BFO ground balance - has this been done?

[Teleno]

It could work well provided the frequencies of the oscillators are not in harmonic relationship. Otherwise they simply lock to the same frequency, and you lose sensitivity. It could work with frequencies in 2:3 ratio. PLL loop filter can be made slow enough to provide for motion type detection.

I've been trying to design a voltage-controlled LC oscilator within the following specs:

Inductance < 150uH
Coil current > 100mA.
Frequency > 100KHz and < 200KHz

This is the best I could come up with (varicap-tunable Colpitts):






As varicap (varactor) I've configured an IRF610 MOSFET with gate connected to source. Its Coss capacitance varies from 2500pF to 750pF as VDs varies from 0 to 12V.

The frequency varies between 112KHz and 116KHz when the control voltage changes from 0V to 12V.

Ideas to improve this? You can use the sim file attached for LTSpice.

[ODM]

Configuring two inductance-shift search oscillators in a single search head will most likely lock them up unless they're running at high Q and at distant, non-related frequencies. Still, they should have minimal coupling to each other.

Slowing down a PLL filter to "lose and lock" reliably for a motion detector is not a simple thing, but certainly possible. Otherwise, like any regular BFO, it's problematic that there are two oscillators that should have their tempco and drift matched. There are some circuits built around thermistor wheatstone bridges and varicaps/zeners to make for trimmer-tuned LC oscillator tempco. For a LC oscillator like a search coil that has widely separated parts this probably wouldn't do unless the thermistors were with the coils.

Using two different search coil frequencies means direct comparison isn't possible. One way would be to try and match the inductance/capacitance ratios to be such that there would be an equal frequency shift to an equal disturbance in both of the paired coils, and use frequency mixing with an intermediate frequency. For example if one coil was at 100kHz and another at 120kHz, when mixing the 100kHz coil with a 20khz signal there would be 80kHz and 120kHz results, latter could be directly mixed with the second coil for beating.

Going with harmonics or multiplying would mean that the resulting multiple would have to have the same frequency shift ratio, so using 100*3 and 150*2 would need frequency shift ratios of 2:3 respectively.

Frankly I'd imagine software frequency detection and list comparison to be the sort-of-easiest way around it, the oscillators wouldn't need any relationship to each other and there would be no manual tuning except toggling calibrate and pumping the coil. Ground balance calibration could be done by raising and lowering the search head while making a table from oscillator A frequency measurements compared to oscillator B frequency measurements. Some extrapolation could probably be done on both ends without losing balance, this would likely be part of curve-fitting or plain moving average on a data point set.

This could work with something like the phase-accumulator based frequency shift detection in my (otherwise unremarkable) avr frequency shift detector. It's still BFO

[Davor]

A simple FM demodulator with great S/N is a pulse counting type. Zero-crossing detector followed by a monostable and a LPF is all it takes.

Regarding VCO-s ... it is not carved in a stone that a signal must be sine. Things become simple if you consider relaxation type oscillators.

[Teleno]

Regarding VCO-s ... it is not carved in a stone that a signal must be sine. Things become simple if you consider relaxation type oscillators.

I have considered that, however, sensitivity to inductance changes is the highest with sine oscillators. So is efficiency, a Colpitts with 100mA in the coil consumes 3mA, a relaxation oscillator would consume > 100mA.

Configuring two inductance-shift search oscillators in a single search head will most likely lock them up unless they're running at high Q and at distant, non-related frequencies. Still, they should have minimal coupling to each other.

Slowing down a PLL filter to "lose and lock" reliably for a motion detector is not a simple thing, but certainly possible. Otherwise, like any regular BFO, it's problematic that there are two oscillators that should have their tempco and drift matched. There are some circuits built around thermistor wheatstone bridges and varicaps/zeners to make for trimmer-tuned LC oscillator tempco. For a LC oscillator like a search coil that has widely separated parts this probably wouldn't do unless the thermistors were with the coils.

Using two different search coil frequencies means direct comparison isn't possible. One way would be to try and match the inductance/capacitance ratios to be such that there would be an equal frequency shift to an equal disturbance in both of the paired coils, and use frequency mixing with an intermediate frequency. For example if one coil was at 100kHz and another at 120kHz, when mixing the 100kHz coil with a 20khz signal there would be 80kHz and 120kHz results, latter could be directly mixed with the second coil for beating.

Going with harmonics or multiplying would mean that the resulting multiple would have to have the same frequency shift ratio, so using 100*3 and 150*2 would need frequency shift ratios of 2:3 respectively.

Frankly I'd imagine software frequency detection and list comparison to be the sort-of-easiest way around it, the oscillators wouldn't need any relationship to each other and there would be no manual tuning except toggling calibrate and pumping the coil. Ground balance calibration could be done by raising and lowering the search head while making a table from oscillator A frequency measurements compared to oscillator B frequency measurements. Some extrapolation could probably be done on both ends without losing balance, this would likely be part of curve-fitting or plain moving average on a data point set.

This could work with something like the phase-accumulator based frequency shift detection in my (otherwise unremarkable) avr frequency shift detector. It's still BFO

I very much appreciate your extensive reply. A lot of food for thought.

My reference for this design is the Falcon M20. It's a frequency-shift detector at 100KHz with a ferrite search head about 1'' (25.4mm) diameter. I want to place two such heads side-by side.

[Skippy]

If you're having a 3:2 oscillator ratio (it seems sensible) how about:
Reference oscillator = 60KHz, detector oscillator = 90KHz.
Have a PLL frequency tripler locked to the 60KHz ref, ie. f = 180KHz, then divide that by 2, to generate 90 KHz.
Mix this with the 90KHz detector signal. Audio difference results, as normal.
The slow time-constants you're wanting from the reference signal can be built into the x3 PLL loop.

Sorry if this is what Carl was describing, I didn't study his post too hard...

[Davor]

That's the general idea.
Although you don't have to divide signals to produce beat, quite contrary - it would provide some gain, and the resulting beat of two squares is a PWM of a perfect triangle. Quite nice sounding actually.

Only bad thing about it is a fact that it is not as simple build as it may initially seem. At that level of complexity you may have a decent PI.

[Teleno]

At that level of complexity you may have a decent PI.

The Falcon M20 can do waht no PI can, which is detecting gold dust. You need a very high frequency to do this.

I've written an Arduino sketch to measure frequencies up to 200KHz and I'll use it as a starting point of a software PLL.

[Gwil]

Maybe a Beat Balance design as in post #7, plus a slow control loop to maintain a constant low beat frequency, avoiding locking. Or maybe not. Just "thinking aloud".