Projects

Zomojo

Tue, 22 Feb 2011 17:59:01 +1100

I started working for Zomojo in November, 2009. Of course, this page is short... Not so much out of laziness so much as the competitive advantage which slips away when you give away the company’s secrets.

Koala (PhD)

Wed, 7 Nov 2007 20:53:07 +1100

My PhD concerns power management in embedded systems. Our thesis is that systems must be modelled correctly, with all of the different facets of the computer’s operation being represented, in order to predict what actions should be taken in order to achieve power management goals.

My PhD work focusses on dynamic voltage scaling as an aspect of this. I build models of a system’s behaviour using performance counter and operating system statistics. That means that, given a prediction of the relevant statistics, we can build a model for how the system will respond to various changes in state (i.e. frequency). It turns out that the trivial models used by many DVFS researchers can lead to very sub-optimal results.

There are a number of publications on the topic. See my publications page.

Update: I finished my PhD in August, 2009. The end result was useful for single-processor systems. Since then, there has been a project to extend this to multi-processor systems.

EnergyMeas

Thu, 26 Jul 2007 16:57:51 +1000

I designed and built a small device which can take accurate measurements of the total energy used by a low power DC system while working on my PhD. The basic design consists of an MSP430 microcontroller, FTDI USB-serial converter, and an MCP3909 energy measurement IC. The latter part is super-cool: 2 phase-matched 16-bit Delta-Sigma ADCs, high-pass filters (for 50/60Hz AC measurements), floating differential inputs, programmable gain amplifier. Pretty much everything you need to be able to do decent current and voltage measurements for energy measurement.

The board has been used to instrument embedded systems, laptops, desktops and USB devices in the context of the power management work at NICTA.

Gumbase

Wed, 25 Jul 2007 03:25:48 +1000

The Gumbase is a base board for the gumstix series of small computers, which integrates a number of useful peripherals, along with component-scale energy monitoring. It was designed so that measurements of the power being used by various components could be measured in a real embedded system. The baseboard provides compact flash, ethernet, a real-time clock and a CAN interface. A substantial amount of energy monitoring hardware is inbuilt: a microcontroller and an MCP3909, but also a number of separate power supplies and associated voltage and current measuring hardware.

The PCBs have been obtained, with the manufacture of the PCBs waiting on my return from the USA.

Sunsprint Tracker

Wed, 18 Jul 2007 06:44:45 +1000

The Sunsprint Tracker is a small MPPT controller for Sunsprint cars. Sunsprint is a model solar car race for high school students. The design was influenced by my experience with the model solar boat race which in which Cranbrook School took part last year (with a very little guidance from me).

The tracker consists of a half-bridge, an AVR Tiny85 microcontroller, and some sensors. The motor is switched using teh half-bridge to some PWM, with the pulse width varying in order to maintain a voltage on the associated capacitors (and their supplying solar panel). The motor is an inductive load, and as a result, it maintains a current. By PWMing the motor, we can control the current flowing through it, and therefore control the current which the solar panel provides (to some smoothing capacitors). Sunsprint regulations stipulate that the cars must have less than 1mF of capacitance, and the tracker has been designed around that restriction.

Cars should show much better torque at slow speeds because the converter will be able to provide much more current (up to 10A) through the motor than if the solar panel were connected directly. In addition, the solar panel is able to run at its maximum power voltage under all light conditions, and at all motor speeds, giving the owners of such a device a clear competitive advantage.

Status: PCBs have been built and a prototype populated populated. The work is on hold while I am completing an Internship at Intel (presently scheduled to finish in December, 2007). Development will continue with the firmware and testing in the new year of 2008. The design will be released on this web site once it has been verified.

DMX Interface

Wed, 18 Jul 2007 06:44:36 +1000

The DMX interface is a compact, device designed to control a DMX-512 lighting network. It consists of a USB-serial converter, microcontroller, RS-485 converters and the massive DMX connectors. The device will be able to able to intercept communications between a lighting desk (or standalone) and the DMX network, modifying the lighting desk commands according to computer instruction. The original motivation for this was to connect a Wii remote to an intelligent light, but other ideas include connecting a lighting setup to an electronic drum kit.

I’ve designed it and got the boards built, but the project is on hold while I’m in the US working for Intel.

Numbat

Thu, 14 Dec 2006 10:31:49 +1100

Numbat was our entry to the RTSS student competition. The competition didn’t really have a whole lot to do with real-time systems, but it was a lot of fun anyway. The idea was that you had to control a simulated robot to navigate a maze, find a piece of cheese (since it was a “micro-mouse”) and navigate back to the start. Leonid, Stefan and I worked on a solution which involved a probability density map of the area. The map was updated by the mouse as it observed its world. We did pretty well, and won third place!

iBox

Mon, 3 Jul 2006 16:57:39 +1000

I designed the iBox over the first half of 2006. Its a PXA270 (~600MHz XScale) based board designed for video surveillance applications. It was the biggest board I’d done at the time, I designed it from the ground up, got JTAG tools working, ported the bootloader and two OSes (Linux and Iguana), wrote drivers for the devices, etc. The trickiest thing about this board is the PXA270 package: its an 0.5mm pitch device, which means you need to use 8 layers and laser drilled microvias. IO-wise it has CF, audio, ethernet, LCD, USB host, RAM, flash, MPEG4 encoder, TVP5150 video decoder, video output, and a random assortment of LEDs, relays, etc.

