BaSzErr

blog:2012:01:31:1 - stop ACTA petition

bartek szurgot — Tue, 31 Jan 2012 15:32:49 +0100

2012.01.31 - stop ACTA petition

and yet the protests against ACTA are not over. though Poland already signed that agreement, still not everything's lost. it still has to be retificated by the Sejm. but wait – there is more! there is one more thing that can be done. when 500k signatures are collected referendum must be held, so that citizens can vote and decide themselfes on the final decitions. such a signatures collectting is taking place since few days now. as for the moment of writing these words about a half of a required signatures is already collected. now is the moment, when your vote counts…

prjs:avr_servo:avr_servo - release v2.0.0

bartek szurgot — Wed, 25 Jan 2012 16:46:57 +0100

AVR Servo

AVR Servo is a multiple RC-servo mechanism controller device. the project aims at creating universal RC-servo controller. why another one? well… i recently needed one and non of those i found on the internet meet my requirements. except for controlling as many servo mechanisms as possible i needed to be able to change maximum and minimum positions of servo mechanisms at run time. additionally i needed to make default positions programmable. having this i can easily create devices, that have only limited movement, without worrying that erroneous high-level software will cause device damage.

all of the software for the uC is written in C++, while the device driver for linux uses new features of C++11 (formerly known as C++0x). thanks to this it combines fast execution times with high-level language features like abstractions.

device is a hard real time solution. it is also very noise resistant.

license

as usual, whole project is released under revised BSD license. feel free to do whatever you want with it.

note that though license does NOT obligate you to provide source code of any changes you make, community always appreciates input… and upgrades, if you'd do some. :)

electronics

the board uses ATmega8 uC as the “core” and wp>MAX232 for RS-232 signal level conversion. uC works at 8MHz. whole board is designed to have a supply voltage of +5V.

schematic diagram follows:

as can be seen hardware handles up to 17 RC-servos¹⁾. there is also an over current protection provided (fuse).

PCB has been created as a two-layer. it makes it more difficult to create at home (though no problem when ordering from PCB-making company), but it gives much “cleaner” routes and better shielding (two layers of “ground”). in order to better handle voltage fluctuations uC and MAX232 are supplied via diode and have own decoupling capacitors.

PCB layout follows:

all of the device schematics is placed in project file, available in download section. it is placed under 'eka' directory.

software

software is placed in 'src' directory, in download packages. it consists of 3 packages:

avr – this is the device code to flash uC with.
tests – these are parts of code for uC, along with test code. this is compiled locally, on linux and can be used for testing parts of code, not directly related to any hardware-related elements.
driver – this is the linux driver to communicate with the device.

linux driver

to build driver in release mode enter src/driver directory and type:

make PROFILE=release

to see what is doing one add the “V=1” option to make.

in order to tou around, use example application (see below). however to make something useful with the driver you'll probably want to create your own application. to do so use headers (gen/release_gcc_normal_default/includes/) and driver's library (gen/release_gcc_normal_default/libs/libservoctrl.a). the classes you are interested in are:

CommDevice – device that will be used to communicate (this code is thread safe).
ServoName – represents servo by the name ('a'..'q').
Servo – class allowing to control selected servo, via provided communication device.

there is also SharedPtrNotNull class. it is std::shared_ptr<> equivalent, that does not allow nullptr to be put inside. thanks to this you no longer need to check if your pointer is valid.

code is trivial. you choose servo mechanism by its name ('a'..'q'), setup device by giving a path to the RS-232 device (in my case – /dev/ttyUSB0²⁾). having this done you can create instance of the Servo class, that gives you direct control over given servo mechanism.

since version 2.0.0 responses from the device are handled via future/promise. this makes reading from port asynchronous. and so it is cheap to send commands and then wait for output on all of them. in case no response is required, it is just ok to ignore the return value(s).

though special means are taken to prevent data loss, some of the packages may arrive broken. it is expected to lost ~1% of commands begin sent to the device. data integrity makes sure that if data is being broken during sending/receiving it will be reported as an invalid and not executed.

