Sunday, July 31, 2011

The age of technology? The age of the interface

When computers were new, they were capable of doing things that were labor intensive: calculating payroll, accounting and billing. People who were able to buy computers were also the people who could spend hours learning them. Early hackers were also engineers, or at least bored students with lots of free time who were happy to learn the minutiae of the system. They were happy to go through large volumes of programming manuals, datasheets. The rewards were rich: you got to be the master of a computing machine that did exactly what was told. Products at this time competed on features alone: new features might have been cumbersome to learn, but the customers didn't mind.

That was the age of technology, and we aren't in it any more.

Now, we're in the age of the interface. In the past few years, we have seen the rise of great products which aren't competing on feature-set alone. The iPhone is a perfect example of this: in functionality alone it was comparable to other devices in the market. What made it appealing was its ease-of-use, it's computer interface. The daily interaction with the device were much better than other cell-phones, and that convinced users to switch. I see this trend in many areas.
  1. Arduino is a microcontroller board and development software that allows you to attach physical objects to computers. The underlying microprocessor is made by Atmel, and it has been around for a lot of years. Arduino's contribution is to package the hardware and software in an easy-to-use, accessible package. You buy the microcontroller board, download their free development software, and can have a programming running within minutes. Arduino makes it trivial to try out microcontroller hacking, and it has led to an explosion in creative hardware hacking. Arduino programs are called 'sketches' and the software makes it easy to share your sketches with others.
  2. Processing is a programming environment that makes it easy to visualize data. Processing is built on Java, and is much easier to learn than Java. Arduino and Processing are very similar. With Processing too, you download free software which runs on all platforms. It allows for fast programming: a child can start programming graphics routines within minutes. Processing makes easy tasks trivial and hard tasks possible, all without having to teach a person the nuances of programming. Processing code compiles into four different programs: one each for Windows, Mac, Linux, and a Java applet. Processing also has an Android mode that allows you to run your processing code directly on an Android device.
  3. Fritzing is a development tool for electronic circuits. You can design the circuit as an physical layout (with bread-board, wires and components), or as a schematic diagram (big boxes rather than pictures). It does a splendid job of making hardware traces that can then be used to fabricate Printed Circuit Boards. Also, you can share your projects with other people by giving them a copy of your .fzz file. Fritizing makes it easy for a student to design a circuit board and have it mass-produced.
All these projects make it easy for a newbie to get started. They all provide software for Linux, Mac, and Windows, and they all make it easy to share your creation with the world. This has resulted in many more people getting involved with electronics and software hacking. It has opened this world up to people who were unwilling to read through thick programming manuals.

People are no longer interested in feature-sets. The feature world has been saturated: people who care for spreadsheet features already have all the features in their spreadsheet. Now, the goal is to make a spreadsheet that is accessible to the rest of the population.

Fewer features, ease of use, and excellent design. Software engineers are learning that good products are easy to use and well designed rather than a random collection of good features.

Image courtesy Jeff Atwood.

    Saturday, July 30, 2011

    GPS datalogger with Arduino and SD card

    I recently finished making a GPS datalogger with an Arduino, a micro-SD card, and an old USB GPS unit. It continuously logs the position from a GPS unit to an SD card.

    The connections are trivial with a 3.3V Arduino, since the micro-SD card has an SPI interface which the ATmega328 supports natively. I used a SiRF GPS unit, which outputs NMEA sentences at 4800bps. This was connected directly to the serial pins on the Arduino. With a 2GB SD card, we can log everything that the GPS outputs for well over a year, so I didn't bother filtering the GPS output.

    The bill of materials is:
    1. 3.3V Arduino, or appropriate level shifters for the SD card
    2. GPS unit with a serial interface
    3. SD card or micro-SD card slot and memory card
    4. Some power source: car 12V plug, or USB, 9V battery
    You can download the entire GPS datalogger Fritzing project. The connections and the schematic are shown below:

    The associated Arduino program is trivial:
    #include "FileLogger.h"
    #define HEADER "Starting GPS logging.\r\n"
    unsigned long header_length = sizeof(HEADER)-1;
    byte header[] = HEADER;
    unsigned long data_length;
    byte gpsdata[128];
    void setup(void) {
      FileLogger::append("data.log", header, header_length);
      data_length = 0;
    void loop(void) {
      if (Serial.available() > 0) {
        gpsdata[data_length] =;
        if(data_length > 127) { 
          FileLogger::append("data.log", gpsdata, data_length);
          data_length = 0;

    The Seeduino Stalker is an excellent Arduino-SD card combination that can be used for this project.

    This GPS datalogger can be powered from a 12V car plug for continuous logging of a car's position. This is useful to monitor the driving habits of your teenage child, or to log the route of a long cross-country drive. Another use is for logging bicycle trips automatically (powered from solar energy and a 9V battery for backup).

    This device requires no user interaction. It starts logging when the device is powered on. If the Arduino has an LED connected to pin 13, it will blink when new data is being written to the card. At the end of a month, you can copy the contents of the SD card to a computer for analysis.

    Most of the Arduino pins are unused, so there is hacking potential.

    Saturday, July 23, 2011

    Whose advice do you trust?

    One of the most difficult things about parenting (and pre-parenting) is knowing whose advice to trust. As a parent, many things are completely new and it is difficult to know the right way to do something. Many people offer to give advice, and while people are well-intentioned, their method might not work for you. The central problem is: whose advice do you trust?

