I recently spoke to Ken Foust from Intel regarding his being the chair of the MIPI Sensor Working Group and the important standards work they are doing there. I also had a conversation with Peter Lefkin, the Managing Director of the MIPI Alliance about I3C.
Foust had worked closely with the MEMS Industry Group (MIG) to survey that industry regarding their thoughts on sensor interface and problems they are having in that area. Companies such as Intel, Qualcom, NXP and TI gave their valuable comments and all of them had the same “pain” points.
Lefkin said that they wanted a new interface for sensors that would be backwards compatible with I2C and SPI buses. “Mobile” products were the main concern. These are automobile communications, wearables, smartphones, tablets and touchscreens and the like. See Figure 1.
MIPI interfaces are used in mobile platforms. The red arrow and circle in the diagram refers to the fact that MIPI Sensor Interface Naming is under final development. (Image courtesy of MIPI Alliance)
The I3C interface was conceived as a possible solution, but there were “pain” points associated with that concept as well. Power and cost were a concern with existing interfaces and would be a problem here too since I3C is conceived as a combination of I2C and SPI . Buffering would need to be built in as well as the capability of storing data while the processor sleeps. I2C is fast and handles this by waking up and goes back to sleep when done, but I2C and SPI are typically best used for clearing large FIFOs and not optimum for sensors. Increased bandwidth was also needed beyond the I2C 400kHz/1MHz going up to 10 to 30 MHz.
Regarding cost, the sensor would be typically off on its own and then get an interrupt on the GPIO with the way today’s Apps Processors function---so more pins are required. Pin requirements are important since high tier mobile SoCs have 10 to 20 I/Os. Most sensors using I2C require one or more interrupts; SPI requires an interrupt and a chip select; some sensors require a dedicated “sleep” signal. See Figure 2.
Too many I/Os and fragmented interfaces are a problem with today’s solutions. (Image courtesy of MIPI Alliance)
This diagram shows what the MEMS industry wants/needs. (Image courtesy of MIPI Alliance)
So the MIPI Group is off with the following goals in mind by 1Q2015:
Define a standardized Sensor Interface
- Consistent within the scope of the MIPI Alliance
Strive to re-use an existing industry interface (in whole or in part)
Flexible and able to evolve
- Enable device interface compatibility between MIPI member companies
Interface features focus on:
- Provide value to adopters and focus on Sensor Vendors
- Anticipate future architectures for mobile and complimentary industry needs
- Unify a fragmented interface industry, in and out of mobile
- Low cost implementation on the sensor side (< 2000 gates)
- Reduced signal count interface
- In-band interrupts
- In-band command codes
- Reduced interface power consumption – support push/pull IO
- Reconfigurable systems – support dynamic addressing
- Increased bandwidth > 10Mbps – Newer Requirement
- Implemented via supported HDR (High Data Rate) Modes
- Data, transcoded ternary symbols for example, transmitted on rising and falling SCL/SDA
Let’s keep an eye on this development because it will be critical to the hopeful ramping up of billions of sensors in the IoT going forward in a cost–effective and expedient way.