As an engineer, you know how fast consumer electronics devices operate and how many tasks they perform in fraction of second. High-speed operation of these devices has made possible the many pleasures of intuitive touch phones and live video streaming, as well as many practical applications such as achieving an ultra-high level of precision to generate energy for Higgs Boson or driving high-speed data for networking and communications applications.
This increasing level of performance and speed of electronic devices has brought many challenges to the design and designer. Safety is a critical factor when designing these devices and requires special attention. Electromagnetic Interference (EMI) is one of the prime threats to user safety and reliable operation of other devices close by.
Once, when I was watching a football match on HDTV, the ball was about to hit the boundary when I received a call on my GSM cell phone, Figure 1 .The TV immediately lost the signal. In surprise, I disconnected the call and the TV signal came back. What happened was a case of electromagnetic interference, or EMI.
Figure 1: Cell Phone Interference with TV Set Operation
Every electrical signal is a combination of electrical and magnetic fields. Any signal which is limited (or bounded) in the time domain is unlimited (or unbounded) in frequency domain and vice versa. All electrical system have signals which carry information, and so they follow particular pattern.
These patterns are defined by the communication protocol used at the receiver and transmitter, which can be located as close as a chip mounted on the same PCB or as far distant as the earth and a satellite. Signals which have a fixed pattern in the time domain will have signal energy spread over a wide frequency range.
Figure 2: Clock Signal in Time and Frequency Domains
For example, Figure 2 shows a 125-MHz clock signal in both the time domain as well as the frequency domain. In time domain, this clock signal has a periodic nature and its energy is spread over a wide range in the frequency domain. Markers in the figure at 125 MHz, 375 MHz, 625 MHz and 875 MHz show four consecutive high energy points in the frequency domain. Peak energy is stored at the fundamental frequency of 125 MHz, which has 10.55 dBm amplitude.
In addition, 375 MHz, 625 MHz and 875 MHz are odd harmonics of the 125-MHz clock signal. This energy is a key contributor to EMI radiation which interferes with other systems running in close proximity. A system which has peak energy at a particular operating frequency is an example of an electromagnetic radiation source.
In the example with the cell phone and TV, the radiation source is the cell phone. When I was watching football on the LCD HDTV, the set top box was tuned to the sports channel in the range of 54 to 890 MHz. When the cell phone is in idle mode – i.e., not being used for calling or texting – it isn’t transmitting data or emitting radiation.
However, when the cell phone receives a call, it begins communicating with the nearest base station. During transmission, the cell phone uses more power and transmits its signal in the GSM band (900 MHz), which causes radiation in the adjacent set top box frequency range (near 890 MHz) as well. Due to this radiation, the set top box is not able to decode the broadcasted TV signal.
This will not happen at your home due to enough margins between operating frequencies and EMI safeguards and guardbands considered in the TV set and cell phone. However, cell phone and TV set designers will have more EMI challenges when 4G transmission is supported, which is in 800-MHz range.
While this kind of disrupted poses little threat of harm when it occurs in consumer devices, imagine what can happen when a Wi-Fi signal interferes with the control operation of a hazardous chemical plant. There have been many accidents noted in history due to EMI, resulting in the world’s leading countries coming forward to regulate the electronic design from an EMI aspect.
Apart from EMI radiation, there is conducted EMI which comes from the traces on the circuit (PC) board, power supply, and from capacitive or inductive coupling that can disturb system operation and the functionality of other devices.
Many international regulatory bodies (IEC, CISPR and EN) have placed a cap on the maximum radiation a system can have, as well as defined, minimum EMI levels which should not affect the system. International standards for EMI are defined based on applications and end-equipments such as military, consumer, industrial, automotive, for example.
High-speed equipment and devices are driven by high-speed clocks. The key contributors to EMI are high-speed clock signals, which are periodic in the time domain. As shown in Figure 3 , a 125-MHz clock signal has peak energy at the fundamental frequency and strong odd-harmonic signals as shown in Figure 2 .
Figure 3 : Peak Energy in Clock Signal
(click here to see enlarged image)
Figure 3 shows a spectrum analyzer plot concentrated on the fundamental frequency for a 125-MHz clock. The peak indicates the energy at the fundamental frequency, which is -5.29 dBm. If this energy at the center frequency is reduced, the radiation due to the clock signal will be decreased as well.In a system where the clock signal drives most of the devices on the board, the improvement on EMI will be significant. Spread spectrum technology uses this same concept to improve system EMI performance.
In spread-spectrum technology, the high-frequency clock signal is modulated over a low frequency. The modulating signal frequency is normally from 30-120 kHz. Due to modulation, the energy stored at the fundamental frequency gets distributed over a wide range of frequencies and reduces the peak, as shown in Figure 4 . Without spread-spectrum techniques, the peak in a 125-MHz clock signal was -5.29 dB.With a ±2% spread, the peak is reduced to -13.37 dB.
Figure 4 : 125 MHz with ±2% Center Spread
(click here to see enlarged image)
Figure 5 : Lexmark Center and Down Spread Profile at 31.7 kHz Modulation Frequency
Modulation Profile : In simple terms, the modulating signal waveform shape represents the modulation profile. There are two well-known profiles: Linear and Lexmark. The Lexmark profile gives better peak reduction compare to the Linear profile. The waveforms in Figure 5 are an example of the Lexmark profile. The Linear profile is similar to triangular waveforms in shape.
Spread Percentage : The peak-to-peak amplitude of the modulating signal represents the spread percentage. The waveforms in Figure 5 capture ±2.5% and -5% spread percentages. The percentage of spread represents the deviation from the nominal signal frequency.
Spread Type : If the nominal signal frequency is at the center of modulation profile, it is called the center spread. If the signal frequency is at the top of the modulation profile, it is called down spread.