example application

after system compiles you have a basic GUI application to toy around. its called simply PanTiltApp. when compiled in release mode it is compiled to file gen/release_gcc_normal_default/pantiltapp/pantiltapp.out. run it with 2 servos connected, that will do pan/tilt, and the camera. now you can move your camera with the mouse (just click on the empty window, that appeared and move mouse around).

protocol

huge advantage of proposed system is its protocol's clarity. it uses plain text to communicate, instead of binary ones. each position is encoded as one byte number.

each command consists of the following form:

(servo)(command)(posH)(posL)(xor)(endl)

where:

servo – name of the servo ('a'..'q')
command – what to do:
1. 's' – set servo to a given position
2. 'l' – set lowest position, that can be set
3. 'h' – set highest position, that can be set
4. 'd' – set default position to put servo in, when device is powered on
posH – high half-byte of the position (as hex)
posL – low half-byte of the position (as hex)
xor – of all of the above bytes, XORed high with low half-bytes (as hex); to make sending commands manually easier, it can be replaced with '?' char, that means “do not check checksum”
endl – any sequence of CRs and/or LFs (i.e. just hit enter on the terminal ;))

uC gives a short response to this: ”(servo)-ok” (on success) or ”(servo)-ERR” (on failure), where ”(servo)” is the servo name send by the client.

as you can see, after flashing uC you can directly play with the device via serial link and a terminal (gtkterm is nice).

download

currently, latest stable version is v2.0.0. – it can be downloaded here. all the software along with electronics schematics are there.

release notes

release notes for each version.

v2.0.0

non-blocking communication on PC side, with future-promise mechanism.
error messages passing interface has been forwarded

v1.1.0

operating up to 17 servos.
added RTS/CTS hardware flow control.

v1.0.0

operating up to 18 servos.
persistent declaration of min/max positions servo cannot overcome.
persistent declaration of default servos position.
watchdog ensuring self-recovery after a failure.
low power consumption.
overcurrent fuse protection.
less then 1.5 deg precision for typical servos.
can go beyond min-max positions of RC mechanisms with longer/shorter impulses.
RS-232, plain-text protocol.

planned improvements

though device already works quite nice, there are still few things i plan to do, namely:

add commands queue, so that commands can be sent in advance.
add possibility to move servos with different speeds.

¹⁾ v1.0.0 controlled 18, but now one pin is used for hardware flow control to prevent data loss.

²⁾ i'm using USB↔RS-232 adapter

blog:2012:01:25:1 - large scale water cooling

bartek szurgot — Wed, 25 Jan 2012 12:36:09 +0100

2012.01.25 - large scale water cooling

i just came across an short, interesting article regarding water-colled data center. but there's more – power source is also taken from natural sources. taking into account that cooling data centers and clusters itself takes about as much power as it takes to run machines themselves, free cooling is nice cost cutter.

blog:2012:01:23:2 - ppl of PL vs. ACTA

bartek szurgot — Mon, 23 Jan 2012 15:18:55 +0100

2012.01.23 - ppl of PL vs. ACTA

few days ago i wrote that ACTA is going to be signed by PL - in just a few days time. in this time ACTA has became one of the major topic in the TV and radio. ppl are fighting to stop scheduled signing of the act. many pages have been already hacked already. among the more interesting ones are attack on platforma.org¹⁾ (polish party that prime minister is from) and deface of premier.gov.pl²⁾ (prime minister's page, on the government server). the second incident's deface looked like this:

as usual, some humor was introduced into the event. see the podcast attached to the message. for outsiders - this is mixed show, formed to look like a parody of martial law's introduction speech by Wojciech Jaruzelski.

