Freescale to stop developing 8 and 16 bit micros

[bombledmonk]

Ran across this and thought it sounded like a big deal..... thoughts?

Freescale is to finsh development of its 8bit and 16bit microcontrollers with a pin compatible 32bit version is coming into production for the same cost.

    Freescale will be shipping its 32bit Kinetis L series microcontrollers using the ARM M0+ Flycatcher core for a little as 30¢ in volume later this year with the same pinout as the 8bit devices. The only area where 8bit development continues is for EMC and ESD protected devices, and there are already test versions of the M0+ devices.

"Once you have an M0+ that's 100% compatible then it doesn't make sense to do antoerh 8bit device and that will be in the next 6 to 9 months," said Geoff Lees, general manager of the industrial and multi-market microcontroller division at Freescale. "For standard devices I can't see any rationale for 8 and 16bit development continuing."

"The products will continue in manufacturing for a long time but I can't see us doing any more serious product development," he said.

Freescale is aiming to move customers over to the 32bit versions with a 'lite' version of the MQX real time operating system. This currently runs in flash but a ROM version is in development that will allow the smalled 8bit-compatible devices to run an RTOS and still have room for applicaiton code.

"The ecosystem is the drving factor," said Lees. "32bit open source software and supprting productivity tools are more an dmore tuned for the ARM IDE development environment. It's going to be very hard for 8bit tool vendros to keep the tools economically up to date."

http://www.digikey.com/techxchange/docs/DOC-2294

[westfw]

Are they going to do an 8pin DIP CM0+ ?  Other DIPs?  That'd be what 'pin compatible" means, wouldn't it?

[andersm]

It's not too long ago that they introduced their ColdFire V1 line, which was also marketed as an 8/16-bit migration path. I guess it's only time until they drop all of their other architectures and become ARM-only.

[free_electron]

Are they going to do an 8pin DIP CM0+ ?  Other DIPs?  That'd be what 'pin compatible" means, wouldn't it?

DIP ?
Dinosaur Invented Package ?
Dead IC Package.

DIP is dead. give it up.

[poptones]

Let's see your surface mount prototyping board.

[free_electron]

Plenty of SO and QFP to dip dapters around. And yes, there SMT prototype boards.

http://www.boardworxsystem.com/

I rarely breadboard anything unless it's some analog muck that needs verification. Straight to PCB.

[mikeselectricstuff]

I'm not bothered about DIP, but often need 5V operation, which is often another casualty of going to smaller devices. Even some of the newer 8 bit PICs use a low core voltage, but hide this to the outside 5V speaking world.

[poptones]

Twelve bucks for a prototyping adapter? yeah right.

[free_electron]

I'm not bothered about DIP, but often need 5V operation, which is often another casualty of going to smaller devices. Even some of the newer 8 bit PICs use a low core voltage, but hide this to the outside 5V speaking world.

That is indeed another casualty... 3.3 volt is the new norm . With device cores running at 1.8,1.5,1.2 and below an i/o rings for fast applications ( memory) running at 2.5.
Also opamps in the rail to rail category rarely go above +- 5 volts working range.

All classic logic is available in 3.3 volts versions ( 74xx and 4xxx series ) and most fpga are 3.3 volts only.
And many other popular 5 volts devices have been ported to a low voltage version. MAX3232 for example. 3.3 volt version of the MAX232

No need to stick to 5 volts. and for the odd circumstance where you can't get around it there are level shifters.

[mikeselectricstuff]

I'm not bothered about DIP, but often need 5V operation, which is often another casualty of going to smaller devices. Even some of the newer 8 bit PICs use a low core voltage, but hide this to the outside 5V speaking world.

That is indeed another casualty... 3.3 volt is the new norm . With device cores running at 1.8,1.5,1.2 and below an i/o rings for fast applications ( memory) running at 2.5.
Also opamps in the rail to rail category rarely go above +- 5 volts working range.

All classic logic is available in 3.3 volts versions ( 74xx and 4xxx series ) and most fpga are 3.3 volts only.
And many other popular 5 volts devices have been ported to a low voltage version. MAX3232 for example. 3.3 volt version of the MAX232

No need to stick to 5 volts. and for the odd circumstance where you can't get around it there are level shifters.

I often have a need for 5V - e.g. when driving a crapload of LEDs - I typically need 4-5V for the LEDs anyway, so 5V everything saves parts.
MOSFETs also switch on harder at 5V. When you have amps, or even tens of amps flying around, the extra noise immunity of 5V doesn't hurt.
I do often use 3V3 parts as well (most commonly whne I run out of RAM or speed on an ATMEGA) ,  but there are often situations where 5V just makes more sense.
It seems a shame that even on parts that use a LV core and have voltage regulation built in, that regulation couldn't go p to 5V instead of 3v3.

