The Detecting Hub

PI detector signal processing question


I've been thinking about what information can actually be extracted from a PI detector beyond simply producing an audio response or target ID.
Suppose a PI detector receives a target response. If you had access to the raw signal rather than just the processed target ID, how much useful information could theoretically be extracted from that response?
For example, could you use things such as:
signal decay characteristics
amplitude
timing
repeated samples
coil position
direction of movement
distance from the target
to build a 2D spatial map of the target response as the coil is swept across it?

I'm particularly interested in whether repeated measurements from slightly different coil positions could be accumulated into something resembling a heat map showing the target's location and approximate shape.

I'm not asking about a particular detector or commercial product — I'm interested in the underlying technical possibilities.
How much information is actually present in a PI target response, and what could realistically be reconstructed from it if you had access to the raw data?

Would there be enough information in successive measurements to distinguish between a single compact target and a larger/irregular target, assuming the coil position and movement were accurately known?

I'll get me coat,

PI detector signal processing question

====================================

Are you serious? Who knows & who fecking cares!

Ask AI.

This is a legitimate signal-processing question, and the answer is genuinely interesting — yes, there's more information in a raw PI (pulse induction) response than typical hobbyist detectors expose, and yes, spatial accumulation into something like a heat map is a real, established technique.

What's in a single PI response

When a PI detector fires a transmit pulse and samples the decaying eddy-current signal, the raw waveform carries several somewhat-independent pieces of information:

Decay curve shape (time constant)
The eddy currents induced in a metal target decay exponentially, but the rate depends on the target's conductivity, size, shape, and (for magnetic materials) permeability. Small/thin/poor-conductor targets decay fast (tens of µs); large or highly conductive targets decay slowly (hundreds of µs to ms). Sampling at multiple delay times after the pulse (instead of one fixed integration gate) lets you fit an approximate decay time constant, or even multiple time constants if the target is a composite object. This is the basis of real "target ID" in PI machines — most consumer units just collapse this into a single number, but the underlying multi-timepoint data is much richer.

Amplitude
Scales with target size, conductivity, orientation, and (strongly, roughly as an inverse power law) distance/depth. Amplitude alone conflates size and depth, which is why it's not sufficient on its own — you need shape/decay info to help disambiguate.

Phase/waveform shape beyond a simple single-exponential
Real targets often aren't perfect single-pole decays. Irregular or multi-part targets (e.g. a cluster of small fragments, or one object with parts at different depths/orientations) can show decay curves that are a sum of exponentials rather than one clean exponential. In principle you can detect non-single-exponential character in the decay and use it as a rough indicator of target complexity — a single compact target should decay cleanly, while a diffuse or irregular one is more likely to show a composite decay signature.

Spatial accumulation — yes, this is real and used

This is essentially what ground-penetrating radar (GPR) and metal-detector array/imaging systems already do, and PI is compatible with the same approach:

Track coil position (via wheel encoder, IMU, or external positioning) as it's swept over the ground.
Log full raw response (not just peak/ID) at each position, ideally with timestamp and multiple decay-time samples.
Bin/interpolate amplitude (and decay-derived features) onto a 2D grid — this produces exactly the kind of heatmap you're describing, similar to a C-scan in GPR.
Because amplitude falls off with distance in a roughly known way (approximately following a dipole-like inverse relationship), the footprint width of the response as the coil passes over is itself informative: a small compact target produces a narrow, sharply peaked response; a larger or elongated target produces a broader, possibly asymmetric or multi-lobed response as the coil sweeps across it.
With known coil geometry and multiple passes/angles (a grid rather than single line), you can attempt a coarse inverse fit — treating the target as one or more dipole-like sources and fitting position/depth/size parameters to the spatial amplitude pattern. This is standard practice in geophysical EMI (electromagnetic induction) surveying for buried metal, and in humanitarian demining research, where multi-pass spatial data plus decay curves are used specifically to help distinguish small ordnance-like objects from clutter, and single objects from fragmented/irregular clusters.
Can you distinguish single-compact vs. large/irregular target?

