I believe strongly in plotting the data that I obtain, because plots tell you a story and they don’t mislead you like summary statistics can.

Often I want to plot data the arduino has acquired, but I find existing solutions don’t suit my needs. So I went out and made my own solution which writes a GNUPlot script in response to plot commands. I integrated the library with my general Arduino Toolbox (WDArduinoLib) and am releasing it under the GNU 2.0 License.

Goals

• Professional graphics
• Graphs manually editable after plotting for change colours etc.
• Arduino outputs data that  other people can defiantly grab and use (just paste into a cloud based GNUPlot terminal)
• Allow embedded linux system to offload the plotting of graphs from arduino collected data
• Easy straight forward API
• Linux, Windows and Mac compatible.
• Quality API documentation

Features

• Configurable output
  • You can use a Serial Port, SD-Card,  Telnet, TCP-IP, IrDA, whatever. So long as it derives from the Arduino Print class (ie has Print and Println()) this library can output a graph to it
• Uses GNUPlot as a server to execute the plot script this API outputs.
  • GNUPlot is the standard for scientific publication, it outputs presentable graphics ready for both websites, and software manuals.
  • GNUPlot has been around forever (1986) and is still actively developed.
  • GNUPlot can run on the Raspberry pi, PC, Mac or even online via http://plotshare.com/
• Extensible
  • Easily over-rideable classes to add new plot types or different output terminal (eg Matlab)
• Built in quality colour palettes designed for graphing.
  • Sample palettes for showing importance, of keeping different coloured categories at the same intensity.
  • Spectral palettes , optimised for on-white graphing.
  • Colour-blind compatible palettes .
  • Can still use arbitrary 24bit RGB colours if you like.
• Unit tested library to support quality software development
• Template driven code designed for maximum data flexibility
• Memory Efficient.
  • Use of progmem strings supported throughout.
• Useful, stable data format that does not rely on someone’s pet project still existing in 5 years time.
• All input, including text, is validated.

Examples

This library is simple to use and produces clean results with minimal configuration. I present several basic examples that show the API in use and the results obtained.

Example: Basic Line Graph

Line plots of time series data are probably the most important feature any graphing API targeted for micro-controllers can offer. This code creates such a plot. It demonstrates the use of multiple line styles.

//main
GNUPlotLineGraph gnuLinePlotter(&Serial);
PlotDampenedSineWaves(gnuLinePlotter);

//method
void PlotDampenedSineWaves(SerialGraph& plot)
{
	//setup graph
	plot.newGraph();
	plot.enableSaveImageFile(true);
	plot.setTitle(F("Dampened Sine Wave"));
	plot.setXAxisName(F("Time (Sec)"));
	plot.setYAxisName(F("Voltage (V)"));

	//setup series
	plot.setSeriesName(0, F("Slow decay"));
	plot.getLineApperance(0)->setLineStyle(LineStyles::SolidLine);
	plot.getLineApperance(0)->setLineWidth(4);
	plot.getLineApperance(0)->setRGB(GraphColorsSchemeBlues::Blue1);


	plot.setSeriesName(1, F("Medium decay"));
	plot.getLineApperance(1)->setLineStyle(LineStyles::DashedLine);
	plot.getLineApperance(1)->setLineWidth(3);
	plot.getLineApperance(1)->setRGB(GraphColorsSchemeBlues::Blue2);

	plot.setSeriesName(2, F("Quick decay"));
	plot.getLineApperance(2)->setLineStyle(LineStyles::DotedLine);
	plot.getLineApperance(2)->setLineWidth(2);
	plot.getLineApperance(2)->setRGB(GraphColorsSchemeBlues::Blue3);
	
        //plot graph
	for(int i=0; i<=720; i+=3)
	{
		float c = cos(DEG_TO_RAD*(float)i); 
		float y1 = Dampen(i, c, 1, 1.0/360.0); 
		float y2 = Dampen(i, c, 1, 1.0/240.0); 
		float y3 = Dampen(i, c, 1, 1.0/180.0); 

		plot.plotDatumXYn(i, y1, y2, y3);
	}
	plot.finishGraph();
}

Example: Plotting Points (Scatter Plot)

A scatter plots, with styles colours and trimmings.

