November 2004, Issue 172

TCP/IP Stack Solution
A Detailed Look at the CMX-MicroNet

As many of you already know, when I’m not putting words into Circuit Cellar magazine I’m moonlighting at EDTP Electronics, which is focused on designing and marketing inexpensive microcontroller-based Ethernet devices from an Internet-based storefront. As a Circuit Cellar columnist and an electronic barkeep (I’m the “support” behind support@edtp.com), I’m in direct contact with many of you. Aside from enjoying shooting the breeze, I also like that I get to know firsthand what you’re interested in and what problems you’re facing. Hopefully, that makes my Circuit Cellar columns interesting to you.

The line of Easy Ethernet devices are all supported by a minimal homebrewed TCP/IP-UDP implementation. The baseline Easy Ethernet products are powered by 40-pin PIC and Atmel microcontrollers that contain between 368 (PIC16F877) and 1,024 bytes (ATmega16) of RAM complemented by 8 (PIC16F877) and 16 KB (ATmega16) of program flash memory. That means the Easy Ethernet Internet protocol and application code must be minimized to leave enough program memory space and RAM for your application code. 

To accommodate those of you who need a bit more application space, I employed the services of Microchip’s next generation of microcontrollers, the PIC18xxxx series. Atmel coders were also considered in the upgrade process because an easy migration path up from the ATmega16 used on the Easy Ethernet AVR is provided by the pin-compatible ATmega32, which doubles the amount of RAM and program flash memory contained in the ATmega16. When moving from the PIC16F877 to the PIC18F452, the Microchip PIC18xxxx coder gains a little more than 1 KB of RAM while matching the Atmel coder’s 32 KB of program flash memory.

I didn’t design the TCP/IP stack for industrial use. The mini stack I wrote is intended to show you that complicated TCP/IP stuff can be done on the cheap with simple code. My simplistic TCP/IP stack works well both in a local LAN environment and on the big wire. However, I’ve seen many of you, especially college students, make modifications and add features to my little garage-built TCP/IP-UDP stack to improve its usability (and to finish those senior projects).   

For more demanding applications, it may be advantageous to deploy a more robust TCP/IP stack. However, in a microcontroller environment, space is always constrained, and you still will be faced with squeezing in just enough stack to support your application. That means the stack you choose or write must be modular. For instance, if you don’t need FTP, there shouldn’t be useless FTP code floating around taking up space in your stack. Another point to consider is the setting up and tuning of the stack. The stack’s knobs and buttons should be located in a logical place. They also should be easy to understand and use.

Writing a serious TCP/IP stack is not child’s play. If you’re not into rolling your own TCP/IP stack, and if you need a bit more than a minimal TCP/IP stack for your application, CMX Systems is a good place to start.