The effect of using spread-spectrum technology on peak reduction depends on the percentage of spread and profile selected. The following example will show the corresponding reduction in radiation due to an increase in spread percentage.
In an example related to IT (information technology) equipment, a CY25100 clock chip is used to drive FPGA and Ethernet chips as shown in Figure 6 . The reference clock output is 25 MHz and has no spread. The 12-MHz signal is generated using an on-chip PLL and can be a spread-spectrum clock.
Figure 6 : CY25100 Application
Initially, the customer used the 125-MHz clock output without spread. For requirements of the European Union, the equipment needs to meet EN 55022 standard limits. After testing a prototype unit of the equipment, 44 dB radiation was found at 125 MHz multiples, thereby exceeding the EN 55022 limit of 40 dB.
After performing a few trials with different spread percentage, a ±2% center spread at 31.7 kHz gave a reduction of 6 dB in overall equipment radiation, enabling the system to meet EN 55022 requirements with a 2-dB margin.
To check the effect of spread spectrum clocking on EMI, the 125-MHz clock peak energy was measured using a characterization board (i.e., with only a CY25100 device on board) in a clean set up with the lowest-possible noise. Without spread spectrum, the clock signal peak measured was -5.29 dB. With ±2% spread, the reduction in clock peak was found to be 8 dB. Table 1 and Figure 7 show the reduction in peak with spread percentage.
Table 1 : Peak measurement with different spread percentage
There is a significant improvement in EMI radiation when using a spread-spectrum clock. In addition, these improvements are available without a time-consuming or costly redesign of the prototype or board re-spin.
Items to check before using a spread-spectrum clock
Jitter : Spread spectrum modulates the high-frequency clock signal at a low frequency, thereby increasing deviation of the clock edge over the period. This results in higher jitter. Measure the jitter of the spread spectrum clock and confirm that the clock receiver can tolerate this added jitter in the spread spectrum clock.
PPM Error : The parts per million (PPM) error parameter is used to measure the accuracy of the clock signal. Due to the spread percentage, the clock frequency will vary from the nominal frequency and so the PPM error will increase. For applications where PPM error specs are very stringent, make sure the PPM error is still within limits after using spread spectrum.
Spread-aware PLLs : Consider an application where a single clock drives multiple clock-receiver chips and hence zero-delay buffers are used to meet the clock-receiver load. If the originally generated clock signal has spread spectrum, verify that the zero delay buffers support the spread by allowing it to pass to the clock receivers to gain full EMI reduction benefit.
If the zero delay buffers do not support spread, the input spread will appear as skew in the clock buffer output.This is known as tracking skew , which can be a point of concern if the application has tight skew specs (such as a synchronization application).
Select spread parameters for a specific application
Increasing the spread percentage increases the reduction in peak energy, but the rate of peak reduction is not constant. As shown in Table 1, when the spread percentage is ±2%, the reduction in peak is 8.08 dB [(-13.37)-(-5.29) dB]. When the spread percentage is increased to ±2.5%, the reduction in peak is 8.82 dB [(-14.11)-(-5.29) dB]. Hence, less than 1-dB peak reduction is observed for a further ±0.5% spread increment.
Spread Type (Down/Center) : This depends on the clock-receiver maximum operating frequency. If your clock receiver can support ±tolerance to the normal frequency, use center spread or down spread. Center spread varies the nominal frequency in both directions: 25 MHz with ±1%, will vary from 24.75 MHz to 25.25 MHz. Down spread changes the nominal frequency on the down side only, ensuring that the maximum frequency will be always the nominal frequency: 25 MHz with -1%, will vary from 24.75 MHz to 25 MHz. In the case where the normal frequency is the maximum frequency, down spread is the right choice.
Modulation Frequency: Many clock generators support a wide range of modulation frequencies around 30 to 120 kHz.The most widely used range is 30 to 60 kHz. Modulation frequency can be anywhere within this range except those frequencies or their multiple which can interfere or couple with other system or device’s operations.
Modulation Profile : The Lexmark profile gives a higher peak reduction compared to the Linear profile. If your clock generator supports the Lexmark profile, choose it.
Most of the standard protocols (such as PCI versions) give recommendations for supported spread-spectrum parameters, which do not adversely affect the communication bit error rate (BER).
Is using a spread-spectrum clock enough to manage EMI?
When an application is a complex system, where multiple data, clock, and address buses are running at different speeds, use different power supplies, and support different communication protocols, using just a spread-spectrum clock for one specific frequency will not completely manage system EMI. Designers need to be cautious about board-design issues as well, such as cross talk and ground loops, because there are multiple sources of this noise.
In such complex systems, there can be protocols which may not function optimally due to tight PPM requirements or jitter specifications. A spread-spectrum clock can help reduce overall radiation due to clocking, but other techniques will be required to compensate for other data or clock signals which do not support spread. However, as shown by the case study, it is always an advantage to have spread-spectrum capabilities available since they reduces the need to shield and filter for many challenging EMI sources.
About the author
Brijesh A Shah is Senior Applications Engineer at Cypress’ Timing Solution Business Unit. He loves to work on Analog circuit applications, and his key expertise lies in clock and power supply applications. He has an MS degree in Microelectronics from BITS, Pillani. Brijesh can be reached at .
Editor's note : Liked this? Want more?
If you are interested in “analog” issues such as signal input/output (sensors and transducer, real-world I/O); interfacing (level shifting, drivers/receivers); the signal chain; signal processing (op amps, filters, ADCs and DACs); and signal integrity, then go to the Planet Analoghome page for the latest in design, technology, trends, products, and news. Also, sign up for our weekly Planet Analog Newsletter .