Standardizing high-speed serial interfaces to manage EMI
Standardization committees are responsible for managing electromagnetic interference (EMI) from interfaces into products that use them. Data rates currently used in high speed serial interfaces exceed many of the operating frequencies used for wireless mobile communication in devices such as smartphones and tablet computers. Careful design is needed in the interface to manage the local electromagnetic radiation from the interface so that interference with any and all local radios is avoided. The most important techniques for managing EMI from high speed digital interfaces are discussed, showing how each contributes toward solving EMI issues.
For any mobile communication device to be successful, all intentional wireless communication signals must not be affected by any radio signals that other circuitry generates unintentionally, and also that the intentional radio signals must not affect the operation of any other circuitry. This is the principle of Mutual Transparency: all components within any mobile communication device coexist while performing their tasks, unaffected by any mutual interference. The operation of any circuit is transparent – meaning not interfering – to the operation of any other circuit. Electromagnetic interference (EMI) problems arise when the mobile device must communicate with remote networks, because the data rates used in modern high speed serial (HSS) interfaces often exceed the wireless communication frequencies used in mobile wireless communications radios.
Specification development committees must pay particular attention to EMI issues, both from the interface operation within the mobile device, and from radio signal coupling into the interface. This responsibility falls onto specification committees because even though the interface might work very well on its own, any interface that is either vulnerable to external interference, or that is noisy to other circuitry, will not provide functioning products. MIPI® Alliance has developed two specifications that pay very close attention to meeting Mutual Transparency, directly supporting the success of mobile products that adopt these interfaces.
Science tells us that whenever electrons are moved around, that radio signals will always be generated. These unintentional radio signals are not desired, so they must be managed at design time to minimize their possibility of becoming electromagnetic interference (EMI). Techniques available to manage EMI at specification development time include:
* Signal amplitude
* Physical Isolation
* Data Rate
* Skew and balance
* Slew Rate control, and
* Waveform Shaping.
These techniques all have different impacts on interface EMI behavior, and are each discussed below.
It is obvious that reducing the interface signal amplitude does lower EMI, however is does so very slowly. This technique is best left as a last resort because when the signal amplitude is cut in half, the corresponding EMI drops by only 6 dB. While this may be enough to get out of an EMI problem which is close to a regulatory limit, cutting the interface signal in half also reduces the receiver operating margin which will lead to errors in interface communication.
One of the more obvious techniques to keep EMI contained is to improve physical isolation. Once a radio signal exists, if it can be kept close to where it is generated then it will not bother anything else. At cellular or wireless LAN frequencies physical isolation is never perfect, and practical isolation values vary between 20 to 40 dB. Figure 1 shows one physical isolation example, measured from a small surface mount package. Realizing this isolation within the product design is usually essential to solving EMI problems. Careful measurement of isolation provided by IC packages and PC layouts is extremely important in the design process.
*See Figure 1
Data rate variation
When one examines the spectrum of the unintentional radio signal from a high speed serial digital interface has distinct properties. From an EMI perspective the most important property is the presence of a spectral null at the data rate and all of its integer multiples, which are clearly visible in Figure 2.
*See Figure 2a
*See Figure 2b
Figure 2. Nulls in the unintentional radio signal spectrum from a high speed serial interface are present at the data rate, and all integer multiples of the data rate. Changing the interface data rate moves all of the spectral nulls. This is a particularly effective technique to reduce EMI within a particular frequency band without any need for filtering. The vertical line demonstrated this improvement for a GPS receiver.
Even if the interface signal is filtered, these spectral nulls will still exist. This is fortunate, because the standardization committee can then choose to allow a change to the interface data rate which can move a spectral null close to a radio receiver band to remove EMI from the interface into the receiver. This technique comes with no cost increase because no filtering components are necessary to realize this improvement. Applying this technique is particularly important for GPS receivers, which must work from extremely tiny signals arriving from multiple orbiting satellites. Figure 2 shows how this technique is used to help protect a GPS receiver, where the standard allows the interface data rate to change from 1.248 Gbps (Fig 2a) to 1.456 Gbps (Fig 2b). GPS receiver interference, at the frequency marked by the vertical line near 1.6 GHz, falls by more than 10 dB.
Skew and balance
The balance of a differential signal refers to the amplitude matching of its two components. Skew refers to any time shift between the two components of a differential signal. These characteristics are largely set by the interface driver circuitry, and are best analyzed together. When signal balance is within 10% of perfect, the effect of signal balance is negligible compared to the EMI impact from skew as seen in Figure 3. From an EMI point of view, then, it is far more important to minimize skew than to be concerned with amplitude balance when designing interface driver circuits.
*See Figure 3
Figure 3. Comparing the EMI impact between signal balance and skew, it is seen that minimizing skew is much more important than getting a near perfect differential signal balance. With even a small skew of 2% UI, the effect of signal balance errors up to 10% become negligible. Perfect signal balance becomes important only if the skew is exactly zero, a condition that is very unlikely.
The bandwidth out to the first spectral null, known as the first spectral lobe, contains all of the information that the interface carries. The presence of additional spectral lobes at higher frequencies, called spectral sidelobes, represent information about the interface’s waveform transitions and not additional information about the data itself. If any EMI is caused from energy in these higher frequency sidelobes, it is possible to suppress this interference by slowing down the waveform transition time, or reducing the transition slew rate. This sets up a trade-off, between the speed of the interface waveform transition vs. the amount of high frequency EMI caused by the interface signals. As long as the interface ‘eye’ stays open in the middle of the unit interval, it is advantageous to use the slowest slew rate from an EMI perspective. The EMI benefit can be very significant, as seen in Figure 4.
