Lee Ritchey; FAQ Ferrite Beads

Lee Ritchey FAQ Why Do Ferrite Beads Show Up in Applications Notes?

 

This is a good question to ask, since these ferrite beads often make PCB layout difficult around very high pin count BGAs, such as FPGAs.

 

It is worth exploring what ferrite beads are as a first step.  In the context of EMI control and power supply engineering these components are not actually beads.  They are, instead, surface mount components much like other chip components such as capacitors and resistors (See Figure 1).  They are available in the same sizes as these other components.

 


Figure 1. A Typical Ferrite Bead Package
Ferrite beads are made from a ferromagnetic material commonly referred to as a ferrite.  This material behaves like an inductor made from a coil of wire.  The feature that is attractive about such a component is that a relatively high inductance can be had in a small package.  Typically, these components are not specified by the amount of inductance they have.  Rather, their impedance at a particular frequency is listed.  As can be seen in Figure 2, the impedance of a ferrite bead is a function of frequency, much like an inductor with the impedance being quite low at low frequencies, rising to a high point and then dropping off.

 

 


Figure 2. Ferrite Bead Impedance vs. Frequency

Why, then, do engineers put ferrite beads in series with the power leads of ICs?  This practice dates back to the earliest days of EMI control, when the containment of EMI was done on a trial and error basis.  A product failed EMI tests, so an EMI technician would explore where the EMI came from by using a near field RF probe.

 

In logic designs, the EMI nearly always came from a large component such as a custom part in a pin grid array (PGA) package.  When a ferrite bead was inserted in the power lead of such a part, the EMI would go away.  The reason given was that this “fast” part was making noise that was getting into the rest of the PCB.  What was actually happening is that the body of the ASIC or IC was radiating the energy because it is a good antenna and the noise voltages related to switching were exciting the case itself.
 

In reality, the circuit illustrated in Figure 3 was created.  The inductance placed an impedance like that shown in Figure 2 in series with the power lead of the IC.  The frequencies involved in EMI range from 30 MHz to 1

 

GHz for most products.  When the IC attempted to draw power at high frequencies from the power supply it was prevented from doing so by the impedance of the ferrite bead.  As a result, there were no high frequencies on the IC package to cause an EMI problem.  This is one of the two ways to control EMI; eliminate the source or eliminate the antenna.  This technique works well as long as the IC or ASIC is not expected to operate with fast edges or fast clocks.

 

The effect of placing the ferrite bead in series with the power lead of the device is to degrade the performance of the power delivery system by increasing its output impedance.  Remember, a power supply is expected to be a voltage source, meaning that no matter how much current is drawn from it, the output voltage remains the same and no matter at what frequency which that current is drawn, the output voltage remains the same.  Said another way, power sources are expected to have zero or very low output impedance at all frequencies in order to do their job properly.

 

Figure 3. IC With Ferrite Bead in Power Lead

Sure enough, the speed of ICs increased to the point that this ferrite bead prevented them from operating properly.  Again, the reason is that the power delivery system output impedance was too high.  The proposed solution was to add a capacitor after the inductor as shown in Figure 4.  This solved the operating problem, but brought back the EMI problem.  More about the EMI problem later.  Now, the method recommended for implementing this circuit was to cut an island in the Vdd plane as shown in Figure 5.

 

Notice that in Figure 4 the capacitor is called a “bypass capacitor” with quotes around it.  The reason for the quotes is to call attention to the fact that this capacitor is not bypassing noise, it is serving a source of high frequency charge, so that the ASIC can again switch rapidly.  A much better name for these capacitors is “coulomb buckets” as they are functioning as local charge storage devices.  Also note that the inductor and the capacitor form a low pass filter, preventing high frequency noise from getting to the ASIC from the power subsystem side of the system.  This is the reason given in most current application notes for placing ferrite beads in series with the power leads of phase locked loops and other “analog” type circuits, including high speed serdes such as Rocket I/O.

Figure 4. IC With Ferrite Bead and Capacitor

 

Figure 5. Isolating an ASIC With A Plane Island

The technique shown in Figure 5 worked as a way to keep noise from the power system out of the power leads  of the ASIC circuit until two things happened.  The first is this noise became high enough in frequency that the ferrite no longer represented a high impedance (see Figure 2) and the capacitor no longer functioned well as a coulomb bucket (See the chapters in “Right The First Time, A Practical Handbook on High Speed PCB and System Design” on capacitor limitations for this.)  When this occurred, the noise was no longer blocked and the function of the circuit in the ASIC was degraded due to a lack of a low impedance source of switching current.

 

That is the reason for the large red X through the drawing in Figure 5.  With the speeds of modern logic this is no longer a safe solution to either noise or EMI.

 

Back to the original question, why do ferrites beads appear so often in applications notes?  It is my experience that there are two reasons for this.  The first reason is the least defendable.  The answer given when the author of the applications note is asked, the reason for the ferrite bead is more often than not, “we’ve always done it this way and if you don’t follow our application note, we won’t guarantee that the circuit will work correctly.”  When the question is turned around by asking if the vendor will guarantee that the circuit will work if the applications note is followed exactly, the answer is still no!   What kind of technical advice is that?  The second reason given is that the ferrite bead is there to block noise in the power subsystem from getting into the sensitive circuit.  I have seen examples of this in actual test circuits.  The noise is blocked, but the circuit performance is likely to be degraded due to poor power delivery to the circuit being “protected.”  Figure 6 shows the output waveform of a 3.125 GB/S serial link with a ferrite bead in the power lead of the output stage.  Figure 7 is the same output with the ferrite bead removed and the power lead connected directly to Vdd.  Inserting the ferrite bead actually made the circuit perform worse than with no ferrite bead.  The circuit for Figure 6 was recommended by the supplier of the part without first checking to see if the advice was sound.  These waveforms were actually taken from an evaluation board supplied by the vendor.  How could such incomplete engineering be done on such important matters?  Good question.

