« Hot-Swappable Power Supply Failures on UPS | Main | Current-Mode Control Articles »

May 05, 2003

Diode Recovery EMI

Question: How do you reduce EMI due to diode reverse recovery?

Original Question: How can the Electromagnetic Interference (EMI) due reverse recovery of output rectifiers be effectively suppressed?

Answer: This has been a problem in power supply design since the days when alloy-junction diodes rectifiers were replaced by "improved" diffused-junction diodes in 50 Hz, 60 Hz, and 400 Hz rectifier circuits.

Things only got worse when switching-mode power supplies came along and the reverse recover of the inductor catch diode became a major EMI source.

Over time, there have been many solutions. First let's look at containing EMI by good layout, then line-frequency rectifier diodes, and then the inductor catch diode in switching-mode power supplies.

Layout: Good layout is simple in concept but not practiced as well it could be. Here is the concept. When a current is established in a conductor, it does not know its return path, so the field reaches out and finds the nearest conductor and establishes an equal and opposite current in it. (Lenz's law: An induced electric current always flows in such a direction that it opposes the change producing it.) If this is your intended return and the loop area is small, then EMI caused by a change in current is minimized.

Another way of looking at this is that a conductor establishes an external field around it and when you change conductors, you want that external field disturbed as little as possible. In rectifier circuits, this often means placing three or more traces as close to occupying the same space as possible -- the diode traces and returns. You also want to distance other conductors, including chassis, heatsinks, ground planes, etc. so they are not possible transient return paths. (Remember, when the electron starts its journey, it has no idea what the desired return path is and reaches out to the nearest conductors to establish the equal and opposite current required by Lenz's law. If this is not the intended return path, then you have EMI.)

For example, in a center tapped transformer secondary with two diode traces, the field transfers from one diode trace to the other diode trace with each sharing the same return. These three traces should come as close to occupying the same space as possible, with the diodes slipped in with as little disturbance as possible to the trace geometry. The effect is that as the current switches from trace to trace, there is little change in the external field, minimizing EMI. Most designers think of laying out parts and then connecting them. A powerful alternate approach is to layout signal and power flow conductors in the best geometry and then slip the parts in with as little disturbance to the geometry as possible. Usually you have to go back and forth between the two layout concepts to get the best layout.

Line Frequency Rectifiers: Back in the 1960's when I designed my first three-phase 400 Hz transformer rectifier sets I used alloy junction 1N538 diodes for the low power outputs. The modules passed EMI tests with no problem. A couple of years later, I got a call from the factory that all my modules were failing EMI tests.

What I found was that the manufacturer had "improved" the diode by using faster diffused junctions.

The initial solution was to add a RC snubber across each diode. Later it was found that by using a poor quality capacitor with low Q at the ringing frequency, the resistor could be eliminated. The final solution was to use the newer diodes with controlled recovery characteristics, but I continued to layout the RC pads across the diodes (with a shorted trace that could be cut for the resistor) in case they were needed. The never were.

Others had similar problems, the initial US military specification for a family of high density avionics 400 Hz 3-phase power supplies contained both switching-mode power supplies and series dissipative regulators. All of the switchers passed EMI tests and all of the series dissipative regulators failed EMI tests. The reason? Everyone knew from the beginning switchers were an EMI problem and this was taken care of in the design, but everyone also knew series dissipative regulators did not generate EMI and no one looked for the EMI problems during design. The reverse recovery spikes of the rectifiers caused them to fail EMI tests.

The solution for the problem now is usually good layout and selection of the rectifier diodes for a soft recovery spike, although snubbers are always a backup solution.

Inductor Catch Diodes: The major reverse recovery problem now is usually the inductor catch diode in switching-mode power supplies and these have been a problem from the beginning.

One of the early solutions, when power transistors were much faster than power diodes, was to place an inductor in series with the power diode with a small fast diode across the inductor with the anode pointing the same direction as the power diode. When the power diode voltage reversed to the blocking state, the inductor would keep reverse current from flowing through the power diode until the diode recovered. In the forward state, the small fast diode would carry the current until the inductor current caught up (the small, fast diode had to have an outstanding surge rating). Games could be played with saturation characteristics of the inductor. Later, lossy ferrites were developed for this purpose.

The next solution was for the semiconductor vendor to match the transistor and diode in the same package. At least one vendor had such a product in the early 1970's and it saved a lot of design grief. An alternate was to match them yourself, a real pain, but it worked. (Remember, the current for the diode spike comes through the switching transistor and this spike also generates EMI.)

Others just slowed down the switch. Just because it switches fast doesn't mean you have to use it that way. This was always part of my design approach.

Now the diode manufacturers shape the reverse recovery to make it softer and generate less EMI. I suspect that this, with good layout, is now the most common solution.

If the above don't work, you can always put an EMI suppression circuit across the diode. I always used to lay out pads across every recovery diode for a 10 ohm carbon resistor and a small ceramic tubular capacitor. If you pick a bad enough capacitor, you don't need the resistor. I heard a lecture from a Bell Labs (remember them?) engineer who said how they used lossy Mylar capacitors for this purpose. You can also use lossy ferrites. The key point here is that you do not want high Q parts for this application, you want them VERY lossy at the EMI frequencies. The Bell Labs engineer put it "you want to find the worse possible parts available."

Since I first published this blog, I have initiated a website that is dedicated exclusively to Snubber Design with its own mailing list. Central to it is Rudy Severns' ebook on the subject. If you are interested to learn more about snubbers, you may find this website at Snubber Design.

Others may have a different perspective on this topic. Comments are always welcome.

Posted by Jerrold Foutz at May 5, 2003 02:25 PM