    After listening to conflicting advice from many people, I have a reliable method of determining whose advice to trust:
    Trust the advice of people with many friends from diverse backgrounds.

    I arrived at this idea after looking at the people who have given us good advice in different scenarios. It seemed that the people with the best advice also had many diverse friends.

    Why? When a person has many diverse friends, it displays two qualities:
    1. Access to a wider variety of human experience.
    2. An ability to handle criticism and a difference of opinion. Being able to hear someone out is good, but being able to patiently hear why you are wrong is a true skill. Being thin-skinned about a topic usually cuts short the conversation before you can see an alternative solution to a problem.
    My friend calls this the "Pagerank" of human advice. While I had not thought of it in those terms, it makes complete sense. People are knowledgeable because of whom they know, not just what they know. When people have a lot of diverse friends, the increase in knowledge is not just linear. It is exponential, as they can relate to how Mary had a similar problem that Bob is talking about, even though Mary and Bob have completely different backgrounds. This allows them to identify patterns of problems and patterns of solutions, which is more valuable than knowing a single solution for a single problem. Also, having a wider set of friends allows you to learn from others' mistakes before you face the same situation.

    Low sensitivity to criticism helps in accepting alternate methods of solving the same problem. Being receptive to new ideas allows us to identify new methods rather than being stuck in a rigidly defined solution. You can measure the receptivity of a person when you decline their suggestion. A person who takes offence when you decline their suggestion has low levels of receptivity. Such people also have the poorest suggestions. The best advice is given by people whose interest is piqued when you say you are going to do things differently. They don't feel offended when someone declines their advice. Indeed, they want to know the alternatives because they actively seek other ways of solving the problem.

    My friend also pointed out that the two qualities listed above might be present in a person lacking diversity among their friends. This could happen when a person has lived in a homogeneous society and does not have access to people of diverse backgrounds. I suspect that in such cases, the people will form lasting friendships with diverse people if they get a chance. I know some people who have grown up in homogeneous societies, and they still have a diverse group of friends for their limited surroundings. Their group might be homogeneous by some respects (ethnic background, religion, etc) but will be completely diverse in others (education, parenting, occupation, age...)

    Sunday, July 17, 2011

    Microcontroller driven LCD panel with voltage measurement, time, EEPROM for computers

    I just finished making a circuit to drive a four line LCD panel from an Arduino. The project started out as a way to drive the LCD to display arbitrary information, but then evolved into a way to tell the time, read and write the EEPROM from the Atmel 328p, and measure voltage. I'm using many existing libraries: the Time library, and the two Controller LCD library.

    The final functionality is:
    1. Commands are sent over USB using vi-line syntax. All commands start with ':' character.
    2. Segments the 40x4 display into 8 segments of 20 chars each. Segment 0 is column 0, line 0. Segment 1 is column 20 line 0. This allows an easy way to display up to eight different status updates. To write "hello world" in segment 3, you write ":s3helloworld\n" to the serial port.
    3. Read and write from EEPROM. The 1028 bytes of EEPROM is chunked into 8 segments of 128 bytes each. Segment 0 cannot be written to, which allows for 128 bytes of ROM. To write a segment 3, you index (:i3), then write (:wMySecretMessage\n). To read from EEPROM, you index first (:i3), then read the full segment (:r).
    4. Time can be read from the device (:t) and the time can be set using Unix time (:T1310349445).
    5. Voltage monitoring at analog pins can be toggled (:a). Voltages are displayed on the LCD (display shows all six pins in decivolts: 5.0v is written 50) and written in raw form on serial (relative to 1023. 1023 = 5V, 0=0V)

    The entire Arduino source code is available as FourLinePanel.pde. It still has some debugging information left in, it prints the value of every new command it receives. This debugging could be removed. It supports a few commands:
    :t Displays the time
    :T<unix time> sets the time
    :l<message>\n Writes message to current position
    :sN<message>\n Writes message starting at segment N
    :a Turns on analog voltage monitoring on LCD and serial
    :r Reads 128 bytes from EEPROM at current location
    :iN Index into segment N of the EEPROM
    :w<data>\n Write data into EEPROM starting at current location

    This requires one Atmega microcontroller, an LCD panel, two 22pF ceramic capacitors, and one 16Mhz crystal. To interface it with USB, a USB to serial cable might be required. In my design, I'm driving the entire project using power from the USB port. This minimises the need for voltage regulation and additional components: power over USB is usually clean. You can download the schematic for the FourLinePanel here.

    To interface with a computer, you read and write to the serial port with 38400 bits per second, no flow control, and no parity bits. An example program to interface this with Python is given below.
    import serial, time
    # Set the time on the FourLinePanel for GMT+5hrs
    # offset from GMT: Add 5hrs
    offset = 5*3600
    port = "/dev/ttyUSB0"
    fourLine = serial.Serial(port, 38400);
    fourLine.write(":T" + 
                   str((int)(time.time() + offset))) 

    There are five spare digital pins that could be used to drive a Real Time Clock or LEDs. These are marked as "Spare Digital (unused)" in the circuit diagram [png], and circuit diagram [ps]. The analog voltage testers can be used as handy ways to measure voltage at the desk when a multimeter isn't handy, or to constantly monitor the voltage drop across a device (like a photovoltaic cell).