//main
GNUPlotPointGraph gnuPointPlotter(&Serial);
PlotPointNoise(gnuPointPlotter);

//method
void PlotPointNoise(SerialGraph& plot)
{
	//setup graph
	plot.newGraph();
	plot.enableSaveImageFile(true);
	plot.setTitle(F("Sine Waves with Noise"));
	plot.setXAxisName(F("Time (Sec)"));
	plot.setYAxisName(F("Voltage (V)"));
	plot.setShowGrid(true);

	//setup series
	plot.setSeriesName(0, F("Pin 1"));
	plot.getLineApperance(0)->setMarkerStyle(LineMarkers::SquareMarker);
	plot.getLineApperance(0)->setRGB(0xff0000);

	plot.setSeriesName(1, F("Pin 2"));
	plot.getLineApperance(1)->setMarkerStyle(LineMarkers::CrossMarker);
	plot.getLineApperance(1)->setRGB(0x00ff00);

	plot.setSeriesName(2, F("Pin 3"));
	plot.getLineApperance(2)->setMarkerStyle(LineMarkers::CircleMarker);
	plot.getLineApperance(2)->setRGB(0x0000ff);
	
        //plot graph
	for(int i=0; i<=720; i+=5)
	{
		float c = cos(DEG_TO_RAD*(float)i); 
		float y1 = c + Dampen(i, ((float)(random(100)-50)) *0.1, 1, 1.0/240.0); 
		float y2 = c + Dampen(i, ((float)(random(100)-50)) *0.1, 1, 1.0/180.0)-2; 
		float y3 = c + Dampen(i, ((float)(random(100)-50)) *0.1, 1, 1.0/90.0)-4; 

		plot.plotDatumXYn(i, y1, y2, y3);
	}
	plot.finishGraph();
}

Example: Plotting Arbitrary Lines

Sometimes your not using time series data, and need to plot lines between arbitrary x/y positions. Many solutions for the arduino breakdown at this point. Here I demonstrate such a plot (based on the logarithmic spiral). Note I am using lines with point markers to show the actual points plotted.

//main
GNUPlotLineGraph gnuLinePlotter(&Serial);
PlotLogarithmicSpiral(gnuLinePlotter);

//method
void PlotLogarithmicSpiral(SerialGraph& plot)
{
	//setup graph
	plot.newGraph();
	plot.enableSaveImageFile(true);
	plot.setTitle(F("Logarithmic Spiral"));

	//setup series
	plot.setSeriesName(0, F("Spiral"));
	plot.getLineApperance(0)->setLineWidth(2);
	plot.getLineApperance(0)->setMarkerStyle(LineMarkers::SolidCircleMarker);
	plot.getLineApperance(0)->setRGB(GraphColorsSchemeSpectral::Spectral7);
		
	//plot graph
	float a = 1;
	float b = 0.1759;
	for(int i=0; i<(int)(10.0*20.0*M_PI); i+=5)
	{
		float t = ((float)i) * 0.1;
		float rt = a*pow(M_E, b* t);
		float x = rt*cos(t);
		float y = rt*sin(t);

		plot.plotDatumXY(x, y);
	}

	plot.finishGraph();
}

Example: Multi Column Bar Graph.

Bar graphs, for histograms and such are easy in this API. The example below shows support for multi column graphs, Note: negative values are also supported. The labels for the X-Axis are plotted as standard x values (the plot command is template driven and can support many different types of data).

//main
GNUPlotBarGraph gnuBarPlotter(&Serial);
PlotSampleBarGraph(gnuBarPlotter);

//method
void PlotSampleBarGraph(SerialGraph& plot)
{
	//setup graph
	plot.newGraph();
	plot.enableSaveImageFile(true);
	plot.setTitle(F("Google Results for Instruments"));
	plot.setXAxisName(F("Instrument"));
	plot.setYAxisName(F("# Google Results (millions)"));
	plot.setShowGrid(true);

	//setup series
	plot.setSeriesName(0, F("play"));
	plot.getLineApperance(0)->setRGB(GraphColorsSchemeOranges::Orange1);
	plot.setSeriesName(1, F("teach"));
	plot.getLineApperance(1)->setRGB(GraphColorsSchemeBlues::Blue1);
	plot.setSeriesName(2, F("like"));
	plot.getLineApperance(2)->setRGB(GraphColorsSchemeGreens::Green1);
	plot.setSeriesName(3, F("hate"));
	plot.getLineApperance(3)->setRGB(GraphColorsSchemeReds::Red1);

	//plot graph
	//Dataset compiled by Google-ing "I X the Y"
	//                instrument	play	teach	like	hate
	plot.plotDatumXYn("Piano",		173,	25,		211,	21);
	plot.plotDatumXYn("Guitar",		163,	22,		193,	27);
	plot.plotDatumXYn("Violin",     41,		2,		43,		1);
	plot.plotDatumXYn("Drums",		30,		8,		83,		9);
	
	plot.finishGraph();
}

Installation

1. Get the library from its github repository.
2. See the official guide on how to install libraries https://www.arduino.cc/en/Guide/Libraries
3. Run the provided example sketch
4. Install GNUPlot v5.0 or later (here).
5. Using the com port viewer, copy the output of sketch into GNUPlot.
6. Enjoy!

