[sci.electronics] microprocessor wanted

fmgst@unix.cis.pitt.edu (Filip Gieszczykiewicz) (04/24/91)

	Greetings.

	I need a microprocessor with the following specifications:

	o	8-bit (16-bit would be nice)
	o	~10MHz clock
	o	ability to address 64K of RAM (more would be nice)
	o	BUILT-IN floating processor (that's the problem)
	o	CHEAP (under $40)

	I realize that the 80486 fits the above specs but not only
	is it out of my budget, it's also a pain to hack.

	Anyone? If not, I would like to use the 8086/8087/glue and
	put them on a small module that I can use as a "chip".
	I want 16 of these "chip"s to fit on one board (that's why
	I worry about space)

	Take care.

	P.S. I have yet to see a uP that does the above :-(
-- 
_______________________________________________________________________________
"The Force will be with you, always." It _is_ with me and has been for 11 years
Filip Gieszczykiewicz  "... a Jedi does it with a mind trick... " ;-)
FMGST@PITTVMS  or  fmgst@unix.cis.pitt.edu "My ideas. ALL MINE!!"

henry@zoo.toronto.edu (Henry Spencer) (04/24/91)

In article <119279@unix.cis.pitt.edu> fmgst@unix.cis.pitt.edu (Filip Gieszczykiewicz) writes:
>	o	BUILT-IN floating processor (that's the problem)
>	o	CHEAP (under $40)

These two requirements are basically incompatible at present.  Built-in
floating-point is a high-end feature, and comes complete with high-end
prices. :-(  I'd be interested to hear of any exceptions...
-- 
And the bean-counter replied,           | Henry Spencer @ U of Toronto Zoology
"beans are more important".             |  henry@zoo.toronto.edu  utzoo!henry

asd@cbnewsj.att.com (Adam S. Denton) (04/25/91)

>In article <119279@unix.cis.pitt.edu> fmgst@unix.cis.pitt.edu (Filip Gieszczykiewicz) writes:
>>	o	BUILT-IN floating processor (that's the problem)
>>	o	CHEAP (under $40)

I would hazard you'll never get it in one chip that cheap until
ten years from now. :-).  However, if you are diligent, you can
get by with two chips.  Motorola used to make
a "floating point ROM" - I believe MC6839 - which interfaced with
the 6800 family.  Basically, it was floating point emulation in
ROM, already coded and debugged for you.  You can get the source
from their BBS too.  Then you wouldn't have to code up floating point
(which sounds about as exhilarating as watching paint dry :-).
Although it'll take more chips, you'll have more spendable dollars too...
6800s are cheep cheep cheep.  And you can burn some EPROMs with code from
the BBS.  If you're truly a nut, you could run all the 6800s off of one
EPROM simultaneously -- sorta a SIMD architecture...naaaa, wouldn't work....

BTW, anybody out there got a description of Intel Hex format they can
e-mail me?  Much appreciated!  (I'll swap moto Srec for it too)

Adam Denton
asd@mtqua.att.com
This sentence is false.

touch@gradient.cis.upenn.edu (Joseph D. Touch) (04/25/91)

In article <1991Apr24.174217.14901@cbnewsj.att.com> asd@cbnewsj.att.com (Adam S. Denton) writes:
>>In article <119279@unix.cis.pitt.edu> fmgst@unix.cis.pitt.edu (Filip Gieszczykiewicz) writes:
>>>	o	BUILT-IN floating processor (that's the problem)
>>>	o	CHEAP (under $40)
>
>get by with two chips.  Motorola used to make
>a "floating point ROM" - I believe MC6839 - which interfaced with
>the 6800 family.  Basically, it was floating point emulation in
>ROM, already coded and debugged for you.  You can get the source

I have one of these ROMs, and the manual.  In fact, it's in a socket
on a 'board' to fit in the slot of a Radio Shack Color Computer (which
uses a 6809 - the ROM is based on a 6809 microprocessor).

I will sell the ROM & manual, and the COCO (first model, 64K,
with disk drive, joystics) for that matter (you can cannibalize 
it for the 6809, a dynamic memory controller 6883, a video chip, 
etc..)

Send me an offer.

