george_d wrote:

> Hello Erik,
> 
> Thanks for the reply.
> 
> We where actually considering ARM for our new generation product's
> MPU, but we thought that the processing power is way overdoing it, so
> were back to 8051.

Correct, if you don't _need_ the power, then stick with the 8051.

  8 bit uC are not going away, but there is a overlap-shift
at the top end ( code size, pin count ), so for a project
that you know needs more than 64K code, or extended peripherals,
or 64+ pins, then the ARM's can now compete with 8 bit.

  That said, you cannot find a socketable ARM, nor anything
under 32 pins (excluding smart cards).
  There are also no ARMs with 24 bit ADCs.....

  In the industrial sector, 80C51 dominate, with companies like
Infineon [new 866/886/888], AnalogDevices, BurrBrown, Philips
all doing 80C51 devices, and all with long records of sourcing.

  Philips are extending their LPC family, up to 64KFlash
- even though they also have ARMs 'down-to' 8KFlash.

  In the ARM sector, there are now ARM9's showing with FLASH,
and doubtless there will be some rationalise of the ARM7
devices...

-jg

Hi Meindert,

Some companies are just better at producing a part longer.  I guess if
one is concerned about 2nd sourcing, that kind of chip manufacture
would be the primary.

By the way, I read on the Microchip board that someone thought that
Microchip might be discontinuing one of the popular PIC18F... chips.

George

Another thought about Erik's reply.

Most controller applications are cheep little 8 bit ones.  Do you
really think the 8 bit MPU is dying out?

George

Hi Grant,

It's been my experience that it takes time to really know an MPU
well.  It's only at that point when one can see if there are bugs in
the chip (see my reply above about the EPROM).  Also one (at least I
do) really needs to know the chip inside and out to take advantage of
the part's futures.

George

Hello Erik,

Thanks for the reply.

We where actually considering ARM for our new generation product's
MPU, but we thought that the processing power is way overdoing it, so
were back to 8051.

George

Hello Jim,

I do agree that there are degrees of freedom to 2nd sourcing.  However,
most of my applications are industrial and are in production for 3 to 6
years, so I need parts that will be around for that long.  Sometimes I
find a close fitting parts and I only need to make a second set of
traces on the PC card for the 2nd sourced part.  Sometimes we buy a
years worth of a sole sourced part, so we have a year's lead-time
incase a part goes out of production.

That actually happened once.  We discovered that our primary MPU was
being phased out.  We bought years worth of parts and designed a new
generation product.  As it turned out, the product was very due for a
new generation anyway.

So if things are well planed (and with some luck), even having a sole
sourced part go pfffit! should not be the end of the world.

George

On 2006-02-20, george_d <gdorian@powis.com> wrote:

> Also, it would make since for a company to stay tooled for one
> micro family.  It's not just a matter of the physical tolling
> but, even more so, the learning curve for the company's
> engineers to learn a new micro.  Many instrumentation
> companies are small, so learning a new micro can be a big
> chunk of time (aka $).

Does it really take that long to learn to use a new micro? I've
used 8-10 different architectures over the past 25 years, and I
don't recall it ever taking more than a few days (maybe a week)
to get reasonably up to speed with a new part.

-- 
Grant Edwards                   grante             Yow!  I wonder if I could
                                  at               ever get started in the
                               visi.com            credit world?

Thanks for your reply.

In general my opinion about 2nd sourcing of micros is this: It strikes
me that many (maybe most) micro applications are consumer based.  That
is, MP3 players, VCRs, and "rice cookers".  These products only
have a 6 to 12 month product life.  So those companies may not be
concerned if a micro is out of production in 12 months.  On the other
extreme, there are companies (like mine) that build instrumentation.
These products can be around for 3 to 6 years.  Also, it would make
since for a company to stay tooled for one micro family.  It's not
just a matter of the physical tolling but, even more so, the learning
curve for the company's engineers to learn a new micro.  Many
instrumentation companies are small, so learning a new micro can be a
big chunk of time (aka $).