    Friday, July 15, 2011

    Who is Secret Labs, and why the Netduino project makes me nervous

    I'm a big fan of the Arduino project. It is a micro-controller board that makes hardware programming fun. It is supported on all platforms, Windows, Mac and Linux, and works equally well on all of them. Arduino has fired off creativity in micro-controller hacking, and one product that caught my attention was the Netduino project. The board looks like the Arduino, but it uses the .NET framework by Microsoft.

    A lot of what I found about Netduino was suspicious, so I thought I'd share this in the hope that someone could clarify the situation.

    For an electronic project, Netduino is too good to be true.
    1. The Netduino contact page is very sparse: it contains a physical address (Secret Labs LLC, 315 Bleecker St #308, New York, NY 10014), an email address (, and a link to the online forums. The Secret Labs LLC website contains exactly this information. Almost nothing is known about them.
    2. The only name associated with Secret Labs is a person called Chris Walker. He is the only moderator of the Netduino forums. This is strange: I would expect Secret Labs to have a lot more than one technical contributor and a lot more people involved.
    3. I couldn't tell any source of funding. Is Secret Labs making money from the Netduino product? Hardware margins are probably too small for this to work. How does Chris Walker pay his rent? Do they have Venture Capital funding? I wasn't aware of open source hardware being a VC magnet.
    4. They have designed and developed the entire product, and it is available for purchase at major electronics stores like Sparkfun. The design is polished and crisp, too crisp for a version 1 product by an unknown company.
    5. The Netduino plus contains an ethernet port, for which they have 5C:86:4A MAC address prefix. Applying for MAC addresses takes some effort, so it is impressive that Secret Labs LLC has done so much groundwork.
    6. All their source code is available from files hosted at There is no public source code repository: no tracking of who made changes, and when. They claim to be an open source project, but if I want to submit improvements, there is no email address or contact in the source code. All source code is copyright Secret Labs LLC. A real hacker would love to get improvements from the community. The source code is not professional: comments are missing, and stale code is commented out when it didn't work. From the look and feel of the source code, I would guess a very small team, and a very hurried release. It was most probably developed in September 2010, which is very close to the release of the .NET Micro framework. 
    7. From a cursory look through Google Maps, 315 Bleecker St. looks like a residential address.  What's worse, the Better Business Bureau lists 318 Bleecker St #308 as a Payment Processing Service called SimplePayClix, with the listing opened in March 18, 2010.
    8. The domain was first created on 16 January, 2009, while Netcraft first saw it in March 2009. This is very surprising. The domain was registered before the Secret Labs LLC was created. Microsoft announced that they would open source the .NET Micro Framework on 16 November, 2009. Did Secret Labs LLC work on a product even before they knew that the framework was going to be open sourced, or were they working on a different design that wasn't released? The owner on is Chris Walker again, with address
    The Netduino project looks very impressive, and if it is the work of a few people, they are very smart and very capable indeed. But it doesn't look anything like a community-driven project. Only one person is visible, and I find it hard to believe that he has developed the hardware, the software, the book, the documentation, the video tutorials, the forums. All in isolation from the hacker community.

    It is entirely possible that Chris Walker is independently wealthy, is an expert electronics and software hacker, is amazing at website design, and has good relations with electronics manufacturers. But I smell a rat.

    Update (23 July 2011): Read the first comment for all the answers.

    More about Chris Walker of Secret Labs: he has a Channel 9 account 'cwalker_secretlabs', created this year. He gave a podcast interview about the Netduino. OReilly is about to publish a book called "Getting Started with Netduino" by Chris Walker, which will be released soon. Chis Walker is listed as the CTO of Secret Labs in one article. There is what he looks like.  Later this year, he is giving a talk titled "Connecting Devices to the Cloud on Open Source Hardware and Software", at Oreilly's OSCON on July 27, 2011. His co-speaker is Colin Miller, who is a Product Manager for the .NET Micro Framework.

    Wednesday, July 13, 2011

    Why is Bombay a target of bomb blasts?

    There were another set of bomb blasts in Bombay. My heart is with the families of the affected people.

    Security in India is pretty bad. There is a culture of lax security.

    In my experience, the police have been unhelpful and unable to enforce laws. The airport police are slightly more diligent than the regular force inside the city, but both are relatively incompetent. At the International airport, I have seen passengers walk away without their bags being scanned through the machines. On one occasion my bag was run through the scanner but there was nobody looking at the scanner to see what I was bringing in. It is easy to see how dangerous, explosive material could have been brought in.

    Inside the city, the police are equally lax. I do not remember approaching the police when there was any problem. The police were a nuisance rather than a service and a convenience. The police were happy to assert their authority and harass students and poor people, while avoiding their responsibility to enforce laws. The porous border with other countries combined with the lax enforcement of laws make it easy for criminals to cause mischief with little consequence.

    A symptom of this is the high number of accidents and traffic fatalities in India. Traffic police are probably the least competent, but the indifference and incompetence of the traffic police points to a systemic problem in law enforcement everywhere: tax evasion, violations of human rights, and a lack of security. Citizens that pay taxes are paying the state to keep them safe, and to guarantee some basic dignity. The police are unable to provide this in Bombay.

    This incident is truly unfortunate, and I share the sense of vulnerability such incidents cause. Unless we can address the core problem: the failure of enforcement of laws, we can never achieve the security that Bombay deserves.

