Q & A with Nicola Rosano, Sr. Strategic FA/System Engineer

Q:  What do you need to know before approaching an EMI challenge?

A:  Troubleshooting EMI requires a strong and solid practical experience. It is not a case of EMC field being defined as a “black magic field”. Some issues can be fixed by using related books on the subject, others can’t and require a fine tuning on the real physical system. The four topics mentioned below identify the needed background to minimize errors on the bench, enabling the user to save huge amount of time. These are: 

• The basics of Electro Magnetic compatibility (x2 books suggested).
• How the electrical noise can be divided in its two main components: Differential mode and Common mode noise. 
• What is the LISN (Line impedance Stabilization Network) and why it’s crucial to measure EMI.
• EMI filter design. The academic background is good, while a dedicated training course at industrial level it’s more effective. Designing EMI filters could appear an easy task if the common switching frequencies are used – say between 50kHz to 200kHz – but when the switching frequency is so high (beyond MHzz like Vicor modules) the designer can meet the components physical limits and other complications which makes the components selection deeper and harder than expected.

Sooner or later, every power supply designer finds out that the main three fields which can cause a return to the bench at the very last moment are: thermal issues, safety-related issues, or a stubborn EMI (electromagnetic interference) problem. Of these, EMI may be the least predictable because it turns out to be a veritable “balloon”. If we try to “push in” the emissions spectrum at one frequency, it “bulges” out at another. So, it happens so often that if we manage to achieve compliance with conducted emission limits, we may find it was at the expense of radiated limits, and so on. EMI in power supplies is a challenging area, partly because a lot of uncharacterized parasitics enter the stage, each vying for attention. So, bench tweaking is not going to be completely avoidable. The rule is the following: in the electromagnetic world, all electrical devices need to coexist with each other and be “good neighbors” — not causing too much interference to others and not being overly sensitive to others’ interference either.

Q:  What are the main drawbacks of having high levels of conducted emissions and radiated emissions?

A:  EMC (Electro Magnetic Compatibility) tree is divided in 4 fields. These are: Conducted Emissions, Radiated Emissions, Susceptibility to Conducted Emissions and Susceptibility to Radiated Emissions.  

The main two fields where SMPS (switch mode power supplies) give problems are the first two: Conducted Emissions and Radiated Emissions. In this context the main drawbacks of having high EMI levels can be collected shortly as follows:  Electromagnetic interference (EMI), Component stress, Compliance issues, Radio frequency interference (RFI), Signal distortion, reduced efficiency. The primary reason for controlling the conducted and radiated EMIis to prevent the emissions from one electrical system from interfering with a second or third system.  Consumers will see this when EMI interferes with radio signals on the infotainment system, causing bad audio or loss of reception.  Operationally excess EMI can degrade the quality of the data signals used within the vehicle, preventing the vehicle from reading all the necessary data for compliance to emissions and safety requirements.

Q:  What are the traditional methods and trade off to fix EMI?

A:  The main traditional methods and tradeoffs to fix EMI can be collected shortly as follows: input & output filters; shielding; grounding techniques; snubber circuits; spread spectrum techniques; soft switching. From these, a crucial attention needs to be addressed on having a proper grounding path as well as the power stage topology selection and its control. If a power stage uses hard switching, the resultant high levels of EMI and switching losses make it difficult or nearly impossibile to achieve high switching frequencies in the MHz range. In addition, if a proper grounding is not guaranteed there is nothing a designer can do to fix EMI on the bench with filters only. The layout is more critical than designing filters. This also explains why cooperation between departments is essential for the overall SMPS development and mass production.

Q:  Which component can we consider as device under test (DUT) for an EMI deep inspection?

A:  To understand this properly, it’s important to have a little introduction about the device under test (DUT). The NBM9280 is a chassis mounted, non-regulated, non-isolated, bidirectional, highly efficient DCDC converter. It has been selected because it has a very high switching frequency as well as is automotive rated.  It has peak efficiency of 99% and it’s mainly used in the 800V/400V DC-to-DC conversion in the automotive industry. It has a very compact dimension of 92 by 80mm and it is able to manage 30kW power flow. The NBM typically acts as interface between the DCDC fast charging station (400V rated and placed externally to the vehicle) and the High Voltage battery (800V rated placed within the vehicle). One of the main advantages of the NBM, current sharing feature apart, is that it allows any OEM to work at 800V battery without modifying the overall powertrain, reducing effort and timing.

Q:  How does the common mode (CM) noise affect the DUT?

A:  If we measure the noise on the metallization in respect to GND (Ground) we found a noise with a square wave shape similar to the one of the high voltages fast switching nodes located internally to the NBM. That noise is relatively fixed in respective of the amount of load. It appears to be almost load independent i.e. the harmonic content stays approximately the same by varying the load. If there is a direct connection between the NBM metallization and the LISN, a ground sniffer can be used. We can find the harmonic content of the plating versus the ground. Also, if we keep that metallization isolated between the ground path and the metallization, we can reduce the CM component reducing the overall noise in the conducted emission test.

Q:  What are the main tricks to fix the common mode (CM) for the DUT?

A:  The use of distributed bypass capacitors help to improve the CM performance by reducing the noise. In the same context, it is possible to prove that by using a distributed placement of capacitors nearby the area at high dv/dt, the CM noise reaches very low levels. Another way (to be added to the previous one), is to reduce the AC noise component by using a ferrite core surrounding the power lines in a way the currents through the wires tend to oppose each other. This approach reduces the ripple and has a very high impact in reducing the EMI spectrum. The external ferrite helps to reduce any unbalance current distribution too. 

Q:  Which filter topology could be used to be CISPR 32 compliant?

A:  The main three filter topologies are: L-type; pi-type, T-type. Any one of these, if correctly designed, can bring the overall noise below the admitted limits. In the example I’ve spoken about, I used the L-type because it is simpler and cheaper to implement on the bench. 

Q:  What are the advantages / disadvantages of using high working frequency?

A:  High frequency, with the same power, is synonyms of small components. So lower weight and lower volume surely is an advantage. On the other side, switch losses tend to increase with the increase in frequency. A good design strategy is to choose high frequency keeping a resonant topology. Resonant topology are advanced SMPS which null the switch losses independently from the frequency getting very high efficiency levels. 

Q:  What are the main lessons learned?

A:  For the case study under investigation, main lessons learned are: 

• Amplitude of odd harmonics are load independent for this DUT. However, even harmonics will increase as the result of loading. 
• Keeping the metallization on NBM9280 isolated from the ground improves the EMI conducted emission result. The increase in the second harmonic component can be mitigated by implementing PI filter. 
• Satisfying CISPR 25-Class 4 & 5 (more stringent limits) still needs more attenuation on the first harmonics. One possible solution is implementing a different ferrite core on HI/PGND which results in more attenuation on the first harmonic. Or considering a double stage filter.
• Working at high switching frequency allows very compact filter dimensions

Nicola Rosano joined Vicor in 2022 as Senior Strategic Application & System Engineer, EMEA, offering technical support and consulting for automotive power systems. Prior to Vicor he worked in the military, defense and space sectors at Thales Alenia Space and Airbus Defense and Space, as well as in the automotive sector at BorgWarner and Stellantis as a Senior HW Power System Engineer. He has several patents and is cultivating an analog/power electronics educational YouTube channel. His current research focus include power electronics, circuits and systems, electronic instrumentation, and engineering education. Rosano received his B.S. degree in 2010 and his M.S. degree cum laude in 2013 in electrical engineering.

Nicola Rosano, Sr. Strategic FA/System Engineer