NXP embraces 28nm FDSOI for MCUs
Lees was speaking at an event to intended to emphasize that the enlarged NXP, after its merger with Freescale, is well place to offer processers, security and software for the Internet of Things in all of its various vertical applications, whether in the consumer, industrial, medical or automotives sectors.
Part of that emphasis is a mobile truck loaded with "show-and-tell" displays that will be driving around events in Europe and making its way to Embedded World, in Nuremburg, Germany at the end of February.
Truck-based exhibition of NXP expertise in IoT
Lees helped build up the microncontroller business at NXP as an early adopter of Cortex-M cores from ARM before changing jobs in 2012 and moving to Freescale to take on a similar task. However, while at Freescale Lees has also been exposed to higher order networking, communications and multimedia processors in the i.MX series and the decision to adopt Samsung’s 28nm FDSOI process.
With the merger of NXP and Freescale Lees has brought an enthusiasm FDSOI into the enlarged company
Prior to the merger Freescale had made the decision that while its i.MX6 series SoC processors were implemented in 40nm bulk planar CMOS the i.MX7 series would be made in 28nm FDSOI (see Freescale, Cisco, Ciena give nod to FDSOI).
Lees said that i.MX8 and i.MX9 would also likely be implemented on FDSOI. However, while these devices are likely to press forward on performance – i.MX8 is set to be based on ARMv8 architectures as implemented Cortex-5X series cores – Lees sees great value for FDSOI at the lower end of performance in IoT applications where the workload may not be high but the energy efficiency and cost are paramount.
28nm FDSOI is the last simple manufacturing node and will be long lived, said Geoff Lees, gm of the MCU business unit at Freescale.
Lees said he would be encouraging the LPC microcontroller part of NXP’s business – MCUs based on Cortex-M series cores – to also adopt 28nm FDSOI. "The 28nm FDSOI is set to be the last simple node to manufacture," said Lees adding that it would be extremely long-lived, stable and, ultimately, low-cost because of that.
One of the reasons Lees favors FDSOI for microcontrollers is the good support provided for analog functions. And in IoT applications and microcontrollers more generally the chips are essentially mixed-signal devices.
"FDSOI is a proven technology that can provide the performance for security in hardware for IoT on 28nm node. It offers better temperature control and lower leakage current compared with bulk CMOS at 28nm," said Lees. "I believe all MCU vendors could move to FDSOI," he added.
Next: Low voltage, non-volatile memory
It is also well suited to low-voltage operation. Being able to scale the voltage right down to near the transistors’ threshold voltages is a boon to energy saving as power consumption varies with the square of the voltage. "28FDSOI has the option to run sub-threshold," said Lees. "Right now we tend to run the FDOSI process at a nominal 0.85V but STMicroelectronics has done a lot of work at 0.5V," he added.
However, such near-threshold operation is not exclusive to FDSOI. Foundry TSMC is offering ultra-low power (ULP) versions of its bulk CMOS processes that operate below 1V at 55, 40 and 28nm. The 16FFC variant of its FinFET process operates at down to 0.55V.
There is also likely to be competition between foundries to provide embedded non-volatile memory at 28nm. Lees reckons its two to three years away for FDSOI with options of perpendicular magnetic RAM and some form of resistive RAM as the most likely to adopted. "They have the advantage of being back-end of line NVMs. We don’t want to change the general transistor structures."
Lees added that there are nine customers with FDSOI designs in manufacture or preparation at Samsung according to his sources. Publicly declared customers include STMicroelectronics, NXP, Cisco, Ciena and Sony. It is possible that Samsung is also fabricating circuits for its own use.
Related links and articles:
News articles:
Freescale, Cisco, Ciena give nod to FDSOI
Samsung is running 28nm FDSOI chip process
Globalfoundries launches own FDSOI processes
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
