Migrating to high-speed USB in stages can reduce the risks associated with completely redesigning a handset’s USB subsystem.
Cellular-service providers continually strive to make more features available to their subscribers, features that will increase the average revenue per user (ARPU) in a market reaching saturation.
Over the last few years, handsets have integrated digital-still-camera (DSC) functionality. Now, most of the handsets found at your local store have cameras equipped as standard. The idea behind this is that subscribers will take pictures and share them with their friends using airtime in the process, thereby creating new revenue streams for the cellular-service providers.
The next trend that’s begun to take off in the last couple of years involves the integration of portable-media-player (PMP) functionality into the handset. As a result, cellular-service providers can charge for music and video content—as well as for airtime—as subscribers download their favorite media using the network.
As mobile handsets continue to integrate features like higher-resolution digital cameras, PMPs, and PDA functionality, subscribers need a convenient method of transferring files to and from the handset and their PCs. The ubiquity of USB—the standard method of data transfer employed in MP3 players (PMPs), digital still cameras (DSCs), flash drives, hard disk drives, etc.—makes it a top candidate.
FULL SPEED VS HIGH SPEED USB
Today, most handsets support full-speed (FS) USB (12 Mbits/s), which is sufficient for small data transfers like address-books. By adding MP3 players and high-resolution digital cameras, FS USB can no longer cut it, considering the large amount of data that now need to be moved between the handset and PC. Consumers are spoiled by the high-speed (HS) USB 480-Mbit/s transfers they enjoy with their dedicated PMPs and DSCs. Having to resort to full-speed transfers when transferring MP3 and picture files will prove to be a disappointing experience.
Compare the difference in device-to-PC data transfer between a HS USB and a FS USB, It takes the high-speed USB device approximately 33 seconds to transfer 105 Mbytes of data from the host PC to the handheld device. The full-speed USB device, on the other hand, takes almost 13 minutes to perform the same transfer.
With cutting-edge, flash-based handheld devices currently supporting up to 8 Gbytes of data storage, it could take over 17 hours using full-speed, versus 44 minutes using high-speed USB to transfer that amount of data. Cutting-edge, hard-drive-based handheld devices support 80 Gbytes; therefore, the transfer time would increase by 10 times to 170 hours using full-speed versus 7.3 hours (440 minutes) using high-speed USB.
Today’s handsets use FS USB for a variety of reasons, including diagnostics and manufacturing testability, as well as modem connectivity. The former affords the handset OEM a convenient method of testing handsets on the manufacturing line to ensure quality, thereby minimizing or eliminating field failures. FS USB’s bandwidth is more than sufficient for these tasks. The latter provides a means for subscribers to use their phones as a modem when connected to a laptop PC to provide wireless Internet access. FS USB provides up to 12 Mbits/s of bandwidth, which is enough (at least in theory) to support existing 2G data standards, such as GSM’s GPRS and EGDE, and CDMA’s 1xEV-DO and 1xEV-DO Rev. A, as well as emerging 3G standards like HSDPA and HSUPA.
One key design issue when upgrading features to support HS USB is that one must abandon field-proven software and come up with a whole new software suite. This takes time and resources, two factors that are in limited supply in the fast-moving handset market.
Because FS USB provides enough bandwidth for these functions, handset OEMs are more inclined to keep the existing approach and simply add HS USB support in the form of a HS USB controller or a PHY, to the design. This adds a high-bandwidth pipe for mass storage requirements only, thereby enabling a better consumer experience with respect to integrated PMP and DSC functionality. Such an approach also enables existing platforms to be more simply upgraded to HS USB as needed by adding this support alongside the FS USB. By introducing HS USB in two stages, OEMs can bring HS USB connectivity for multimedia data transfers to market much quicker than would be the case with a complete redesign around HS USB.
Another reason to add HS USB in this fashion is the limited number of endpoints in current HS USB controllers. In PC applications, HS USB controllers generally have a specific application defined and require a small number of endpoints (i.e., as few as 4 or 8 endpoints are sufficient for most applications). Mobile handsets use USB to provide many functions, so the demand for endpoints increases dramatically—12 or 16 or even upwards of 20 endpoints have been suggested. Examples of functions supported by mobile handsets (each requiring single or multiple endpoints) include: Mass Storage, Media Transfer Protocol (MTP), Modem (CDC), Device Management, Object Exchange (OBEX) and Debug/Test. For this reason, handset designers can effectively support more endpoints using both FS and HS USB datapaths than is possible with a HS USB datapath alone.
FULL-SPEED AND HIGH-SPEED TOGETHER
So how do you add HS USB functionality to an existing handset OEM design that supports FS USB? Obviously, it can’t simply be added as a separate entity. Otherwise you’d need two mini- or mirco-USB connectors on the handset, which would only serve to increase cost and confusion among consumers. The most cost-effective approach, which also reduces subscriber confusion, is to merge the two USB data pipes together onto a single connector (Fig. 1).
Every designer who has designed with high-speed signals knows that while you might get away with the FS link operating sufficiently, the HS link will never work. That’s because the FS traces act as a stub and antenna to the HS transmission lines, causing a severe degradation in signal quality and a closing of the signal eye. Such a configuration also assumes that both the HS and FS USB outputs can support some kind of tri-state mode to achieve this design (i.e., the HS signals are tri-stated while FS is in operation, and vice versa). This is something that most USB devices don’t support today—such is also the case in traditional applications such as PCs. With PCs, multiple connectors are the norm, so there’s never a need to merge multiple USB signals onto a single connector.
