On Fri, 24 Feb 2006 10:28:55 +0000, Pete Fenelon wrote:

>Ian Bell <ruffrecords@yahoo.com> wrote:
>> 
>> I always thought the PIC might be a handy device. But I have used micros (4,
>> 8 and 16 bit) professionally for over 20 years and the stupidly odd
>> terminology the PIC uses for things known commonly by other names simply
>> alienates me from the parts.
>> 
>
>The PIC is a pile of I/O lines and some peripherals with just enough
>glue in the middle to hold it together as a CPU. Nobody chooses one
>because of the instruction set or architectre...  For a long time I
>reckoned the only thing that kept Microchip going was Keeloq. 
>
>Every time Microchip have tried to make a "more modern PIC" they've hit
>the wall - PIC17 was a monstrosity which even the firm's own engineers
>don't like to talk about; PIC18 is less gruesome but still too expensive
>and unpleasant to program; PIC30 has a (relatively) nice architecture
>but was too little, too late - conventional micros and cheap fixed-point
>DSPs outperform it for less money. 

I think PIC addresses a different market. They are designed as logic
replacement, not as general purpose CPUs, hence the unsuitability for e.g.
the C language. For logic replacement, an ARM would be way overkill, asuming
it even could do it. For thing like coffee machines, hairdryers etc. a PIC is
a very good solution. 

Mat Nieuwenhoven

Jonathan Kirwan wrote:
> I have yet to see a better designed and crafted use of 16 bits for an
> instruction.  The JSR you mention is itself a thing of beauty and
> power -- it's utility exceeds mere subroutine calls and is useful for
> coroutines and more.
> 
> For embedded processors, I can accept the idea that better might
> possibly be done.  For example, I'm attracted to the idea of using one
> of the registers as a constant generator, combined with the addressing
> modes to provide commonly found constants.  The loss of a register
> might be well compensated -- and I think an analysis of a large number
> of real embedded applications should be done and would help to decide
> if so.  The MSP-430, for example, takes this approach -- but
> sacrifices too much in going over the top to support 16 registers in
> each and every instruction and thereby severely cripples its
> addressing modes.

  If you are a student of processor design, this
one is worth a look.

http://forums.parallax.com/forums/attach.aspx?a=5975

  Bizare rollout aside, this started life as a FPGA core,
and uses 32 bit opcodes, so has room for two 9 bit register fields.

  Similar to the Intel 196, that blurred register/ram

  Present silicon has a 512W memory limit
(per core, of which there are 8!), but it looks like
the CALL can go further, in future.

  -jg

On Sat, 25 Feb 2006 07:56:03 +1300 in comp.arch.embedded, Jim
Granville <no.spam@designtools.co.nz> wrote:

>Dave Hansen wrote:
>> 
>> In the applications where I've used them, they were chosen for their
>> mix of I/O and their environmental specifications.  And their price.
>> As the software guy, I was always looking for better alternatives, but
>> everything cheaper didn't have the I/O we needed, and everything that
>> did was _way_ more expensive.  I remember one project I asked
>> purchasing to get a quote for a Cygnal (now SiLabs) part (F300 IIRC).
>> They came back with about $6.50 at 10k quantities, which was about 5x
>> the Microchip part.
>
>  That must have been a long time ago ? - In feb 2006 the quote their 

Long enough that I don't remember exactly when.  Maybe 3 or 4 years
ago.  As I said, it was Cygnal then.  And the incident sticks out
because it was extreme.  I don't doubt their prices have come down.

>much bigger USB devices as  >= $1.76/10K, for the 32x and >= $3.42/10K 
>for the 34x.
>
>  Having said that, when you work out the PIC Average selling price,
>it is still very close to $1 - so they clearly deliver shiploads of
>vanilla devices...

It just seemed funny that Microchip almost always had exactly the part
we needed, with nothing missing, and very little (if any) extra.  

Regards,
                                        -=Dave

-- 
Change is inevitable, progress is not.

Dave Hansen wrote:
> 
> In the applications where I've used them, they were chosen for their
> mix of I/O and their environmental specifications.  And their price.
> As the software guy, I was always looking for better alternatives, but
> everything cheaper didn't have the I/O we needed, and everything that
> did was _way_ more expensive.  I remember one project I asked
> purchasing to get a quote for a Cygnal (now SiLabs) part (F300 IIRC).
> They came back with about $6.50 at 10k quantities, which was about 5x
> the Microchip part.