Joe Touch
Dept of Computer and Information Science
University of Pennsylvania
touch@cis.upenn.edu

<LEEK@QUCDN.QueensU.CA> (04/25/91)

In article <119279@unix.cis.pitt.edu>, fmgst@unix.cis.pitt.edu (Filip
Gieszczykiewicz) says:
>
>        Greetings.
>
>        I need a microprocessor with the following specifications:
>
>        o       8-bit (16-bit would be nice)
>        o       ~10MHz clock
>        o       ability to address 64K of RAM (more would be nice)
>        o       BUILT-IN floating processor (that's the problem)
>        o       CHEAP (under $40)

Closest match is a 10MHz 68000.  Cost $10.  Hook it up with a 68881 floating
point unit. I paid $50 for a new 20MHz chip from a mail order place, so a
slower one might be cheaper.  Mind you that the 68000 does not have a
coprocessor interface, but can emulate that in software.  The 68881/882
manual has a section on that.

Cost:~$60  you get  16/32 bit uP, 16M addressing space, 10MHz CPU, 20MHz FPU...

>        Anyone? If not, I would like to use the 8086/8087/glue and

Doesn't that cost > $40 per pair ?  You could have a 16Mhz 68020 and 20Mhz
68881 for $100 and they are much faster than 8086/8087.  The 68881 have more
trig. functions too.

>        put them on a small module that I can use as a "chip".
>        I want 16 of these "chip"s to fit on one board (that's why
>        I worry about space)

How much space per "chip" ?  You could stack a second board, couldn't you ?
I think I can fit the 020/881 onto a 2" by 4" space - slightly bigger than
8086/8087/glue.
>        P.S. I have yet to see a uP that does the above :-(

Price is the only barrier left...  However if you relax the floating point
part, you can get microcontrollers well below $40.

>"The Force will be with you, always." It _is_ with me and has been for 11
>years
>Filip Gieszczykiewicz  "... a Jedi does it with a mind trick... " ;-)
>FMGST@PITTVMS  or  fmgst@unix.cis.pitt.edu "My ideas. ALL MINE!!"

K. C. Lee
Elec. Eng. Grad. Student

P.S. for the chef/priest/student/chip designer/PC user etc in you...
"May the sauce/cross/course/MOS/DOS/... be with you !!"   :)

whit@milton.u.washington.edu (John Whitmore) (04/25/91)

In article <119279@unix.cis.pitt.edu> fmgst@unix.cis.pitt.edu (Filip Gieszczykiewicz) writes:
>	I need a microprocessor with the following specifications:

>	o	8-bit (16-bit would be nice)
>	o	~10MHz clock
	Um... is this a maximum, because of memory speed or battery
power requirements, or a minimum?  Or do you care more about MIPS than
clock speed?
>	o	ability to address 64K of RAM (more would be nice)
>	o	BUILT-IN floating processor (that's the problem)
>	o	CHEAP (under $40)

>	I realize that the 80486 fits the above specs
	as does the 68040

	Almost any inexpensive CPU with floating point and relatively
high throughput will be a DSP; the so-called 'microprocessors' are 
usually more logic controllers than floating point engines.  Prices
on DSPs vary, as do the particular floating point capabilities (the
least expensive units actually are FIXED POINT, which is ever so
slightly different from floating point).
	The TMS320C25, with 64k data +64k program memory space,
runs at 25 MHz and costs circa $90.  Motorola's 5600x and 96xxx
offerings, Analog Devices, and even AT&T, will undoubtedly be
able to fill your needs (all except maybe price...).
	The multiply/accumulate time on the TMS320C25 is 100 ns;
could you perhaps use half as many processors, if each is twice
the speed you were aiming for?

	John Whitmore

<LEEK@QUCDN.QueensU.CA> (04/25/91)

In article <119279@unix.cis.pitt.edu>, fmgst@unix.cis.pitt.edu (Filip
Gieszczykiewicz) says:
>
>        I need a microprocessor with the following specifications:
>
>        o       8-bit (16-bit would be nice)
>        o       ~10MHz clock
>        o       ability to address 64K of RAM (more would be nice)
>        o       BUILT-IN floating processor (that's the problem)
>        o       CHEAP (under $40)

I got a second solution to the problem if floating point is the main purpose
of the "chips".  68020 has a coprocessor interface and can be hooked up to 8
coprocessors. (All one need to do is to do address decoding for them.) Get 2
of them to control 16 of the 68881 math coprocessors.

16MHz  68020: $50 x 2   \  works out to be about $60 per "chip"
20Mhz  68881: $50 x 16  /

This solution can be cheaper if you opted for 16MHz 881 parts and ask for
a discount for 16 chips.