 

Figure 6 3.125 GB/S Serdes Output with Ferrite

    Figure 7 3.125 GB/S Serdes Output without Ferrite

 

When the second reason is given, namely to block noise from the power subsystem, this is treating a symptom, not dealing with the problem.  The problem is that there is noise in the power subsystem because it was not designed correctly.  Chapters 32 to 37 in the above mentioned book cover how to do a good power delivery system design.  Several of the references listed below also lend insight into how to do this.

 

My experience has been that the use of ferrite beads is either a knee jerk reaction or a band aid.  In 30+ years of designing high speed computing systems and networking products, I have never used a ferrite bead in the power lead of a device, whether it is a phase locked loop or an “analog” circuit.  Instead, I have determined what the “ripple” requirements of a circuit are and designed the power delivery system to meet this requirement.

 

What should the vendor have done to insure that its application note correctly advised its customers?

 

The first thing an IC vendor needs to do is understand the power delivery needs of each IC.  This includes maximum delta I that the circuit may demand of the power delivery system as well as at what frequencies and the maximum allowable delta V (ripple).  Without this, it is not possible to design a power delivery system.

 

The next thing a vendor needs to do is to advise users on how to create a functional power delivery system.  Any time that there is a temptation to add a ferrite bead in the power lead of a device, four things must be demonstrated.  These are:

  1. There is a problem that can be solved by the use of a ferrite bead.
  2. The ferrite bead actually solves the problem
  3. The ferrite bead does not cause a new problem such as that illustrated in Figure 6.
  4. Using a ferrite bead is the best way to solve the problem

My experience is that, after step 1 & 2, ferrite beads are eliminated from the design.

 

What should an engineer do when he or she encounters an application note recommending ferrite beads?

 

Whenever I encounter an applications note that recommends the use of ferrite beads I call the author and ask that the four steps above be demonstrated.  In no case have I found one where going through these steps results in agreement that the ferrite bead was a good choice.

 

If the vendor still insists that the ferrite beads are required, insist on seeing a test circuit in which the component is used exactly the way it is intended to be used in the new design.  If no test circuit is available it is good to be suspicious.  In one case when I was having trouble getting a microprocessor to work properly I asked to see the test circuit used to arrive at the application notes and specifications for the part.  I was told there was none and never had been one.  To this I responded, how do find out if the part works correctly?  The response was, “we give them to our customers and they tell us if they work”!

 

If there is no test circuit, be suspicious.

 

Why are there so many applications notes that contain ferrite beads if they are not a good solution?

 

As mentioned earlier, many times ferrite beads are included in applications notes “because we have always done it that way”.  Why, then, haven’t they caused problems?  The answer is that prior to 130 nanometer ICs coming onto the market, most circuits ran slow enough that there wasn’t the need for a very low impedance power source at high frequencies.  This is a case of succeeding in spite of bad habits rather than because of good engineering practices.  The same thing is true of the use of bypass capacitors.

 

Because users have succeeded in spite of poorly engineered applications notes there had been little incentive to put any resources into making sure they are technically correct.  As with FPGA packages, it is only after a number of customers fail and start to complain, or worse, switch to the competition that resources are applied to creating applications notes that correctly characterize parts and how to use them.

 

130 nanometer IC processes have resulted in ICs that are as fast as or faster than the high speed logic know as ECL which all engineers seem to know requires good high speed design practices.  Fortunately, there is an abundance of good design information available to make this transition.

 

EMI containment.

 

EMI containment is a major reason for ferrite beads in power leads of ICs.  As mentioned earlier, this solves the EMI problem by preventing the part from switching fast.  Again, this is fine as long as the part does not need to switch fast.  Virtually all modern ICs need to switch fast to do their job.  Placing ferrite beads in their power leads stops the fast switching and the EMI.  The same result can be achieved by turning off the power.  A solution is required that does not preventing proper function.  Again, there are two ways to contain EMI, turn off the source or eliminate the radiating surface or antenna.  For most products turning off the source is not a choice.  All that remains is to avoid making unwanted antennas.

 

What things function as antennas?

  • Most BGA packages
  • Memory SIMMS and DIMMS
  • Any two PCBs joined by a connector
  • Cases that have been connected to logic ground at more than one place
  • Unshielded wires exiting a case

 

To deal with these antennas, see the EMI article in Volume 3 Page 2 published by Speeding Edge.

 

References:

  1. Ritchey & Zasio, “Right the First Time, A Practical Handbook on High Speed PCB and System Design”, Speeding Edge, 2003.
  2. Hubing, Todd H. etal, “Power Bus Decoupling on Multilayer Printed Circuit Boards” IEEE Transactions on Electromagnetic Compatibility, Vol. 37, NO 2, May 1995.
  3. Greim, Michael C.  High-end Digital Systems Give a Thumbs Down to Rules of Thumb”   EDN, June 5, 2000.
  4. Smith, etal, “Power Distribution System Design Methodology and Capacitor Selection for Modern CMOS Technology”  Published by Sun Microsystems, 1999.
  5. Ritchey, Lee W. “What is EMI, Where does it come from and How Can It be Controlled”, Current Source, Volume1, Issue No 3, Summer 2005, available from Speeding Edge, www.speedingedge.com.

 

Credits:

Figures 1 & 2 Taiyo Yuden

Figures 6 & 7 Mahi Networks