  That must have been a long time ago ? - In feb 2006 the quote their 
much bigger USB devices as  >= $1.76/10K, for the 32x and >= $3.42/10K 
for the 34x.

  Having said that, when you work out the PIC Average selling price,
it is still very close to $1 - so they clearly deliver shiploads of
vanilla devices...

-jg

On Fri, 24 Feb 2006 14:11:43 GMT, Tauno Voipio
<tauno.voipio@INVALIDiki.fi> wrote:

>Pete Fenelon wrote:
>> Everett M. Greene <mojaveg@mojaveg.iwvisp.com> wrote:
>> 
>>>>The MSP430 won't teach you too many bad habits and it's a nice clean
>>>>instruction set. (It'll also give you a good heads-up should you ever
>>>>fall back through a timewarp and need to program PDP-11s). 
>>>
>>>The Motorola 68xx family is a lot closer to the PDP-11.
>> 
>> 
>> 68xxx maybe; 68xx doesn't have enough registers or addressing modes to
>> feel like a PDP-11 -- at least not to me ;)
>
>6800 is kinda correct: It was all of PDP-11 Motorola could
>squeeze on the silicon of that day. Motorola was using PDP-11's
>for their system products, and the resemblance is not accidental.
>
>Of course, most of the good ideas of PDP are missing:
>
>  - multiple general registers,
>  - homogenous addressing modes,
>  - subroutine call with register linkage
>
>The worst property of a 6800 is the bottleneck created
>by the accumulators and a single addressing register (X).
>
>There are other PDP-like architectures: M68k and Hitachi H8.
>IMHO, both have lost some of the beautiful simple completeness
>of the genuine PDP-11.

I have yet to see a better designed and crafted use of 16 bits for an
instruction.  The JSR you mention is itself a thing of beauty and
power -- it's utility exceeds mere subroutine calls and is useful for
coroutines and more.

For embedded processors, I can accept the idea that better might
possibly be done.  For example, I'm attracted to the idea of using one
of the registers as a constant generator, combined with the addressing
modes to provide commonly found constants.  The loss of a register
might be well compensated -- and I think an analysis of a large number
of real embedded applications should be done and would help to decide
if so.  The MSP-430, for example, takes this approach -- but
sacrifices too much in going over the top to support 16 registers in
each and every instruction and thereby severely cripples its
addressing modes.

Jon

Pete Fenelon <pete@fenelon.com> writes:
> Everett M. Greene <mojaveg@mojaveg.iwvisp.com> wrote:
> >> The MSP430 won't teach you too many bad habits and it's a nice clean
> >> instruction set. (It'll also give you a good heads-up should you ever
> >> fall back through a timewarp and need to program PDP-11s). 
> > 
> > The Motorola 68xx family is a lot closer to the PDP-11.
> 
> 68xxx maybe; 68xx doesn't have enough registers or addressing modes to
> feel like a PDP-11 -- at least not to me ;)

Quite true but wasn't the original poster looking for
something simple for learning purposes?

On Fri, 24 Feb 2006 14:33:23 +0000, the renowned Pete Fenelon
<pete@fenelon.com> wrote:

>Ian Bell <ruffrecords@yahoo.com> wrote:
>>> The PIC is a pile of I/O lines and some peripherals with just enough
>>> glue in the middle to hold it together as a CPU. Nobody chooses one
>>> because of the instruction set or architectre...  For a long time I
>>> reckoned the only thing that kept Microchip going was Keeloq.
>>> 
>> 
>> I do know of other volume applications. They are very aggressive on price
>> for high volume applications. I found one once in a hair dryer, one in an
>> electric toothbrush and we actually used one once in a CO2 detector design.
>> 
>> But I agree, they are crap devices.
>> 
>
>I don't doubt that there are *many* ultra-low-margin applications that
>use PICs (and many of them could probably get by with a 555 timer and no
>micro, but everything's got to have a micro in it these days, right?!) but
>Keeloq seems to be the one truly profitable line of business Microchip
>have. Then again they have some good points; sitting right next door to
>Intel's fabs in Chandler basically means that as Intel scrap one line
>Microchip can take it over ;)
>
>I'm not quite sure what they're trying to achieve with PIC24, either -
>cutting the fixed-point DSP end off PIC30 doesn't seem to be much of
>a saving?
>
>pete

It's all about marketing and money, not engineering and gates. It can
increase their profits even if each one costs them *more* to make. 


