How EMI is eliminated in a high-speed digital interface, seven methods are shared with you

Small size and low cost high-speed serial (HSS) interface is particularly valuable for mobile devices with small size, low power consumption and light weight. Electromagnetic interference (EMI) occurs when mobile devices have to communicate with the remote network, because modern HSS usually uses a higher frequency of wireless communication than mobile devices.

The science of electromagnetic compatibility tells us (according to Maxwell's equation) that when electrons move, radio frequency signals will definitely be generated. In the design, seven main technologies can be used to manage EMI, including isolation, signal amplitude, offset range, data rate, signal balance, swing rate control and waveform shaping. These technologies have different functions, and we will discuss them one by one.

Quarantine

Physical isolation is probably the most obvious technology. For RF signal, if it can be "shielded", it will not interfere with any other signal. Although isolation will never be perfect, and at cellular or wireless LAN frequencies, the actual isolation decibel value is between 20 and 40 dB. To achieve this level of isolation solution EMI problem is usually necessary. Therefore, it is very important to carefully measure the isolation between IC packaging and PCB layout.

Figure 1. An isolation cover for RF encapsulation

Signal amplitude

Reducing the amplitude of the interface signal will certainly reduce EMI, but the effect is not significant. If the signal amplitude is halved, EMI will only be reduced by 6dB. This may be enough to get rid of the locking problem. However, this method also reduces the margin of the receiver and may lead to interface errors. Based on this, it is best to use this as the last resort to respond to the EMI problem.

Drift and balance

Drift is the time shift between two components of a differential signal. Balance is the amplitude matching between two components. These two parameters are basically determined by the interface drive circuit, and it is better to analyze them together. As shown in Figure 2, when the signal balance is within 10%, the EMI impact ratio caused by drift and the exact value of signal balance are not so important. This means that from the perspective of EMI, when designing interface drive circuits, minimizing drift is far more effective than focusing on amplitude balance.

Figure 2: Group comparison of signal balance and drift

The chart shows that managing drift is much more important than obtaining a very closed signal balance. Even at 2% UI signal balance error of up to 10% is negligible in the drift process. The signal balance becomes important (unlikely) only when the drift is 100% zero.

Data transfer rate

The RF spectrum of digital signals has different characteristics. From the perspective of EMI, the most important thing is the spectrum zero of data rate and its integer multiplier. Figure 3 clearly shows the zero values of these spectra.

These zeros exist independently in any signal filter. It is a practical choice to remove the value EMI entering the receiver by changing the data rate rather than moving the zero value of the spectrum to the frequency band of the RF receiver. This is particularly important for GPS receivers, which must recognize the extremely weak signals sent by multiple satellites. Figure 3 shows this technology to help protect the GPS receiver. The data rate has changed from 1.248Gbps (Figure 3a) to 1.456Gbps (Figure 3b).

（a）

（b）

Figure 3: Changing the interface data rate will move the spectrum zero value. This is a particularly effective method to reduce EMI in a specific frequency band without any filtering.

Yaw rate

All necessary information carried by the interface is located on the main spectral valve. The spectrum side lobe carries the data waveform transformation information, not the data itself. In short, EMI of side lobes generated by energy (the frequency of these side lobes is higher than the data rate) can be suppressed by reducing the swing rate of each waveform conversion. This is effective because the total bandwidth of the unexpected RF signal is not controlled by the data rate, but by the fastest conversion (edge) of the data waveform.

Figure 4a (top) shows that this technology does affect the "eye diagram" of interface signals. Although the width of the fully open eyes narrows, the separation between the top and bottom of the eyes is not affected. This is the price that must be paid for using this filtering technology.

Please note that the swing rate control will only reduce the amplitude of the side lobe. Any influence on the main lobe can be ignored. This has both advantages and disadvantages: it means that the swing rate control will not dilute the data content. The disadvantage is that when the interference frequency comes from the main lobe, the technology will be invalid. For this reason, in other applications, M-PHY's MIPIAllianceDigRFSM, people tend to use each channel to work at a lower data rate rather than a higher data rate.

（a）

（b）

Figure 4: The effect of swing rate control on the side lobe with higher differential signal frequency: definition of the top) eye image edge conversion time; Bottom) and transform the corresponding spectrum.

Waveform shaping

The direct method to control the pressure swing rate is to adjust the charge and discharge capacitance of the current source. This results in the linear transformation in Figure 3 and Figure 5a below. Other waveforms do affect EMI values, with good and bad results. For example, Figure 5b simply shows the exponential waveform effect obtained by RC through filtering. In fact, EMI has become worse here. The reason is that at the beginning of any conversion, the exponential waveform forms a sharp angle, even if the end of any conversion is smooth. But at the end of the conversion, the infringement has occurred.

Figure 5c shows that when all sharp corners are removed from the interface waveform, the limiting performance of the spectral clamp is greatly improved. The removal of sharp corners is the main goal of corrugated plastics, so it is sometimes referred to as waveform curvature limitation.

（a）

（b）

（c）

Figure 5: Signal transformation has different waveform shape EMI signal spectrum changes: a) linear transformation, b) exponential transformation, and c) filtered waveform. Exponential transformation actually suppresses its worst EMI capability.

Technical combination boxing

All EMI management technologies begin with maximizing physical isolation. In addition to isolation, different technologies will be adopted according to the specific problems encountered by the Interface Standardization Committee. The following is an example of two standards from the announcement MIPI.

The M-PHY specification of MIPI Alliance is a standard HSS link using low amplitude differential signals. Since the data transmission rate is higher than that of many cellular and other wireless communication frequencies, the combination of data rate selection, swing rate control and drift boundary is used to reduce EMI at the internal (including possible monolithic) RF receiver input. Figure 6 is an example of this improvement.

Figure 6: MIPI Alliance's M-PHY interface combines drift boundary and swing rate control technology to reduce high-frequency EMI. Fig. 4 Spectrum in result b comparison.

MIPI RF Front End Alliance (RFFE) has different problems, and its technology management is also different from EMI. Even if the RFFE interface works close to the sensitive RF input, the application also needs a large single-ended signal. The combination of technologies used here first uses the lowest data transfer rate consistent with the application requirements. Then, we control the curvature of the interface waveform to ensure that the EMI operating frequency of anything is only lower than the local RF operating frequency. Figure 7 is an example to demonstrate its effect.

（a）

（b）

Figure 7: The RFFE interface of MIPI Alliance combines data rate selection and waveform shaping technology to control the unnecessary RF signal frequency band below the main wireless communication frequency band: (top) 26MHz data rate makes most signal energy at low frequency, while (bottom) realizes a small amount of curvature control at the beginning and end of each conversion, which greatly improves EMI suppression performance.

summary

The designed EMI management is a key component to realize the mutual transparency of the interface and receiver in mobile devices. The standardization committee that defines these interfaces, such as the MIPI Alliance, would better control this capability.

When emphasizing mutual transparency, the experience gained in the development of M-PHY and RFFE interface specifications shows that the EMI of interface specifications is reduced. In short, some technologies are very effective, while others are not. So far, the most effective technology is good physical isolation. The second is to limit the allowable drift of the differential signal and avoid using the drift RC filter EMI that may lead to exponential interface waveform. In order to reduce the sharp corners on the interface waveform, the waveform shaping technology is particularly effective.

Selecting the data rate is a technology that does not require filtering. Because it comes from digital waveform EMI, there is a spectrum zero for data rate and all integer multiples. It is also very effective to place these zeros near the relevant frequency band. Finally, but of course, this is not important. This is for the amplitude of the mouth waveform. This technology has a negligible impact on EMI.