TI MSP430

On 2006-04-08, larwe <zwsdotcom@gmail.com> wrote:

>> That quantity factor makes a big difference.  My customer had
>> no problem with a waaaaay over-specified MSP430F149 as a combo
>> RTC and I/O controller at quantities of a hundred per year.
>> For that
>
> You didn't provide the punchline - did he look at AVR and
> compare prices at all?
>
> MSP430 is a truly elegant architecture, there's no denying it.
> von Neumann simplicity, totally transparent handling of
> registers and addressing modes; I really can't find much to
> fault with it (except maybe the way info memory is handled).
> But:
>
>> software and sensor circuitry.  A few dollars here and there
>> to make the engineers comfortable is no problem when you're
>
> How much "comfort" did the F149 earn you, out of interest?
> I've worked a lot with AVR in my real life, and MSP430 mostly
> at work. While AVR is not internally as tidy as MSP430, it's
> really not bad

As long as you close your eyes and force yourself not to look
at the code a compiler has to generate for pointer-intensive
stuff, or for 16-32bit math, or for anything involving the
stack pointer.  It's _ugly_.  Due to the mostly 1-cycle
execution in the AVR, it's not really a lot slower, but it
hurts my eyes when I look at it.  It also requires 25-50% more
code space than the '430 for the functions I've benchmarked.

> (especially compared with a horror story like the PICmicro).

The PIC is truly crap.  

> And both parts are quite C-friendly, if this tickles your
> pickle.

I wouldn't consider the AVR to be C-friendly.  It's not as
hostile as the 8051, but it's far from friendly.

For instance, consider the contortions the compiler has to go
through to reserve space on the stack for auto variables. IIRC,
it has to disable interrupts while it does multiple 8-bit I/O
operations because the stack pointer isn't directly accessible
as a register.

The dearth of registers that can be used as pointers is also
rather C-hostile.

-- 
Grant Edwards                   grante             Yow!  PARDON me, am I
                                  at               speaking ENGLISH?
                               visi.com            

Grant Edwards wrote:

> > I've worked a lot with AVR in my real life, and MSP430 mostly
> > at work. While AVR is not internally as tidy as MSP430, it's
> > really not bad
>
> As long as you close your eyes and force yourself not to look
> at the code a compiler has to generate for pointer-intensive

Oh, AVR ain't so bad. Most of the irritations stem from the different
handling of r16-r31 and the fact that X,Y,Z are (a) different from one
another, and (b) the only 16-bit regs. If they had been able to
implement it such that every register was part of a register pair (and
every register pair able to act as a pointer), it would have been much
better.

But that's where a 16-bit instruction word (and not being a slave to
the RISC idea of one word per instruction) comes in handy.

> through to reserve space on the stack for auto variables. IIRC,
> it has to disable interrupts while it does multiple 8-bit I/O
> operations because the stack pointer isn't directly accessible
> as a register.

There are ways around this, e.g. using a separate stack implemented in
software.

> The dearth of registers that can be used as pointers is also
> rather C-hostile.

Well, yes - I didn't say it was perfect.

On 8 Apr 2006 17:55:22 -0700, "larwe" <zwsdotcom@gmail.com> wrote:

>Right. But these are distribution prices. They don't reflect the direct
>pricing at all. Clearly TI's prices to distris are closer to the
>break-even edge than Atmel's, so neither TI nor the distri have much
>room to discount deeply on quantity.

     The fact that TI is willing to make their microcontrollers
available at reasonable prices in small volumes indicates their
interest in dealing with small companies.  This ties in with the low
cost development tools and free seminars.  While this policy may not
seem important to large companies, it is reason to take notice by
those of us in the low-volume high value-added segment.

Gary Reichlinger wrote:

>      The fact that TI is willing to make their microcontrollers
> available at reasonable prices in small volumes indicates their
> interest in dealing with small companies.  This ties in with the low

It's more or less an established fact that Atmel is not very interested
in doing business with small accounts. Samples, support, low prices are
just unavailable for anyone smaller than <$x>.

I threw the price fact into the mix because some of us moonlight at day
jobs where we work on high-volume consumer junk. Of course our real
lives in our home labs are not the same, but it's always good to have
more data points.

larwe wrote:
<paste>
> The MSP430 prices are annoying, though. AVRs come down to
> as little as 1/4th the price of an equivalent TI part.

> Jim Granville wrote:
> 
>>That can happen, where one supplier is able to get better
>>"column movement" than another.
>>Plus lead free 'stock clearing' can also affect prices...
> 
> 
> We buy direct in both cases, and we only spec in RoHS-compliant parts.

... then it might be prudent to not mention that Atmel's price is 1/4
that of an equivalent part. Someone is likely to get their wrist 
slapped, and your price might vaporise... :)

  IME uC normally follow a 'fairly well behaved' price curve,
and where they diverge greatly, be wary of the sustainability of that
price....

-jg

On 8 Apr 2006 17:43:20 -0700, "larwe" <zwsdotcom@gmail.com> wrote:

>"Now, it's an obscure but well-recognized ritual amongst engineers and
>computer scientists to gauge the architecture of any new device on the
>basis of its similarity to the PDP-11(1). 

     Do you guys belong to AARP?  

