Group Actions
Engagement
  • Author Author: txbeech
  • Date Created: Date Created
  • Views 1636 views
  • Likes 0 likes
  • Comments 5 comments
Related
Recommended

Is a 32 bit or 8 bit MCU better for your application?

txbeech

When beginning a project that you know will contain an MCU there are so many options that it may be overwhelming.  One of the biggest questions is what size MCU should you use?  I will release a series of articles that will shed some light on this issue faced by so many designers and developers.


Uncovering the differences and architectures

The 2 ends of the scale for bit size of microcontrollers are 8 bit and 32 bit. Bit size in this case means that the MCU processes an 8 or 32 bit data word at a time and also dictates the register, address and bus size.  Both have advantages and disadvantages and throughout the series I will talk about both. The most common architectures within these categories is 8051 architecture for the 8 bit and ARM architecture for the 32 bit.  There seems to be a popular belief that 8 bit MCUs are on their way out. With the release of new products such as the EFM8 MCUs, I have heard many people question using an 8 bit in their design.  Well, not only will this series distinguish the best cases for each MCU but also shed some light on why the 8 bit with its 8051 architecture still thrives in many applications.

image

 

First, let’s discuss some of the more general and obvious differences, namely size, cost, and ease of use.  Before we begin there are fine lines that can be drawn between which is better for a certain application.  If a systems demands >64 KB of RAM then the choice to use a 32 bit MCU is an easy one to make.  If the system is ultra-cost sensitive then using an 8 bit MCU is the correct decision. However for the application where this clear line can’t be drawn there are deeper things that must be considered.

 

Ease of use and cost

image

The picture is very exaggerated but it points out the general truth that the ARM core, which most 32 bit devices use, is easier to use compared with the 8051 core, which the majority of 8 bit MCU use. The ARM core 32 bit MCUs utilize familiar compilers, have a long list of available libraries and the perhaps the most important of all, have unified memory, make coding on the ARM an easier task than on an 8051. However, you pay for this as the price for ARM based devices are normally higher than the 8051. The most aggressively priced 8 bit 8051s out there hit ridiculous lows and can be bought for cents. But, with ease of use comes a quicker time to market.  For some products time to market is a deciding factor in its success therefore for these end products paying a bit more for the 32 bit can be well worth it.

 

Size

An advantage of the 8 bit devices is that they are generally smaller devices. This becomes an enormous edge over the 32 bit if the final product is space constrained. If you were designing a wearable like a watch, an 8 bit 8051 could allow the device to be smaller and have the same functionality compared to a wearable with a 32 bit. To give the edge even more to the 8 bit, some manufacturers like Silicon Labs use chip scale packages (CSP) with their 8 bit devices which decreases the size significantly. Their CSP 8051 is 1.66x1.78 mm2 which is one of the smallest on the market. This compared to their smallest 32 bit, the Tiny Gecko which is 4x4 mm2 is 4 times the size!


image

 

A theme you will see multiple times in this series is that knowing your application and final design is perhaps the most important thing when choosing the MCU for you!

  • Nice post.

     

    Over the years I have built some very impressive implementations using 8-bit microcomputers.

    I used an old 8085 to control a 400 by 400 pixel display overlay for a FLIR camera on which I put real time helicopter attitude information.

    I used a second 8085 to collect the data and calculate a full six degree of freedom control for the FLIR camera so that it was always following the flight vector.

    Granted, I used an AMD math co-processor for the trig functions, but the system was perfectly fine for search and rescue work for my customer.

     

    DAB

  • in reply to johnbeetem

    It 's ture Mark, NXP is producing the smart family of devices; there is also one smartest of the LPC series that has less pins and less features, but it includes a customisable system so you can assign to the pins only some of all the internal MCU I/O features.

     

    Enrico

  • in reply to michaelkellett

    Looking over his previous posts I would say... Yes.

  • Mark - do you work for Silabs ?

    In any discussion of 8 v 32 bit micros you should at least mention Microchip (world's biggest player in 8 bit micros), Atmel (8 bit AVR is probably the best known modern 8 bit architecture) and the market leaders (ST, TI, NXP) in 32 bit ARMs.

    Physical size is, for most applications, not  a key determinant in processor architecture choice - you can buy an ARM from Freescale in a 1.99 x 1.94mm package (MKL02Z32CAF4R).

     

    If I'm choosing  a micro (or any other key component) for a project I start off with the things it must have to meet the project requirements, once I have a short list on that basis I can refine it on the basis of cost. time to market, reliability, other project "nice to haves" and so on. Processor core notional bit width very rarely appears directly in such considerations but may be a factor indirectly, for example: integer maths performance will be affected by register width which may matter in some applications.

     

    MK

  • The NXP LPC1101LVUK and LPC1102LVUK (ARM Cortex-M0) have a 25-ball 2.17 mm x 2.32 mm package.  That's the smallest I've seen for a 32-bit microcontroller.