¹⁾ site is back, as of writing this post

²⁾ currently site is dead

blog:2012:01:23:1 - /proc/mem exploit

bartek szurgot — Mon, 23 Jan 2012 15:01:28 +0100

2012.01.23 - /proc/mem exploit

few days ago i came across information about error in the way kernel handled access to mem file of the processes, when running sudo applications. didn't have to wait long to see working exploit. description is worth reading for at least two reasons - it's up to date event and each step is shown and explained in the detail. another nice thing, except for the exploit itself, is the source of the shellcode (both 64-bit and 32-bit).

enjoy reading, have a nice time patching… ;)

blog:2012:01:20:1 - SOPA, PIPA, ACTA...

bartek szurgot — Fri, 20 Jan 2012 15:03:19 +0100

2012.01.20 - SOPA, PIPA, ACTA...

there is a lot of discussion on SOPA and PIPA – the open doors for internet censorship. this directly influence USA, though in a way gives a green light to the governments of other countries. mean while, i just found out, that on the local (i.e. European) grounds ACTA is going to be signed by my own country on the 26th of january this year! in a very few days…

after acceptance of the Patriot Act in the USA in 2001 i was shocked, that such a radical freedom restriction can be accepted by citizens. looks like it was just the beginning of freedom cutting down project. having all this in mind, idea of launching private communication satellites sounds better and better. it may appear that quite soon Russia or China will become the best place to live, after all.

one more thing that bothers me is the possibility that government of the USA can actually introduce such a right as a (poor) countermeasure to rapid knowledge transfer.

links:electronics - Raspbery, TP-Link WR703N

bartek szurgot — Thu, 19 Jan 2012 17:00:18 +0100

electronics

Raspberry Pi – very cheap hardware for development (ARM).
TP-Link WR703N – nice (MIPS) router, that can be used as a base for hardware projects.
http://www.pandaboard.org – interesting base board (ARM) for advanced mobile applications.
http://felektr.katalogi.pl (PL) – internet forum i appear in from time to time. guys from there have helped me with conceptual work on some electronic parts of TIER robot – thanks for your time! :)
http://avr.elektroda.eu/?q=node/8 (PL) – site dedicated to AVR microcontrollers.
http://www.kmitl.ac.th/~kswichit%20 – 8051 programmers, uCs, etc…

blog:2012:01:19:1 - base hardware

bartek szurgot — Thu, 19 Jan 2012 16:55:36 +0100

2012.01.19 -- base hardware

few days ago i've ordered very small router with integrated WiFi. not that i needed new router - its just the price/functionality factor. just take a look at the TL-WR703N's parameters. when you strip it from its casing you'll see nice, small piece of hardware (image to the right). USB 2.0, Ethernet 100Mbps, WiFi/n… and it all comes for ~22$, with free shipping. looks nice - we'll see what can be done with it, when it finally arrives. :)

being happy, waiting, and browsing i came along another interesting project - Raspberry Pi board. this in turn is a 25$, ARM board. it has enough computational power (along with GPU) to play BluRay movies. all powered with micro USB or 4xAA.

there is one more nice feature of goth devices - very low power consumption! when it comes to that, Intel's x86* family is still far behind (also counting Atom CPUs). wojek recalled recently Total Annihilation and its ARM vs. Core conflict. looks like it came true, though not that literally as in the game… ;)

as i've said many times before - we live in an interesting times… :)

blog:2012:01:13:1 - monitor on channel -1

bartek szurgot — Fri, 13 Jan 2012 10:47:40 +0100

2012.01.13 - monitor on channel -1

since some time there was a but in driver for intel's wireless chip, in linux kernel. when you setup monitor device (ex. mon0) on specific channel (say: N), when injecting it says:

mon0 is on channel -1, but the AP uses channel N.

it didn't bother to fix this any way, until one day… and it appeared that whole fix means just executing few commands. in case you spot the same problem, see this fix description (or this one). funny thing is that it boils down to applying 2 trivial patches on the kernel driver: fixing '-1' channel issue and one more.

few weeks ago wojek told me that the newest Ubuntu release already has a patched kernel. what makes me sad, is that it took so so long to make it official, vanilla kernel part…

blog:2012:2012 - monitor on channel -1

bartek szurgot — Fri, 13 Jan 2012 10:37:10 +0100

2012

2012.01.13 – monitor on channel -1