[req] Need suggestions for a micro

Jul 20, 2004 23 Replies

I have a client who is currently using a NEC V-25 at 16 MHz. This micro is no longer adequate for several reasons: (a) it's no longer manufactured, and (b) his expanded application requirements need more computing power.



Can anyone suggest a replacement. Here's what we need:


  • At least 20 address lines


  • Can't be "PC-based", including PC-104. (ie it can't resemble an x86 any closer than the V-25.)

  • 16-bit preferred


  • at least 33 MHz or better (66 MHz preferred)

  • reasonably priced and well supported development tools. (perhaps like AVRGCC for Atmel from avrfreaks.net)

Any help appreciated Sean


Huh? What kind of requirement is that? Why would they rule out an otherwise nice and applicable platform based on it being too PC-like a design? Is there any more rational point to this requirement than "used that last time, hated it, don't want to see anything like it soon"?

As in: at least, at most, or exactly 16 bit?

Given you haven't homed in on any particular architecture yet, specifying MHz'es up front is quite silly. Heck, even Intel finally had to accept the fact that there's no such thing as "real GHz" --- after having spent millions of dollars advertising there was.

Hans-Bernhard Broeker (broeker@physik.rwth-aachen.de) Even if all the snow were burnt, ashes would remain.

That's a very odd requirement. May I ask why it can't resemble an x86?

So you don't want to look at cheap 32-bit parts?

Well, specifying MHz is rather meaningless unless you're also specifying a CPU architecture.

I would recommend that you a look at the Renesas (nee Hitachi) H8 series. The low-end 32-bit ones (H8/300H "value" series) start at $4 for a 25MHz part with 128KB flash and 4KB RAM. IIRC, the newest "SX" version of the core runs at 50MHz.

Development tools are free: Gnu gcc and binutils.

You should also take a look at some of the low-end ARM based uControllers.

Grant Edwards grante Yow! I want another at RE-WRITE on my CEASAR visi.com SALAD!!

Sean,

Yes, I can suggest a replacement: a 683xx or Coldfire processor. It has a clean architecture, no messy segmentation to deal with, and a plentiful supply of tool vendors: SDS in Illinois, HI-TECH,ourselves and others whose names I can't bring to mind at present.

The registers are 32 bit but you can use them 16bit-wise; use the ".w" specifier on instructions.

I think it can meet that.

That really depends on whether you mean supported or maintained. If you want decent support, a vendor is going to have to charge at the very minimum $500 to be able to afford this without charging extra. If you mean "currently maintained", that is a different question.

If you want to go for GCC tools, you'll have to frequent GCC user groups to get help.

Not at all,

Dave Bardon, Avocet

If you're interested in another x86, I've had no problems with the RDC R8822. We've got 2 boards based on the design, one of which replaces our V-25 board. The R8822 is a pin-compatible replacement for the AMD 186ED. It runs at 40mhz and will interface directly with cheap DRAM.

For a non-x86 processor, my vote to an ARM, e.g. Atmel AT91 -series.

The question of MHz is here somewhat moot, a x86 uses the clocks very differently compared to real RISC processors (like ARM).

There are full GNU toolsets (GCC, assembler, linker, debugger and utilities) for ARM running on different host systems.

Tauno Voipio tauno voipio (at) iki fi

The reason is that getting approvals and certifications for PC-based equipment in my client's market is more difficult than for proprietary boards.

Agreed. These 2 items are purely for better performance.

Ideally we'd like a V-25 running 3 or 4 times as fast as currently, if such a thing existed. Actually I've already recommended an ARM-7 running at 66 MHz. However there's several drawbacks to this; he needs a 5-volt part and ARM vendors seem to be moving to 3V; also getting the ARM-7 GNU toolchain up and running seems like quite an effort. My approach would be to get an eval board with a fast micro, pull the V-25, RAM, EEPROM and EPROMs out of his current board and try replacing it with the eval board running a firmware port in flash.

Sean

Not all ARM variants are moving to 3V, look at the new Motorola MAC71x1 family : this has on-chip regulators, and a comprehensive temperature spec as well. How much code space do you need ? - the single chip ARMs uC come from

32K to 512K on chip FLASH

-jg

Thanks for the help, I'll check it out.

Currently I'm using close to 100k bytes of code, but I also need

256k RAM.

Sean

Have a look at the Toshiba TLCS-900, this is heaven to write code for, its what a Z80 would be if it was 32 bits.

remember LD A,(HL), well now its LD A,(XHL+)

where XHL is 32 bit reg with auto-incr......

If it's going to be a new design, you might try a core module instead of designing from scratch.

formatting link

formatting link

MAC71x1

If you're going to address more than 65 kbytes of memory, you should seriously think about using a 32 bit part, to facilitate the unavoidable address arithmetic.

The ARM T variants (e.g. ARM7TDMI) are able to run 16 bit (Thumb) code for saving on code space.

The low-voltage core of the AT91 ARM consumes so little power that a simple linear regulator is more than sufficient.

Tauno Voipio tauno voipio (at) iki fi

I may be a bit dense today, but to me that sounds like you're saying it's not your client itself who's gone nuts, but actually the entire market they're working in? It's one thing to say "Ordinary PCs won't cut it in our market, because ...", but going out of your way to effectively forbid anything that smells like a PC does seem excessive.

Someone call the "no discrimination" amendment (5th?) on the persons ruling that market, please. ;->

Hans-Bernhard Broeker (broeker@physik.rwth-aachen.de) Even if all the snow were burnt, ashes would remain.

Perhaps Sean can give us more information on what market area this is, and if there is an historical basis for the aversion ?

Such an assertion did raise eyebrows, but it may have been something that smells like a PC, could run something that smells like windoze ?

-jg

I agree -- in fact I've suggested PC-104 several times. However, the client insists on using his own proprietary boards with a "proprietary operating system" (essentially a round-robin task scheduler). I'm not involved with the approvals so I just took his word for it.

Sean

I can't say too much about it except that it's a specialty security application and he tells me that any platform which could even _potentially_ run any Microsoft o/s would instantly alienate most of his customer base.

Sean

It's not just ARM vendors. The entire universe is migrating to lower core voltages. Many ARM parts have 5V-tolerant I/Os, can you consider that as an option? You can supply the core and I/O ring voltages (1.8V and 3.3V typically) with small linear regulators; the currents involved are miniscule.

Not really, at least - not any more. is an excellent resource.

Should be able to do this with an ARM part that has 5V I/Os...

I was able years ago to pursuade a client to use PC104 and it has worked well for them. But what is happening is that feature creap in the available PC104 CPU from the vendor (Megatel) has kept the price climbing as the mfg. drops the older/simpler boards for more MHz and MB. Result is I keep having to defend that decision (for free), but I'm always looking for cheaper alternatives. - RM

Are these Motorola ARMs actually available ? I found some datasheets for them, but could not find anyone who lists them on their site. Can anyone recommend a supplier from where one might be able to buy a couple ?

Regards Anton Erasmus

If you want x86 Try AMD They have embedded stuff like the 186EM. They also have the Elan series 386 and up.

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required