By Axel Streicher
I remember a time — not too many years ago — when the idea of “16-bit is dead in body electronics” was widely promoted in the automotive industry. Or should I say some would have liked 16-bit to be dead? Following the “more is better” theme that we take for granted in the PC industry, 32-bit microcontrollers seemed to be the solution to all problems. But as they say, not all that glitters is gold!
In June, Strategy Analytics released its “Automotive Semiconductor Demand Forecast,” which suggests the global demand for 16-bit microcontrollers will grow by 52% from 2009-2014, representing a market value of close to $400M. In the same year, 32-bit microcontrollers will represent a market of approximately $360M. So, who said 16-bit is dead?
Where will all of these 16-bit controllers go? The majority of that growth will come from emerging markets such as China and India where the “affordable car” segment is driving low system cost rather than more and more performance requirements. Typically, 16-bit microcontroller designs allow for better code density, more cost-efficient power supplies, lower cost development tools, etc. while still meeting “Zero Defect” quality requirements.
| Freescale is a global leader in 16-bit automotive microcontrollers, and has just announced the S12-core based 16-bit MC9S12G family of microcontrollers, which is focused on low cost, high performance and low pin count for automotive applications. Supporting standard body network interfaces such as CAN and LIN, the S12G family offers up to 128kB of on-chip Flash memory, up to 8kB on-chip SRAM and 25MHz bus frequency.
Get the details on the S12G: 16-bit Microcontroller for Body Electronics Applications on the Freescale automotive web site. You can also download the S12G Family Fact Sheet. |
|
Via RSS
13 Comments
Ummm…Freescale did when they left 16 bit out of the “controller continuum”.
The 16-bit S12 device families are focused on the specific requirements of the automotive market, so is the S12G family that we announced recently. The Controller Continuum, on the other hand, enables engineers to scale between 8-bit and 32-bit for industrial applications, with our Flexis MCUs providing compatible device duos. In industrial, 16-bit is key for motor control applications and power management and we continue to invest in that product line. In fact, Freescale just recently announced the MC56F82xx DSC for power conversion.
Well, since you included a link to this article again, I will comment again.
Why can’t a 32 bit processor do the job of a 16 bit and do it better? I say better because with the S12 line, you need to play paging games (memory models) once you go above 64K.
It’s not like the the S12 is inexpensive.
I think the answer is that S12 customers have working code that is reliable, and just don’t want to re-code.
Technology has a zombie effect, and while 16 bit IS dead, it still walks.
I agree with JimD. The S12 isn’t much more than a HC11 on steroids. The HC11 was great in its day – but that was a quarter century ago.
There are a lot of low cost 32 bit MCU’s available now. Freescale’s Coldfire V1′s, the new Kinetis range etc – not to mention the many offerings from other vendors.
If 16 bit isn’t dead it should be.
PaulK
Wow! Jim and Paul better tell Microchip that their huge success based on 8 and 16 BIT mcus is imaginary and they should just pack it up!
The lack of interest in the S12′s ability to process multi valued logic demonstrates that the Galileo effect is alive and well in the church of western engineering. Our earth revolves around T and F! In order to get more T and F we need more bits! More bits and more T and F! Never enough bits!!!!
Groan…..
Well, I don’t think that we mentioned 8-bit, which is Microchips forte.
However, even Microchip is pushing 32 bit, and for a very good reason. It is not uncommon to have SD card support and even TCP/IP support, not to mention an RTOS in a bare metal app, and these are made more effective with more than 16 bits of address space. If you are doing software, you would instantly understand that not messing around with paging and memory models to get past a 16 bit address space will save money in software development, usually the largest budget item.
I mostly do new designs, and unless it is truly a toaster application, in which case I would use an 8 bit, I would not consider a 16 bit. I guessing that you have not been confronted with the issue of “memory models” lately, nor remember the wonder of going to a 32 bit flat address model.
In the Freescale world, I don’t see much if any discussion of new designs using 16 bit.
And yes, 32 bits IS enough, and 16 bits is NOT enough.
It’s not just an arbitrary whim or a marketing thang, it’s a real need to address more than 64 K. Also consider that 32K of SDRAM on board is becoming common – further squeezing the 64K address space.
So that is why the 16 bit is dead to a large market segment of the bare metal world.
And if they are not dead, please kill them now.
16 bit is not dead. When it comes to reliable issues, you have to (cyclically) check you whole hardware for integrity, correct configuration, correct operation etc.
Do this with a 32 bit machine!
OK – 8 bitters are nowadays for the toasters. 32 bitters are for the luxury generation that does not want to bother with code efficiency, optimized algorithms and so on. 16 bitters are for the tough guys – an endangered species
There is a real need in high-volume, mid-size applications (eg. anything using an EC motor of any control principle) where 16 bitters are the perfect fit: fast enough, enough RAM and ROM, reasonable prize. Though – I’m not speaking of the S12 but of the so-called DSC class – MC56*, dsPIC and TI *C2000. Though TI thinks that 32 bits would be fine. But that’s another story.