    Image courtesy: blog on Indian police

    Sunday, July 10, 2011

    Dual HD 44780 Controller Liquid Crystal Library for Arduino

    I came across a four row LCD panel while browsing at an electronics store: the model name was WM-C4004A. I wasn't sure if I could drive it through an Arduino but decided to try. It was easy to drive once I hunted down the datasheet for the WM C4004A, a dual HD44780 controller, and understood the controller.

    There is a good Arduino library for driving HD 44780 compatible controllers and I modified the LiquidCrystal library to drive two separate controllers. The usage is identical to the LiquidCrystal library. You specify the Register Select (RS) and enable bits for both controllers, the data pins, and the rows and column per controller. Then you can seek() and write() through Arduino.

    My modifications are available as the  LiquidCrystalDual library. Download the library: LiquidCrystalDual.tar.gz or Unzip and place the contents in the libraries/ directory in the arduino installation.

    Here is a small example program (there is a detailed program in the examples/ menu in the Arduino IDE).

    #include <LiquidCrystalDual.h>
    // 2 controllers: RS, RW, EN1, EN2, D4, D5, D6, D7
    // We have the RS (RS=12) line, *no* RW line (RW=-1),
    // two enable lines (EN1=6, EN2=7),
    // four data lines (D4=11, D5=10, D6=9, D7=8)
    LiquidCrystalDual lcd(12, -1, 6, 7, 11, 10, 9, 8);
    void setup()
      // I have 40 characters and two lines per controller.
      // Move to the first column of second row.
      lcd.print("Line 1: Hello");
      lcd.setCursor(0, 1);
      lcd.print("Line 2: World");
      // Move to the second controller
      lcd.setCursor(0, 2);
      lcd.print("Line 3: Four...");
      lcd.setCursor(0, 3);
      lcd.print("Line 4! Lines!");
    void loop()
      // Do nothing.

    Friday, July 08, 2011

    What determines the success of a country: Nurturing creators

    Why does a single area dominate one industry? Detroit with cars in the 1960s, Silicon companies in California in the 1980s, .... Before the computer revolution, there was no centre of electronics. A lot of work was done in the UK, Japan, the US, and other countries. And yet, most of the computer design is now centred in the US.

    I am beginning to think that the core reason why the US forged ahead with computers was that it had a large number of initial hobbyists and tinkerers. Many individuals tried different ideas. Most of the ideas made no commercial sense, but there were enough ideas that reached commercial success. The initial tinkering group found themselves employed at companies that were successful. Apple started out as a tinkering collection, with Steve Woz designing motherboards for sale for $666.66. He had worked previously at Hewlett Packard, and brought his impressive tinkering skills to bear on the Apple I and the Apple II. Dell, Compaq, Intel, they all had a core group of tinkering engineers who had a commercially viable idea. Everyone was producing things, and most were of dubious value. But for every hundred bad ideas, there was one Apple or one Microsoft. And that hit-rate was enough to sustain entire generations of engineers. The UK had some tinkerers but their total number was small compared to those in the US, so while they had some success, they couldn't sustain a lead.

    Something similar happened with the Web 2.0 boom. For a while, every college student in the US was making a Web 2.0 mashup, armed with an Apache server and a passing knowledge of Javascript. Not everything worked, but enough ideas were bounced around to give rise to a Twitter, a Facebook, a LinkedIn, a Youtube. All these sites were possible elsewhere.

    This is happening in the mobile world, all over again. College students in the US are making apps for their iPhone and Android devices. Most of them are failures, but enough are succeeding in the San Francisco area to create a virtuous cycle. Many college students are getting together to form two-person startups around their iPhone app.

    I think this tinkering and production of ideas is the reason for phenomenal American success in computer hardware, Web 2.0, and mobile applications. One of reasons for this is cultural: American individuality recognises the lone wolf who is trying something new, even if the idea isn't mainstream. The other is the large payoff to a successful venture. The third is the general risk-loving attitude: students willingly take a year off from University to try out a startup, working families try a startup even if they know it might cause some hardship if the venture fails. College students are happy working on some idea they love.

    All these factors contribute to an environment where tinkering and idea creation is nurtured and idea creators are rewarded. Even if most projects fail, enough of them succeed that the job market is always looking for engineers and other tinkerers. Many people working at Apple or Google today have a varied history of trying out different ideas and learning from their mistakes.

    I see this trend most clearly in hobbyist hacking involving the Arduino or other micro controllers. Again, a lot of tinkering is being done in the West. The difference is that there are a significant numbers of Chinese hobbyists involved. At first, this was because the manufacturing had moved off to China, and the Chinese engineers were exposed to a lot of Western products and ideas. Now, we see many Chinese engineers tinkering and experimenting. Unlike the Web or mobile applications, electronics hacking doesn't require an excellent command of English, and some Chinese electronic designs are approaching the best in their class.

    It seems to me that countries should focus on nurturing creators. It pays off in the long run.

    Tuesday, July 05, 2011

    Fun with Assembly Language

    This is a quick page about my adventures with Linux assembly language, with links to resources, and some source code.


    I won't fool you into thinking that assembly language is a useful tool: it isn't. By and large, if you want to get stuff done, you're much better off learning Python, C or Java. Assembly language is great if you're a tinkerer, and want to know what is going on inside your computer. Many people have written interesting tiny programs using assembly, which might be handy on embedded environments. Finally, there are some features that you can only get through assembly: low-level architecture features, getting around compiler limitations and understanding languages themselves. For me, assembly is just for fun!


