phemt colpitts: beads

That does sound really cool, no pun intended. Never heard of anything like this.

I'd be tempted to use an RF DAC to hit it with a broadband burst containing all of the expected/possible sinusoids, since it sounds like the frequencies are known in advance (right?) Then, switch to receive and capture X number of milliseconds' worth of the entire spectrum with an RF ADC at the same sample rate. The resulting buffer could then be correlated against the original transmitted data, hopefully yielding a nice set of peaks where matches were found.

Of course, I'm sure that's the first thing the company tried, so it must be a lot harder than that. This sort of problem usually comes down to finding ways to narrow the measurement bandwidth as aggressively as possible. My scheme would work at *some* level of oversampling, I'm sure, but whether it would have been feasible 10 years ago, I don't know.

Gotta say, though, rigging up 480 varactors would be a long way down my list of Stuff to Try Before Giving Up. Is there a patent that shows that block diagram?

-- john, KE5FX

It was fun stuff to work on.

Exactly what they do. It's all hidden in an FPGA and I never saw that code.

There are too many to hit them all at once. It hits a bunch of different bands. The spectra overlap and sometimes resonators will still be ringing from the last hit too - maybe even out of phase so the new hit turns them off. The details are trade secret and I don't know them in any case.

Actually they drift up to 10% depending on temperature. You also get spuria from local AM stations, wall warts, etc, so there's a substantial effort in error correction (this is the main bit of code I wrote). After error correction it's possible to figure out which spikes are real, and by knowing the coding to shift them back into the correct location. The magnitude of the shift also estimates the temperature.

Pretty much. But the raw data is pretty hard to get, and the error correction is very pretty too.

It's operationally necessary to return a reading in well under a second. Many samples are rapidly spoiled if warmed up - some must be discarded if they rise to -80C more than twice. No-one wants to spend too long in those fat mitts either.

It was never once done manually. It's surprisingly cost-effective also.

A number of patents, not sure what diagrams you'll find interesting.

CH

Well, you could pre-bias the crystal with 100 kV. Turn that off, and you'd get a nice ~1-V sine wave right away. ;)

Cheers

Phil Hobbs

I do that the other way round for calibrating DLVAs (detector/log video amplifiers). If you add a parallel inductor, you can make a crystal oscillator run close to the series resonant frequency, so that the mechanical vibration is maximized. Then when you switch it off by grounding the base of the transistor, you get just the free oscillation of the crystal. (You take the output from a small resistor in series with the crystal.)

Your average 80-MHz crystal decays about 1 dB per millisecond, which is very convenient for a ring-down calibrator like that.

Cheers

Phil Hobbs

Fun. For uniform expansion, the inductance goes like the CTE.

However, the coil is much stronger in the radial direction than the axial, so I could believe Poisson's ratio for plastic (which tends to be about 0.5, i.e. approximately constant volume) could stretch the coil out axially enough to do that.

L = a**2 N**2 / (9a + 10b) (L in uH, a & b in inches)

dL/dT = 2a da/dT N**2 / (9a + 10b) - a**2 N**2 (9 da/dT + 10 db/dT) /(9a + 10b)**2

So if a is constant, the Poisson thing about triples db/dT.

In that case

(I'm sure that Kevin could have done that much better in SuperSpice.) ;)

Cheers

Phil Hobbs