As for Microchip;  Until a few years ago I used to be a Microchip
junky.  But I keep finding bugs.  Bugs in their documentation, chips,
IDE, emulator, and other things.  The last straw for me was when I read
some where that if you program a cell in the on-chip EEPROM, you should
rewrite the cells on ether side of it, because they might be affected
by the write of the first cell.

It's hard enough to design a controller, and the last thing I need is
to have unreliable development equipment, chips, documentation, etc.

On Tue, 14 Feb 2006 18:39:52 +0100, "Meindert Sprang"
<ms@NOJUNKcustomORSPAMware.nl> wrote:

>"george_d" <gdorian@powis.com> wrote in message
>news:1139933701.226319.283970@g44g2000cwa.googlegroups.com...
>> That's Silicon labs.  I know them.
>>
>> They make some pretty amazing chips, but they don't have any 2nd
>> sources, and my company is apprehensive about no 2nd sources.
>> George
>
>I hear that a lot. And I wonder if needing second sources is really a valid
>argument. It is used many times by 8051 "addicts" but there are so many
>processors around that do not have second sources. See for instance DSP's. I
>hear noboday complain about that. And look at the amount of PIC's used in
>automotive; no second source.

This "2nd source" term is probably a conflated mush of more exact
ideas.

There is a 2nd source meaning that is strict, in that the processor
packaging and pin layout, electrical specifications of various kinds,
peripheral mix and their programming methods, CPU instructions, etc.,
are equivalent for the purposes of the product.  In that it can simply
be dropped in, from a different vendor, and work properly with binary
production images used with existing production tools, unchanged.

In that sense, I think the 8051 core may offer some limited options of
2nd sourcing where very few other possibilities exist.  But then, I
haven't directly experienced such a project outside of hobbyist ones
using very basic, generic cores with only standard features and DIP
packaging (which makes sense, for boards where the board maker isn't
sure which processor someone intends to drop in.  However, at least it
is possible to consider.

There are other facets to what 2nd source means to each, from FABing
to embedded development toolset investments.

And yes, so long as companies have a demonstrated record in providing
what your business requires, then even a single source is often just
fine.  The Microchip PIC is an excellent example.  Their transition
from providing them in million-lots to rice cooker makers, into
deciding to provide them to hobbyists and smaller businesses was
significant.  But they have carried it off with extreme competence and
it shows in the loyalty they have earned over time.

Jon

Meindert Sprang wrote:

> "george_d" <gdorian@powis.com> wrote in message
> news:1139933701.226319.283970@g44g2000cwa.googlegroups.com...
> 
>>That's Silicon labs.  I know them.
>>
>>They make some pretty amazing chips, but they don't have any 2nd
>>sources, and my company is apprehensive about no 2nd sources.
>>George
> 
> 
> I hear that a lot. And I wonder if needing second sources is really a valid
> argument. It is used many times by 8051 "addicts" but there are so many
> processors around that do not have second sources. See for instance DSP's. I
> hear noboday complain about that. And look at the amount of PIC's used in
> automotive; no second source.

There are multiple aspects to 'Second sourced' -
a) Some worry about an exact physical replacement, and for them, yes you 
can get that, if you are prepared (not surprisingly) to work with a 
subset of all parts. PLCC44 and DIP20 are the sweet spots here.

b) Some worry about functional replacement, and a PCB respin is not
nearly as risky as a core change - smallish PCB respins are actually
common. For those, you need "Same core, Similar Size package"
- and that definition of second source widens up the candidate list.

eg 80C51 with 24 bit ADC :

Silabs, TI, ADi, and even TDK all have vey similar resourced
devices

c) Then there is 'Second Fab sourcing' - you will often see larger
suppliers mention that they have more than one FAB qualified to run 
their devices. That gets close to having another supplier, and
covers physical risks to FABs, as well as reduces process EOL risks.

-jg