    I do all my programming on Linux, and it is one of the best environments to learn assembly. The tools are quite good, and there is some lovely documentation. If you don't already have Linux, get a copy of Knoppix, which is a live CD that lets you try out Linux without modifying your computer setup. Here is a list of books that I highly recommend.
    1. [PGU] - Programming From the Ground Up, by Jonathan Bartlett: an excellent introduction to x86 assembly language on Linux. The entire book is free for download. This one book is all you need, initially.
    2. [PAL] - Professional Assembly Language: another excellent introduction.
    3. [IPM] - Intel Programming Manuals: complete, in-depth information on architecture and assembly language. You're probably interested in Vol 1, 2A, 2B, 3A and 3B. It is meant as a reference: not a cover-to-cover read.
    4. [APM] - AMD x86-64 Programming Manuals: complete, in-depth information from AMD. Get all the volumes.

    Fun with Assembly

    So you got yourself a copy of [PGU], and want a challenge?
    1. We try to rewrite the first program from [PGU] as 01_exitValue.s. The expected return value is 999, but you get something else.
      1. Can you guess what the return value is without running the program?
      2. Can you explain why the return value is not 999?
      3. Can you write the program so that this problem can be caught by the assembler?
      4. 01_exitValue_Solution.s.
    2. We try to rewrite the second program from [PGU], using our understanding from the earlier problem, as 02_maximumValue.s. The expected return value is 214, but you get something else.
      1. Can you guess what the return value is without running the program?
      2. Can you explain why the return value is not 214?
      3. Can you write the program so that this problem can be caught by the assembler?
      4. 02_maximumValue_Solution.s.
    3. Here's an example for why you might need assembly language. Your mission is breaking into a program to steal a secret key. Some experienced hackers have isolated where the secret key is being passed to a secret function.
      1. secretFunction(char *useless, char *secret_key) calls validate() immediately upon starting. You job is to print secret_key inside validate. As an example, the file 03_keyIsHere.o contains the secret function that has the second argument as a key. You are only allowed to write a validate() method in a separate .s or .c file. Try not to modify the original object file. You are assured that the key is exactly 13 characters long. Try writing an assembly solution, and compile and run with gcc 03_keyIsHere_Solution.s 03_keyIsHere.o -o 03_keyIsHere; ./03_keyIsHere
      2. If the validate was called in the end, how does it change your solution?
      3. Can you use this trick to guess the local variables of secretFunction?
      4. In case the object file doesn't work for you, 03_keyIsHere.c contains the C source. The solution must not modify it.
      5. 03_keyIsHere_Solution.s Compile and run with gcc 03_keyIsHere_Solution.s 03_keyIsHere.c -o 03_keyIsHere; ./03_keyIsHere
    4. This is a more sophisticated example compared to the previous one. Having moved on in your career, you are faced with a new program that calls printf, and you have to inject code without a recompilation. The program 04_crackMe executes happily on its own. You know that printf is being called in it, and that on the first invocation, the calling function has a secret-key as its second parameter.
      1. secretFunction(char *useless, char *secret_key) calls printf() immediately upon starting. You job is to print secret_key. Most probably, you'll want to inject code into printf. The solution does not require modifying the original binary. You are assured that the key is exactly 13 characters long.
      2. Can you do this while continuing to print the original messages?
      3. In case the object file doesn't work for you, 04_crackMe.c contains the C source. Compile and run with gcc 04_crackMe.c -o 04_crackMe; ./04_crackMe . Remember that the solution does not require modifying the 04_crackMe binary or the C source code.
      4. 04_crackMe_Solution.s. Instructions on how to run it are inside the solution.
      5. Sufficiently pleased? Good, now carry it over to the next level by visiting this awesome page on making 13 equal to 17.

    Cost effective (cheap!) USB Serial cable for any project

    If you've been playing with microcontrollers, or hacking any sort of hardware, you know that it is great to be able to communicate with a computer. A usb cable is often the best option, since usb serial drivers are ubiquitous and easy to work with. However, cheap usb cables are difficult to come by. On this page, you will find instructions to build a very low cost USB serial cable, which can also supply +5V to a board. This is ideal for breadboard arduinos, but will also work great for any microcontroller or TTL circuit.

    All instructions here are provided solely as a guide. If you miswire the cable, you might end up burning your computer, your electronic gadgets, or your house. I do not take personal responsibility for your actions.

    Cost, parts, and capability

    You need a Nokia CA-42 cable, a soldering iron and some method of determining continuity. A multimeter would be handy, but is not required. Assuming you have the tools, the CA-42 cable sells for $3 including free shipping from many sellers on eBay and Amazon. Since the cost of the cable is so low, online fraud is unlikely. The total cost of the project is $3, and you don't need any extra parts.
    The final cable works on Linux, without requiring any extra drivers. Most cheap CA-42 cables are made using a PL-2303 chip, which is a cheap RS232 convertor. For data transmission, you only get two lines: RX and TX, but this is usually sufficient. You get +3V and +5V, which are very well regulated and clean sources of power. These could be directly attached to microcontroller boards. I have seen drivers for PL-2303 on Windows and Mac, and you should search for these over the Internet. You might need to mask the device ID, since the cable might announce itself as a Nokia phone cable. You don't need any drivers if you only want to use the cable as a power source.

