Alternate Art Aperture

[Allan_ecker]

What frequency are you using?

300MHz has a wavelength of about one meter, so it might be a good(ish) choice..

But you're going to have one hell of a time dealing with reception and bandwidth. A good ADC (one with high enough frequency to read zero crossings directly off a 300MHz signal with any accuracy) will probably be fewer than 12 bits and require some special surface mount packaging.

Theoretically, you -can- use an ADC directly and pick up the signals, turning them into waveforms which can then be digitally munched into position data, but getting an ADC to pick up the type of information you're interested in would be problematic at best.

A better choice might be to get a PLL with a pin to access the "phase detector" output and then connect that to an ADC. What you've created is a primitve accelerometer. When a signal changes phase, the PLL's phase detector will momentarily spike up. With some practice, you can probably use this spike to measure movement over time, and what's better, you can low pass filter this thing to make it more paletable to typical low-bandwidth ADCs.

Of course, even recieving a 300MHz signal and scaling it up can be problematic, as you might imagine. Look up stuff related to radio kits and the like; most radio kits are about three hundred times lower frequency than your signals of interest, though, so do some research into RF recievers.

Basically, the frequencies you'd need to make "small distance" measurements would very likely require special equipment, but assuming you can overcome the -very- high frequencies involved, it can be done. This project is deceptively complex; the concept is simple but at the frequencies you're talking about the implementation isn't. Maybe you could practice with recieving AM radio (1MHz), then FM (100MHz), and finally 300MHz. Actually, you could get relatively good resolution at 100MHz, so taking apart an old FM radio might not be out of the question!

[Randyg]

(lots of stuff)

I already know this design won't work, so this post is solely for theoretical value... I've known it wouldn't work since thinking about it for 30 seconds after I first posted it. the idea works, just not the accuracy needed. :)


I wasn't trying to digitize the waveform directly... Rather, digitize an analog voltage proportional to the time between the zero crossings. Since only one in some number of cycles will have all 3 zero crossings line up, the first step would be gating on that to only look at the useful times. Then, for each pair of signals, use the polarity...
+ -
+ gnd +v
- -v gnd
and run it into a lowpass filter. Perfectly in phase signals will have 0v on the filter, then the more out of phase the signals are, the greater the voltage will be, either positive or negative depending on which leads which. (they don't cancel out since the first step only lets this work for part of the wave)

this trivializes the ADC part of it, since the ADC only needs to sample whenever I want a reading, and can take as long as it wants. The choice of 12 bits was for suitable accuracy... If I were directly digitizing the waveform, I'd be using a 1 bit adc aka a comparator. :)


The problem is engineering the polarity comparator to have a suitably square output, that responds in a fraction of the period of the signals. To do this you really only need two very high gain amplifiers, one inverting, one not... linearity is completely irrelevant; the goal is to get as square as possible. since the output won't be square, just calibrate the software to curve the readings of the ADC to linearize it.


A PLL was among the first ideas I had, but I want absolute positioning, not relative positioning... the drift/cumulative error would be unacceptable.


And, no, this isn't deceptively complex. I knew it was very complex already. :)



The problem is the time the amplifiers/comparators/etc have to work with is constant with respect to frequency... the required response time is only based on the difference in distance to the transmitters, and on a small scale like this, trying to measure even a tenth of an inch is not feasable. To measure a tenth of an inch, the time difference would only be 8 picoseconds.

This design would work well on a larger scale... If I wanted to know where I was in a building or something, this would work... it just won't work at this small of a scale. I might build the where-am-I-in-the-building tracker at 1mhz for the fun of it, but there's a huge difference between 1 foot accuracy and 1/100th inch accuracy...

Right idea, wrong application...



--Randy

[Allan_ecker]

Oooo. That turns into a real RF problem then. Nobody in my projects class came up with that one!