Can't remember the last time I designed a DIP in though. In terms of prototyping difficulty, the difference between DIP and SO isn't as big as SO and smaller pitch or quad packages. 
Apart from DIP switches - how come nobody makes SMD DIP switches with sensible sized actuators. The old argument was they needed to be tape-sealed for flux wash, but with no-clean, this argument isn't really valid any more. 

[gregariz]

"For standard devices I can't see any rationale for 8 and 16bit development continuing."

That sort of thinking is why Arizona Microchip and Atmel have both kicked their ass. But I guess it means no new HC08/11's which is a bit of a shame to see the end of the Motorola line.

[free_electron]

I often have a need for 5V - e.g. when driving a crapload of LEDs - I typically need 4-5V for the LEDs anyway, so 5V everything saves parts.

you driv ethem directly from i/o pins ? For led chains i typically use an open drain driver like an uln28xx or i slap in 2n7002 mosfets. work like  charm.
there are special low voltage mosfets for this kind of app.
if lots of current is involved you need a real gate driver cicuit anyway.

It seems a shame that even on parts that use a LV core and have voltage regulation built in, that regulation couldn't go p to 5V instead of 3v3.

power dissipation in the package... they design these things to be low power , and then you burn off 'waste-power' using an ldo to go from 5 to 1.8 ... so you just shot your 'low power' to shreds... for a 5 volt power supply and a core voltage of 1.8 volts . lets say the core consumes 1.8 milliwatt... while burning off 3.2 to get there ... efficiency : zilch ...

From 3.3 you only burn of 1.5... 3.2 to  1.5 is more than 50% gain.

If only they would use an on board switcher ( switchcap or external coil) then it would get interesting.

Can't remember the last time I designed a DIP in though. In terms of prototyping difficulty, the difference between DIP and SO isn't as big as SO and smaller pitch or quad packages. 

indeed. SO is no harder than dip. TSSOP is a bit more annoying...

how come nobody makes SMD DIP switches with sensible sized actuators.

you mean the ones that stick out as opposed to the annoying recessed ones ?
i believe APEM and some german company have those. they are black body with white levers that stick out. they are a bit more expensie as they come with a piece that needs removing after assembly ( the bits that stick out prevent P&P robots from picking them up. so they have a transparent cover so the robot can handle them. this thing just pops off after assembly. )

[poptones]

I really don't understand the continuing with 8 bit devices either. Microchip (for example) has a 32 bit chip starting at 2 bucks. I realize you can get others for 75 cents, but considering the economies of scale it seems like if they made all 32 bit devices the price tag could go even lower.

[mikeselectricstuff]

I often have a need for 5V - e.g. when driving a crapload of LEDs - I typically need 4-5V for the LEDs anyway, so 5V everything saves parts.

you driv ethem directly from i/o pins ?

No, Constant-current  LED drivers, usually Macroblock MBI5031. Sometimes hundreds of them.

It seems a shame that even on parts that use a LV core and have voltage regulation built in, that regulation couldn't go p to 5V instead of 3v3.

power dissipation in the package... they design these things to be low power , and then you burn off 'waste-power' using an ldo to go from 5 to 1.8 ... so you just shot your 'low power' to shreds... for a 5 volt power supply and a core voltage of 1.8 volts . lets say the core consumes 1.8 milliwatt... while burning off 3.2 to get there ... efficiency : zilch ...

These are typically low power devices, only a few mA running - if you're wanting 5V chances are you don't are about a few mW.

From 3.3 you only burn of 1.5... 3.2 to  1.5 is more than 50% gain.

If only they would use an on board switcher ( switchcap or external coil) then it would get interesting.

Efficiency gain of a switcher can be minimal at low levels due to the current to drive the SMPS

[gregariz]

I really don't understand the continuing with 8 bit devices either. Microchip (for example) has a 32 bit chip starting at 2 bucks. I realize you can get others for 75 cents, but considering the economies of scale it seems like if they made all 32 bit devices the price tag could go even lower.

People have been predicting the death of 8 bit for as long as I can remember. It just provides people with options. I don't think its so much a bom cost issue - more a manufacturing cost issue and a simplicity option. If I were doing production of a medium volume power controller item with alot of pth parts I may simply not want to tool up for smt. Or I may be wanting to make use of an old smt line that cant place fine sub mm pitches. In which case a 8 pin pic10 dip for 30c from Microchip is the way to go.