Best regards, 
Spehro Pefhany
-- 
"it's the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

On Fri, 24 Feb 2006 11:28:30 +0000 in comp.arch.embedded, Ian Bell
<ruffrecords@yahoo.com> wrote:

>Pete Fenelon wrote:
>
[...]
>> 
>> The PIC is a pile of I/O lines and some peripherals with just enough
>> glue in the middle to hold it together as a CPU. Nobody chooses one
>> because of the instruction set or architectre...  For a long time I

In the applications where I've used them, they were chosen for their
mix of I/O and their environmental specifications.  And their price.
As the software guy, I was always looking for better alternatives, but
everything cheaper didn't have the I/O we needed, and everything that
did was _way_ more expensive.  I remember one project I asked
purchasing to get a quote for a Cygnal (now SiLabs) part (F300 IIRC).
They came back with about $6.50 at 10k quantities, which was about 5x
the Microchip part.

I've always had good luck with their support as well.

>> reckoned the only thing that kept Microchip going was Keeloq.

Actually, I never used that.

>> 
>
>I do know of other volume applications. They are very aggressive on price
>for high volume applications. I found one once in a hair dryer, one in an
>electric toothbrush and we actually used one once in a CO2 detector design.
>
>But I agree, they are crap devices.

Actually, they're good, solid, reliable devices with a hideous
programming architecture.  Cheap doesn't hurt either, especially in
high-volume applications that can tolerate a little extra NRE.

Regards,
                                        -=Dave

-- 
Change is inevitable, progress is not.

Ian Bell <ruffrecords@yahoo.com> wrote:
>> The PIC is a pile of I/O lines and some peripherals with just enough
>> glue in the middle to hold it together as a CPU. Nobody chooses one
>> because of the instruction set or architectre...  For a long time I
>> reckoned the only thing that kept Microchip going was Keeloq.
>> 
> 
> I do know of other volume applications. They are very aggressive on price
> for high volume applications. I found one once in a hair dryer, one in an
> electric toothbrush and we actually used one once in a CO2 detector design.
> 
> But I agree, they are crap devices.
> 

I don't doubt that there are *many* ultra-low-margin applications that
use PICs (and many of them could probably get by with a 555 timer and no
micro, but everything's got to have a micro in it these days, right?!) but
Keeloq seems to be the one truly profitable line of business Microchip
have. Then again they have some good points; sitting right next door to
Intel's fabs in Chandler basically means that as Intel scrap one line
Microchip can take it over ;)

I'm not quite sure what they're trying to achieve with PIC24, either -
cutting the fixed-point DSP end off PIC30 doesn't seem to be much of
a saving?

pete
-- 
pete@fenelon.com [Support no2id.net: working to destroy Blair's ID card fraud]

Pete Fenelon wrote:
> Everett M. Greene <mojaveg@mojaveg.iwvisp.com> wrote:
> 
>>>The MSP430 won't teach you too many bad habits and it's a nice clean
>>>instruction set. (It'll also give you a good heads-up should you ever
>>>fall back through a timewarp and need to program PDP-11s). 
>>
>>The Motorola 68xx family is a lot closer to the PDP-11.
> 
> 
> 68xxx maybe; 68xx doesn't have enough registers or addressing modes to
> feel like a PDP-11 -- at least not to me ;)

6800 is kinda correct: It was all of PDP-11 Motorola could
squeeze on the silicon of that day. Motorola was using PDP-11's
for their system products, and the resemblance is not accidental.

Of course, most of the good ideas of PDP are missing:

  - multiple general registers,
  - homogenous addressing modes,
  - subroutine call with register linkage

The worst property of a 6800 is the bottleneck created
by the accumulators and a single addressing register (X).

There are other PDP-like architectures: M68k and Hitachi H8.
IMHO, both have lost some of the beautiful simple completeness
of the genuine PDP-11.

-- 

Tauno Voipio
tauno voipio (at) iki fi