelength = 1m, the two signals are 180 degrees out of phase and will cancel out each other. This is not just a theoretical nicety, it actually happens!
The solution is height scanning. It can be shown that, for every frequency in the range 30 – 1000MHz, there is an antenna height at which the two signals are in phase, roughly resulting in a 5dB increase in signal strength. This increase is allowed for in the limit levels specified in the standards. - The above is an account of what happens with just one reflection. Any kind of screened room (not anechoic) is literally a box of reflections and resonances which will make any measurement of signal strength quite meaningless. Do not use a screened room!
Solutions
The integrity of radiated emissions measurements is all about the test site. It makes no difference how sophisticated or expensive your receiver/analyzer is. The real source of error or measurement uncertainty is the site. Even with a good site that is clear of reflecting surfaces but with a ground plane, errors of up to 18dB will be experienced if height scanning is not employed. In a screened room, the errors will be off the scale.
So focus your budget on the test site. A chamber or a test cell are the best solutions, if you can afford them. But, if the budgets are restricted (and they always are!), then create the best OATS that your facilities allow, and use an emissions reference source (ERS) to calibrate the site “as is.” This will enable the characteristics of the site to be accurately measured and allow you to generate correction factors that can be applied to the results from your DUT. And the process can be entirely automated with advanced software solutions that are currently available.
Another advantage of this technique is that the requirement for height scanning is avoided.
Ambient noise is the final issue to contend with. A problem with ambient noise is that it suffers significant short-term fluctuations which can mask DUT emissions. Again, software solutions can help stabilize ambient noise while the DUT is switched off, thereby permitting the measurement of DUT emissions on subsequent scans.