    Make it

    The construction is trivial. This is what the cable looks like (image courtesy Timothy Small):

    The pinout, from left to right is:
    1. This is body of the cable, and is not connected
    2. The missing pin, no connection
    3. No connection
    4. +3.3V, might be as low as +3V
    5. No connection
    6. RX
    7. TX
    8. GND
    It might help to view the pin out diagram (hosted at

    To create the cable, cut the head near the top of the cable, as shown. When you cut the cable, try to match the color of the wire with the pinout given above. I've found the following colors, though you should check your own cable:
    • Black: Ground
    • Red: 3V
    • White: RX data
    • Blue: TX data
    • Green: no connection
    Now, the cable has five wires: two are used for data (TX, RX), two are for power (+3.3, GND). There is an extra wire in there, which is useless for our purposes. So we repurpose this wire to carry the +5V connection from the USB port. In my cables, this wire has always been green, though you should check which wire, if any, is unused in your cable. The head of the cable comes off quite easily if you hold it and twist the metal USB connector as though it were a key. Connect the far right metal lead from the USB connector to the green wire. The green wire should be disconnected from the body, but the +5V connect should not be disconnected from the body: the PL-2303 requires +5V power to work.

    Here is the final soldered assembly. In this image, the usb connector is the correct side up. For your reference, here is pin-out for the USB A connector. The soldering is done to the +5V wire.



    The +5V and ground can be used to power an arduino or most Atmel Atmega AVR microprocessors. The TX wire (colored blue in my cable) goes into the TXD pin of microcontrollers (Arduino: digital 1), and the RX wire (colored white in my cable) goes into the RXD pin of microcontrollers (Arduino: digital 0). I've tested the power using an oscilloscope: the power in my cables is very clean. This means that you can avoid using a 7805 and other circuitry that would provide a clean +5V signal.

    Credit: The idea for this cable is due to the the ARM9 Linkstation hackers.

    Making your Linux computer silent

    These are some notes I made back before Solid State Disks were popular. If I were doing this today, I'd get an Intel Atom processor with SSD disks, and no fans. These tips might still be relevant for desktop computers.

    While you may not realize it, your computer is capable of being silent, or at least quieter. I collect some tips here which might help in reaching this goal.
    • Buy a good power supply. A lot of the noise comes from power supply fans and other fans in the system. While the CPU fan is usually rather silent, some power supply fans are very noisy. If you want to make your existing fan quieter, a good way is to blow the dust out. Dust stuck in the fans makes noise. You could use a bicycle pump to force air through the fan after switching off the computer. Your mileage may vary. CPU heatsink fans have the tendancy to get very dirty, and dirt gets accumulated in the heatsink fins. I've found cleaning these fins to make quite a dramatic change in the amount of noise from the heatsink fan.
    • Use noflushd: It is a user level daemon that prevents disk writes from spinning disks back up.
    • Don't use journaling filesystems. Ext3, reiserfs, xfs, jfs are all journaling filesystems and commit data to disk without going through any cache. While journaling filesystems are cooler and have much better recovery and consistency: if you have a noisy disk: they will make it spin almost continuously. Turn off journaling with ext3, or change the filesystem if you can. Here is a quick way to convert an ext3 partition to ext2. This will work even if it is the / filesystem, though nothing is guaranteed. First, turn off all the services possible, and try to get almost no processes running. Then edit /etc/fstab so that the ext3 filesystems are changed to ext2. Next, run the command
      vi /etc/fstab  # <- change /dev/sdaXX from ext3 to ext2
      mount -o remount,ro /dev/sdaXX
      tune2fs -O ^has_journal /dev/sdaXX
      mount -t ext2 -o remount,rw /dev/sdaXX
      After these commands, you might want to reboot, and see if the filesystem is now ext2. It worked for my iMac, where I could not reboot to another Linux distribution. Again, your mileage may vary.
    • Turn off as many services as you can. If you don't need mysql running all the time, switch it off. A lot of services want to write logs, and you can either change their configuration to disable log writing, or look at the next step.
    • If you have directories in which frequent writes are required, consider mounting them as a tmpfs. This is a new filesystem in Linux which keeps data on RAM, and not on disk. This means that if the machine goes down, your data is lost. To avoid losing data, you might want to copy the data out to real disk when switching off the machine. This is a great way to write log files, or put images from the webcam that you take every second or so. After all, when you have the most recent image, you might not be so interested in the older photos. If your kernel is too old, you might want to use the ramdisk module, but tmpfs is a better choice if you have it.