In article <1144537383.517473.281170@g10g2000cwb.googlegroups.com>, 
zwsdotcom@gmail.com says...
> 
> Mark Borgerson wrote:
> 
> > That quantity factor makes a big difference.   My customer had no
> > problem with a waaaaay over-specified MSP430F149 as a combo RTC
> > and I/O controller  at quantities of a hundred per year.  For that
> 
> You didn't provide the punchline - did he look at AVR and compare
> prices at all?

At the time I was coming from the 68K world, and the MSP430 architecture
appealed to me.  I was also using it as a RTC running off a 40KHz xtal
when the rest of the instrument was asleep.  One spec was that the
system had to draw less than 100uA in sleep mode.  That meant paying
very careful attention to the power supply and finding an RTC/IO 
controller that would run on very low  current.

I also made good use of the MSP430 12-bit ADC for some auxilliary
sensors that were not connected to the 16-Bit ADC.
> 
> MSP430 is a truly elegant architecture, there's no denying it. von
> Neumann simplicity, totally transparent handling of registers and
> addressing modes; I really can't find much to fault with it (except
> maybe the way info memory is handled). But:
> 
> > software and sensor circuitry.  A few dollars here and there
> > to make the engineers comfortable  is no problem when you're
> 
> How much "comfort" did the F149 earn you, out of interest? I've worked
> a lot with AVR in my real life, and MSP430 mostly at work. While AVR is
> not internally as tidy as MSP430, it's really not bad (especially
> compared with a horror story like the PICmicro). And both parts are
> quite C-friendly, if this tickles your pickle.

Been there and done that with the PICs.  Never again.  (I'm both
busy enough and close enough to retirement that I would turn down
jobs where using a PIC was part of the requirement.)

I don't think I ever used more than 1/4 the code space of the 
MSP430F149.  I did use a lot of the RAM, though.  The MSP is
the interface to a data flash chip which   sent and received
data in 1056-byte blocks.   It was much easier to have the
full block buffered on the MSP and sent to the ARM as required.
A chip with less than 2K RAM would have required  a more complex
approach.
> 
> > a different approach to component selection.    I think I'll stay
> > in the low-volume, high value-added arena where I don't have to
> > count the bytes and pennies quite so carefully!   ;-)
> 
> The dollars add up pretty quickly. I can't even make a decision that's
> worth less than $250K these days. (These numbers are frightening, by
> the way).


I don't think I've ever worked on an oceanographic project that sold 
more than 2000 units.  The oceanographic instrument market just isn't 
that large. However,  it is the right market for a self-taught embedded 
systems developer with an MS in chemical oceanography.   I've tried
Apple II peripherals and software  and Macintosh software (back in the 
80's----grey beard is itching like mad now!!!)  but wasn't nearly as 
happy in that field.

Mark Borgerson

On 8 Apr, in article
     <1144543400.381853.139720@i40g2000cwc.googlegroups.com>
     zwsdotcom@gmail.com "larwe" wrote:

>Marc Ramsey wrote:
>
>> > MSP430 is a truly elegant architecture, there's no denying it. von
>>
>> Of course, the architecture is a simplified version of the TMS9900 CPU,
>> which was derived from the TI 990 minicomputer, which was in turn a
>> rip-off of the PDP-11 ISA, with slight changes (fewer addressing modes,
>
>LOL. You know, writing this most recent book is so amusing. I think I
>could probably use 75% of it as responses in c.a.e.
>
>"Now, it's an obscure but well-recognized ritual amongst engineers and
>computer scientists to gauge the architecture of any new device on the
>basis of its similarity to the PDP-11(1). The strongest term of
>approbation you can use for a CPU design is to say "It's just like a
>PDP-11!". Exactly why this is universally regarded as a Good Thing is
>not exactly clear, but in any case, this epithet is frequently applied
>to the MSP430. (I've always felt that this is the same sort of
>statement as saying "My 2007 Mercedes convertible is just like a 1965
>International Harvester Scout light truck. They both have pneumatic
>tires and a removable roof!"). What I think these people probably
>mean is that the MSP430 has a very nice orthogonal instruction set and
>simple memory addressing scheme.

That is my take as well, which is why I think a lot of people also
liked 68xxx, 6809, H8 and a few others.

>(1) - If I ever design a microprocessor, I am going to print a line
>drawing of a complete PDP-11/20 front panel on the top of the package."

At least you did not say 11/34 or 11/60! Says he who would prefer 11/05 or
11/10 front panel from memories of having a couple under his bench in
cut down racks as test machines, linked to the departments 11/45 later
11/70 for code writing and download the code from. Had some nice inhouse
steps then.

Anyway 11/05 front panel might be easier to draw on a smaller chip :-^

-- 
Paul Carpenter          | paul@pcserviceselectronics.co.uk
<http://www.pcserviceselectronics.co.uk/>    PC Services
<http://www.gnuh8.org.uk/>              GNU H8 & mailing list info
<http://www.badweb.org.uk/>             For those web sites you hate

Gary Reichlinger wrote:

> >basis of its similarity to the PDP-11(1).
>
>      Do you guys belong to AARP?

Not yet, but the curmudgeonliness flux density is high in the
engineering field, so it's forgivable that you would ask the question
:)

Jim Granville wrote:

> ... then it might be prudent to not mention that Atmel's price is 1/4
> that of an equivalent part. Someone is likely to get their wrist

Doubt it. Atmel wants our business. TI is happy enough to have our
business, but not actively seeking to grow it.