It is very possible to get the output generated by this library to GNUPlot automatically. How you achieve this is up to you.

It is possible to direct the output to a SD card. See the arduino help page on the file class, to create you own file on a SD card. Then pass the open file object in to the constructor for your plotting class (where I used Serial in the examples).

Documentation

Full documentation available here.

License

GNU 2.0

State

Most Features are well tested, but I don’t warrant or accept liability for any of the functionality in any way.

NB: some dynamic memory usage , refrain from using malloc during plotting operations to prevent possible heap fragmentation.

• Dynamic memory usage is a design decision I made weighing pro’s and con’s of the situation.
• Given ‘typical usage’ scenarios of the API it should not cause heap fragmentation.

Road-Map

May happen:

• Second x/y axis support
• Matlab support
• 3d plots
• Spectrogram
• Support for cloud based rendering for graphs
• Sage support
• Thousands separator (right now I could not deploy it because of a bug in the windows version of GNUPlot).

Won’t Happen:

• Real-time visualisation (plenty of other projects to do that already)

Solution in this video.

This particular puzzle is from The Moscow Puzzles. This book is awesome, sold millions of copies, and is one of the best “brain teaser” books ever (first released back in the 1950’s).
It’s not a “modern self help book” so nobody is trying to market it, which means it can typically be purchased for around $4. Not bad for a book so full of great puzzles and rustic Russian charm.

There are two solutions. I will leave them for you to find.

This particular puzzle is from The Moscow Puzzles. This book is awesome, sold millions of copies, and is one of the best “brain teaser” books ever (first released back in the 1950’s).
It’s not a “modern self help book” so nobody is trying to market it, which means it can typically be purchased for around $4. Not bad for a book so full of great puzzles and rustic Russian charm.

Testing this out on screen with a big colour palate would be kinda boring, so I made some software to take a large image and convert it to use one of every colour.
I am using  4096 * 4096  = 16,777,216 (or 2²⁴) images to do this.

I played around with a few algorithms, but found the best results came from a very simple idea:

1. Start with a image containing all colours.
2. Inspect two pixels
3. Swap the two pixels, if that would make If the image more like the target.
4. Repeat steps 2 – 4, until the image looks good.

Immediately I was stunned by the results.

• The Images actually worked, they looked like the source image.
• Parts of the image which where pulled toward a different colour and showed a very interesting (and accurate) image segmentation,
• The algorithm handled both photographs and solid colour cartoons well.

Example Images:

Firstly these files, produced by this method are are 50mb, and don’t compress well. I scaled them down for the website. If you want the verbatim image; you will have to download the software and make your own. (BTW: Did you notice the unusual beverage flavor in the vending machine pic, when you see it…).

Download the software:

Install file:  here
Oneclick: here

Screen Shot

Software usage note(s):

• Application works on high DPI screens.
• Some images prefer that I choose one of the pixels in a structured sequential fashion, others prefer both to be picked at random.
  • I solved this by doing n iterative rounds, followed by x minutes of random swapping.
  • The software lets you set these options.
  • Iterative rounds produce good results very quickly, but will leave some lines on cartoons.
  • Random rounds are slow to improve the image. A minute is a good buffer to clean up an image, but consider 40+ minutes for cleaner results.
• Uses many threads, and hogs the system bus. Will give the CPU a quick workout.
• Can be used in a command line mode, Suit batch processing.
  • To do this, provide one optional parameter (directory with images).
  • Finds all pictures in directory that don’t have “_done” or “_ignore” in the name.
  • Saves a new image with _done added to the file name.
  • Overrides that image without prompting!
• Save as bitmap If you don’t want the jpeg compression to alter the validity of the “every single colour” image.

On images heavily biased to one colour, the iterative pass may exhaust that colour to the top of its run.
eg..

In this case, set “Iteration Runs” to 0. This will disable the iterative passes and go strait to random passes.

Contact / Feedback

[contact-form subject=’All Colour Software (via website)’][contact-field label=’Name’ type=’name’ required=’1’/][contact-field label=’Email’ type=’email’ required=’1’/][contact-field label=’Type of comment’ type=’select’ options=’Support,Bug Report,Feature Request,Sales Enquiry,Other’/][contact-field label=’Comment’ type=’textarea’ required=’1’/][/contact-form]

Select the text in yellow line below to see the answer.

This particular puzzle is from a great book that I got for free when a local library discarded it.
Teaching Middle School Mathematics [Brumbaugh, Ortiz & Gresham]

Select the text in yellow line below to see the answer.

This particular puzzle is from a great book that I got for free when a local library discarded it.
Teaching Middle School Mathematics [Brumbaugh, Ortiz & Gresham]