    Programming the NXT to communicate with Linux

    This is Part 3 of a three-part series on programming the Lego NXT through Linux.
    1. Part 1 to setup the NXT-bluetooth connection
    2. Part 2 to setup your programming environment
    If you have followed along, you are now ready to program in NXC.
    Let's write a simple C program for Linux, and an NXC program for the NXT that communicate through Bluetooth. The interface of this program is totally dopey and unprofessional, but it demonstrates how to send information to the NXT and get information from it. The ideas are borrowed by notes from the CPSC 3720 Software Engineering course at the University of Lethbridge, in particular this handout. Here are the steps you need to follow.
    • Install Bluetooth libraries and GTK libraries on Linux. On Ubuntu, the command is:
      $ sudo apt-get install libbluetooth-dev \
             libgtk2.0-dev libc6-dev make gcc
    • Make the Tribot from the Mindstorms quickstart. This is the first project, the quick start.
    • Ensure that B is connected to the right motor, and C to the left motor. Leave A unconnected.
    • Now download my nxtBluetooth.tar.gz package to demonstrate Linux Bluetooth communication. Unpack it, and go through the files.
    • There are two programs: nx_soldier, which is the NX's program. It listens on Bluetooth and obeys the commands of the commander. The pc_commander is a GTK application that allows you to specify commands that the soldier will faithfully carry out.
    • There is a single place you have to edit to specify your NXT's address. Remember the times you changed 11:22:33:44:55:66 address above? Well, you need to open the pc_commander.c file and change this address to that of your NXT again.
    • Attach your NXT using the USB cable. Compile the programs with
      $ grep ADDRESS pc_commander.c       
      $ # Make sure you have the right address
      $ make nx_soldier.rxe
      $ make pc_commander && ./pc_commander
    • Run the nx_soldier program on the NXT.
    • Run the pc_commander on the Linux machine. Press the buttons to make the soldier move according to your wishes.
    • When you ask the soldier for his status, he always replies he is fine.
    • This program makes a good starting point: the code is reasonably well documented, and easy to adapt. You could try putting the bumper sensor on the TriBot, and modifying the nx_soldier so that he replies with "I am stuck" when it has hit a wall.

    Setting up the Lego programming environment in Linux

    This is Part 2 of programming the Lego NXT using Linux. Part 1 covers the Lego-Linux bluetooth connection.

    If you're like me, you've used the graphical NXT programming environment for about twenty minutes before wrist pain and frustration sets in. I did not enjoy the graphical programming at all, and the fact that it did not work on Linux was another factor. Well, there is a very nice alternative, which is John Hansen's NXC language. It is a C-like language which you can program the NXT in. You'll hear a lot of references to BrixCC, which is an IDE for NBC and NXC. It doesn't exist for Linux, but any text editor works quite well. All features are supported, and you can write code that is more compact and faster than the graphical tool. Also, you can use NXC to allow your robot to communicate with your Linux computer over bluetooth! This is what you need to do to start using NXC.
    • Visit the excellent NXC beta release page . There is no stable release as of March 2009, but the beta is pretty rock solid. Forget the source code, it is made in Delphi and there is little documentation on how to compile it under Linux. Get the binary distribution. It consists of some very spare documentation and the nbc executable. Despite its name, it is also the NXC compiler. This will only work on x86 Linux.
    • Run nbc -help to learn the options. Really, this program needs a man page!
    • Visit the NXC documentation page and get the lovely NXC_Guide.pdf, a guide on how to program using NXC. Download it but don't look at it yet. It is a reference, not an introduction.
    • The real document you're looking for is Danny Benedettelli's NXC tutorial, which is far more useful than the guide. Download it, print it out, read it through and through.
    • To set up the USB communication, download shell script to perform USB settings for Linux communication with NXT.
    • Run it by itself as follows 
      $ chmod +x ./;
      $ ./
      This will prompt you for the root password when required. The file follows instructions from the page about Linux USB setup with NXT, so all credit to them. Feel free to edit the file if you suspect something went wrong. The script will prompt you with any errors that it encountered.
    • In case you aren't certain if the setup worked, plug in your NXT. You should see the word "USB" in the top left of the NXT display, and the output of the command ls -l /dev/lego* should show at least one device. My output is given as a reference
      $ ls -l /dev/lego*
      lrwxrwxrwx 1 root root 15 2009-04-19 17:45 /dev/legonxt-2-1 -> bus/usb/002/042
    • Write a sample program using Danny's guide.
    • Compile your program with nbc your_program.nxc -d -S=usb, where your_program.nxc is the name of your program.
    • Now the program has been compiled and loaded on to the NXT. It should appear in the "My Files" -> "Software Files" area of the NXT, from where you can run it!
    • There, that wasn't so hard! As you learn NXC, you'll realize that it is feature complete, and is a much more elegant way of programming the NXT.

    Setting up NXT-Bluetooth support on Linux

    So you want your NXT device to talk back to your Linux machine? Perhaps you want to control your robot through your computer. As a first step, you have to set up the Linux - NXT bluetooth connection. My setup is explained below.
    • Ubuntu Gutsy (7.10) Linux, though these instructions are not version dependant. The only instructions that depend on the distribution are the ones related to package installation (apt-get). If you know your distribution's package manager, you should have no trouble.
    • An x86 computer. Typically, you'll want a laptop with bluetooth support, so you can walk around with your NXT construction. A desktop will do just fine too. You need an x86 because the nbc binary only works on an x86.
    • A Lego Mindstorm NXT.
    • Bluetooth hardware on your linux machine.
    • The Linux packages: bluez-utils, python-bluez, bluez-hcidump, bluez-gnome, bluetooth
    Make sure you have all the packages installed. On Ubuntu, you can run this command.

    sudo apt-get install bluez-utils \
         python-bluez bluez-hcidump bluez-gnome bluetooth

    Now, once all the packages are installed, make sure that your bluetooth hardware is supported and is running fine. One way to do this is to check for the bluetooth icon in the taskbar of Gnome.

