Bats! Status

 

Status of the Bat Project

The Bat Project ran in anger for over two years. I sometimes wondered what it was like to have free time - this project grew and grew, and proved to be rather more complex than I first imagined. No real problems there, rather my perspective of the project moved from a purely academic exercise towards a rather more pragmatic business view... Bear in mind, I do have a full-time day job (and a family!). To do this I was getting up at 4 or 5am, working until 6:30, getting the kids ready for school, waking my wife at 7, getting the train at 7:20, working for an hour on the train, working on it for an hour at lunch and again on the train home, and if I could, doing a couple of hours late in the night after everyone else wass in bed. Luckily, I only need about 6 hours or less sleep a night... Least that sounds a bit over the top, I did have 20 days off for sailing that year, plus we spent some time skiing in Åre (pronounced "Oar-er") in Northern Sweden, so it wasn't all work...

The initial idea was to produce a hand-held unit that allows a field worker to fully analyse a bat's calls and hopefully to use AI techniques to suggest what type of bat it might be. Many PhD students have tried and failed to make this approach work, but it takes more than a few failed theses to daunt me (stupidly).

Originally, I approached my alma mater, Southampton University, for advice. Due to the necessity of making money from anything (even probably making the tea), they were not very helpful, however they did point me in the general direction of other academic researchers who have been extremely helpful. One, it turns out, lived adjacent to my in-laws' farm in East Sussex, and these kindly folk aided me greatly, mostly in the area of functional requirements. Others helped me by allowing me to bounce technical ideas off them, and provided much welcome constructive criticism.

The unit became rather awesome. It contained the very latest (at that time) low power digital signal processor, running at over 150MHz, highly accurate digitising circuitry, a beautiful high-resolution graphics display, loads of memory, an expandable file system, a USB interface and much, much more... Oh, and it was only slightly larger than current top-end detectors... actually, it was smaller than many, and anyway, if offered facilities that are so far beyond anything then on the market, or even proposed, it's all bit shocking...

I had some ideas that are unique in this field, and which have never been considered practical before. As my device was very much state-of-the-art, many of the components were newly in production or only available as samples. PCB design is critical at these sorts of frequencies, and the software and hardware development environments are very very expensive - the ICE alone was USD 2,000, and the whole environment easily exceeded USD 25,000 (compilers, debuggers, ICE, JTAG programmers, arbitary function generators, digital storage 'scopes, ultra-high speed logic analyser, etc.). All the technology was low or micro-power surface mount, so I had to invest in very good SMT reflow equipment (another USD 5,000) - you can't use normal soldering irons for this stuff, many of the devices have pins that are small fractions of a millimetre apart. The PCB was a 4-layer 6thou design with a significant number of VSLI devices, including the DSP, RAM, FLASH, CPLD etc, apart from the ADC/DACs and analogue circuitry. I also converted some outbuildings into a lab with a clean supply.

The mathematics involved in the DSP, and the development of real-time filtering and analysis software that can capture, process, analyse and display up to 500,000 samples a second, was a challenge. The filtering allowed a number of extremely clever tricks, such as compensating for transducer non-linearity, time-expansion, frequency division and heterodyning. Not only that, but it allows you to take a previously recorded signal and analyse it as if it was a live bat, even if that signal has been frequency shifted, expanded or divided and converted by another unit.

The graphics display provided the user interface as well as displaying analysis results. Its high-resolution and efficiently backlit. Real-time spectral analysis and sonograms could be displayed, allowing zooming, panning, analysis and correction for a number of influences. You could select sections of calls, process them, store and recall them, play them back as human audible signals or reconvert them back to ultrasound. Selected data could also be stored and recalled at will, or downloaded to another processor via a USB link - the device appeared as a Windows disk to the PC.

Software updates were available over the USB cable from a PC, and can be downloaded from the web. This allowed completely new functionality to be added to the device at any time. Factory diagnostic access was via on-board JTAG interfaces.

Some mark II devices were taken into the field in SE Asia in March/April The devices will be taken into high-humidity jungle conditions for up to a month, and if they survive that, anything is possible. As I tried to engineer everything to IP65/NEMA4 or better, I hoped that there will be no problems. However, they could easily be eaten by a monitor lizard, or dropped from a great height...

So there you are. What started off as a simple idea one spring evening with the kids, has turned into a fairly major engineering exercise with as much, if not more, software and mathematical content as electronic engineering. A neat way to stretch my (highly limited) skills and (very tired) "intellect"...

If you have any questions about this project, drop me a line or use the guest book...

Back

Valid XHTML 1.0!