How to choose a suitable MCU for a project?

[reiniheini]

How to select a MCU

Hi there!

i'd like to know if there is a way to select a MCU on a given problem!

The discribed way should have a result which indicates the significant properties of the MCU.
I mean properties like: numbers of counters, hardware multiplier......

I hope you'll understand my question.
I think my English isn't so good!

Thanks in advance

Bye REINI

 

[Analyzer]

I am sorry, may be i didnt understand your question.You may find out useful information to choose best fit mcu in manufacturer catalogues.For example MicroChip sends a catalogue and data cd's freely.If this is not your answer please elaborate it.

Analyzer

 

[reiniheini]

Hi!

thanks for answering me!

My problem is not yet to find a special MCU of a specifiy manufacturer.

I would like to know how I can predict what features a MCU must have to fit to my problem.
In fact the main problem is which MCU do i need to solve a FFT operation and a neural net in a given time period. (many addition and multiply operations)
So i hope you'll understand that the peripheral options of the cpu are not so important. I think it's more a question which core is the best!

Bye Reini

 

[Analyzer]

@Reini,

You even can not predict it, you have a problem then you develop answers to you project.If i understand you, you would like to know how to choose MCU type.You need more experience about this subject then you can choose it automatically.

Analyzer.

 

[psubiaco]

How to select a micro??
If you need to design a microcontroller card for high volume production, two things must be kept in mind:
microcontroller cost and programming time.

For simple boards for those an 8 bit microcontroller is sufficient, a good way to achieve a cheap and fast solution is to use Holtek microcontrollers.
You can find all details on the site www.holtek.com : i suppose that holtek is a Microchip competitor, with RISC microcontroller that are a little better than microchip, cheaper, and also they provide an In-Circuit Emulator at very very low cost, with IDE interface, C/ASM compiler with source level debugger.
Also, programming time is 5-6 times faster than PIC microcontroller!

Are these good reasons to select a micro instead of another one??

 

[Analyzer]

Hi,

You maybe right but other important reason to choose a device is reachanable.I can't find devices such as holtek's devices in my country.So, if i choose this, i should always order at their online shop.This is no good you know.

Analyzer

 

[psubiaco]

Hi,

You maybe right but other important reason to choose a device is reachanable.I can't find devices such as holtek's devices in my country.So, if i choose this, i should always order at their online shop.This is no good you know.

Analyzer

Ok, you are right! So the best microcontroller is 8051 because it is widely product by any semiconductor factory.

Regards. Paolo

 

[Antharax]

choosing the best uc for your application depends on allot of things

-max power usage your application allows
(use MSP430 and derivates, Xemics for ultra low power systems)
(8051 is a Risc and consumes allot of power, although there are ...)

-availability and cost
(AVR's and PICs are pretty powerfull things but cost allot, that might be some overkill for some applications) Normally most types can be delivered in large quantities (100k or more) but you should check this if required!! Also don't choose an "end of life" uC because this could cause the obvious problem that you cannot buy them any more when your design is finished.

-designtime
when everybody in your company has experiance with AVR it might be better not to choose an PIC. Then you can ask ppl in your company. Also most of the time mainstream uC (8051, AVR, PIC, motorola's, ...)have better compilers (also free ports from gcc)

-"numbers of counters, hardware multiplier"
depends on what you're using it for! When you need raw computing power, you'd better choose a DSP. Otherwise an simple uC is good.
Memory is a bigger issue for uC. Programmemory is most of the time sufficient (you can get a big application in a 16k uC) Ram might be a problem, depending how you use it and how much is available.

and then there are some other issues like
Do you want to be able to reprogram or is it a final version of the code
Do you need In Circuit Programmability
Do you want to be able to write a bootloader (self programming should be possible)
Do you need UART, USB, Ethernet, I2C, ...
Do you need hardware timers
Do you need a watchdog (I'd say yes ;-) )
And i'm sure there are allot more issues

Choosing the correct uC isn't a simple problem ;-)

good luck
Antharax

 

[atmel2000]

Antharax has made a good response.

I'd like to EMPHASIZE end of the life.

If you want to produce this product more than 2-3 years you
have to check from how many years this micro is on the market.
It's very important!

For example if you select AT90S8515 or AT90S8535 - wrong!
they are almost obsolete!

There is a big list of obsolete or almost obsolete micros and you
have to verify this instead of working for nothing.

Now we are doing this: after 2 years we have to rewrite all the
software and change the pcb for more than 20 products because
someone has selected the wrong micro!!!

Other example is ATTINYxx - I don't want to say the number. We
received Atmel's CD in 1999 - it was there as preliminary information
and in the next CD from 2000 it was obsolete ;-(((

Anyway, I love Atmel!

I have a list of 30+ other points of view but this is for detailed discussion.

regards, atmel2000

 

[ME]

When you choose a MCU it is also important to look at the compilers and/or assemblers availible for the MCU.
How good are the compilers and at what cost. If you already have a compiler that support specific MCU's, then maybe you should choose one of these MCU's.

 

[Analyzer]

Yea,
I also think the cost, effectiveness, accesibility, technical support, design and coding time for mcu, development kits price, and its life are very important to choose an mcu.So, if you have enough experience you can easily compare devices which is suitable for your project.If you have very big project and no experience try to ask in technical support sections of great companies.Rgds,

Analyzer.

 

[Rayengine]

Your budget has not been well defined.

In traditional and small quautity design, many peoples will choose 8051 families and PIC series. However, for some application like embedded LCD controller, many designer will choose Japanese one. For huge quautities design, now it will consider Japanese and also Taiwanese.

BR

 

[reiniheini]

Hi there!

Thanks to all of you!
The post you've worte was quite useful to me!

At this point I have to say that I'm well introduced to Mircochips PIC -controllers. But as you said, one point of my decision is the power consumption of the MCU. And the Pics consume more power then I can accept.
So I took a look to the MSP series from TI. They seem to be very powerful controllers at very low power consumption.

Today I know more about the features the controller must have.
It must be able to calculate more than 1000 multipier and 1000 addition operation in less than 3 seconds. PICs are tooo slow.....

So I think you'll understand that even if you are well experienced it's not easy to decide which MCU to use.
A new controller family means: Buy new
development environment
develop with a new system needs more time
and so on

So my actual question means. How can I know if one controller can fit my requirements before i buy it...

After I've talked to many people I know it's moreover a dataflow problem.
This could also solved by a DSP.

So im now on the point where i have to decide which controller familiy is best for my requirements an then i have to look on the price too.

OK That's enough at this point.

I hope this little discussion helped not only myself.
Again I have to thank all of you

Bye reiniheini

 

[elcielo]

the Eight-bit Processors Special Report PDF

the Eight-bit Processors Special Report PDF

**broken link removed**

AVR selection guide

**broken link removed**