The amount of glue and board space is drammatically reduced - only 2 CPU's +
glue + 16 881 chip + 2 decoder chip.  SHould be able to fit on an AT style
expansion card with no problem...

>        P.S. I have yet to see a uP that does the above :-(

Then look at the problem differently.  Use a FPU instead of a CPU. :)

>_
>"The Force will be with you, always." It _is_ with me and has been for 11
>years
>Filip Gieszczykiewicz  "... a Jedi does it with a mind trick... " ;-)
>FMGST@PITTVMS  or  fmgst@unix.cis.pitt.edu "My ideas. ALL MINE!!"

K. C. Lee
Elec. Eng. Grad. Student

jjohnson@cs.ucf.edu (John Jjohnson) (04/25/91)

fmgst@unix.cis.pitt.edu (Filip Gieszczykiewicz) writes: <..stuff deleted..>

>	I need a microprocessor with the following specifications:
>	o	8-bit (16-bit would be nice)
>	o	~10MHz clock
>	o	ability to address 64K of RAM (more would be nice)
>	o	BUILT-IN floating processor (that's the problem)
>	o	CHEAP (under $40)

>	I want 16 of these "chip"s to fit on one board (that's why
>	I worry about space)

Check into some of the DSP chips that have been around. I believe
one of the DSPs from NEC (7720 or 77230) used a 24-bit (non-standard)
floating point format and was real inexpensive. Plenty of dynamic
range for most control jobs.

AT&T has two 32-bit floating point chips: the DSP32 (NMOS) and the
much faster DSP32C (CMOS). I think you can get the NMOS part real cheap.

I *think* Texas Instruments has been working on a cheap floating point
DSP. The number TMS320C31 rings a bell as being a TMS320C30 with fewer
ports and buses, or something like that.

Of course, Motorola's DSP96001 & 96002 will be real expensive for a while,
so don't hold your breath for them.

Analog Devices may have released their ADSP-21000 signal processor by now,
but I haven't heard. (Has anyone?) They may have a low cost version of it.

I think OKI Semiconductor (Japan) was selling development kits for PC's
that included a board using a DSP with a 22-bit floating point format.
(I may have this mixed up with the NEC parts.)

If you only have an integer processor, U.S. Software (Seattle?) sells
re-entrant floating point libraries for several families. I've heard
their code is good and there are no royalty fees.

Good Luck. Mark Johnson (jjohnson@eola.cs.ucf.edu)

rick@electro.com (Rick Kozak) (04/25/91)

In article <1991Apr24.154831.9757@zoo.toronto.edu> henry@zoo.toronto.edu (Henry Spencer) writes:
>In article <119279@unix.cis.pitt.edu> fmgst@unix.cis.pitt.edu (Filip Gieszczykiewicz) writes:
>>	o	BUILT-IN floating processor (that's the problem)
>>	o	CHEAP (under $40)
>
>These two requirements are basically incompatible at present.  Built-in
>floating-point is a high-end feature, and comes complete with high-end
>prices. :-(  I'd be interested to hear of any exceptions...
>-- 

The Motorola 56000, although not a floating point processor itself, has av
available a 2K floating point library that Motorola claims executes most
instructions faster than a 25MHz 68882. Depending on quantity, the 56K isn't
all that much more than $40.

rick
.

bmw@isgtec.uucp (Bruce M. Walker) (04/26/91)

In article <119279@unix.cis.pitt.edu> fmgst@unix.cis.pitt.edu (Filip Gieszczykiewicz) needs:
< 	o	8-bit (16-bit would be nice)
< 	o	~10MHz clock
< 	o	ability to address 64K of RAM (more would be nice)
< 	o	BUILT-IN floating processor (that's the problem)
< 	o	CHEAP (under $40)

You need a DSP.  Look at AT&T's offerings and Analog Devices ADSP
series.  The Analog Devices advertise (in EET) at $40 prices.  I have
used TI's fixed-point devices, and they have a nice FP jobby, but it is
way out of your price range.

Either way, expect byzantine instructions sets that make the 80x86
series look orthogonal.  Most (maybe all?) have C compilers.

