At the beginning of December in Paris I had the opportunity to make a presentation to a very impressive audience, technical gurus from companies contributing to MIPI Alliance specification were here, including ST-Microelectronics, Intel, Qualcomm, TI, Toshiba, Nokia, Samsung, to name a few.
The goal of my presentation was to review the adoption level of the main MIPI Alliance Specifications. If you take a look at the organization chart below, you will see that close to 15 specification working groups are active (the one group that is not reflected in this diagram is the Sensor group).
When you buy a smartphone or other mobile device, you probably don’t know that various MIPI Alliance specifications are integrated in several ICs – from the Application Processor to the Image Sensor, as well as the RFIC or Display Controller. When examining a smartphone, you will not see which MIPI specifications are used, unlike with USB or HDMI, or even with SATA and PCI Express if you look inside a PC. MIPI Alliance specifications are defined as Chip to Chip, and the only way to assert which MIPI specification is used in a mobile device is to look at the block diagram of the various IC integrated in the system, if the IC chip maker publicly released it… but this is not always the case.
Another approach is to monitor the MIPI Alliance IP sales generated the previous year. IPnest has done this job since 2011, we have observed that MIPI IP license revenues have seen a fourfold growth between 2010 and 2012, to reach $20M, attracting many IP vendors, including Cadence and Synopsys. In the mobile segment, known to be extremely competitive, the key players, and certainly the Application Processor chip makers, tend to use internally designed functions rather than sourcing IP externally (think about Broadcom, Qualcomm, STM, Intel etc.). IP sales growth is a good indicator of MIPI Alliance specification usage, as well as obtaining information directly from chip makers. If I tell you that the #1 chip maker in mobile wireless segment, Qualcomm, is also the chip maker integrating the greatest number of MIPI specifications, it helps us understanding that the MIPI specification adoption rate is high.
My goal is to review the adoption status for:
- Camera (CSI-2, CSI-3), Display (DSI, DSI-2)
- Chip to Chip specifications: Low Latency Interface (LLI), Super Speed Integrated Circuit (SSIC), Mobile Express (M-PCIe) and UniPort (UniPro + M-PHY)
- Universal Flash Storage (UFS) and Modem to AP chip to chip (DigRFv3, DigRFv4)
Our goal was also to evaluate the adoption for mobile specific specs like RF Front End (RFFE), Battery Interface (BIF) or Serial low power inter-chip Media Bus (SlimBus), some of these specifications are not necessarily complex enough to attract IP vendors.
The first high speed serial specification (DSI, CSI-2 and DigRF v3) was released to operate with a PHYsical layer running up to 1.5 Gbps per lane, the D-PHY. The M-PHY was defined later, operating at 1.5 Gbps (gear 1), 3 Gbps (gear 2) and 6 Gbps (gear 3), and to be used with the following controllers:
- CSI-3 + UniPro
- DSI-2 + UniPro
- USB 3.0
- PCI Express
- DigRF v4
Review of the adoption status means checking if a certain MIPI Alliance specification is widely used within the industry, and if it’s not the case, what can be the reason? The first reason can be technology related: trying to implement M-PHY on a 90nm technology node can be a show stopper. The second reason can be due to a specific market, were using a different protocol standard other than MIPI is more straightforward (for example DisplayPort in the PC market).
Camera Serial Interface
CSI implementation is complete in the wireless mobile segment, and it’s almost impossible to find any Application Processor (AP) not supporting CSI. This is true for CSI-2, used with D-PHY. If we try to identify any AP supporting CSI-3, we can’t find the information about released AP supporting the specification. There could be several reasons why CSI-3 is not yet implemented:
- Too early: CSI-3 is planned to be integrated in the future on AP not yet released.
- Technology limitation for the CMOS image censor IC chip maker: implementing M-PHY on a mature technology node could be an issue. Then, using more lanes of D-PHY (or higher baud rate) can help avoiding implementing M-PHY.
Even if moving to CSI-3 will take longer than expected, CSI as a specification is widely used, and here to stay. Moreover, the pervasion of MIPI CSI in Automotive or Consumer Electronic is likely to happen. (See: See new companies list on MIPI web site).
Display Serial Interface
DSI implementation is also massive in the wireless mobile segment, once again DSI rather than DSI-2. The Display market is different from the image sensor market, as other protocols are available, like DisplayPort, MHL or HDMI for external display interface or embeddedDisplayPort (eDP) or even LVDS for internal. Thanks to a large MIPI Ecosystem (IP vendors, VIP vendors, Test equipment, etc.), DSI is dominant in the wireless mobile segment. Because the need for higher bandwidth (to support even higher HD format) is more crucial than for camera, we could see a faster adoption for DSI-2. There is a drawback for using DSI-2, linked to the Display controller technology: the chip form factor (long rectangle) makes it more difficult to implement M-PHY, in fact a topology issue. Nevertheless, Application Processor chip makers plan to move to MIPI specification post a DSI solution.