*See Figure 4a
*See Figure 4b
Figure 4. Taking a longer time to transition the interface signal, called slew rate control, reduces the magnitude of higher frequency sidelobes from a differential signal: a) definitio
n of edge transition time on the eye diagram; b) corresponding spectra from the transitions shown in part a).
Figure 4a illustrates that this technique does impact the ‘eye diagram’ of the interface signal. The separation of the eye top and bottom is not affected, though the width of the eye that is fully open is reduced. This is a necessary price for using this technique.
Using Slew rate control on the interface signal has a negligible effect on the main spectral lobe. This is because slew rate control only reduces magnitude of the higher frequency sidelobes. This is both good and bad: it is good because slew rate control does not reduce the data content and it is bad if the frequency of interference is coming from the mainlobe because then any slew rate manipulation is ineffective from an EMI perspective. For this reason the MIPI Alliance DigRFSM committee, when addressing higher data rate requirements, selected the preference to use multiple lanes of M-PHY® with each lane running at lower data rates, instead of using one M-PHY lane operating at a higher data rate.
The shape of the waveform transition also has an impact on the EMI properties of the interface. For example, one straightforward way to implement slew rate control is to vary the current source values used for charging and discharging capacitances along the interface connections. Ideally this provides straight-line transitions like those seen in Figure 4 and also in Figure 5a. Other waveform shapes also impact EMI values, some for better and some for worse. In another example, Figure 5b shows the effect of an exponential waveform that results from simple R-C filtering of the interface signal. Here the EMI is actually worse than that seen from straight line transitions. The reason is that the exponential waveform has a sharp corner when any transition starts, even though the finish of any transition is very smooth. The sharp corner at the start of each transition causes more EMI problems than the smooth transition at each transition end solves.
*See Figure 5a
*See Figure 5b
*See Figure 5c
Figure 5. Different waveform shapes for interface signal transitions can have a large impact on EMI management: a) linear transitions, b) exponential transitions, and c) curvature limited waveform. Exponential transitions actually cause the worst for EMI interference, while curvature limited transitions are by far the best at suppressing EMI.
When all sharp corners are removed from the interface waveform, the spectral containment is greatly improved. Eliminating all sharp corners from the interface waveform, referred to as waveform curvature limiting, is the objective of waveform shaping. One example of a curvature limited interface waveform is shown in Figure 5c.
Good EMI management techniques all begin with maximizing the physical isolation of the application. Once good isolation performance is achieved additional techniques can be used as necessary, depending on the particular issues encountered by the interface standardization committee. Two examples from published MIPI standards are presented below.
The MIPI Alliance M-PHY specification is a HSS link that uses differential signaling with small amplitudes. Because M-PHY interface data rates are higher than many cellular communication frequencies, a combination of data rate selection, slew rate control, and bounds on skew are used to reduce EMI present at the input of internal radio receivers. Figure 6 provides one example of EMI improvements available with this combined technique. The interface spectrum in Figure 6 can be directly compared to that in Figure 4c to note these additional EMI suppression results.
*See Figure 6
Figure 6. The MIPI Alliance M-PHY interface Working Group has selected to combines slew rate control with bounds on signal skew to manage EMI reduction at high frequencies. Compare this result with the spectra in Figure 4b.
The RFFE interface Working Group within the MIPI Alliance has different issues to address, and uses different techniques to manage the EMI from that interface. Single-ended signal waveforms with large amplitude are needed by the RFFE application, which causes additional EMI problems because this interface operates very close to sensitive radio inputs. In this case the RFFE Working Group adopted a technique combination where the lowest data rate is used that is consistent with the interface communication needs. Curvature control is then used on the interface waveform to insure that any EMI is suppressed below the operating frequencies used by the internal wireless communications. Figure 7 provides one example of how effective this technique combination is for reducing EMI from the interface.
*See Figure 7a&b
Figure 7. Combining a low data rate with waveform shaping through curvature limiting allows the MIPI Alliance RFFE interface to suppress unintentional EMI at frequencies used by the major wireless communication bands: a) the 26 MHz data rate keeps most of the EMI energy at low frequencies; b) a small amount of curvature added at each end of every interface waveform transition provides dramatic additional reduction in EMI.
Managing EMI through actions of specification committees and working groups by design is an important component to achieving Mutual Transparency between interfaces and wireless communications links which are both present in mobile devices. The MIPI Alliance is a leader in recognizing this need and implementing it in interfaces used within mobile devices.
Learning which EMI mitigation techniques are either very effective, or not so effective, in establishing Mutual Transparency is an important result from deliberations within the MIPI Alliance M-PHY and RFFE interface working groups during specification development. Good physical isolation is the most effective technique by far. Following this, major improvements come from bounding of allowable skew for differential signals, and then avoiding exponential shaped interface waveforms which are the natural result from simple R-C filtering of the interface signal. Another particularly effective technique for minimizing EMI at higher frequencies is using waveform shaping to reduce sharp corners in the interface waveform.
Allowing selection of the interface data rate from a small list is a technique that does not require filtering and can provide spectrally local EMI suppression. Placement of spectral nulls near frequency bands of concern is an extremely effective EMI suppression technique. This weakest technique to reduce EMI is lowering the amplitude of the interface waveform.