2016-06-12

An ARM experience

Well... we're back at writing.
Two years ago NXP lauched a 8pin dip ARM with 4K FLASH and 1K RAM, lots of people started doing things with it, as most other ARM processors (even Cortex M0) are multi-leaded-smd-monsters. This one, on the other side, looks like an ATTiny85 in its PDIP8.
This weekend decided to try it out. I just followed the herd, although not the common route. I followed this tutorial and bought this "starter kit".
I used one LED and the two switches that come in the bundle to wire one LED output as the example and one switch for the RESET and the other to force the bootloader.

I used two machines for the LPCXpresso installation (you do have to register to get a free non limiting license, but if you can't be bothered for this processor you don't need to). You will need to install some extra packages, follow the procedure in this link. I managed to have LPCXpresso running both in OpenSuse 13.2 and Leap 42. In Leap you need to install the 32bit packages instead of the x64. Download the LPC810 Codebase example from here, import the project and build all. No joke, it is that simple.
Since I'm running linux, the alternative loader is this one. Like most linux command line programs, it is crude and simple. Just download, extract and type make.
Just copy the example command line with the your hex program (it is in the Release directory in the workspace project folder).
and yes... it blinks!
The processor has lots of embedded resources but only 4k of code might be the limiting factor. Still it is very interesting, in particular if you can save more power than an ATTiny and when needed produce way more MIPS.
For a first post after such a long hiatus it was not bad... more to come :-)


2015-09-06

Building the "new" MSPGCC from TI and RedHat

Ok, after a system update (opensuse 12.3 to opensuse 13.1) I was without my MSP430 compiler and launchpad connection. Also since the last time a new compiler version appeared (from 4.6.3 to 4.8.0) so I thought I could build (install) a new one.
TI also got together with RedHat to build a new opensource msp-gcc compiler. It took them quite some time but we've got them.
For those that like to try linux installers (beware: sometimes things go up in flames and you don't know why.) you can download the installer from here and run it as
su.

If that doesn't work for you... you can always build from source, these instructions (in spanish) appear to unlock the code .

On the other hand I also tried the install from source! After a couple of attempts I manged to do it, here's how:
- download the sources from here and untar it. (in the following commands remove the "" in the comand line;
- create a new directory like "mkdir build";
- "cd build";
- "../sources/tools/configure -target=msp430-elf -prefix=/opt/cross/msp430"
- "make"
- go and take a looooooooooong coffee (might as well go to work and come back later).
- "sudo make install".

A few caveats I found out the hard way:
- the target must be specified as msp430-elf otherwise gdb won't compile;
- the prefix can be other, it is the directory where all the files will be placed.

I haven't checked actually using the cross compiler for anything but I'll try soon. Also there are some heather files that must be placed in some directory (according to TI) but they don't tell you how and where... I will also have to check that...
Furthermore it appears that mspdebug is not built... also need to check that...
But the compiler runs and gdb also.

Using TI Launchpad in linux

Texas Instruments made a good effort in having free and community tool chain for their family of MSP430 low power microcontrollers. Before the tool chain had to be build by the user and that had the usual problems (dependencies, headers, etc).

You will need to register with TI in order to download the installer (apparently you can download the source without registering, current version (for this post) is 3.04.05.01. You need to change the permissions on that file (chmod 777), then run it.
The launchpad is TI low cost entry point development tool, for 10USD you get two processors, a PCB with embedded programmer and debugger and you can also get plenty of add on modules. The new launchpad has a different interface to the first ones (one row of connectors to two rows on each side), in order to support processors with more IO.

Once the tool is installed you can go to the examples directory and search one for your launchpad. There are still some issues, but they can be alternatives.
I didn't manage to connect the gdb_agent_console to gdb and without this connection debugging and programming are impossible. The previous tool chain needed another agent to interface to gdb (mspdebug), this one still works reasonably well.
You still need to program the processor with mspdebug, then enter gdb mode. In another terminal start gdb, connect to target remote on the correct port and then continue.

2014-05-24

Current increase for MAX6957

This will be a mostly Analog Electronics post...

In a project I made with friends we had a Raspberry Pi driving a MAX6957 for driving some LEDs. The MAX6957 was connected to the SPI bus of the RPI and the driving functions were developed in python (if you're interested in this you can post a question in the comments but if google it it's faster).

We were driving a few piranha RGB LED but then we bought a bigger LED (something like this one). In order to keep the software simple and unchanged I wanted to use the MAX to drive the bigger LED. If the MAX drives the LEDs with a current source that can be stepped (in 16 equal current steps), so I would have measure the current of the MAX and amplify it for the new LED.
This is what I've done:



First I have a current mirror, as the MAX sinks current I used PNP transistors, on the other side I have a small resistor to convert current into voltage.
This voltage drives a voltage to current converter with a protection for over current (in case the MAX is removed, goes crazy or it's output is miss configured). Then repeat for each of the 3 channels.