I could probably live without the S12 (I had limited fun supporting a new safety-related application during the last years), but really not without the DSCs. 32 bitters of any kind are either fast, but not applicable for the extended temperature range, or not faster than a DSC of the current generation.
Yes, in the last five years, 32-bit controllers have seen a big automotive push for engine management, advanced safety systems, infotainment, body control modules, gateways – the domain of 32-bit micros, no doubt. Only last week at Electronica in Munich, Germany, Freescale launched the Qorivva family: A broad range of 32-bit controllers for automotive, based on Power Architecture cores and 55nm CMOS technology. However, let us not overlook the huge world of applications where small footprint, fine Flash granularity, lowest power, low cost development systems are key: peripheral body applications for example, where ever increasing frequencies, megabytes of Flash, FlexRay or Ethernet are not on the requirement list! This is what the 16-bit S12G family is all about.
Well, I’m not sure that s12 will win on price vs flexis. I’m not a corporate purchasing agent, and I don’t buy parts in these huge quanites.
It certainly will over PPC.
There is another thing to consider – using the same family of MCU’s over your product line has an advantage in terms of both purchasing and development.
Even if it is slight overkill in some appications, it could still work out to be more economical in the long run on both counts.
What more, flexis has a much broader range of flash granularity and foot prints. Whats more, the same development skill can be leveraged to the equivelent 8-bit parts as well. With a bit of forethought (which CodeWarrior enables)the exact same code can run on both familys (no, I dont’ work for freecale! but I do use flexis and it is a wonderful thing that you can do this)
ARM Cortex M0 parts are 32 processors with a 16 bit instruction set. They are available in reduced pin count packages for under $1. To my understanding the M0 processor core requires less transistors than an 8051 core.
Explain again why I would ever want to select a 16 bit processor for a new design? If you have an extensive code base you need to maintain then sure, but otherwise what benefit does a 16 processor give you over an inexpensive 32 bit ARM processor?
I am really looking forward to using the NXP LPC43XX parts in my new designs. These combine an M4 core(DSP extentions) with an M0 core on a single die. With a full complement of peripherals and lots of memory these part are supposed to sell for under $10. This is a great way to partition up a signal processing/machine control application. Let the M0 deal with the outside world while the M4 does all the serious number crunching.
http://ics.nxp.com/products/lpc4000/lpc43xx/
Hi Phil, Freescale ships more than 100 million per year of our S12 16-bit microcontrollers to the automotive market. There is an extensive code base and hundreds of “development years” of experience with our global customers.
The automotive industry thinks and lives in very different development and product life cycles than other markets. The S12 16-bit microcontrollers are part of the Freescale Product Longevity program and will be available for a minimum of 15 years after product launch, meeting the extended supply requirements of the automotive industry.
Also, many automotive systems have not changed much over the years. Zero failure operation and car maker approval have highest priority, paired with the demand for cost reduction from generation to generation. And so, there are more factors in the equation than only the 16-bit or 32-bit question – we have not yet even talked about 8-bit controllers!
No doubt, many automotive applications see the need for more performance, more Flash, more communication interfaces, multicore MCUs, embedded graphics, etc. There are more details on our automotive website that show how Freescale addresses these requirements with the growing Qorivva families of 32-bit microcontrollers based on Power Architecture or the ARM based i.MX application processor family of devices.
Hope this helps. Let me know if you need additional info.
Cheers,
Axel
Axel,
So you see, you are really agreeing with those who say 16 bit is dead.
Sure, with a huge legacy code base you will sell 100M units, but that says little about the future or new designs.
Let me quote the Freescale web site:
“With this announcement, plans call for every NASCAR Sprint Cup Series race team to use a control system with Freescale’s advanced 32-bit Power Architecture® based engine management processors at its core, beginning in 2012.These same processors power millions of today’s most energy efficient engines.”
Didn’t hear anything about 16 bit here. Sounds to me like it’s Freescale saying 16 bit is dead.
8 bit is a different story. Low end applications will still be using 8 bit fore quite some time.
But we are discussing 16 bit here.
Other than address range, there is very little technical reason to be concerned with the bit-ness of an MCU. This is especially true when considering a 32 bit MCU for an application previously/typically handled by an 8 or 16 bit MCU. Bit-ness is becoming as meaningless as MIPS. There is so much overlap in the performance of available MCU devices on the market today that for the majority of applications the decision tends to be based on the integration of peripherals to match the application and possibly power consumption (again, depending on the application). Features like DMA and automated peripheral inter-operation have crept into the lowly 8 bit and 16 bit MCU offerings giving them as good real time performance than many 32 MCU’s for ‘simple’ applications. This is a general statement and there are always going to be exceptions on either side, but think as a broad, general statement: bit-ness is meaningless.
2 Trackbacks
[...] timing of this inquiry also coincides with Axel Streicher’s article asking “Who said 16-bit is dead?” Axel makes a similar observation about 16-bit processors. I would have liked to have seen him [...]
[...] timing of this inquiry also coincides with Axel Streicher’s article asking “Who said 16-bit is dead?” Axel makes a similar observation about 16-bit processors. I would have liked to have seen [...]