Wombat Pulsed Oscillation

[Tec]

The ‘Wombat Pulsed Oscillation’ uses a mono coil and capacitor (tank circuit) oscillating freely, similar to a good old BFO or a ‘VLF’ detector.


Unlike a BFO or VLF detector, it uses a ‘time of flight’ chip, the TDC7200, to measure frequency and amplitude change with unbelievable precision.

It resolves changes of 50 ps (picoseconds). At 5 kHz; that’s ~0.25 parts per million, on a single wave cycle.
(The TDC chip is used in laser rangefinders; its precision equals the time light takes to travel 1.5 cm.)


By letting the coil ring and measuring the change in frequency and amplitude, we get a clean ‘reactive’ and ‘resistive’ change (frequency SHIFT and current LOSS).


On the Lab-Bench, sensitivity is extreme.
Unfortunately, this doesn’t translate to real searching in the field because, as many of us know, ground is magnetic and/or conductive. In particular, wet salty sand looks like a small target, and soils around here typically look like iron.


However, the clean separation between shift and loss allows clean discrimination between iron and non-iron targets, as well as between targets of different conductivity.



The Wombat PO contains two ‘XIAO’ modules: high-power Arduino-compatible boards that would probably exceed the total computing power of planet Earth in the seventies, but now cost $10. They handle sound processing and a web interface, so the target ID angle can be displayed on any mobile phone.


Here is an interesting graphic showing how targets responds to different frequencies:




(Developed with AI)

[Byko]

Interesting idea. A few questions:
- The polar plot angles match 90° + atan(ωτ) exactly (124.6/152.8/168.0° at 10 kHz), so it looks calculated. Is there a plot from real coins?
- VLF IB lets you hear the full approach and departure at 1-2 ms resolution, which is how iron is told from foil. Are a few periods (~5-10) per 5..10 kHz ping enough to resolve consecutive samples?
- At 5 kHz, VLF IB is poor on small chains, which is why multi-frequency helps. How would multi-frequency work with this method?
- How do stability, sensitivity and discrimination compare with VLF IB?
- What sets 0°? The true reference is TX current; with TX voltage, ferrite calibration sets the zero. How is the drifting air baseline tracked here?
- Why does the ping stop abruptly instead of decaying?

Scope captures and drift data would help.

Based on the graphics, it looks like a theoretical idea rather than a successfully tested operating principle.

[Tec]

This is a 25cm coil running at 5.5kHz , I want to find silver coins, in soil:

Pings per second = 331.
Cylces per ping 12

It is a 'Motion' detector. Compares recent samples against the ~ 3 second average.

This is a recording of 20-cent-AUD coin, vs Iron , on park ground at 15cm, multiple passes over the target



The Ground is iron-shifty. We can see the shift-dip (blue plot) in the 20c but it is on top of the ground shift.

Here is the same data in a polar plot. X axis is 'shift', Y axis is 'Loss' Each sweep is plotted as a coloured line. The points represent the peak Loss sample. The peak Loss sample, is where the angle is calculated from the loss and shift value.

Note the AUD 5-cent piece (middle) is hidden and dominated by the ground-shift to the right.





The sound algorithm is improved by detecting a target 'loss bump' and then calculating the shift and loss angle using all of the signal in the entire bump (envelope) that last about 0.3 a second as we swing over a target. Rather than just the single point where the loss was at a peak. This algorithm is called 'delta-fit' below.

The big coin ID is cleaner, but the small target is still indistinguishable from iron unless we carefully sweep over it without wobbling the coil up and down relative to the ground.

[Byko]

Thanks for sharing your development results. It's impressive work, and an unconventional approach. Australia is well known for its heavily mineralized soil, which makes it a tough proving ground.