Reducing EMI in digital systems using spread spectrum clock generators (Part 1 of 2)
Over the past couple of decades, more and more applications are going digital. Implementation of digital systems is very simple as it is entirely about logic; however, complexity increases exponentially with signal speed, specifically clock synchronization, setup and hold time, jitter, and so on.
These problems affect the functionality of not only the individual subsystem, but also cause electromagnetic interference (EMI) when high frequency devices are operating in close proximity. Figure 1 shows a typical example of EMI caused by a DVD player on TV reception.
Figure 1: EMI Effect on Television Reception
EMI is an undesirable system response due to either electromagnetic conduction or radiation emitted from an external source. This undesirable response or disturbance may interrupt and degrade the effective performance of any electronic system and might cause a complete system failure. Controlling electromagnetic interference (EMI) in any electronic system, therefore, has become an important design issue for electronic system designers.
Most of the problems faced while designing a digital system are directly or indirectly related to the clock of the system. Being the highest frequency signal, high slew rate and periodic nature (usually 50% duty cycle), clock signals become the largest contributor and primary source of EMI.
Furthermore, increasing speed requirements result in radiation with higher electromagnetic energy. To keep control of this radiation, there are several regulatory agencies across the globe which manage various EMI standards to ensure that any electronic equipment does not cause problems to the functioning of other devices.
These agencies set the limit on the maximum allowed radiation emission, which may vary from one country to another. Note that the maximum allowed radiation does not refer to the averaged emission but rather to peak emission. Any single frequency violating this limit will cause a device to fail compliance testing.
Multiple ways have been devised to address EMI and reduce radiated emissions. These include shielding, filtering, isolation, ferrite beads, slew-rate control, and good PCB layout using added power layer and ground planes. These methods can be used individually or in conjunction with others.
While shielding seems to be a relatively simple approach to reducing EMI, it is a mechanical implementation which is expensive and not at all suitable for portable and handheld equipment. Filtering and low slew rate may be an effective approach at low frequencies but not at the signal transitions rates being implemented today. Precise PCB layout techniques, for their part, tend to be time consuming and unique to a system, meaning that one kind of layout technique used in a system may not transfer exactly to another system.
Spread spectrum clocking is another method which can be used effectively to bring down EMI radiations. This article specifically discusses how Spread Spectrum Clock Generators can be used to cut down EMI radiations.
Spread spectrum clock generators
With Spread Spectrum Clocking, the concentrated energy of the narrowband clock signal is spread out over a wider bandwidth, reducing the radiated peak emission. Spread spectrum clocking can be visualized as frequency modulation of the input reference clock with controlled frequency deviation (∆f) and modulation rate where the output modulated clock sweeps its frequency repeatedly with time between two fixed frequency points as shown in Figure 2.
Figure 2: Frequency modulation of clock signal and EMI reduction:
(upper) modulated clock signal; (lower) output spectrum
Since the total energy contained in the signal remains constant and is distributed over a range of frequencies, the peak emission at any particular frequency is reduced. As the frequency band is made wider, the peak energy is reduced more. A peak EMI reduction of approximately 2dB to 18dB can be achieved using this technique. Such clock generators which generate Spread Spectrum (SS) clock are called Spread Spectrum Clock Generators (SSCG).
The biggest advantage of using spread spectrum techniques is that other timing, data, address, and control signals, which are commonly synchronized with and derived from the source clock, are also modulated, contributing a prominent EMI reduction throughout the system. Low cost and portability between different kinds of systems are some of the major advantages of spread spectrum clocking.
Conventional digital clocks have a very high Q factor, which means that all of the energy at that frequency is concentrated in a very narrow bandwidth, resulting in a higher energy peak. When viewed in the frequency domain for spectral density, one can clearly observe a taller narrow peak at the centre frequency and other relatively smaller, but again narrow, peaks on either side located at the harmonic frequencies.
SSCGs use the approach of reducing the peak energy in the clock by increasing the clock bandwidth and lowering the Q factor. SSCGs take a narrow-band digital clock signal to the input and generate an output clock which sweeps between a controlled start and stop frequency, at a precise modulation rate. In practical applications, the clock frequency is modulated with a modulation rate of 30 kHz to 120 kHz. This modulation rate is selected such that it stays high above the audio band, to avoid any interference with audio frequencies and not cause the system to suffer from any kind of tracking (e.g. setup, hold) problems.
The reduction in EMI is directly proportional to the spread amount of clock. The spread amount is usually quantified in terms of a percentage and is defined as the ratio of difference between the two boundary frequencies (Δf) to the clock target frequency (fo). Figure 3 shows EMI radiation with different spread amounts.
Figure 3: EMI reduction with Increasing spread amount
In most systems, it is the harmonics of the fundamental frequency which create problems. Fortunately, SSCG reduces EMI not only in the fundamental clock frequency, but also attenuates the radiation from harmonic frequencies as well. In fact, the attenuation of peak energy is more prominent at higher-order harmonics compared with attenuation at the fundamental frequency.
This happens because, for a fixed spread amount, the frequency band becomes wider at higher frequency values (i.e. at the harmonics which are just an integer multiplication of the center frequency) hence causing more reduction in radiated energy, Figure 4.
Figure 4: EMI reduction vs. harmonics
Selection of the spread profile also plays a vital in determining the amount of reduction in peak-energy content with SS technique. The spread profile is nothing but the envelope of the frequency variation of modulated signal (spread clock) with respect to time. A triangular profile is easy to implement from a design point of view, but the spectrum produced using this profile exhibits side lobes which are approximately 1-2dB higher than the center part, as shown in Figure 5a.
Figure 5: Comparison between triangular and "Hershey Kiss" profiles:
(upper) linear spread profile and output spectrum;
(lower) Hershey Kiss spread profile and output spectrum
As discussed earlier, a device will fail the EMI standard, even if one frequency component falls out of the maximum allowed radiation limit. Thus, a triangular spread profile which contains the peak emission in side lobes of the spectrum may violate the spec under certain operating conditions.
An almost-flat spectrum can be obtained by with more EMI reduction using a "Hershey Kiss®" spread profile (see Figure 5b). The Hershey Kiss spread profile has a radically distinctive shape, where the clock frequency sweeps at a higher rate near the start- and end-frequency points and is slowed down at the center. Because of the higher rate of frequency change near the two boundary points, the two side lobes are attenuated, and the reduced energy is distributed over the center flat portion of the spectrum.
This results in a dramatic change by approximately flattening the complete energy spectrum. As shown in figure, the Hershey kiss spread profile has provided a further 1.13dB reduction. This reduction can be higher based on the actual frequency values.
[Part 2 will look at types of spread (down, center, and up); and precautions to be aware of when using a spread spectrum clock (jitter, and also effects on PLL).]
About the authors
Ashish Kumar is presently working with Cypress Semiconductor India Pvt. Ltd. as a Senior Product Engineer. His interests are in making hobby electronic projects, debugging circuit boards, and dealing with complex analog and digital circuits.
Pushek Madaan is currently working with Cypress Semiconductor India Pvt. Ltd. as a Senior Application Engineer. His interests lay in designing embedded system applications in C and assembly languages, working with analog and digital circuits, developing GUIs in C# and, above all, enjoying adventure sports. Pushek can be reached at email@example.com.