One piece of circuitry that I like on this board is the overlay generation circuit. It is done using the clock output of the TVP5159. This feeds the clock input on the SPI interface on the PXA270, divided down by two. The output of the SPI port controls an analogue switch which switches between the input video and a fixed voltage signal (i.e. some shade of grey, controlled by a DAC consisting of a PWM signal and a filter). We then set the SPI port up to DMA from a 384x576 bitfield, with the appropriate bits set to the input signal (i.e. all the video synchronisation comes from the input signal). This setup gives a reasonably high-resolution overlay at low cost (for high-cost, high quality, the IME6510 can be used to generate an overlay).

NSLU2

Thu, 1 Jun 2006 18:05:57 +1000

We needed a replacement board for our advanced operating systems course at UNSW. I was involved with putting together a new platform, based on some work by my friend Scott McDaid. Its essentially an IXP420 based single board computer. The IXP420 is an XScale, and therefore ARM compatible, which means that it fits in well with the ERTOS focus on embedded systems. Scott ported L4 and got network drivers up and running. Godfrey Van Der Linden got the SOS framework up and running for the AOS students. I worked with David Johnson to put the hardware together. David designed a small interface board which provides a terminal, power and reset via USB, making the NSLU2s much more practical to use. We have built 30 of these devices and they have now been used for two sessions with very positive feedback from the students.

For more information, see the pages on the AOS site which relate to the lab equipment.

NICTA Hardware Lab

Fri, 1 Jul 2005 10:31:44 +1000

In July 2005, the Kensington node of NICTA moved into its new home in the L5 building (223 ANZAC Parade). The move afforded our research group a significant amount more space. That space allowed us to set up a dedicated SMT soldering and hardware facility.

I worked with the architects to design the fit out of the lab, purchased and installed the various equipment, made various improvements to the facility, and was its prime user for the first part of its operational life.

Its major features include:
• ERSA SMT60A soldering station
• ERSA Digital 2000 station and tech/power tools
• ERSA X-Tool desoldering tool
• ERSA IR550 reflow rework station (allows rework and BGA placement)
• Fume extraction
• Assorted tools, components, and supplies (solders, fluxes, wires, glues, ...)

Sunswift 3

Sat, 1 Jan 2005 10:31:10 +1100

Sunswift 3 was the culmination of all of the ideas that we’d had over the past few races. Tim Wardrop did a lot of the mechanical engineering behind it. I was the chief electrical guy, but had a lot to do with the mechanical ideas. Between the two of us, we designed most of the car (although there were a lot of other people doing lots of the detailed work, and building it).

My main contributions to the project were:
• Guiding technical discussions, organising meetings, general butt-kicker.
• Array design and implementation
• CFD simulations (with Sam and Tim) - also needs its own page. We ran some of the biggest models in Australia at the time (15 milion elements) over a three month period on the laboratory computers in CSE. We ran about 100 models during that time, which allowed us to optimise a number of aspects of the car’s aerodynamic design as well as confirm various intuitions. I was also suggested a bunch of the ideas that ended up in the shape. The car’s excellent aerodynamic performance (a CdA of 0.11 for a 2 seat solar car) is exceptional.
• Structural spat design (with Tim) and manufacture (these really need a page to themselves - they are a very complex carbon fibre part that is manufactured in one piece and comes out of the mould with ready-to-go aero surfaces. They take the driving, braking, cornering, etc, loads). This included the way in which the wheels are mounted inside the spats.
• Various mechanical ideas for the structure: kick-boards, fillets and sealing system, bulkhead sealing system.
• The steering system (which failed during the pre-WSC testing, but not because of the overall design, as was proven by the team in the trans-continental attempt).
• General car wiring
• Battery monitoring system (with Scotty)
• Battery pack selection and design.
• MPPT selection, implementation.
• Tyre pressure monitoring
• Various random sensors (GPS, current, temperature, etc, etc)
• Strategy and telemetry software and data management (scandalhub)

PLEB 2

Wed, 2 Jun 2004 16:58:05 +1000

PLEB 2 was the basis of my PhD work for several years. I was involved in the design. I built the initial prototypes by hand (including placement of the PXA255 BGA using the school’s IR rework station - something they hadn’t tried before). I also built the necessary JTAG software, ported U-boot (the bootloader) and ported Linux 2.4, Linux 2.6 and L4/Iguana.

Rugrat

Thu, 1 Mar 2001 11:43:33 +1100

Rugrat was one of my first real electronics projects and I designed it with Luke Macphearson. It was a robot designed to run java programs on a Dallas (now Maxim) TINI board. Luke wrote the software, and I wrote the hardware. Java was the teaching language at CSE at the time, and the idea was that it could be used as a teaching platform. It could follow lines, run a web server, etc, etc.

Sunswift Electronics

Mon, 1 Jan 2001 11:38:31 +1100

So, there’s a lot to write about Sunswift’s electronics. Consider this a placeholder (which may eventually point to its own open-source project).

Sunswift 2.4

Sat, 1 Jan 2000 11:39:13 +1100

Sunswift 2.4 was essentially a solar array. We damaged Sunswift 2.3’s array and needed to replace it. For nearly two years we worked to build a new array. We came up with a novel technique which has a lot of merit, but needs a lot of refinement. There’s a publication about it on my web page. Consider this a placeholder.

Sunswift 2.3

Fri, 1 Jan 1999 11:38:44 +1100

Sunswift 2.3 was my first solar car project, and completed 1.5 races. It entered the world solar challenge in 1999, and made it half way through the 2000 Sunrace before being run off the road by an enthusiastic onlooker. Bugger.

Again, a placeholder. There’s a lot more to write about this.