Can we expect the pervasion of MIPI DSI into other market segments? There are certainly opportunities to implement DSI into Industrial, Medical or Automotive applications, even if PC like applications may tend to use DisplayPort or eDP protocol.
Chip to Chip Communication Protocol
2013 was too early to see potential adoption for SSIC or M-PCIe, even if IP vendors reported a strong interest from the market to know more about these new protocols. The communication link between the AP and the Modem, or with a combo chip integrating WLAN, WiGig, etc., has been supported by HSIC in the past (2007-2011) or by Low Latency Interface (LLI) until now. According to our research with IP vendors, UniPort, based on UniPro + M-PHY, has not seen a demand which can be comparable to LLI, nevertheless UniPort may have been internally developed by chip makers, rather than externally sourced.
Because chip-to-chip communication is a strong need, not only within the wireless mobile segment, we expect specification like SSIC, M-PCIe and UniPort to see adoption in the near future. If you want to forecast which will be dominant out of these above mentioned protocols, the answer will be linked with their respective ecosystem: a protocol supported by a wider ecosystem (including the S/W installed base) may be preferred to a smarter- or more technically efficient protocol. In other words, a company being in the past active in the PC or PC peripheral segments may prefer using M-PCIe or SSIC, simply because this company can benefit from a strong expertize of these protocols and tend to reuse existing hardware IP or software pieces, instead of learning a new protocol. On the other hand, a seasoned chip maker developing ICs for the wireless mobile segment may use a chip-to-chip MIPI protocol…
Universal Flash Storage
UFS specification is based on UFS (Host or Device) + UniPro + M-PHY, and used to interface with Flash memory device within a smartphone. UFS is expected to be the widely adopted specification based on UniPro and M-PHY to be. In fact, UFS is the only MIPI Alliance specification that supports Mass storage; an Application Processor chip maker as no other choice than using eMMC (non MIPI) or UFS.
To be clear: the word “commodity” is absolutely not derogatory term. The set of specifications like Battery Interface (BIF), RF Front End (RFFE), System Power management Interface (SPMI), Debug and so on, are certainly filling a gap. Nevertheless, these specifications are mobile phone specific, and not complex enough to justify a third party like an IP vendor to invest to develop and market it.
MIPI Ecosystem: large and growing!
When analyzing the MIPI Alliance Ecosystem, we see that adoption for some of the specifications like CSI-2 or DSI is nearing 100%, and the adoption for other specifications like UFS is growing fast. If you look at the above picture ( Source: “MIPI Ecosystem Survey” 2012), you realize that a large ecosystem has been built, made of Design and Verification IP vendors, test equipment or test service providers, with maybe some weakness in the “Software” category. The existence of such an Ecosystem is vital: when an EDA vendor like Cadence or Synopsys invests to develop Design and Verification IP, this vendor is becoming a MIPI Alliance promoter, helping the standard to be widely known and later used. The fact that the wireless mobile chip maker leader, Qualcomm, is also the company claiming to integrate the most MIPI specifications is a very strong driver for MIPI Alliance specification adoption!
Taking a look at MIPI Alliance’s web site under New Members, you will see that more than 50 new companies have joined the Alliance between May and October 2013. New member companies are coming from Asia, North America and Europe, and active in various market segments, including Consumer Electronics, Medical or Industrial segments. If you consider that, in July 2012, the MIPI Alliance was counting 235 companies, 50 new companies joining in just five months is a huge accomplishment.
What else could be done to increase MIPI Alliance adoption of specifications? MIPI Alliance events organized around the world are restricted to MIPI members only, so the Alliance can be perceived (by non-members) as working in a close system… One solution could be to host a MIPI Alliance world tour, similar to PCI-SIG at the emergence of PCI Express, and see if it make sense to open MIPI Alliance meetings to non-members.
Previously, the MIPI Alliance specifications was perceived as very complex, and some people are still asking to IP vendors if they can provide them with “MIPI”: they have heard about MIPI but don’t realize that the word is synonymous with a complete set of technical specifications. Perhaps an evangelization campaign could help educate the part of the industry that may need it. To support such a campaign, the creation of white papers and the dissemination of contributed articles in the press would certainly help.