So now that I have my frame data in neat frame by frame arrays I only need to build some test hardware and write the code for my micro processor. The hardware part was easy – I simply used some stripboard to solder a classic one resistor per led matrix and since the MSP430 value line can only sink and source about 2 mA per pin I also added some npn transistor buffers to the columns for protection.

Now for the interesting part. Since I wanted to keep current consumption as low as possible I decided to drive my matrix one led at a time just like a progressive scan TV functions. With enough speed its is possible to drive a quite large display this way while keeping current consumption relatively low. This method is usually used to drive single color led matrices but  I wanted at least 8 bits of color depth so instead of just scanning the whole display and either turning the led on or off  accordingly I figured that I might get away with one PWM cycle per led.

My matrix was gonna be 5×10 so max brightness of it was gonna be 1/50 of the maximum brightness of the led, well at least in theory.

On to the code part. I implemented my ideas on the MSP430 Launchpad ( the MSPg2553 to be exact ) .  My algorythm looked something like this:

Timer0 has 2 compare match interrupts ( CCR0 and CCR1 ) and it can be configured to count up to CCR0 and then to reset itself. At CCR0 I take the value of first pixel of the frame and set CCR1 to equal it.  I also set the corresponding led to be lit. At CCR1 I set the led to be off and increment the pixel counter so that at CCR0 the next led will be lit and the corresponding value will be written to the CCR1 register.

This will continually cycle through the entire matrix and thus draw a frame. I also count the number of times the frame has been drawn and after the correct amount of times I increment the frame counter  so the next frame will be drawn. All this is done with iterative for loops because that’s how interrupts function.

My frame data is in arrays and selecting an array in a for loop is trickier than you might expect because you  need to employ pointers. Now most programmers will run in fear when you mention pointers but I didn’t find them to me difficult at all I thought they were rather neat and useful.

Here is all that is needed for my project.

I have an array of pointers that point to arrays. Yo dawg, right ? And to initialize this array of pointers I simply point the elements to the arrays I need. Simple really.

What usually confuses people is the syntax and there really is two things you need to know – ‘*register’ equals the value of the register being pointed to and ‘& register’ equals the address of the register. An example

And finally some videos of it in action.

Some friend of a friend is a pixel artist and did this animation for this display. I also gifted this to a someone for her birthday, hence the box.

 

After a bit of delay or laser-cut parts finally arrived. They are a mixture of aluminium and steel parts. I spent most of the day cleaning the components from bits of molten metal, a side effect of the laser cutting process. Next step will be to send them off to be anodized for protection and aesthetic purposes. This is all very exiting as this brings us ever closer to a complete robot platform.

Unfortunately not everything has been satisfactory. This is the second time we have ordered our laser cut components from FinEst Steel and both of the times we had problems with poor quality controll. The first time one of the sharp leg “bones” was misshapen from what seem like a CNC belt slipping. We did not recive a replacement. At the time we believed this to be an isolated incident and made no mention of it here.

This time one of the ordered components was not cut and in its place we received an overabundance of a different, similar component. I have contacted them about the problem and we shall see how it gets handled this time.

LEDs are cool. There is no doubt about it. But you know what is even cooler? A lot of leds. As a fun learning exercise I decided to attempt a LED dot matrix display with some limitations to make it interesting. Namely it needs to be capable of at least 8 bits of grey-scale and each pixel needs to be individually addressable. Why? Why for animations of course. Also the total current consumption of the device needs to be less than 10mA at 3V.

Ever since playing Mass Effect 2 I have been thinking of 8bit fire animations. After a couple of hours of searching the web for fire GIF-s I found a suitable animation. I happened to have 50 Led-s laying around so I decided that my resolution was going to be 5×10 pixels. Obviously the animation I found is much larger than that. Nothing a little image manipulation can’t fix. Using Photoshop I re-sized the animation, converted it to 8-bit gray-scale and extracted the frames into separate bitmaps.

That was the easy part. In order for those animations to be useful to me they need to be converted into arrays of brightness levels. Since I didn’t believe it to be an extremely important for me to learn all about GIF encoding I wrote this python script to handle the conversion.

Here is my code. It should be portable and should handle any sequence of images that follow the naming convention of ‘prefix+number(0..n).bmp’

requirements: pyton 2.7 & PIL(Python Imaging Library)

and here is some sample data that I got

Because of lazyness I used notepad++ to delete the last comma of the array (search and replace ‘,}’ with ‘}’)

Next post will explain why my script expands the original values to 12 bits and how pointers and arrays can made to play nice.

When it comes to electronic temperature measurement one only really has two options thermistors or thermocouples. Both means have their place in the arsenal of an electrical engineer but they fundamentally differ in the type of temperature measurement that they enable. One measures relative temperature and the other measures absolute temperature.

Because thermocoples can only measure relative temperature their interface circuits tend to be more complex and as such more expensive than thermistor circuits. For my reflow oven project I decided to go with thermistors as the interface circuitry only needs a ADC and resistor.


Thermistors are really nothing more than resistors that have been engineered to have a maximal resistance variance over a temperature range. There are two type of thermistors NTCs and PTCs. NTCs (Negative Temp. Coefficient) as the name states has a negative temperature coefficient and are most often used for temperature measurement. PTCs usually serve as resetable fuses.

Measuring resistance with a micro controller poses some interesting challenges. Normally one would use a known constant current source and measure the voltage drop on the resistor but something cleverer can be done with resistors. Since the temp-resistance graph for thermistors is roughly exponential and the output voltage-resistor graph of a voltage divider resembles the reciprocal of negative exponential graph, combining them yields something quite surprising. Namely we see a nearly linear temperature – voltage graph. No need for look up tables.

I had trouble finding thermistors that met my strict specs of being able to handle 300C and not costing 10 euros. Finally my friend Jaanus gave me one cheap thermistor that according to the manufacturer only did up to 150C. Some testing revealed that this thermistor could handle temperatures in excess of 250C and when compared to a FLUKE thermometer was only off by a few degrees.

A word on using thermistors in ranges that are out of spec.

First of all in order to use the thermistor effectively a proper selection of the accompanying resistor needs to be made. This is easiest with a reference thermometer. Since the voltage – temperature graph is linear only near the middle it makes the most sense to choose the second resistor to be equal to the resistance of the thermistor in the middle of the temperature range you are interested in.

Secondly since you are now using components that no longer are in spec a recalibration needs to be done. Since we now can presume a linear proportionality we only need to take calibration measurements at the extremes of the temperature range we are interested in. It is also important to note that the smaller the range, the more accurate the reading will be.

Join us next time as we discuss the programming of a simplistic PID controller and microcontrollers in general.