The MCB2100 ARM7 board from Keil

This is the MCB2100 development board from Keil:




The board itself is pretty simple. It contains an LPC2129 microcontroller from NXP (Philips). One standard 20-pin JTAG connection, one ETM connection, 2 serial ports and 2 CAN ports. There are also leds and buttons and the usual stuff on the board



There is also a ULINK USB-jtag in the package :)




The included CDROM contains a "light" version of Keil uVision for ARM that works very well with this JTAG unit. Note however that you might need to download the latest drivers from Keil website, the old ones sometime BSOD on laptops :(

---

The schematics below are copied from the included PDF documents:





While the board is pretty basic by itself, itis easy to connect it to add-on or self made stuff. Consider for example the "TV-Based Oscilloscope" project by Ilya Mamontov:

TDA1543, dead on arrival?

I have hard time getting my TDA1543 to work. The DAC is controlled from an ATTiny11 running on the slow internal oscillator (1MHZ), which means that it is unlikely to be a timing problem.

This is the start sequence

This is a whole sample (in stereo, right channel is set to 0xFFFF):

I tried multpile ICs, all show the same behaviour:

Aref is at 1.9-2.3 volt

AOr/AOl are both (!) at 0.9-1.1

VCC is regulated 5.0, but I am seeing 4.9 and 5.1 p-p sometimes

I wonder if all my ICs where DOA?

Tool Hell

I was checking out the common ARM development toolchains. Seems like the whole industry of full of crapware. If these are the successful ones, how bad are the less successful tools??


I tested IAR System Embedded Workbench (they have a free Starter version). It couldn't even get simple things such as memory layout right. Not to mention that the simulator freaked out as soon as I decided a 8000 bytes heap was too much for a MCU with a total of 8 KB RAM (LPC2104).

I must also mention that their C compiler sucks big time. Without any prior knowledge of ARM7 assembler, I was able to make hand-code things to go 50% faster and become 100% smaller.

Then I looked into Keil uVision. I have seen Jell-O deserts that are more stable :(

I have to admit that uvision have a really good ARM simulator. I specially enjoyed the "Logic Analyzer" function :)

Next was Rowly's CrossWorks. It uses GCC, but has its own IDE above it that take care of the configuration-madness.

I could not get the simulator to work. For some reason it gives me data aborts during the startup sequence (i.e. in crosswords own Philips_LPC21xx_startup.s). Turns out the simulator have hard time understanding that 0x4000000 is actually a valid memory address and is in fact writable. And the simulator is completely undocumented, so I cant figure out how to fix this :(

Then we have the hurdle of homebrew GCC based toolchains. The more famous ones are WinARM, YAGARTO, GnuArm, and DevKitPro for the Game Boy Advance people. While the GCC itself is very solid, the project configuration and things like setting up GDB and simulator will drive you made. I am sure there are tons of documents and great tutorials on this subject out there, but the ones I tried did not help me much.

So how hard is it to create a toolchain that works correctly out of the box?? With the amount of money these guys charge, why cant they test the damn thing before shipping?

Note I haven't tried the ARM Develop Suite yet, so I am giving it the benefit of doubt. After all, there must be _one_ good toolchain out there given the popularity of ARM?

 

Here we go

This is the first blog post, and I guess I should start by explaining why this page is here...

This is a technical blog. You will read about things like ARM7 and FPGAs here. You will not read or see anything (or hear or smell) about Britney Spears here [I think the two last words just improved my page rank].

First of all, I needed some sort of log book, a place to keep my ideas alive. Furthermore, I am hoping to come in contact with people that have similar interests.