--
         "Remember, only *you* can prevent emacs!"
bmw@isgtec.uucp  [ ..uunet!utai!lsuc!isgtec!bmw ]  Bruce Walker

robf@mcs213j.cs.umr.edu (Rob Fugina) (04/29/91)

In article <119279@unix.cis.pitt.edu> fmgst@unix.cis.pitt.edu (Filip Gieszczykiewicz) writes:
>	I need a microprocessor with the following specifications:
>	o	8-bit (16-bit would be nice)
>	o	~10MHz clock
>	o	ability to address 64K of RAM (more would be nice)
>	o	BUILT-IN floating processor (that's the problem)
>	o	CHEAP (under $40)

I don't have a solid answer for you, but somebody else may be able to
finish it.  I think I heard from somebody once that the uP in the TI99/4A
had a uP with a built-in floating point processor.  If this is true,
somebody else should be able to tell you what uP it was.  The only thing
I wouldn't be sure about would be the 10MHz clock...

Rob  robf@cs.umr.edu

rwa@cs.athabascau.ca (Ross Alexander) (04/29/91)

robf@mcs213j.cs.umr.edu (Rob Fugina) writes:

>In article <119279@unix.cis.pitt.edu> fmgst@unix.cis.pitt.edu (Filip Gieszczykiewicz) writes:
>>	I need a microprocessor with the following specifications:
>>	o	BUILT-IN floating processor (that's the problem)
>>	o	CHEAP (under $40)
>I don't have a solid answer for you, but somebody else may be able to
>finish it.  I think I heard from somebody once that the uP in the TI99/4A
>had a uP with a built-in floating point processor.  If this is true,

It's not.  Sorry.  The cpu in a TI99/4 is a TI9900, and it definitely
doesn't have a floating instruction set.  Good grief, it just barely
has an integer instruction set :-P.  I hacked on the 9900 for a while,
and it was not pleasant.

-- 
Ross Alexander    rwa@cs.athabascau.ca    (403) 675 6311    ve6pdq
		`You were s'posed to laugh!' -- Zippy

graeme@labtam.labtam.oz (Graeme Gill) (04/30/91)

In article <119279@unix.cis.pitt.edu> fmgst@unix.cis.pitt.edu (Filip Gieszczykiewicz) writes:
>	I need a microprocessor with the following specifications:
>	o	8-bit (16-bit would be nice)
>	o	~10MHz clock
>	o	ability to address 64K of RAM (more would be nice)
>	o	BUILT-IN floating processor (that's the problem)
>	o	CHEAP (under $40)


If you are interested in a resonably recent processor, take a look at the
Intel 80960SB. 

	Summary:	(from the data book)

	32 bit internal, 16 bit external data.
	32 bit multiplexed address bus
	5 Mips at 16 Mhz
	4 Direct interrupt pins
	F.P. performance: (in usec)

			32 bit	64 bit
	Add		0.6	0.8
	Subtract	0.6	0.8
	Multiply	1.1	2.0
	Divide		2.0	4.5
	
	Price - talk to Intel. I think I heard a price of $25 in quantity
	mentioned at some stage (But that might have been the 80960SA).
	A GNU based C compiler is available.

	Graeme Gill
	Labtam Australia

deene@ace.austin.ibm.com (05/03/91)

> Subject: Re: microprocessor wanted

> >>	o	BUILT-IN floating processor (that's the problem)
> >>	o	CHEAP (under $40)
> >I don't have a solid answer for you, but somebody else may be able to
> >finish it.  I think I heard from somebody once that the uP in the TI99/4A
> >had a uP with a built-in floating point processor.  If this is true,
> 
> It's not.  Sorry.  The cpu in a TI99/4 is a TI9900, and it definitely
> doesn't have a floating instruction set.  Good grief, it just barely
> has an integer instruction set :-P.  I hacked on the 9900 for a while,
> and it was not pleasant.

Easy now...  The 9900 was quite a machine for its time, but ho-hum today.
I worked on the 9900 design team and when it was initially available, the
king of the microprocessor hill was the lowly 8080 - 8 bits and no hw
multiply/divide.  The 9900 was a full 16 bitter with separate TTL
compatible address and data busses and had hardware multiply and divide.
Its architecture was very similar to the PDP-11 except that registers
were in memory.  

The 99/4A was slated to use an 8-bit bus version (9985, years before the
8088) of the 9900, but TI had problems with the on-board oscillator/clock
system.  So the 9900 was substituded which caused mucho cost problems.
The 9900 family also had a single chip microcomputer version with 
on-board rom and ram called the 9940.  It pushed technology  and had
significant development problems.

Deene Ogden
(Boy, I wish we had patented more of our work.)  




It's too bad the TI 990, 99/4A, and 9900 were not big successes.