ISOLATING FULL SPEED AND HIGH SPEED SIGNAL TRACES
Engineers must, therefore, completely isolate FS and HS USB signal traces. Today, the best approach toward bringing HS USB to mobile handsets is to add the HS path and multiplex it with the existing FS path via a signal switch (Fig. 2).
While this seems simple enough, it can actually cause some real issues when it comes to HS signal integrity. In fact, it may cause failures in USB compliance testing. Even though there are switches on the market intended specifically for HS USB applications, they will diminish the quality of the eye to some extent, and in some cases, to the point of failing compliance. Many things must be considered when selecting a switch as well as when laying out a board, but first consider the ideal HS USB datapath—one without a switch at all.
When looking at a HS USB datapath, the board designer has control over several aspects that must be optimized to create a clean eye diagram. First, the trace impedance of the D+ and D- lines must be 45 Ω to match the internal impedance seen at the input of the receiving device’s D+ and D- pins. This creates the appropriate voltage divider to give a compliant HS logic HIGH of 400 mV. Another aspect is the trace-length matching of the D+ and D- traces. Provided there are no other complications, for example ESD or EMI protection devices, this should provide a clean eye diagram (Fig. 3).
When the switch is inserted into the datapath, distortion will occur. What type of distortion, and to what extent, depends on the characteristics of the switch. The first step is to look for the switching speed of the switch. The switch must be able to handle switching at 480 Mbits/s (which is equivalent to 240 MHz) to be compliant with HS USB. If it’s not, the switch should not be considered as an option. Most likely, if the switch states that it’s intended for HS USB use, this won’t be an issue.
The next characteristic, which is probably the most important albeit overlooked, is the series resistance (RON) of the switch. The higher the series resistance, the more the eye gets squashed. This ends up being the biggest headache when attempting to gain USB-IF certification.
Consider the following example of how a higher series resistance can affect the eye diagram. Let’s say switch A has a typical series resistance of 5 Ω and switch B has a typical series resistance of 10 Ω. In the case of switch A, the total trace series resistance would be 50 Ω instead of 45 Ω. When doing a simple voltage divider, this gives a logic HIGH of 379 mV, instead of the required 400 mV. The spec provides a 10% tolerance to the 400-mV requirement, so a 360-mV logic HIGH is still within spec. When switch B is inserted, it adds 10 Ω of series resistance, giving a total of 55 Ω of trace impedance. This results in a logic HIGH of 360 mV, leaving no margin for error. It’s unrealistic to expect this to be compliant, considering that additional inaccuracies exist in the termination resistors as well as in the trace impedance.
The lower left eye diagram in Figure 3 has had the trace series resistance increased by 10 Ω in the signal path. Notice that the upper and lower boundaries have been squashed due to the added trace series resistance. This is a passing eye diagram; however, there is far less margin for error.
Even if the voltage levels fall into the acceptable range with the added switch resistance—and after taking into account the various tolerances—switches can affect the eye diagram in another way. The switch will also add capacitance to the traces, which slows down the edge rates (rising and falling edges). This can result in the corner of the eye diagram keep-out area getting clipped, causing the eye to fail.
For example, say that switch A has 5 pF of capacitance and switch B has 15 pF of capacitance. The area around the keep-out area of the eye (the margin) will reduce by as much as 50% with the additional 10 pF of capacitance that switch B adds over switch A. Today, typical switches have between 6 and 15 pF of capacitance when the switch is on. The top right eye diagram in Figure 3 illustrates an eye diagram that’s been distorted by adding 15 pF of capacitance.
If the switch only added series resistance or capacitance, chances are it would not cause any issues. However, the reality of the situation is that a switch adds both, and this combination can cause real problems with the eye diagram. Let’s say the most ideal switch has a low series resistance and capacitance. The low series resistance will cause the top and bottom levels to move closer toward the center of the eye, giving less of a margin for error. The capacitance will make the transitions slower, cutting into the eye keep-out area and ultimately causing the HS USB signal-integrity tests to fail. This is shown in lower right eye diagram of Figure 3.
Picking a switch with low mix of RON and CON characteristics becomes essential to achieve a successful switch-based design.
Another factor to consider in such a design is knowing when to switch between the FS and HS USB paths. Today, this is mostly done in software—subscribers must, for example, manually select whether they want to use Mass Storage or Modem mode. The system processor (baseband or applications) then enables the correct signal path. The default mode is usually FS USB mode, since it’s the mode used for diagnostic and manufacturing testing at the factory. The whole approach is cumbersome and highly undesirable, though, because it complicates ease-of-use and potentially will lead to increased support calls. Thus, over time, handset designers will want this controlled without any user intervention, which will mean moving to a fully converged architecture.
In time, it’s certain that handset designers will make the full migration to support a single USB path, which will allow FS and HS USB to coexist. Time will be spent optimizing the software for this solution, thereby creating more elegant and optimized designs. Product architectures will arise which integrate a sufficient number of endpoints to support handset applications. Until then, handset designers wishing to support HS USB today, as well as quickly bring to market products to support a satisfying consumer experience, will rely on implementing HS USB in stages.
In any case, for such a design to be successful, designers must consider the RON and CON parameters of the switches they choose. Following these guidelines will prevent from having to spend time debugging the USB connection, and allow mobile-handset designers to provide HS USB functionality to the market quickly. As a result, subscribers can use their handsets as they would any of their consumer-based multimedia functions.