With accurately known coil position and enough spatial sampling density, yes, this is realistic in principle:

Amplitude footprint shape: compact target → tight, roughly radially symmetric response profile; irregular/larger target → elongated, multi-lobed, or asymmetric footprint.
Decay curve consistency across the footprint: for a single compact target, the decay time constant should stay fairly consistent as you sweep over it (just amplitude changes with distance/offset). For a cluster or irregular object, you'd expect the apparent decay characteristics to vary as different parts of the object dominate the response from different coil positions.
Multi-angle passes: a truly compact/symmetric target gives a consistent response independent of sweep direction; an elongated or irregular one gives directionally-dependent response width, similar to how magnetometer surveys distinguish point sources from linear/distributed ones.

This is essentially a 2D-to-parametric inversion problem — same math family as source localization in geophysics, magnetics, or seismic imaging, just with a very simple diffusive/EMI forward model instead of wave propagation.

The realistic limits
Depth/size are fundamentally coupled in a single measurement (a big deep target and small shallow target can look similar) — spatial sampling is exactly what helps break that ambiguity, since footprint width scales with depth in a way amplitude alone doesn't capture.
Decay-curve target ID is inherently coarse — many different real-world objects share similar conductivity-geometry decay signatures, so classification is probabilistic, not exact.
Soil mineralization / ground effects add noise that has to be characterized and subtracted, especially for the fine decay-shape features.
None of this requires exotic technique — it's the same category of processing used in ground-penetrating radar C-scans, EMI-based UXO discrimination, and induction-balance imaging arrays in landmine/demining research literature, if you want search terms for further reading.

So: yes, considerably more can be extracted than a single-number "ID," and yes, position-tagged raw waveform accumulation into a 2D map — with the potential to distinguish compact vs. irregular targets — is both theoretically sound and something that already exists in adjacent fields.

Thanks

CLAUDE

My initial thought, after "That is a bit random!", was "Probably along the lines of ground-penetrating radar, but I do not have a foggiest idea how".

So well done to Mudripper on such a comprehensive reply which I am sure the OP will appreciate, even if 90% of it went over my head,.

Tech_leaker wrote: ↑Mon Aug 24, 2026 5:45 pm PI detector signal processing question


I've been thinking about what information can actually be extracted from a PI detector beyond simply producing an audio response or target ID.
Suppose a PI detector receives a target response. If you had access to the raw signal rather than just the processed target ID, how much useful information could theoretically be extracted from that response?
For example, could you use things such as:
signal decay characteristics
amplitude
timing
repeated samples
coil position
direction of movement
distance from the target
to build a 2D spatial map of the target response as the coil is swept across it?

I'm particularly interested in whether repeated measurements from slightly different coil positions could be accumulated into something resembling a heat map showing the target's location and approximate shape.

I'm not asking about a particular detector or commercial product — I'm interested in the underlying technical possibilities.
How much information is actually present in a PI target response, and what could realistically be reconstructed from it if you had access to the raw data?

Would there be enough information in successive measurements to distinguish between a single compact target and a larger/irregular target, assuming the coil position and movement were accurately known?

You're not the first person to wonder about this sort of thing:

https://www.noktadetectors.com/metal-de ... venio-pro/

What would potentially be a step forward would be a search coil with multiple transceivers operating simultaneously, to build up a genuine 3D representation (ie as a phased array). https://en.wikipedia.org/wiki/Phased_array

Mudripper wrote: ↑Mon Aug 24, 2026 10:02 pm ........

Thanks

CLAUDE

I think that the OP was constructed with the help of Claude or one of his cousins!

HammeredDeus wrote: ↑Tue Aug 25, 2026 4:30 pm I think that the OP was constructed with the help of Claude or one of his cousins!

Indeed,i cant pretend to know that much if at all,funnye nough i discovered Claude by listening to DETECTORMODS on youtube a PI modder,repairer and experimenter

MR

As i said it before, the future is in the soil reading who can master that will own the market.

We may see in the future:

Supra mega fast processors

Probably AI assisted discrimination

Multi-frequency VLF and pulse induction tech will continue to improve ground balancing in mineralized/salt soil, which is the practical bottleneck that limits depth more than raw sensitivity does.

If you can penetrate deeper in the soil you will reach unlimited riches (dream dream dream)

Forgot to mention you will have to dig

MR

HammeredDeus wrote: ↑Tue Aug 25, 2026 4:30 pm I think that the OP was constructed with the help of Claude or one of his cousins!

tech leaker has form
Remember his more VDI numbers = better thread?

Thank you to everyone that has replied to this question.

You have all helped me greatly and i appreciate your replies and the time taken to post the replies.