    You should be able to click on that icon and get a list of preferences. In case you are not sure if you device is getting recognized and used, check for lines similar to the ones below in the output of dmesg
    [243203.060226] Bluetooth: HCI USB driver ver 2.9
    [243203.063512] usbcore: registered new interface driver hci_usb

    Now that you have a working bluetooth setup, you need to pair it with the NXT brick. First, you need to ensure that the serial service is started by selecting it in the preferences dialog from the bluetooth icon.

    Now turn on your NXT Brick, and enter bluetooth settings, and turn bluetooth on. You can verify that bluetooth is turned on by looking in the top left of the NXT display. It should have a bluetooth icon. Now run the command sudo hcitool scan from a console. The output should contain a line for your NXT device. My device is called Snow, and I get the following output.

    Scanning ...
     11:22:33:44:55:66 Snow

    Great!. This is definite proof that your bluetooth device works, and that it can see your NXT. Time to set them up for pairing. Edit the file /etc/bluetooth/rfcomm.conf to contain the following block:

    rfcomm0 {
            # Automatically bind the device at startup
            bind yes;
            # Bluetooth address of the device
            device 11:22:33:44:55:66;
            # RFCOMM channel for the connection
            channel 1;
            # Description of the connection
            comment "Snow Mindstorm";
    Change the address 11:22:33:44:55:66 to match the output of your hcitool scan command that you made in the previous step.  Now run the command sudo /etc/init.d/bluetooth restart, and then rfcomm.  The output should contain your device's address.  A sample session is shown here:

    $ sudo /etc/init.d/bluetooth restart
       * Restarting bluetooth            [ OK ]
    $ rfcomm
    rfcomm0: 11:22:33:44:55:66 channel 1 clean

    Now, you need to pair your device. This is done with the following command
    $ sudo l2ping 11:22:33:44:55:66
    This will prompt the NXT to beep happily, asking for a passphrase.   The  number "1234" is already entered, so just accept it by pressing the  orange button.  Now your computer will want to know this passphrase.   Click on the button in the notification that pops up (example shown  below), and enter the same number as shown.  This is a one-time  procedure for each computer-NXT combination.

    If you have a multi-seat (or multi-terminal) setup, this notification might be going to another seat, so check all notification areas on every terminal. If you don't know what multi-seat or multi-terminal setup is, ignore the previous statement. If all went well, your ping command should be able to send packets back and forth to the NXT. A happy sample output is shown below.

    Ping: 11:22:33:44:55:66 from 00:09:DD:50:67:CE (data size 44) ...
    4 bytes from 11:22:33:44:55:66 id 0 time 49.95ms
    4 bytes from 11:22:33:44:55:66 id 1 time 85.71ms
    4 bytes from 11:22:33:44:55:66 id 2 time 27.80ms
    4 bytes from 11:22:33:44:55:66 id 3 time 54.80ms
    4 bytes from 11:22:33:44:55:66 id 4 time 65.81ms
    4 bytes from 11:22:33:44:55:66 id 5 time 47.80ms
    4 bytes from 11:22:33:44:55:66 id 6 time 84.79ms
    4 bytes from 11:22:33:44:55:66 id 7 time 45.81ms
    Great!  So now your NXT is paired with the computer.  In order to communicate with the desktop, you need to download the NXT Python library.   Once you have downloaded and installed it, enter the sample program,  and you should be able to get the NXT's name and ID.  If you have any  problems at this stage, try rebooting the NXT brick, it sometimes helps.   After running the sample program, you should see your hostname in the  "My Contacts" in the NXT's bluetooth menu.  This verifies that you can  connect with the NXT, and have the NXT talk back to the desktop  computer.

    (This is the stable home of the page that was earlier at

    Sunday, July 03, 2011

    Hiking: Portola Redwoods State Park, July 4, 2010

    This was our first time camping and hiking at the Portola Redwood State Park. The drive to Portola from San Jose is almost as long as the drive to Big Basin - about 1 hour. The park is nice, but we like Big Basin and Henry Cowell better.
    There's a short trail starting behind the camp office, the Sequoia Nature trail I think, that takes you to a small stream with a bridge running across it. This is a great place to come by for a late evening walk from the campground.
    The main hike marked in the map was on the Slate Creek trail. Took us a while to get to the trailhead, as you can see. We walked over 9 miles, spanning an elevation gain of 1000 - 1200 feet. A large map of the hike can be found here.

    Friday, July 01, 2011

    Hiking: Henry Coe State Park, Jun 26, 2010

    The campgrounds at Henry Coe are very nice. There are just a few campgrounds, so it's a lot quieter, and the view is great. On a nearly-full-moon Friday night in late June, it was spectacular. The facilities are primitive, but well maintained.
    We started hiking early, expecting most of the trails to be exposed to the sun, but our path was reasonably well shaded by trees and large shrubs. Okay - here's the 6.6 mile loop we did: started up the Monument Trail, made a left on Hobbs road, right on the Frog Lake Trail (there were many tiny, camouflaged frogs jumping all over!), right on the Middle Ridge trail, right on the Fish Trail, and then back to the Camp Headquarters via the Corral trail. The route was pretty well-marked all along.
    Oh, and apart from being on the lookout for poison oak, also be on the lookout for ticks. Light colored clothing is probably a good idea - makes them easier to spot and de-tick.
    A large map of the hike can be found here. The route covered roughly 1800 feet of elevation gain over a 6.6 mile stretch, but the uphill stretches are interspersed with easier stretches. Walking the same route in an anti-clockwise direction would probably be a lot more strenous.