[poptones]

Wow

[tom66]

8 bit will be around for a long time in ultra low cost applications like consumer gear. There was some audio amp with a neat VFD display of some kind that used a 4-bit micro to do the job. I just wait for single-instruction micros  .

[EEVblog]

Wow

Yeah, but how much are the feeders?

Dave.

[deephaven]

I'm not bothered about DIP, but often need 5V operation, which is often another casualty of going to smaller devices. Even some of the newer 8 bit PICs use a low core voltage, but hide this to the outside 5V speaking world.

I was frustrated to see that the max3313 I used in a 5V design didn't work down to 3.3V for another design I had. This is a single RS232 TX and RX with a working voltage range of 4.5-5.5V. For my 3V3 design, I couldn't find a single channel version so had to go to a MAX3232 which was a bigger package.

[mikeselectricstuff]

Wow

Yeah, but how much are the feeders?

Dave.

It s not hard to find an old pick.& place under $5K if you have room for a big one , but feeders can easily cost more or be impossible to find - as long as there are some of that model in use, users will want feeders (you can never have too many feeders). Also bear in mind spares are likely to be unobtanium. You really need to know which old models are still viable and which suffer from expensive mechanical issues later in life. 

[tom66]

My workplace uses an ancient Europlacer... it's pretty worn out and spares are really difficult to find. It also requires a Windows 95 computer to work.

[Bored@Work]

I really don't understand the continuing with 8 bit devices either. Microchip (for example) has a 32 bit chip starting at 2 bucks. I realize you can get others for 75 cents, but considering the economies of scale it seems like if they made all 32 bit devices the price tag could go even lower.

We had that discussion before, it gets boring. They are easier to use, easer to understand and more predictable. The development environments are easier and simpler, and well understood, too. The devices and the development environments have gone through many iterations.  All this is a bonus when you need to do high-reliable, "works all the time" stuff. If you can grasp the whole thing, how everything works together you are less likely to make errors.

Of course, kids these days need 10 GByte memory to just get a led blinking, but not without a realtime OS running some virtual machine using some stinking interpreter executing some code that was clicked together with a code wizard. Because that is modern and new, and typing in real lines of code is for old people. Oh, and it is best when the whole thing is Enterprisified, Webified and Cloudified.

[amyk]

8 bit will be around for a long time in ultra low cost applications like consumer gear. There was some audio amp with a neat VFD display of some kind that used a 4-bit micro to do the job. I just wait for single-instruction micros  .

There are probably more 4-bit ones being made than 8, but they're mostly used in very mundane applications that people never think about and are large-volume, bare die mask ROM parts. Musical greeting cards (and other cheap music-type toys), 4-function calculators, various small appliances, digital watches, and just about any sort of toy that needs rudimentary logic but where discrete components or an ASIC would be relatively expensive --- these minimal MCUs are in the ~$0.01 range. I don't see 8-bit taking over that market, and doubt 32-bit will replace 8: many Chinese MP3/4 players are still being made containing a 6502, 8051, or Z80 core, and they're perfectly capable of doing what they do.

[jerry507]

I really don't understand the continuing with 8 bit devices either. Microchip (for example) has a 32 bit chip starting at 2 bucks. I realize you can get others for 75 cents, but considering the economies of scale it seems like if they made all 32 bit devices the price tag could go even lower.

We had that discussion before, it gets boring. They are easier to use, easer to understand and more predictable. The development environments are easier and simpler, and well understood, too. The devices and the development environments have gone through many iterations.  All this is a bonus when you need to do high-reliable, "works all the time" stuff. If you can grasp the whole thing, how everything works together you are less likely to make errors.

Of course, kids these days need 10 GByte memory to just get a led blinking, but not without a realtime OS running some virtual machine using some stinking interpreter executing some code that was clicked together with a code wizard. Because that is modern and new, and typing in real lines of code is for old people. Oh, and it is best when the whole thing is Enterprisified, Webified and Cloudified.

This attitude is just as ridiculous as the people who feel the need to use that 32bit micro for blinking LEDs. The people who claim 8bit just needs to die are mirrored by the people who think 32bit micros are for inferior engineers.

Reality will continue to be that 8bit micros are getting outpaced by what we want to do with embedded systems, but just like 4bit micros they will continue to survive and sell in huge quantities. Thank God we're now seeing 32bit micros become absolutely dirt cheap so we can do more and more for the same price without having to offload many of the advanced capabilities like wireless to some annoying ASIC.

The ecosystem is diversifying to the betterment of everyone.

[T4P]

The transistors in a processor is more efficient at 90nm, therefore the 3.3V requirement (If that's what i've been led to believe)